Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Characterizing the genetic and environmental contributions to ocular and central nervous system health
(USC Thesis Other)
Characterizing the genetic and environmental contributions to ocular and central nervous system health
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
i
CHARACTERIZING THE GENETIC AND ENVIRONMENTAL CONTRIBUTIONS
TO OCULAR AND CENTRAL NERVOUS SYSTEM HEALTH
by
Darryl Reth Nousome
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(EPIDEMIOLOGY)
August 2019
ii
Dedication
To my parents, Sophia Var and Wood Nousome, who crossed the ocean and survived a
new world as refugees. Thank you for instilling in me the value of education and always
supporting me in this process. To my family, I am privileged to be here when you could
not be.
iii
Acknowledgements
I wanted to thank my advisor Dr. Roberta Mckean-Cowdin for her mentoring and
support. Your guidance, both professionally and personally, has been invaluable. To my
committee members, Dr. Jim Gauderman, Dr. Xuejuan Jiang, and Dr. Grace Richter,
thank you for all your insight and encouragement.
To my friends, thank you for your years of support. Lastly, Sydney, my friend and wife,
your drive inspires me every day and that drive made this dissertation possible.
iv
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables vi
List of Figures viii
Abbreviations x
Abstract xiii
Chapter 1. Introduction 1
Specific Aims 15
REFERENCES 16
Chapter 2. Retinal Nerve Fiber Layer Thickness in Normal Latinos, Chinese
Americans, and African Americans: A Multiethnic Study of Population-based Data
24
ABSTRACT 24
MATERIAL AND METHODS 26
Study design and participants 26
Demographic and clinical data 27
Optical coherence tomography imaging 28
Inclusion and exclusion criteria 28
Statistical methods 29
RESULTS 30
DISCUSSION 34
REFERENCES 60
Chapter 3. A Genome-Wide Association and Admixture Study of Retinal Nerve
Fiber Layer Thickness in a Latino Population 63
ABSTRACT 63
MATERIALS AND METHODS 66
Study design and participants 66
Optical coherence tomography imaging 67
Genotyping and imputation 68
Statistical methods 69
v
RESULTS 73
Study population 73
Genome-wide association mapping results 73
Heritability and Admixture Mapping 75
DISCUSSION 78
REFERENCES 97
Chapter 4. Evaluation of the Main and Joint Associations between Ambient Air
Pollution and Retinal Vein Occlusion 101
ABSTRACT 101
MATERIAL AND METHODS 104
Study design and participants 104
Vein occlusion measures 105
Air pollution estimation 105
Genotyping and gene-environment interaction 107
Statistical methods 108
RESULTS 108
Study population 108
Effects of air pollutants and RVO 110
DISCUSSION 110
REFERENCES 128
Chapter 5. Discussion 132
Summary of findings 132
Future directions 134
Public health impact 136
REFERENCES 138
vi
List of Tables
Table 2.1 Analytical Cohort Comparisons Between the Full Dataset to Dataset with
Complete Covariate Data 40
Table 2.2. Comparison of Optic Nerve Head Parameters and Retinal Nerve Fiber Layer
Thickness Measurements Across the Three Study Ethnicities. 42
Table 2.3. Univariable and Multivariable Regression Results of Average RNFL
Thickness. 43
Table 2.4. Stepwise Forward Regression Coefficients of Covariates on Ethnicity Effect
with Average RNFL Thickness 45
Supplementary Table 2.1: Analytical Cohort Comparisons Between Full AFEDS Cohort
and Cohort with Complete Covariate Data 46
Supplementary Table 2.2. Sensitivity Analysis of Optic Nerve Head Parameters and
Retinal Nerve Fiber Layer Thickness Measurements Across the Three Study Ethnicities
47
Supplementary Table 2.3 Comparison of Retinal Nerve Fiber Layer Thickness
Measurements Across the Three Study Ethnicities at the Fifth Percentile 48
Supplementary Table 2.4. Multivariable Linear Mixed Regression Models Across
Quadrants Using the Model Developed for Average RNFL Thickness 49
Supplementary Table 2.5. Semipartial Correlation Coefficients for Effect of Overall
Model and Individual Variables on Contribution to RNFL Thickness Using Linear Mixed
Models 50
Table 3.1. Analytical Cohort of the Los Angeles Latino Eye Study Compared to
Participants with Imaging Data Only 83
Table 3.2. Top Summary Results for Association Mapping of Variants and RNFL
Thickness Across All Quadrants in the Los Angeles Latino Eye Study. 84
Table 3.3. Overall Genetic Heritability Estimates of RNFL Thickness and Effect of
Global Ancestry on RNFL Thickness in the Los Angeles Latino Eye Study 85
Table 3.4. Summary Results of Local African Ancestry Associated with RNFL thickness
in the Los Angeles Latino Eye Study 86
Table 3.5. Validation of Top SNPs P-values and the Association with RNFL Thickness in
the Los Angeles Latino Eye Study and the Chinese American Eye Study 87
vii
Table 4.1. Demographic characteristics of Any RVO and non-RVO in the Combined
Multiethnic LALES and CHES 119
Table 4.2. Associations Between Air Pollutants and RVO, Adjusted for Age and Sex
Only 120
Table 4.3. Multivariable Associations Between Air Pollutants and RVO Adjusted for
Age, Sex, Hypertension, Diabetes, and Smoking Status 121
Table 4.4. Joint Air Pollutant and Thrombosis Genetic Variants and Risk of RVO 122
Supplementary Table 4.1. Geocoding Match Stratified by Ethnicity 124
Supplementary Table 4.3. Distribution of Air Pollutants Combined and Stratified by
Ethnicity 126
Supplementary Table 4.4. Sensitivity Analysis of Univariable Model of RVO and Air
Pollutants Including all Geocoding Qualities 127
viii
List of Figures
Figure 1.1. The retina is an extension of the CNS 2
Figure 1.2. OCT in comparison to other imaging technologies 3
Figure 1.3. Multilayered view of the retina 5
Figure 2.1. Flow Chart of the Overall Study 51
Figure 2.2. Line Series Plot of RNFL Thickness by Clock-Hour Stratified by Ethnicity 52
Figure 2.3. A) Linear Mixed Regression Line Plots of Average RNFL Thickness with
Age Across Three Ethnicities. B) Linear Mixed Regression Line Plots of Quadrant
Specific RNFL Thickness with Age Across Three Ethnicities 54
Figure 2.4. Clock Hour of RNFL Thickness Decrease for Each 10-year Increase in Age
Separated by Ethnicity 56
Supplementary Figure 2.1. Line Series Plot of RNFL Thickness Stratified by Ethnicity in
Participants with Suspect or Untreated Glaucoma 57
Supplementary Figure 2.2. Clock-Hour of RNFL Thickness Decrease for Each 10-year
Increase in Age Separated by Ethnicity in Participants with Suspect or Untreated
Glaucoma 59
Figure 3.1. A) Global Ancestry for All Participants in the Los Angeles Latino Eye Study.
B) Local Ancestry for Three Randomly Selected Individuals. 88
Figure 3.2. Manhattan Genome-wide Association Plot for A) Nasal Quadrant B) Inferior
Quadrant RNFL Thickness 90
Figure 3.3. Manhattan Gene-Based Plot for A) Nasal Quadrant and B) Temporal
Quadrant RNFL Thickness 91
Figure 3.4. Manhattan Local Ancestry Plot for Inferior Quadrant RNFL Thickness and
Local African Ancestry 92
Supplementary Figure 3.1. Manhattan A) Genome-Wide Association Plot for Average
RNFL Thickness and B) Gene-Based Plot for Average RNFL Thickness 93
Supplementary Figure 3.3. Manhattan Genome-wide Association Plot for Temporal
Quadrant RNFL Thickness 95
Supplementary Figure 3.4. Manhattan Gene-Based Plot for Inferior Quadrant RNFL
Thickness 96
ix
Figure 4.1. A) NO2 and B) PM2.5 Concentration in the Combined Multiethnic Los
Angeles Latino Eye Study and Chinese American Eye Study Population 123
x
Abbreviations
AD Alzheimer’s disease
AFEDS African American Eye Disease Study
AL Axial length
AMD Age-related macular degeneration
BMI Body mass index
BP Base pair
BRVO Branch retinal vein occlusion
CDR Cup-disc area ratio
CHES Chinese American Eye Study
CHR Chromosome
CRVO Central retinal vein occlusion
DALY Disability-adjusted life years
DM Diabetes mellitus
DR Diabetic retinopathy
EA Effect allele
ELM External limiting membrane
FDA United States Food and Drug Administration
FUMA Functional Mapping and Annotation
GCL Ganglion cell layer
GCTA Genome-wide Complete Trait Analysis
GRM Genetic related matrices
GWAS Genome-wide association studies
xi
GxE Gene-environment
ILM Internal limiting membrane
INL Inner nuclear layer
IPL Inner plexiform layer
LALES Los Angeles Latino Eye Study
LMM Linear mixed models
MAGMA Multi-marker Analysis of GenoMic Annotation
MS Multiple sclerosis
NEI National Eye Institute
NO2 Nitrogen dioxide
O3 Ozone
OCNS Ocular central nervous system
OCT Optical Coherence Tomography
OD Right
ONH Optic nerve head
ONL Outer nuclear layer
OPL Outer plexiform layer
OS Left
PBWT Positional Burrows-Wheeler Transform
PD Parkinson’s disease
PM Particulate matter
PM10 Particulate matter ≥10µm
PM2.5 Particulate matter >2.5µm & <10µm
xii
RGC retinal ganglion cells
RNFL Retinal nerve fiber layer
RNFL Retinal nerve fiber layer
RPE Reginal pigment epithelium
RVO Retinal vein occlusion
SD Spectral domain
SR
2
Semipartial correlation coefficients
TD Time domain
VCDR Vertical cup-disc ratio
xiii
Abstract
Population-based studies were conducted to identify demographic, clinical, and
genetic determinants of ocular central nervous system (OCNS) health. We assessed
various aspects of the OCNS through the retina by examining the optic nerve head
(ONH), retinal nerve fiber layer (RNFL) thickness, and retinal vein occlusions (RVO) in
a multiethnic population.
In the first study, we pooled the largest number of racial/ethnic minorities to
describe the distribution of the ONH and RNFL thickness. Chinese-Americans had higher
RNFL thickness across most clock hours compared to their African-American and Latino
counterparts, after adjusting for age and sex. Chinese participants also had larger overall
disc area, cup-to-disc ratios, and vertical cup-to-disc ratios than their African-American
and Latino counterparts. Age and ethnicity explained the greatest proportion of variance
in RNFL thickness. These findings suggest that African-American and Latino eyes are
aged by approximately 20 and 5 years, respectively, with RNFL thickness compared to
Chinese-Americans.
In the second study, we utilized whole genome genotyping genetic data and 1)
determined heritability of RNFL thickness, 2) identified single genetic variants associated
with RNFL thickness, and 3) performed admixture mapping of RNFL thickness in the
Latino population. Genetic heritability of RNFL thickness ranged from 10%-46% across
all quadrants. Using standard association mapping, two genetic variants were associated
with RNFL thickness in the nasal and inferior quadrant. Using admixture mapping, we
identified that in the 17q21 loci, increasing proportions of African ancestry were
associated with thinning of the RNFL.
xiv
Lastly, we examined the relationship between ambient air pollution and traffic
density on the risk of RVO. There were no statistically significantly associations of the
main effects of air pollutants and risk of RVO, but replicated the hypertension and age
associations with RVO. The association between traffic density and risk of RVO was
modified by a fibrinogen (FGG) genetic variant.
There have been no population-based studies to examine the environmental and
genetic factors associated with these measures of OCNS health. These findings address
the gaps in the literature regarding clinical, demographic, and genetic factors of OCNS
health as measured through the retina in a multiethnic population of Latinos, Chinese-
Americans, and African-Americans.
1
Chapter 1. Introduction
Neurodegeneration refers to decline of the ocular central nervous system (OCNS),
either structurally or functionally, of neuronal cells. While neurological diseases and
ophthalmologic conditions may appear as distinct, disparate events, epidemiological
studies and molecular biology suggest overlap between their etiology (1).
Neurodegenerative diseases include epilepsy, dementia and Alzheimer’s disease (AD),
Parkinson’s disease (PD), multiple sclerosis (MS), and migraines. Ophthalmologic
disorders, like glaucoma, diabetic retinopathy (DR), and age-related macular
degeneration (AMD) are diseases associated with degeneration of the optic nerve and
retina, tissues that are extensions of brain tissue and considered part of the OCNS.
By finding common determinants pathways that underlie these disorders of the
OCNS, which together constitute a large public health burden, including visual
impairment, decreased quality of life, and increased mortality, we may better prevent
their occurrence. Previous epidemiologic studies of the OCNS rely on comparing
diseased to non-diseased participants to ascertain risk factors, but by identifying genetic
and environmental determinants of the healthy OCNS, we can provide insight onto their
natural course, and may even provide predictive value on when disorders arise. This
dissertation proposes three projects to identify the genetic and environmental
determinants of disorders and neurodegeneration within the OCNS system. The first
project will use optical coherence tomography (OCT) to examine the differences of and
determinants of retinal structures in a healthy population using a multiethnic population-
based study. The second project will expand on OCT imaging to determine the genetic
contribution toward OCNS structure. The third project will assess the variation in air
2
pollution and determine if host genetic variation modifies the effect of air pollution on
retinal structures.
Components of the OCNS
The OCNS is composed of the
brain, the spinal cord, and the optic
nerve. There are two classes of cells in
the nervous system, neurons, which
process and transmit signals, and glia,
which support the neurons with
structural and metabolic functions.
Anatomically and developmentally, the
eye, and specifically, the retina, is an
extension of the OCNS. The retina consists of retinal ganglion cells (RGCs), neurons that
form the optic nerve; when light hits the eye, the first layer of cells, rods and cones,
convert the information into signals that passes to RGCs, seen in Figure 1, used from the
Salk Institute (2). The retina is a window to the CNS and linking the eye to the CNS
could improve the understanding of the two systems, especially when dysfunction arises.
The eye as a window to the ocular central nervous system
Changes that affect the brain, including neuropsychiatric changes or cognitive
decline, may be first recognized through changes in the eye. This association has been
pursued as the retina is an extension of the brain and OCNS. Specifically, RGCs within
the retina, are extensions from the diencephalon of the brain (3). The optic nerve, or the
central location that transmits light and images to the brain, consists mainly of axons and
Figure 1.1. The retina is an extension
of the CNS
3
other glial cells. Furthermore, the immune responses that have been characterized in the
eye, are similar to that seen in the brain and spinal cord (3).
To characterize the changes in the OCNS, studies have used magnetic resonance
imaging (MRI) to view white matter and grey matter volumes, in both normal and
diseased populations. MRI is an imaging technique that relies on magnetic fields unique
to specific cells and tissues. While used routinely as a diagnostic tool, epidemiologic
studies have also relied on MRI to provide an overview and insight of the OCNS. For
instance, studies have examined cognitive decline in those with AD and have found
changes in MRI volumes of gray matter in the brain. (4). Others have assessed variation
of MRI assessed measures, specifically white matter volume (axons), in normal
populations and identified changes associated air pollution or other host characteristics
(5–8). Genome-wide association studies (GWAS) in MRI derived brain volumes have
shown brain volume variation is associated with genetic variation, suggesting that these
genetic risk variants can affect normal brain tissue, including whole brain and white
matter volumes (9). One challenge in conducting MRI in epidemiological population-
based studies is performing MRI in a cost and time-effective manner. OCT, Therefore,
imaging and examination of the eye could provide better insight, at a level suitable for
population-based epidemiological studies, that
MRI or other imaging modalities alone could not
provide.
Figure 1.2. OCT in comparison to
other imaging technologies
4
Ocular imaging-optical coherence tomography imaging
OCT is a relatively new imaging technique that works similarly to ultrasound by
using and reflecting light instead of sound to perform scanning of tissue. The technology
works in tissue due to heterogeneity of the medium, which has different refractive index.
The technology was adapted for use in the retina in 1991 (10,11). Like MRI and other
ultrasound techniques, it can be conducted in vivo, but provides much higher resolution,
at ~10 um resolution (Figure 1, adapted from http://obel.ee.uwa.edu.au).
Generally, the principal of OCT is based on interferometry, which relies on
combining information from multiple waves and then detecting the differences between
them. Two waves with the same frequency will generate a larger wave (addition of
waves), and two waves in opposite phase will subtract. Light travels from a source and is
then split into a reference arm and a sample arm. The reference arm directs the source
light to a reference mirror, and the light from the sample arm directs its source light to the
sample (retina). Both directed light returns to a detector, where their information is
combined to generate an image (12). An amplitude scan, or A-scan, is created, which is
the optical reflectance through the depth of the tissue (13). The A-scan is the primary
form of scanning, which measures a direct line from the light source to the retina.
Multiple A-scans can be combined across a tissue to present a B-scan, which is a cross-
sectional image of the tissue of interest (13).
OCT technologies vary, with earlier instruments relying on time domain (TD)
OCT and newer instruments using spectral domain (SD) OCT. The technologies are
similar except for a few variations. TD-OCT relies on a moving reference mirror that
oscillates and takes multiple scans to account for the differing depths of tissue (11). The
5
newer SD-OCT approach uses all frequencies of light with a stationary reference mirror
and can acquire all information in one A-scan (14).
While optical parameters, including
resolution and scanning speed, play an
important role in characterizing the retina,
the multilayered portion of the eye and the
heterogeneity of the individual tissues
within each layer that possesses different
reflective properties complicate this process
of imaging. OCT imaging can provide relatively straightforward quantitative measures of
the retina and the surrounding structures, but there remain challenges in determining
these various layers of the retina. The retinal layers are composed of 10 layers that consist
of (starting from the outer most layer) the reginal pigment epithelium (RPE),
photoreceptor layer, external limiting membrane (ELM), outer nuclear layer (ONL), outer
plexiform layer (OPL), inner nuclear layer (INL), inner plexiform layer (IPL), ganglion
cell layer (GCL), and retinal nerve fiber layer (RNFL), and internal limiting membrane
(ILM), as seen in Figure 3 adapted from Kafieh (15,16).
OCT in epidemiology
There are many reports and epidemiological studies of OCT that have
characterized disorders of the OCNS. In ophthalmological studies, OCT is used routinely
for many optic conditions, including diagnosing glaucoma through examining
deterioration of the RNFL. (17,18). Similarly, in studies of neuropsychiatric disorders,
OCT was used to characterize the RNFL and found to be thinner as cognitive scores
Figure 1.3. Multilayered view of the
retina
6
decreased in those with AD (19–21). In other OCT studies of the OCNS, thinning of the
RNFL was seen in a meta-analysis of those with MS compared to healthy controls (22).
While 10 layers of the retina are identifiable, the focus of most analyses are the
“RNFL”, which is a combination of the ILM, GCL, and actual RNFL. These parameters
vary depending on the manufacturer and the respective algorithms used. Segmentation is
the process that delineates layers during imaging and analysis of OCT. Segmentation,
therefore, is not only an imaging consideration, but a computational effort that relies on
algorithm development to draw boundaries and better describe these individual layers.
For instance, one algorithm will search for borders of each layer by searching for the first
highly reflective increase after preprocessing by scaling individual pixel intensities,
known as an adaptive thresholding technique (23). There are concerns with the
differences in imaging and/or segmentation that occur when devices come from different
manufacturers that rely on different algorithms. Future studies will require better
segmentation protocols to better delineate and examine retinal topography. Generally,
there are differences between SD and TD-OCT imagers and their respective methods in
imaging and segmentation (24). This can differ between imagers using the same platform.
For instance, Carl-Zeiss will perform both types of scans, but their segmentation
algorithm only combines the GCL and IPL, while the Optovue protocol includes RNFL
in their combined GCL and IPL layers (25).
The human retina contains ~1 million RGCs, and these cells are concentrated
within the macula (26). Studies have shown that OCNS conditions, like glaucoma, MS,
and AD, are associated with degeneration of the GCC and circumpapillary RNFL
(17,27,28). There are fewer studies of normal populations, which are important in
7
examining the retinal structures unrelated to disease process. These studies have shown
that age, gender, and BMI may be predictors of lowered numbers of RGCs measured by
the RNFL (29–33). The limitations of these studies are that they tend to be restricted to
case-control studies, are not population-based normal studies, and/or small sample sizes.
Epidemiology of ophthalmologic disorders and neurological disease
Neurological diseases and ophthalmologic conditions disorders, account for a
substantial burden of disease. The World Health Organization estimates that
neuropsychiatric diseases, including epilepsy, dementia and AD, PD, MS, and migraines
are 2% of the global burden of disease (34,35). There will be an estimated 13.2 million
and 1.5 million people living with AD and PD, respectively in the United States by 2050
(34). These neuropsychiatric diseases account for 10.4 disability-adjusted life years
(DALY), which represents the number of years of life due to a disease (36).
While ophthalmologic disorders, especially those related to the central nervous
system, have lower DALYs, they tend to occur in the presence of other co-morbidities
like visual impairment (36). These diseases include glaucoma, AMD, and DR and affect
3 million, 2 million, and 7.7 million cases in the United States, respectively. (37).
Glaucoma is the leading cause of global blindness, and it is estimated that there will be
111.8 million people diagnosed with some form of glaucoma in 2040 (38). Globally, it is
estimated that 295 million people are visually impaired, and ~15% of these can be
attributed to the mentioned conditions (39).
Specifically, one retinal cardiovascular disorder is retinal vein occlusion (RVO).
RVO is a retinal vascular disease that may cause sudden, unilateral loss of vision. Like
thrombosis in other regions, the pathogenesis of RVO may be due to three host systemic
8
changes known as Virchow’s triad: venous stasis, vessel wall damage, and
hypercoagulability (40). There are two types of RVO, branch retinal vein occlusion
(BRVO) and central retinal vein occlusion (CRVO). While BRVO is more common,
CRVO tends to be more symptomatic, which can present with the sudden loss of vision.
BRVO may be asymptomatic, however, patients may have blurred vision if the occlusion
occurs near a visual field. Additionally, visual loss can result from the associated macular
edema and macular ischemia (41). The prevalence of any RVO is estimated at 5.20 per
1000 for any RVO, 4.42 per 1000 for BRVO and 0.80 per 1000 for CRVO (42). The
highest prevalence of RVO has been found among Hispanics (sex and age adjusted
incidence rate = 6.9 per 1000), followed by Asians, Blacks, and then whites (3.7 per
1000) (42) with an estimated 16 million adults affected globally.
Determinants of ophthalmologic disorders and neurological disease of OCNS
Multiple cohorts and studies have been assembled to identify risk factors in
ophthalmologic disorders of the OCNS. Analyses of glaucoma have found host risk
factors that include variations in race, intraocular pressure, systemic hypertension,
perfusion pressure, and family history of glaucoma (43,44). In studies of AMD, large
meta-analyses that included data across multiple continents consistently found several
risk factors, including age, smoking, family history, BMI, cardiovascular disease,
hypertension, and plasma fibrinogen as risk factors (45). Similarly, consortiums of DR,
have identified that risk factors include longer duration of diabetes and poor glycemic
and blood pressure control (46). In other neuropsychiatric disorders, AD is positively
associated with hypertension, diabetes, and fewer years of education and negatively
associated with use of nonsteroidal anti-inflammatory drugs, wine consumption, coffee
9
consumption, and regular physical activity (47,48). The risk factors for RVO are similar
to other vascular diseases, including being hypertensive, being diabetic, higher body mass
index, and smoking (49,50).
GWAS are the standard approach to agnostically identify whether many genetic
variants across the entire genome may be associated with disease. GWAS and other
molecular biology approaches have been utilized to examine various OCNS outcomes,
including AD, PD, DR, glaucoma, and AMD. These studies combine tens of thousands of
cases and compare them to a healthy set of controls to detect potential variants associated
with risk. These studies have been successful in identifying common variants that
contribute small increases in risk (48,51–54). In some instances, there may be one gene,
like APOE, that is responsible for a large proportion of risk of AD (48). In studies of
AMD, 34 loci have been discovered that are associated with increased risk, though a
problem that persists across studies is that loci function remain unknown (55).
While the genomic loci have been discovered using GWAS or other candidate
gene studies, further functional follow-up will need to be completed to understand their
role in OCNS diseases. Additional studies may benefit from testing for gene-environment
(GxE) interactions between an exposure and potential effect modification by a genetic
variant on an outcome. This may allow for identification of a subset of participants that
may be at higher risk out a disorder (56). Additionally, potential functional genetic
variants may be missed with GWAS because of the agnostic nature of its study design.
For instance, a previous GxE study in understanding coagulation suggests that association
between PM10 and fibrinogen in blood is modified by the genotype of a fibrinogen gene,
FGB. Subjects homozygous for the minor allele had a fibrinogen response eightfold
10
higher compared to subjects homozygous for the major allele (57). Therefore, it is
important to link identified environmental and genetic risk factors of these disorders on
their function in normal tissue and healthy participants.
Air Pollution and the OCNS
There is substantial evidence that exposure to air pollution is associated with poor
health outcomes, including, premature all-cause mortality, as well as mortality due to
heart disease, cognitive deficits, brain abnormalities, and a host of other outcomes
(58,59). Air pollution is a likely inflammatory compound as a risk factor for many OCNS
diseases. Air pollution can be described as the complex composed of particulate matter
(PM), ozone, carbon monoxide, sulfur oxides, nitrogen oxides, and other gases, volatiles
organic compounds (including benzene), and metals (like lead, manganese, and iron).
These compounds come from both human-made and natural sources. PM within air
pollution is defined by its size (where PM2.5 represents particles smaller than 2.5 µm). In
the United States, an ~29 million people are exposed to PM10, and ~88 million people are
exposed to PM2.5 (60).
Neurodegenerative disorders and cognitive decline have been examined in
relation to the effects of air pollution. Observational studies of animals living in highly
polluted areas have lesions that resemble AD-like pathology, where neuritic plaques and
neurofibrillary tangles were observed in post-mortem dogs and mice (61–63). In human
epidemiological studies, traffic-related pollutants were associated with decreased
cognition in men and women (64,65). These effects are not restricted to older adults, as
children living in higher regions of NO2 had lower scores of motor function and working
memory (66,67).
11
There are multiple pathways that could be responsible for the role of air pollutants
and its effect on the brain and overall OCNS. The constituents of air pollution are able to
reach and impact the OCNS (68,69). Generally, two mechanisms can be described (70).
The first is systemic inflammation caused by indirect exposure, where circulating
cytokines affect peripheral innate immune cells and move on to the OCNS (71). Well
characterized cytokines, like TNFα, IL-1β, NF-Kβ, are able to translocate to the OCNS,
following air pollution signaling induced in order parts of the body (72,73). Once part of
the OCNS, these cytokines cause neuroinflammation, neurotoxicity, and cerebrovascular
damage (74,75). The other mechanism of air pollution on the OCNS is through direct
exposures. Fine particulate matter may be inhaled and later translocated to circulation
system or move directly to the brain (76,77). The nasal olfactory pathway is the area of
entry, and PM can reach the brainstem and hippocampus (78). PM was seen in post-
mortem brains, specifically the frontal lobe, in healthy adults who had died suddenly
(79). Once part of the OCNS, PM can cause neuroinflammation or neurotoxicity by
inducing innate immunity or begin the production of pro-inflammatory cytokines in the
olfactory systems (80,81).
Recent studies suggest that air pollution may affect certain cells specifically,
namely glial cells, like astroglia and microglia (68,70). Microglia are the immune system
within the OCNS, these cells are macrophages that respond to foreign compounds in the
brain (82). Microglia act in response to PM, both in vitro and via systemic inflammation.
(73). Microglia change in both neuropsychiatric and ophthalmologic disorders, including
AD and glaucoma (83,84). When activated, microglia can induce cell apoptosis of
ganglion cells. The other type of cells, astroglia, provide structural support, supply
12
nutrients to the nervous tissue, and help repair the brain in trauma by walling off
damaged areas during inflammation and injury. There is evidence to suggests that ozone
exposures are associated with increases in inflammatory cytokine production in
astrocytes (85).
The effects of environmental and genetic contributions to OCNS disorders and
OCNS structural changes stem from changes in to the number of RGCs, either measured
through the RNFL or the RGC complex in the macula. As shown previously, there are
multiple pathways through which air pollution manages to affect the OCNS. Specifically,
the retinal structures may be prone to damage by oxidative stress due to systemic
inflammation. Free radicals created through adsorbed pollutants can attack the membrane
of retinal and induce lasting damage because the retina is a sensitive structure (86–88).
While a thrombosis is the underlying pathology of RVO, thrombotic events are
also the underlying pathology for other cardiovascular events, including coronary heart
disease and stroke. The presence of RVO is also associated with an increased risk of
cardiovascular mortality and stroke (89). Cardiovascular disease is the leading cause of
death in the United States (90), and recent studies have suggested the role of ambient air
pollution and its association with cardiovascular disease, and specifically thrombosis
(59,91–93). The components of ambient air pollution, including PM, ozone, nitrogen
oxides, and others, may induce pollution-mediated thrombosis and cause additional
epigenetic changes (93). Acute exposure to pollutants, specifically PM, is associated with
changes in fibrinogen and other coagulation proteins, seen both in controlled
experimental and observational studies (94–96). Long term epidemiological studies have
shown similar changes in the process of clotting, however additional studies have
13
suggested long-term systemic inflammation caused by air pollution on changes in
vasculature and fibrinogen (97). Another mechanism of air pollution on the
cardiovascular system is the role of oxidative stress (98). Air pollutants induce oxidative
stress when inhaled by triggering an initial inflammatory response and then causing
secondary systemic inflammatory effect (99), which disrupts vasodilatation and inhibition
of platelets (98). Overall, these pollutants are may disrupt haemostasis, the process of
maintaining and preserving blood circulation, and systemically shifts the vasculature
towards a thrombotic nature (93). This suggests that identifying genetic and
environmental determinants of the OCNS structures as seen in the retina may provide
more insight onto their natural course and may even provide predictive value on their
manifestations as disorders.
Shared risk factors
Although a crude understanding of the disorders affecting the OCNS might
suggest that there is no relation between ophthalmologic diseases and neuropsychiatric
conditions, there is some evidence that they may be linked. Meta-analysis of 29 studies
suggested that those with PD have decreased risk of any cancer (100). Age, race, and
gender are common demographic characteristics linked to all OCNS outcomes, and while
not likely causal factors themselves, they may serve as proxies for other shared
modalities. There is a role of smoking, and the components of smoking, like polycyclic
aromatic hydrocarbons, that are associated with risk of PD, AMD, and RVO (42,45,101).
There are numerous genes and genetic pathways associated with risk that are
shared between all OCNS disorders, including genes like ATM and CDKN2B-AS1, which
are mutated in many cancers, PD, and glaucoma (52,102,103). Recent studies have
14
identified the APOE gene mutation in AMD, which is shared between AMD and AD
(104). Another common pathway between all OCNS disorders is the complexity
inflammatory responses. In cases of these diseases, degenerative tissue are stimulants of
inflammation, leading to local and systemic damage (105–107). These studies have
identified potential genetic variation in inflammatory response genes on their risk of AD,
glioma, and AMD, respectively.
There are no large population-based epidemiological studies that have examined
the genetic and environmental determinants of the OCNS as measured through retinal
health. This dissertation proposes to understand the contributions of genetic susceptibility
and air pollution on the OCNS structures and risk of diseases. The aims of each project
are outlined below.
15
Specific Aims
Study 1: Clinical determinants and race/ethnicity are associated with OCT
measured retinal structures in a multiethnic population without eye disease.
Aim 1: Determine if clinical and demographic determinants, including gender, age, BMI,
lipid levels, diabetes, and race/ethnicity are associated with various measures of the
OCNS as measured through OCT
Aim 2: Determine whether the association with clinical determinants differs across ethnic
groups African Americans, Chinese Americans, and Latino Americans
Study 2: Genetic variation is associated with OCT measured retinal structures in a
multiethnic population without eye disease.
Aim 1: Perform GWAS to determine if genetic variation is associated with OCNS as
measured through OCT in a Latino American population
Aim 2: Determine if genetic variation, specifically exonic variation, is associated with
OCT measured OCNS structures in a Chinese American population
Project 3: Ambient air pollutants, PM2.5, PM10, ozone, NO2, and traffic density, are
associated with RVO. The risk of RVO is modified by host genetic variation.
Aim 1: Ascertain air pollutant and traffic density levels of a multiethnic population
Aim 2: Examine the relationship between air pollutants and RVO
Aim 3: Determine effect modification between genetic variation and air pollution on risk
of RVO
16
REFERENCES
1. Plun-Favreau H, Lewis PA, Hardy J, et al. Cancer and neurodegeneration:
Between the devil and the deep blue sea. PLoS Genet. 2010;6(12):1–8.
2. Salk News. From eye to brain: Salk researchers map functional connections
between retinal neurons at single-cell resolution - Salk Institute for Biological
Studies. (https://www.salk.edu/news-release/from-eye-to-brain-salk-researchers-
map-functional-connections-between-retinal-neurons-at-single-cell-resolution/).
(Accessed May 23, 2017)
3. London A, Benhar I, Schwartz M. The retina as a window to the brain - From eye
research to CNS disorders. Nat. Rev. Neurol. 2013;9(1):44–53.
4. Mungas D, Jagust WJ, Reed BR, et al. MRI predictors of cognition in subcortical
ischemic vascular disease and Alzheimer’s disease. Neurology. 2001;57(12):2229–
2235.
5. Calderón-Garcidueas L, Mora-Tiscareño A, Styner M, et al. White matter
hyperintensities, systemic inflammation, brain growth, and cognitive functions in
children exposed to air pollution. J. Alzheimer’s Dis. 2012;31(1):183–191.
6. Deary IJ, Bastin ME, Pattie A, et al. White matter integrity and cognition in
childhood and old age. Neurology. 2006;66(4):505–512.
7. Chen JC, Wang X, Wellenius GA, et al. Ambient air pollution and neurotoxicity
on brain structure: Evidence from women’s health initiative memory study. Ann.
Neurol. 2015;78(3):466–476.
8. Knopman DS, Mosley TH, Catellier DJ, et al. Fourteen-year longitudinal study of
vascular risk factors, APOE genotype, and cognition: The ARIC MRI Study.
Alzheimer’s Dement. 2009;5(3):207–214.
9. Terwisscha Van Scheltinga AF, Bakker SC, Van Haren NEM, et al. Genetic
schizophrenia risk variants jointly modulate total brain and white matter volume.
Biol. Psychiatry. 2013;73(6):525–531.
10. Hee MR, Izatt JA, Swanson EA, et al. Optical Coherence Tomography of the
Human Retina. Arch. Ophthalmol. 1995;113(3):325–332.
11. Huang D, Swanson EA, Lin CP, et al. Optical coherence tomography. Science.
1991;254(5035):1178–81.
12. Introduction to OCT - OBEL.
(http://obel.ee.uwa.edu.au/research/fundamentals/introduction-oct/). (Accessed
May 15, 2017)
13. Schuman JS. Spectral domain optical coherence tomography for glaucoma (an
AOS thesis). Trans. Am. Ophthalmol. Soc. 2008;106:426–58.
14. Fujimoto JG. Optical coherence tomography for ultrahigh resolution in vivo
imaging. Nat. Biotechnol. 2003;21(11):1361–1367.
15. Kafieh R, Rabbani H, Kermani S. A review of algorithms for segmentation of
17
optical coherence tomography from retina. J. Med. Signals Sens. 2013;3(1):45–60.
16. Remington LA. Clinical Anatomy of the Visual System. 2011 588 p.
17. Firat PG, Doganay S, Demirel EE, et al. Comparison of ganglion cell and retinal
nerve fiber layer thickness in primary open-angle glaucoma and normal tension
glaucoma with spectral-domain OCT. Graefe’s Arch. Clin. Exp. Ophthalmol.
2013;251(3):831–838.
18. Knight OJ, Girkin CA, Budenz DL, et al. Effect of race, age, and axial length on
optic nerve head parameters and retinal nerve fiber layer thickness measured by
cirrus HD-OCT. Arch. Ophthalmol. 2012;130(3):312–318.
19. Ascaso FJ, Cruz N, Modrego PJ, et al. Retinal alterations in mild cognitive
impairment and Alzheimer’s disease: An optical coherence tomography study. J.
Neurol. 2014;261(8):1522–1530.
20. P.K. I, O. A, T. T, et al. Relationship between cognitive impairment and retinal
morphological and visual functional abnormalities in Alzheimer disease. J. Neuro-
Ophthalmology. 2006;26(1):18–24.
21. Paquet C, Boissonnot M, Roger F, et al. Abnormal retinal thickness in patients
with mild cognitive impairment and Alzheimer’s disease. Neurosci. Lett.
2007;420(2):97–99.
22. Petzold A, de Boer JF, Schippling S, et al. Optical coherence tomography in
multiple sclerosis: A systematic review and meta-analysis. Lancet Neurol.
2010;9(9):921–932.
23. Ishikawa H, Stein DM, Wollstein G, et al. Macular segmentation with optical
coherence tomography. Investig. Ophthalmol. Vis. Sci. 2005;46(6):2012–2017.
24. G. M, C. A, P. M, et al. Comparison of retinal thickness measurements and
segmentation performance of four different spectral and time domain OCT devices
in neovascular age-related macular degeneration. Br. J. Ophthalmol.
2009;93(11):1453–1460.
25. Rao HL, Zangwill LM, Weinreb RN, et al. Comparison of different spectral
domain optical coherence tomography scanning areas for glaucoma diagnosis.
Ophthalmology. 2010;117(9):1692-1699.e1.
26. Curcio CA, Allen KA. Topography of ganglion cells in human retina. J. Comp.
Neurol. 1990;300(1):5–25.
27. Larrea BA, Iztueta MG, Indart LM, et al. Early axonal damage detection by
ganglion cell complex analysis with optical coherence tomography in nonarteritic
anterior ischaemic optic neuropathy. Graefe’s Arch. Clin. Exp. Ophthalmol.
2014;252(11):1839–1846.
28. Cheung CYL, Ong YT, Hilal S, et al. Retinal ganglion cell analysis using high-
definition optical coherence tomography in patients with mild cognitive
impairment and alzheimer’s disease. J. Alzheimer’s Dis. 2015;45(1):45–56.
29. Varma R, Skaf M, Barron E. Retinal nerve fiber layer thickness in normal human
18
eyes. Ophthalmology. 1996;103(12):2114–2119.
30. Oshitari T, Hanawa K, Adachi-Usami E. Changes of macular and RNFL
thicknesses measured by Stratus OCT in patients with early stage diabetes. Eye.
2009;23(4):884–889.
31. Parikh RS, Parikh SR, Sekhar GC, et al. Normal age-related decay of retinal nerve
fiber layer thickness. Ophthalmology. 2007;114(5):921–6.
32. Khawaja AP, Chan MPY, Garway-Heath DF, et al. Associations with retinal nerve
fiber layer measures in the EPIC-Norfolk Eye Study. Invest. Ophthalmol. Vis. Sci.
2013;54(7):5028–34.
33. Budenz DL, Anderson DR, Varma R, et al. Determinants of Normal Retinal Nerve
Fiber Layer Thickness Measured by Stratus OCT. Ophthalmology.
2007;114(6):1046–1052.
34. Hebert LE, Scherr PA, Bienias JL, et al. Alzheimer disease in the US population:
Prevalence estimates using the 2000 census. Arch. Neurol. 2003;60(8):1119–1122.
35. Neurological disorders: public health challenges GBD STUDIES AND THEIR
KEY RESULTS.
(http://www.who.int/mental_health/neurology/chapter_2_neuro_disorders_public_
h_challenges.pdf?ua=1). (Accessed May 12, 2017)
36. Murray CJL, Vos T, Lozano R, et al. Disability-adjusted life years (DALYs) for
291 diseases and injuries in 21 regions, 1990–2010: a systematic analysis for the
Global Burden of Disease Study 2010. Lancet. 2012;380(9859):2197–2223.
37. NEI. Prevalence of Adult Vision Impairment and Age-Related Eye Diseases in
America--2002. 2002 (Accessed May 12,
2017).(https://nei.nih.gov/eyedata/adultvision_usa). (Accessed May 12, 2017)
38. Tham YC, Li X, Wong TY, et al. Global prevalence of glaucoma and projections
of glaucoma burden through 2040: A systematic review and meta-analysis.
Ophthalmology. 2014;121(11):2081–2090.
39. Pascolini D MS. Global estimates of visual impairment. Br J Ophthalmol.
2010;96:614–618.
40. Yau JWY, Lee P, Wong TY, et al. Retinal vein occlusion: An approach to
diagnosis, systemic risk factors and management. Intern. Med. J.
2008;38(12):904–910.
41. Clarhson JC. Natural history and clinical management of central retinal vein
occlusion. Arch. Ophthalmol. 1997;115(4):486–491.
42. Rogers S, McIntosh RL, Cheung N, et al. The Prevalence of Retinal Vein
Occlusion: Pooled Data from Population Studies from the United States, Europe,
Asia, and Australia. Ophthalmology. 2010;117(2):313–9.e1.
43. Bonomi L, Marchini G, Marraffa M, et al. Vascular risk factors for primary open
angle glaucoma: The Egna-Neumarkt Study. Ophthalmology. 2000;107(7):1287–
1293.
19
44. Leske MC, Connell AMS, Wu SY, et al. Risk Factors for Open-angle Glaucoma:
The Barbados Eye Study. Arch. Ophthalmol. 1995;113(7):918–924.
45. Smith W, Assink J, Klein R, et al. Risk factors for age-related macular
degeneration: Pooled findings from three continents. Ophthalmology.
2001;108(4):697–704.
46. Yau JWY, Rogers SL, Kawasaki R, et al. Global prevalence and major risk factors
of diabetic retinopathy. Diabetes Care. 2012;35(3):556–564.
47. R.N. K, G.E. M, R. A, et al. Alzheimer’s disease and vascular dementia in
developing countries: prevalence, management, and risk factors. Lancet Neurol.
2008;7(9):812–826.
48. Lindsay J, Laurin D, Verreault R, et al. Risk factors for Alzheimer’s disease: a
prospective analysis from the Canadian Study of Health and Aging. Am. J.
Epidemiol. 2002;156(5):445–53.
49. Rehak M, Wiedemann P. Retinal vein thrombosis: Pathogenesis and management.
J. Thromb. Haemost. 2010;8(9):1886–1894.
50. Risk factors for central retinal vein occlusion. The Eye Disease Case-Control
Study Group. Arch. Ophthalmol. (Chicago, Ill. 1960). 1996;114(5):545–54.
51. Lambert JC, Ibrahim-Verbaas CA, Harold D, et al. Meta-analysis of 74,046
individuals identifies 11 new susceptibility loci for Alzheimer’s disease. Nat.
Genet. 2013;45(12):1452–1458.
52. Hysi PG, Cheng CY, Springelkamp H, et al. Genome-wide analysis of multi-
ancestry cohorts identifies new loci influencing intraocular pressure and
susceptibility to glaucoma. Nat. Genet. 2014;46(10):1126–1130.
53. Harold D, Abraham R, Hollingworth P, et al. Genome-wide association study
identifies variants at CLU and PICALM associated with Alzheimer’s disease. Nat.
Genet. 2009;41(10):1088–1093.
54. Liew G, Klein R, Wong TY. The Role of Genetics in Susceptibility to Diabetic
Retinopathy. Int. Ophthalmol. Clin. 2009;49(2):35–52.
55. Fritsche LG, Igl W, Bailey JNC, et al. A large genome-wide association study of
age-related macular degeneration highlights contributions of rare and common
variants. Nat. Genet. 2016;48(2):134–143.
56. Gauderman WJ, Zhang P, Morrison JL, et al. Finding novel genes by testing G × E
interactions in a genome-wide association study. Genet. Epidemiol.
2013;37(6):603–613.
57. Peters A, Greven S, Heid IM, et al. Fibrinogen genes modify the fibrinogen
response to ambient particulate matter. Am. J. Respir. Crit. Care Med.
2009;179(6):484–491.
58. Krewski D, Burnett RT, Gapstur SM, et al. Ischemic Heart Disease Mortality and
Long-Term Exposure to Source-Related Components of U.S. Fine Particle Air
Pollution. Environ. Health Perspect. 2015;124(6):785–794.
20
59. Lelieveld J, Evans JS, Fnais M, et al. The contribution of outdoor air pollution
sources to premature mortality on a global scale. Nature. 2015;525(7569):367–
371.
60. Seagrave JC, McDonald JD, Bedrick E, et al. Lung toxicity of ambient particulate
matter from southeastern U.S. sites with different contributing sources:
Relationships between composition and effects. Environ. Health Perspect.
2006;114(9):1387–1393.
61. Calderón-Garcidueñas L, Reed W, Maronpot RR, et al. Brain inflammation and
Alzheimer’s-like pathology in individuals exposed to severe air pollution. Toxicol.
Pathol. 2004;32(6):650–658.
62. Yan W, Yun Y, Ku T, et al. NO2 inhalation promotes Alzheimer’s disease-like
progression: Cyclooxygenase-2-derived prostaglandin E2 modulation and
monoacylglycerol lipase inhibition-targeted medication. Sci. Rep.
2016;6(1):22429.
63. Fonken LK, Xu X, Weil ZM, et al. Air pollution impairs cognition, provokes
depressive-like behaviors and alters hippocampal cytokine expression and
morphology. Mol. Psychiatry. 2011;16(10):987–995.
64. Power MC, Weisskopf MG, Alexeeff SE, et al. Traffic-related air pollution and
cognitive function in a cohort of older men. Environ. Health Perspect.
2011;119(5):682–687.
65. Weuve J, Puett RC, Schwartz J, et al. Exposure to particulate air pollution and
cognitive decline in older women. Arch. Intern. Med. 2012;172(3):219–227.
66. Freire C, Ramos R, Puertas R, et al. Association of traffic-related air pollution with
cognitive development in children. J. Epidemiol. Community Health.
2010;64(3):223–8.
67. Calderón-Garcidueñas L, Mora-Tiscareño A, Ontiveros E, et al. Air pollution,
cognitive deficits and brain abnormalities: a pilot study with children and dogs.
Brain Cogn. 2008;68(2):117–27.
68. Block ML, Elder A, Auten RL, et al. The outdoor air pollution and brain health
workshop. Neurotoxicology. 2012;33(5):972–84.
69. Brunekreef B, Holgate ST. Air pollution and health. Lancet.
2002;360(9341):1233–1242.
70. Block ML, Calderón-Garcidueñas L. Air pollution: mechanisms of
neuroinflammation and CNS disease. Trends Neurosci. 2009;32(9):506–16.
71. Dantzer R., O’Connor J. C., Freund G. G., et al. From inflammation to sickness
and depression: When the immune system subjugates the brain. Nat. Rev.
Neurosci. 2008;Vol. 9(1):46-56 pp.
72. Tamagawa E, Van Eeden SF. Impaired lung function and risk for stroke: Role of
the systemic inflammation response? Chest. 2006;130(6):1631–1633.
73. Qin L, Block ML, Liu Y, et al. Systemic LPS Causes Chronic Neuroinflammation
21
and progressive neurodegeneration. Glia. 2007;55(14):1416–1425.
74. Rivest S, Lacroix S, Vallières L, et al. How the blood talks to the brain
parenchyma and the paraventricular nucleus of the hypothalamus during systemic
inflammatory and infectious stimuli. Proc. Soc. Exp. Biol. Med. 2000;223(1):22–
38.
75. Manousakis G, Jensen MB, Chacon MR, et al. The interface between stroke and
infectious disease: Infectious diseases leading to stroke and infections
complicating stroke. Curr. Neurol. Neurosci. Rep. 2009;9(1):28–34.
76. Oberdörster G, Utell MJ. Ultrafine particles in the urban air: to the respiratory
tract--and beyond? Environ. Health Perspect. 2002;110(8):A440-1.
77. Oberdörster G, Sharp Z, Atudorei V, et al. Translocation of inhaled ultrafine
particles to the brain. In: Inhalation Toxicology. 2004:437–445.
78. Wang B, Feng WY, Wang M, et al. Transport of intranasally instilled fine Fe2O3
particles into the brain: Micro-distribution, chemical states, and histopathological
observation. Biol. Trace Elem. Res. 2007;118(3):233–243.
79. Calderón-Garcidueñas L, Solt AC, Henríquez-Roldán C, et al. Long-term air
pollution exposure is associated with neuroinflammation, an altered innate immune
response, disruption of the blood-brain barrier, ultrafine particulate deposition, and
accumulation of amyloid β-42 and α-synuclein in children and young adult.
Toxicol. Pathol. 2008;36(2):289–310.
80. Block ML, Zecca L, Hong J-S. Microglia-mediated neurotoxicity: uncovering the
molecular mechanisms. Nat. Rev. Neurosci. 2007;8(1):57–69.
81. Tin-Tin-Win-Shwe, Mitsushima D, Yamamoto S, et al. Changes in
neurotransmitter levels and proinflammatory cytokine mRNA expressions in the
mice olfactory bulb following nanoparticle exposure. Toxicol. Appl. Pharmacol.
2008;226(2):192–198.
82. Nimmerjahn A, Kirchhoff F, Helmchen F. Neuroscience: Resting microglial cells
are highly dynamic surveillants of brain parenchyma in vivo. Science (80-. ).
2005;308(5726):1314–1318.
83. Prinz M, Priller J. Microglia and brain macrophages in the molecular age: from
origin to neuropsychiatric disease. Nat. Rev. Neurosci. 2014;15(5):300–312.
84. Wang JW, Chen S Da, Zhang XL, et al. Retinal microglia in glaucoma. J.
Glaucoma. 2016;25(5):459–465.
85. Araneda S, Commin L, Atlagich M, et al. VEGF overexpression in the astroglial
cells of rat brainstem following ozone exposure. Neurotoxicology.
2008;29(6):920–927.
86. Demontis GC, Longoni B, Marchiafava PL. Molecular steps involved in light-
induced oxidative damage to retinal rods. Investig. Ophthalmol. Vis. Sci.
2002;43(7):2421–2427.
87. Heil K, Pearson D, Carell T. Chemical investigation of light induced DNA
22
bipyrimidine damage and repair. Chem. Soc. Rev. 2011;40(8):4271–4278.
88. Contín MA, Benedetto MM, Quinteros-Quintana ML, et al. Light pollution: The
possible consequences of excessive illumination on retina. Eye. 2016;30(2):255–
263.
89. Hayreh SS, Zimmerman B, McCarthy MJ, et al. Systemic diseases associated with
various types of retinal vein occlusion. Am. J. Ophthalmol. 2001;131(1):61–77.
90. Roger VL, Go AS, Lloyd-Jones DM, et al. Heart Disease and Stroke Statistics—
2012 Update. Circulation. 2011;125(1):e2–e220.
91. Brook RD, Rajagopalan S, Pope CA, et al. Particulate Matter Air Pollution and
Cardiovascular Disease. Circulation. 2010;121(21):2331–2378.
92. Brook RD, Franklin B, Cascio W, et al. Air pollution and cardiovascular disease:
A statement for healthcare professionals from the expert panel on population and
prevention science of the American Heart Association. Circulation.
2004;109(21):2655–2671.
93. Robertson S, Miller MR. Ambient air pollution and thrombosis. Part. Fibre
Toxicol. 2018;15(1):1.
94. Mills NL, Törnqvist H, Robinson SD, et al. Diesel exhaust inhalation causes
vascular dysfunction and impaired endogenous fibrinolysis. Circulation.
2005;112(25):3930–3936.
95. Hildebrandt K, Rückerl R, Koenig W, et al. Short-term effects of air pollution: A
panel study of blood markers in patients with chronic pulmonary disease. Part.
Fibre Toxicol. 2009;6:25.
96. Baccarelli A, Zanobetti A, Martinelli I, et al. Effects of exposure to air pollution on
blood coagulation. J. Thromb. Haemost. 2007;5(2):252–260.
97. Viehmann A, Hertel S, Fuks K, et al. Long-term residential exposure to urban air
pollution, and repeated measures of systemic blood markers of inflammation and
coagulation. Occup. Environ. Med. 2015;72(9):656–663.
98. Miller MR, Shaw CA, Langrish JP. From particles to patients: Oxidative stress and
the cardiovascular effects of air pollution. Future Cardiol. 2012;8(4):577–602.
99. Seaton A, Godden D, MacNee W, et al. Particulate air pollution and acute health
effects. Lancet. 1995;345(8943):176–178.
100. Bajaj A, Driver JA, Schernhammer ES. Parkinson’s disease and cancer risk: A
systematic review and meta-analysis. Cancer Causes Control. 2010;21(5):697–
707.
101. Checkoway H, Powers K, Smith-Weller T, et al. Parkinson’s disease risks
associated with cigarette smoking, alcohol consumption, and caffeine intake. Am.
J. Epidemiol. 2002;155(8):732–738.
102. Burchell VS, Gandhi S, Deas E, et al. Targeting mitochondrial dysfunction in
neurodegenerative disease: Part I. Expert Opin. Ther. Targets. 2010;14(4):369–
385.
23
103. Garber JE, Offit K. Hereditary cancer predisposition syndromes. J. Clin. Oncol.
2005;23(2):276–292.
104. Baird PN, Richardson AJ, Robman LD, et al. Apolipoprotein (APOE) gene is
associated with progression of age-related macular degeneration (AMD). Hum.
Mutat. 2006;27(4):337–342.
105. Akiyama H, Barger S, Barnum S, et al. Inflammation and Alzheimer’s disease.
Neurobiol. Aging. 2000;21(3):383–421.
106. Klein RJ, Zeiss C, Chew EY, et al. Complement factor H polymorphism in age-
related macular degeneration. Science. 2005;308(5720):385–9.
107. Amirian E, Liu Y, Scheurer ME, et al. Genetic variants in inflammation pathway
genes and asthma in glioma susceptibility. Neuro. Oncol. 2010;12(5):444–452.
24
Chapter 2. Retinal Nerve Fiber Layer Thickness in Normal Latinos, Chinese
Americans, and African Americans: A Multiethnic Study of Population-based Data
ABSTRACT
Changes in the optic nerve head (ONH) and thinning of the retinal nerve fiber
layer (RNFL) are associated with aging and optic neuropathies. The goal was to compare
a multiethnic population to normative databases of RNFL thickness and ONH measures
and identify their determinants. We pooled 11,585 eyes from 6,133 participants from
three population-based cohort studies of older (≥ 50 years) Latinos, Chinese Americans,
and African Americans from Los Angeles. We measured the ONH using the Cirrus HD-
OCT 4000 from participants without ocular diseases. Linear mixed regression analysis
was used to identify associations between RNFL thickness and potential determinants.
The mean age of the study participants was 58.8 (SD=7.9) years and 63.7% were female.
Chinese Americans had higher RNFL thickness across most clock hours (P < 0.05)
compared to their African-American and Latino counterparts, after adjusting for age and
sex. Chinese participants also had larger disc area, cup-to-disc ratios, and vertical cup-to-
disc ratios than their African-American and Latino counterparts. Age and ethnicity
explained the most proportion of variance in RNFL thickness. This analysis of OCT
scans from the largest number of racial/ethnic minorities describes the distribution of
RNFL thickness and ONH and investigates demographic and clinical factors associated
with RNFL thickness.
25
Optical coherence tomography (OCT) has allowed high-resolution visualization
and quantification of the structure of the optic nerve head (ONH) and the peripapillary
retinal nerve fiber layer (RNFL) (1). Changes in the ONH and thinning of the RNFL are
associated with aging and optic neuropathies, including glaucoma, conditions that are
distinguished by degeneration of the ganglion cell axons and changes of the neuroretinal
rim (2–4). With newer hardware and software, OCT devices have allowed for automatic
segmentation of images and subsequent detection of many of these abnormalities.
Currently, commercial OCT devices use built-in normative database as the reference for
detecting RNFL abnormalities. However, these normative databases have limited
representations from minority populations (5). The U.S. Food and Drug Administration
(FDA) report (6) on normative databases of OCT compared manufacturers’ FDA clinical
trials and identified that these studies consisted of sample sizes of 200-500 participants.
Problematic across these normative database studies were the lack of standardization of
protocol and exclusionary criteria across manufacturers, and this limits the interpretation
and further comparisons across OCT studies.
Additional studies using OCT have suggested that compared to whites, Hispanics
and Asians have higher RNFL thickness overall while the evidence for RNFL thickness
in African Americans compared to other ethnicities have been mixed (7,8). Limitations of
these multiethnic studies are their small sample sizes and lack of quadrant specific reports
which may not describe the visual field deficits that occur throughout the retina. For
instance, Hispanics and Blacks, despite their high susceptibility for glaucoma and
evidence of racial/ethnic differences in ONH and RNFL thickness, are underrepresented
in many OCT normative studies. Studies have also shown that racial/ethnic differences in
26
optic disc area (9–12) and factors such as presence of myopia (13) may compromise the
diagnostic ability of OCT based on comparison with age-matched reference database.
Additionally, other OCT measured ONH parameters, including the neuroretinal rim area,
may better distinguish normal from glaucomatous eyes highlighting the need to
understand the differences of additional measures by race/ethnicity (14).
To gain an improved understanding of biological mechanisms underlying
complex optic neuropathies and the changes in RNFL thickness, further studies are
needed to assess the influence of factors such as ethnicity and myopia on the ONH and
RNFL in aging adults and identify additional demographic and clinical factors that may
also be associated with ONH measures and RNFL thickness. The aims of this study were
to provide a robust assessment of 1) ethnic differences in RNFL thickness and retinal rim
in understudied minority Americans, and 2) demographics and clinical factors, such as
age, axial length (AL), BMI, and diabetes, that may be associated with RNFL thickness
and rim area. We created a large database (N=11,585 eyes from 4,472 individuals) of
optic disc scans obtained through Cirrus HD-OCT 4000 from participants of three
population-based cohort studies that were designed by the same investigative team and
shared similar study protocols, the Los Angeles Latino Eye Study (LALES), the Chinese
American Eye Study (CHES), and the African American Eye Disease Study (AFEDS).
MATERIAL AND METHODS
Study design and participants
We pooled epidemiological, ocular, and OCT imaging data from three similarly-
designed population-based studies of eye disease conducted in Los Angeles, California
by the same investigative team: the 8-year follow up study of LALES (LALES III)
27
conducted between 2010-2014, CHES conducted between 2009-2013, and AFEDS
conducted between 2014-2018. The methods for each study have been described
previously (15–17). Participants were eligible for each respective study depending on
self-reported ethnicity, being 40 years and older for LALES baseline and AFEDS and 50
years and older for CHES, living within respective census tracts of the cities. Trained
interviewers completed in-home interviews, and participants were then invited to
complete clinical examinations by ophthalmic technicians and ophthalmologists. The
studies of OCT measured RNFL were completed between 2010-2016. Written informed
consent was obtained for all participants. Study approval was obtained from the Los
Angeles County/University of Southern California Medical Center Institutional Review
Board and adhered to the tenets of the Declaration of Helsinki.
Demographic and clinical data
All studies conducted standardized, comprehensive ocular examinations, which
collected ophthalmic clinical and demographic information, including gender, age,
height, weight, waist-hip ratio, hypertension, diabetes mellitus (DM), and axial length
(AL). Random blood glucose and glycosylated hemoglobin were measured using the
Hemocue B-Glucose Analyzer (Hemocue Inc., Lake Forest, CA) and the DCA 2000+
System (Bayer Corporation, Tarrytown, NY), respectively. Participants were considered
to have DM if 1) the participant reported a history of diabetes and was being treated with
medication or insulin 2) HbA1c measured at 6.5% or higher. AL was measured using an
ultrasonic A-scan/pachymeter DGH 4000B SBH IOL Computation module (DGH Tech
Inc., Exton, PA).
28
Optical coherence tomography imaging
OCT imaging of the RNFL and ONH was performed using the Cirrus HD-OCT
4000 (Carl Zeiss Meditec, Dublin, CA, USA). In the AFEDS population, the Cirrus HD-
OCT 4000 was replaced with a newer Zeiss AngioPlex OCT in April 2016. We compared
the AFEDS population for those collected before and after this time point
(Supplementary Table 2.1). Trained ophthalmic technicians collected Optic Disc Cube
200x200 scans for the right (OD) and left (OS) eye separately. The Cirrus algorithm
identifies the center of the optic disc and automatically places a 3.46 mm 360º circle. To
calculate RNFL thickness, the anterior and posterior boundaries of the RNFL are
delineated first. The system then calculates the RNFL thickness by counting the number
of pixels between the anterior and posterior boundaries along each point on the circle of
the A-scan. RNFL thickness measurements are categorized into clock hour, quadrants,
and by aggregate average. To assess ONH measures, the algorithm identifies the
termination of Bruch’s membrane to determine the disc edge. The software then extracts
the rim width around the optic disc by measuring the thickness of the neuroretinal tissue
in the optic nerve. We then extracted the ONH parameters, including disc area, rim area,
CDR (cup-to-disc area ratio), VCDR (vertical cup-to-disc ratio), and cup volume. Disc
area represents the surface area measured by the optic disc outline. Rim area is the
surface area of the cup area minus the disc area. CDR measures the ratio of the cup area
to the disc area, while VCDR is the ratio between the longest vertical line across the cup
to the longest vertical line of the optic disc. Cup volume measures the total RNFL volume
in the rim of the optic disc. Segmentation data were extracted for all subsequent analyses.
Inclusion and exclusion criteria
29
A scan was retained if it had signal strength ≥7 out of 10, as recommended by the
manufacturer for high quality (18,19). To make inferences on a normal population, we
also excluded participants that had a diagnosis of glaucoma, diabetic retinopathy, macular
degeneration, and cataract related phenotypes, determined at the clinic examination of
each study using ICD-9 definitions. Additionally, we excluded participants with any
reported history of glaucoma or cataract procedures, including glaucoma drain tube,
glaucoma laser surgery, and cataract extraction. These quality control procedures ensured
high-quality images used for downstream segmentation for the RNFL thickness and ONH
values. To exclude scans with potential any segmentation errors, scans were excluded if
any values for RNFL thickness were equal to zero. We assessed symmetry across eyes
and excluded participants if the differences between eyes was greater than 30µm. If
participants had multiple quality scans that passed, only one eye was randomly selected
(Figure 2.1).
Statistical methods
We considered demographic variables including age, sex, BMI, waist-hip ratio,
smoking, high density lipoprotein, low density lipoprotein, hypertension, and diabetic
status. These variables have been shown or suspected to be associated with RNFL
thickness in prior studies (20,21). Ocular variables included axial length, intraocular
pressure, refractive error, disc area, and signal strength. Statistical analyses were
performed using R (Version 3.5.2). Counts, proportions, means, and standard deviations
were used to summarize the demographic and clinical characteristics. We used linear
mixed models (LMM) to assess the relationship between continuous RNFL thickness
measures and demographic and clinical variables while allowing for the relatedness
30
between eyes by including a random effect for each participant. We performed
univariable analyses for all variables and constructed a final multivariable model
including all variables that were significant at P < 0.1. Estimated marginal means with
95% confidence intervals were also calculated to compare the differences of ONH and
RNFL thickness across difference ethnicity and demographic characteristics after
adjusting for other covariates. We used quantile regression to compare the effects of
ethnicity on RNFL thickness at fifth percentile. We performed analyses at the fifth
percentile to be comparable to the OCT normative studies (7). Standardized coefficients
were estimated for a better comparison of the effects of different factors on outcomes
such as RNFL thickness. Additionally, we calculated semipartial correlation coefficients
(SR
2
) to estimate the unique contribution one variable to the variation in outcomes such
as RNFL thickness. We additionally plotted clock-hour- and quadrant-specific RNFL
thickness adjusting for age and sex. Plots were directly connected lines across each
estimated clock-hour and quadrant after adjusting for age and sex. Sensitivity analyses
were performed using two subsets of this final dataset including analyzing 1) higher
quality scans using a signal strength ³9 and 2) randomly selecting one eye per participant
and running regression models without a random intercept. There were no material
differences in either sensitivity analyses of the OCT measured outcomes after adjusting
for age and gender (Supplementary Table 2.2).
RESULTS
Across the three eye cohorts, there were 13,520 participants that were over >50
years old (Figure 2.1). Among them, 4,210 had ocular diseases, a history of glaucoma,
history of cataract procedures, or extreme refractive error. Among the remaining 9,310
31
healthy individuals free of ocular diseases, 6,133 (65.9%) had OCT scans with a signal
strength of 7 or greater and were used for our primary analyses. Those that were excluded
based on signal strength were older and more likely to be from the AFEDS and LALES
cohorts. Complete data on clinical, demographic, and OCT variables were available for
4,472 healthy participants for multivariable analyses of associations with RNFL thickness
and rim area (Figure 2.1). Those with healthy eyes were comparable to those with healthy
eyes and high-quality OCT and additionally to those with complete demographic/clinical
characteristics and ONH/RNFL thickness parameters (Table 2.1). In participants healthy
eyes with high quality OCT data, Chinese-American constituted most of the participants
followed by Latinos and African Americans. The mean age of the study population was
60.3 (SD=7.5). As expected, there were more women in this study (n=3,923, 63.9%).
We observed ethnic specific differences across all RNFL and most ONH
measures after adjusting for age and gender (Table 2.2). In general, Chinese Americans’
RNFL thickness was greater than their African American counterparts and the ethnic-
difference was greater in the temporal quadrant (mean difference=13.38 µm), however,
Chinese Americans had the thinnest nasal RNFL compared to both Latinos and African
Americans (both pairwise P < 0.05). In Chinese Americans, temporal quadrants were
thicker than their nasal quadrants opposite what was observed in Hispanic whites and
African Americans. Chinese American participants also had larger rim area, disc area,
cup-to-disc ratios, and cup volumes than their African-American and Latino counterparts,
adjusting for pairwise comparisons using a Tukey test (Table 2.2).
Figure 2 present both the average and the fifth percentile RNFL thickness by
clock-hour for each ethnicity after adjusting for age and gender. In general, Chinese-
32
American participants had thicker RNFL across most clock hours. These differences were
present when considering the fifth percentile across ethnicity groups (group level quantile
regression P < 0.05), though pairwise comparisons could not be made using quantile
regression between each ethnicity group (Supplementary Table 2.3).
Both average and quadrant-specific RNFL thickness were lower in older ages
(Figure 2.4). respectively. Per each 10-year increase in age, average RNFL thickness was
2.8, 2.5, and 2.5 µm thinner in Chinese Americans, African Americans, and Latinos (age
trend P < 0.05); in the rate of age-related loss of RNFL. For quadrant-specific RNFL,
age-related RNFL loss in the entire cohort was more pronounced in the inferior and
superior quadrants (4.8 and 4.6 µm per each 10-year increase in age, respectively) than in
the temporal and nasal quadrants (1.1 and 1.3 µm, respectively). When stratified by
ethnicity, (Figure 2.3), there were ethnic-specific differences in age-related loss in
average RNFL (interaction P = 0.04), the nasal quadrant (interaction P = 0.01), and
marginally significant differences in the inferior quadrant (interaction P = 0.09).
Figure 2.4 summarizes the clock hour rate of RNFL thickness decrease for each
10-year increase in age by ethnicity group. The greatest change in age across all groups is
within Latinos, where there is a 5.93 µm decrease in RNFL thickness in the clock hour
seven region. The change in CHES and AFEDS is 4.47 and 3.51 µm, respectively. These
results are additionally stratified and split into clock hours in Figure 2.5, where the largest
decreases were observed in Latinos in the superior quadrants compared to their Chinese-
and African-American counterparts.
To examine the potential role that glaucoma has in our multiethnic population, the
results are separated into those with suspect or undiagnosed glaucoma and additionally
33
stratified by racial/ethnic group. The average RNFL thickness in those with suspect or
undiagnosed glaucoma is similar to the RNFL thickness in healthy eyes at the fifth
percentile (Supplementary Figure 2.1). Additionally, the rate of change of RNFL
thickness in those suspect or undiagnosed glaucoma is highest in Latinos (Supplementary
Figure 2.2).
Table 2.3 summarizes the univariable and multivariable linear mixed regression
model of all factors on average RNFL. Using the univariable model, after adjusting for
age, ethnicity, and gender, the covariates waist-hip ratio, hypertension, being diabetic,
axial length, disc area, and signal strength of scans were significantly associated with
average RNFL thickness (P < 0.05). A shorter axial length was associated with lower
average RNFL thickness. All statistically significant variables were combined into one
multivariable model. The final model was a backward stepwise regression which
removed all variables where P > 0.05. The final model for average RNFL thickness
included age, sex, ethnicity, hypertension, being diabetic, axial length, disc area, and
signal strength. This multivariable model was across all quadrants, and these patterns
were observed in the same direction across all quadrants (Supplementary Table 2.3).
These factors explained between 7% to 25% of the variance in quadrant-specific RNFL
thickness and 17% of the variance in average RNFL thickness (Supplementary Table
2.4). Age ethnicity, and disc area explained the most proportion of the variance in RNFL,
however the proportions of variances explained were quadrant specific as well.
Table 2.4 summarizes only the ethnicity group effect when the statistically
significant covariates are added in various orders. The model including axial length, age,
34
sex, and signal strength changes the regression coefficients for the ethnicity effects at a
greater percent change than a model with disc area, age, sex, and signal strength.
DISCUSSION
The present pooled analysis is the largest population-based study evaluating the
normative values of and determinants of RNFL thickness and ONH measured using OCT
in a healthy population. This study is comprised a multiethnic population from the
AFEDS, CHES, and LALES, which are three of the largest population-based studies of
ocular conditions in African Americans, Chinese Americans, and Latino Americans,
respectively.
In this study, we confirm the associations between age and all OCT measures
(4,8,11). Average RNFL thickness decreased approximately 3 µm per each year after
adjusting for age, gender, and additional covariates. A longitudinal study suggested that
0.14 µm of the RNFL thickness was lost per year (22). Others have observed between a
0.15-0.28 µm decrease per year when using a cross-sectional design (23,24). Our study
observed a similar or slightly higher decrease in RNFL thickness, and this was consistent
across all three ethnic groups. This could be due to the previous studies lacking the ethnic
specific measures, which many studies are not able to recruit. Additionally, this could be
due to our sample recruitment was population-based and may have included a slightly
older population than other groups, however, the average age across all studies were
similar (~55 years). One major difference between the current study and all previous ones
is that this study’s recruitment is done at the neighborhood level, not a convenience
sampling frame. If studies using convenience samples report less loss of the RNFL, this
could be due to overall healthier participants being including in studies. Regardless, in all
35
studies the age-related loss appears to follow a consistent linear decreasing pattern for
each year increase in age, when we compared a linear model to a loess smoothed plot.
In our analysis, we identified sizable differences across of RNFL thickness by
race/ethnicity, even after accounting for the potential confounding by age and gender.
Across most quadrants and when in aggregate, we find that Chinese participants have the
thickest RNFL except for the nasal quadrant. Those with African descent have the thinner
RNFL than their other counterparts, except for the nasal quadrant. Other studies have
used quadrant-specific and average RNFL to diagnose optic neuropathies and have found
them to aid in building prediction models (25,26), and we highlight the need to show
quadrant and ethnic specific results here since there appear to be differences by ethnic
group at the quadrant level.
After adjusting for age and sex, the average RNFL thickness was estimated to be
96.87 µm and 91.21 µm in Chinese Americans and African Americans, respectively.
Examining this in context with our regression model, this suggests that this difference is
equal to a shift of approximately 20-years of age-related loss between the two groups.
While we are not positioned to make comparisons to European populations, our findings
are consistent with previous findings where those with Asian and Hispanic descent are
significantly different than those with African descent (3,7,27). When comparing
Hispanic and Asian participants, the evidence appears to be mixed (3,7,8). Our analysis
mirrors the NEI study on the prevalence of glaucoma, where Blacks have the highest
rates of glaucoma (28). The prevalence rate at age 75 for Hispanics is similar to the
prevalence of glaucoma at age 65 for Blacks, which mirrors the loss of the RNFL.
Additionally, with the present study, we additionally compared our study sample to those
36
we excluded for suspect or untreated glaucoma and identified an average RNFL thickness
decrease of 4 µm for each year in age. The loss is still substantially different by race even
in this diseased population, suggesting there may be additionally demographic, ocular, or
genetic factors that have not been considered.
Disc area has been suggested to bias the results of RNFL thickness (29), and with
the current study, we were able to control for differences of disc area using the measured
OCT values. Disc area was largest in the Chinese population compared to both the
Latinos and African Americans, and this difference remained after adjusting for age and
sex. Disc area is positively correlated with RNFL thickness, and this effect was consistent
by ethnicity. Other studies in participants of African descent, have identified a larger disc
area than this cohort and a similar cup-disc ratio, however, these results were not adjusted
for age, and were (30). When incorporated into a multivariable regression model, the
effect of disc area attenuated the differences of RNFL thickness by ethnicity.
When axial length is incorporated into the multivariable model, increasing axial
length is associated with decreased RNFL thickness, and additionally magnified the
ethnic specific differences between our three populations. Previous studies have shown
that longer axial length is associated with thinner RNFL, and this association remained
even after adjustment for smaller disc and rim area associated with longer axial length
(7,31). Axial length is suggested to be artifacts of axial-length-related ocular
magnification, which must be accounted for when diagnosing potential conditions of the
optic nerve. When stratified by ethnic group, axial length was significantly associated
with average and all quadrant RNFL in all ethnic groups. These results were consistent,
even using the axial length cut point at 26mm (32).
37
Our results are in line with other ethnic-specific studies for Chinese participants,
including The Singapore Epidemiology of Eye Diseases Study, where the average and
fifth percentile values are similar to our Chinese participants (33). We observe different
average values for our Hispanic and African population as compared to Girkin et al,
where the results in the present analysis are attenuated (8). However, their results are not
adjusted for age and additional covariates.
The normative population used to construct the original Cirrus OCT database
consists of 284 participants of European, Chinese, African, and Hispanic descent (3),
however this population consisted primarily of participants of European descent. Our
comparisons are in relationship to this normative database. Regarding the ONH and its
structures, we identified significant differences by ethnicity in all ONH areas examined,
including the rim area, disc area, CDR, VCDR, and cup volume. Our findings contrast
with previous studies that have stated that those with African descent have the largest
disc area (3,8,9,11,12,31), as we observe that those with Chinese descent have the largest
disc areas. This was also seen regarding CDR, where previous studies have identified
those with African descent as having greater CDR (3,34,35). We observed a statistically
significant difference among our studies where those with Chinese descent had the
highest CDR, however the magnitude of this difference was not pronounced as previous
findings. This result remained after adjusting for axial length to account for potential
differences in rate of myopia.
In building a normative database, the key findings are what constitutes the normal
range of participants. In the present study, we also identify that the lowest quantiles are
different across ethnicities. For instances, using the extreme (i.e. fifth percentile) values
38
for Chinese eyes differ from African eyes and using these values for any downstream
interpretations, with potential implications for diagnoses of glaucoma and other optic
neuropathies.
The normative database studies of OCT, including that of RTVue, Spectralis, and
Cirrus, have attempted to retain a “normal” population by excluding those with active
ocular diseases (including glaucoma and diabetic retinopathy) and additional neurological
conditions or other comorbidities (36). Our study attempted to use similar exclusionary
criteria, without knowing the explicit definition some of the criteria may differ. There is
generally good agreement across different instruments (37), which allow us to make
comparisons to this normative database.
Overall, through this study, we identify a discordance between established
normative values and differences in rate of change by each year increase. Many of
previous studies are convenience clinic-based studies. One of the likely reasons is that
our study includes three population-based studies while most others are recruited directly
from a clinic population. This could indicate some level of selection bias, since these
people might already be seeking eye care. This would limit the external generalizability
of the results of all other OCT RNFL thickness studies. These normative databases are
adequately powered to detect association for participants of European decent, but likely
were not able to do for any other ethnicity. The authors of the normative database also
state that their participants of Chinese descent were from Hong Kong and might not be
representative of the Chinese population.
Our findings include age and ethnicity being the largest contributors to the
variability of RNFL thickness. While other variables were found to be statistically
39
significant, these were the largest contributors to RNFL. The rate of glaucoma is high
among those with Hispanics with Mexican ancestry (38),and this suggests that the genetic
contributions of retinal structure cannot be understated. ONH parameters are highly
heritable (39), and studies in this Hispanic population found that there is are genes
associated with VCDR (40,41) highlighting the need to understand the genetic
contributions to the retina.
There are many strengths to this study, most importantly, that we utilized the data
from multiple population-based eye studies and obtained a very large sample size of OCT
characterized eyes. Because we conducted this analysis in a population-based setting, our
conclusions are generalizable to a multiethnic population. Limitations of the current study
include the older age of this population and that the Hispanic population primary consists
of Mexican-Americans. These limitations would only restrict the external generalizability
of our study.
We highlight the need for an ethnic specific normative database. In conclusion,
this study provides additional evidence on the relationship between demographic and
clinical determinants to ONH and RNFL thickness. These findings should help identify
factors in not only the imaging parameters but also aid in diagnosing optic neuropathies.
40
Table 2.1 Analytical Cohort Comparisons Between the Full Dataset to Dataset with Complete Covariate Data
Variable Healthy eyes + OCT
(n=4,472)
Healthy eyes without OCT
(n=4,538)
OCT versus no
OCT
n (%) n (%) P
Female 2942 (64.8) 3035 (63.6) 0.22
Age
< 0.001
50-59 2446 (53.9) 2144 (44.9)
60-69 1588 (34.9) 1709 (35.8)
70-79 446 (9.8) 736 (15.4)
80+ 58 (1.3) 183 (3.8)
Race/ethnicity
< 0.001
Chinese
Americans
2260 (49.8) 1021 (21.4)
Hispanics 1246 (27.5) 1341 (28.1)
African
Americans
1032 (22.7) 2410 (50.5)
BMI
Normal 1679 (37.0) 1205 (26.4) < 0.001
Overweight 1537 (33.9) 1606 (35.2)
Obese 1322 (29.1) 1755 (38.4)
Diabetic 749 (16.5) 997 (23.2) < 0.001
Ever Smokers 1235 (27.2) 1662 (37.3) < 0.001
Hypertensive 1167 (25.7) 1546 (32.9) < 0.001
Variable Mean (SD) Mean (SD)
Waist-Hip Ratio 0.88 (0.1) 0.87 (0.1) 0.41
HDL 52.47 (17.4) 50.65 (15.7) < 0.001
LDL 104.71 (33.8) 104.94 (33.1) 0.72
41
Axial Length 23.58 (1.2) 23.55 (1.1) 0.17
IOP 14.84 (2.7) 14.73 (2.7) 0.04
Abbreviations: HbA1c, Hemoglobin A1c; HDL, high density lipoprotein, LDL, low density lipoprotein; IOP, intraocular pressure
42
Table 2.2. Comparison of Optic Nerve Head Parameters and Retinal Nerve Fiber Layer Thickness Measurements Across the Three
Study Ethnicities
Variable All Ethnicities
Combined
Mean (SE)
Chinese
Americans
Mean (SE)
Latinos
Mean (SE)
African
Americans
Mean (SE)
Pairwise differences by
Ethnicity
a
Optic Nerve
Head
Rim Area 1.32 (0.003) 1.33 (0.004) 1.31 (0.005) 1.34 (0.007) A-L
Disc Area 2.02 (0.005) 2.06 (0.007) 1.98 (0.008) 1.99 (0.01) C-A C-L
Average CDR 0.55 (0.002) 0.56 (0.003) 0.55 (0.003) 0.53 (0.004) C-A C-L A-L
VCDR 0.51 (0.002) 0.51 (0.002) 0.51 (0.003) 0.5 (0.004)
Cup Volume 0.18 (0.002) 0.18 (0.003) 0.18 (0.003) 0.19 (0.004) C-A
RNFL
Average RNFL 95.07 (0.12) 96.76 (0.18) 95.17 (0.21) 91.21 (0.26) C-A C-L A-L
Temporal RNFL 65.71 (0.15) 71.53 (0.2) 62.33 (0.23) 58.15 (0.29) C-A C-L A-L
Superior RNFL 118.47 (0.19) 119.73 (0.28) 119.1 (0.34) 114.76 (0.42) C-A A-L
Nasal RNFL 71.21 (0.13) 69.77 (0.19) 72.94 (0.23) 71.72 (0.29) C-A C-L A-L
Inferior RNFL 124.89 (0.21) 126.05 (0.3) 126.36
(0.36)
120.16 (0.44) C-A A-L
Abbreviations: CDR, Cup-disc ratio; VCDR, vertical cup-disc ratio; C-A, Significant difference between CHES and AFEDS (Tukey’s
P < 0.05); C-L, Significant difference between CHES and AFEDS (Tukey’s P < 0.05); C-L, Significant difference between CHES and
LALES (Tukey’s P < 0.05); A-L, Significant difference between AFEDS and LALES (Tukey’s P < 0.05); AFEDS, African American
Eye Disease Study; CHES, Chinese-American Eye Disease Study; LALES, Los Angeles Latino Eye Study
a
P values were estimated from linear mixed model after adjusting for age and sex.
43
Table 2.3. Univariable and Multivariable Regression Results of Average RNFL Thickness.
Variable Average RNFL thickness
Model 1
a
Average RNFL thickness
Model 2
b
β (SE) P β (SE) P
Age
60 vs 50 -2.8 (0.3) <0.001 -2.05 (0.28) <0.001
70 vs 50 -5.84 (0.48) <0.001 -4.91 (0.45) <0.001
80 vs 50 -7.16 (1.24) <0.001 -5.94 (1.14) <0.001
Sex 1.15 (0.29) <0.001 0.73 (0.27) 0.01
BMI
Overweight vs normal 0.22 (0.35) 0.53
Obese vs normal -0.09 (0.42) 0.83
Waist-Hip Ratio -4.41 (2.01) 0.03
Ever Smokers 0.1 (0.34) 0.77
HDL 0 (0.01) 0.68
LDL 0.01 (0) 0.09
Hypertension -0.71 (0.32) 0.03 -0.62 (0.3) 0.04
Diabetic -0.95 (0.37) 0.01 -0.85 (0.35) 0.02
Ethnicity
African Americans vs. Chinese Americans -5.55 (0.35) <0.001 -5.4 (0.33) <0.001
Latinos vs Chinese Americans -1.45 (0.33) <0.001 -1.73 (0.32) <0.001
Axial Length -1.24 (0.09) <0.001 -0.96 (0.09) <0.001
IOP -0.08 (0.04) 0.06
Disc Area 4.54 (0.24) <0.001 4.01 (0.23) <0.001
Disc Signal 1.18 (0.07) <0.001 1.06 (0.07) <0.001
Abbreviations: HbA1c, Hemoglobin A1c; HDL, high density lipoprotein, LDL, low density lipoprotein; IOP, intraocular pressure
a
Model 1-Effects are adjusted for age, gender, and ethnicity.
44
b
Model 2- Beta effects include all variables discovered in model 1, with P < 0.1 and using a backward stepwise model to exclude all
variables with P > 0.05.
45
Table 2.4. Stepwise Forward Regression Coefficients of Covariates on Ethnicity Effect with Average RNFL Thickness
Average RNFL Thickness Effect
a
LALES vs CHES
β
AFEDS vs CHES
β
Study Only -2.09 -5.97
Study, Age, Sex -1.75 -5.54
Study, Age, Sex, Axial Length -2.41 -5.77
Study, Age, Sex, Disc Area -1.34 -5.27
Study, Age, Sex, Axial Length, Disc Area -1.93 -5.49
All covariates -1.78 -5.39
All covariates without Disc Area and Axial
Length
-1.62 -5.45
Abbreviations: AFEDS, African American Eye Disease Study; CHES, Chinese-American Eye Disease Study; LALES, Los Angeles
Latino Eye Study
a
All covariates include all variables discovered in the multivariable regression model including hypertension, diabetes status, and disc
signal.
46
Supplementary Table 2.1: Analytical Cohort Comparisons Between Full AFEDS Cohort and Cohort with Complete Covariate Data
Variable Full AFEDS OCT AFEDS
n (%) n (%) P
Female 2237 (64.9) 844 (64.4) 0.72
Age
0.02
50-59 1495 (43.4) 623 (47.5)
60-69 1276 (37.1) 478 (36.4)
70-79 568 (16.5) 182 (13.9)
80+ 103 (3.0) 28 (2.1)
BMI
0.05
Normal 663 (19.7) 217 (16.8)
Normal 1156 (34.4) 449 (34.7)
Overweight 1540 (45.9) 628 (48.5)
Obese 629 (19.3) 215 (17.2)
Ever Smokers 1515 (45.4) 589 (45.4) 1.00
Hypertensive 1048 (30.6) 376 (28.9) 0.29
Variable Mean (SD) Mean (SD)
Waist-Hip Ratio 0.88 (0.1) 0.87 (0.1) 0.38
HDL 52.47 (17.4) 50.63 (15.7) <0.0001
LDL 104.71 (33.8) 105 (33.1) 0.65
Axial Length 23.58 (1.2) 23.55 (1.1) 0.21
IOP 14.84 (2.7) 14.73 (2.7) 0.03
Abbreviations: AFEDS, African American Eye Disease Study; OCT, optical coherence tomography; HbA1c, Hemoglobin A1c; HDL,
high density lipoprotein, LDL, low density lipoprotein; IOP, intraocular pressure
47
Supplementary Table 2.2. Sensitivity Analysis of Optic Nerve Head Parameters and Retinal Nerve Fiber Layer Thickness
Measurements Across the Three Study Ethnicities
Variable All Ethnicities
Combined
Mean (SE)
Chinese Americans
Mean (SE)
Latinos
Mean (SE)
African Americans
Mean (SE)
Pairwise
differences by
Ethnicity
a
Sensitivity Analysis 1
Rim Area 1.34 (0.004) 1.34 (0.006) 1.32 (0.007) 1.35 (0.009) 0.007
Disc Area 2.04 (0.007) 2.09 (0.01) 1.98 (0.012) 2.01 (0.015) <0.0001
CD Ratio 0.55 (0.002) 0.56 (0.003) 0.54 (0.004) 0.53 (0.005) <0.0001
Average RNFL 96.81 (0.17) 98.43 (0.23) 96.91 (0.3) 92.81 (0.36) <0.0001
VCD Ratio 0.51 (0.002) 0.51 (0.003) 0.5 (0.004) 0.5 (0.005) 0.08
Sensitivity Analysis 2
Rim Area 1.32 (0.004) 1.32 (0.005) 1.31 (0.007) 1.34 (0.008) 0.004
Disc Area 2.02 (0.006) 2.05 (0.008) 1.97 (0.011) 2 (0.012) <0.0001
CD Ratio 0.55 (0.002) 0.55 (0.003) 0.54 (0.004) 0.53 (0.004) <0.0001
Average RNFL 95.53 (0.14) 97.17 (0.2) 95.69 (0.27) 91.72 (0.3) <0.0001
VCD Ratio 0.51 (0.002) 0.51 (0.003) 0.51 (0.004) 0.5 (0.004) 0.28
Abbreviations: CD, Cup-disc; VCD, Vertical Cup-disc; AFEDS, African American Eye Disease Study; CHES, Chinese-American Eye
Disease Study; LALES, Los Angeles Latino Eye Study
a
P values were estimated from linear mixed model after adjusting for age and sex.
48
Supplementary Table 2.3 Comparison of Retinal Nerve Fiber Layer Thickness Measurements Across the Three Study Ethnicities at
the Fifth Percentile
Variable All Ethnicities Combined Chinese Americans Latinos African Americans P for Quantile Regression
5th Percentile (um) (um) (um) (um)
Average RNFL 79 82 80 74 <0.001
Temporal RNFL 48 55 49 43 <0.001
Superior RNFL 92 93 94 88 <0.001
Nasal RNFL 55 54 58 55 <0.001
Inferior RNFL 97 99 100 91 <0.001
P-values were estimated quantile regression after adjusting for age and sex.
49
Supplementary Table 2.4. Multivariable Linear Mixed Regression Models Across Quadrants Using the Model Developed for Average
RNFL Thickness
Variable Temporal
Superior
Nasal
Inferior
Beta (SE) P Beta (SE) P Beta (SE) P Beta (SE) P
Age
60 vs 50 -0.92
(0.32)
0.00 -3.14
(0.46)
<0.0001 -0.5
(0.32)
0.1174 -3.49 (0.48) <0.0001
70 vs 50 -1.75
(0.52)
0.00 -7.95
(0.74)
<0.0001 -1.72
(0.52)
8.00E-
04
-8.1 (0.77) <0.0001
80 vs 50 -1.41
(1.33)
0.29 -9.37
(1.89)
<0.0001 0.38
(1.32)
0.7702 -13.45
(1.97)
<0.0001
Sex 2.19 (0.32) <0.0001 -0.54
(0.45)
0.2233 -0.41
(0.31)
0.1906 1.39 (0.47) 0.0028
Hypertension -0.21
(0.35)
0.54 -1.06
(0.49)
0.0304 0 (0.34) 0.9937 -1.2 (0.51) 0.0198
Diabetic -0.34
(0.41)
0.41 -0.86
(0.57)
0.1307 -1.05
(0.4)
0.0084 -1.15 (0.6) 0.0544
Ethnicity
African Americans vs. Chinese
Americans
-13.2
(0.38)
<0.0001 -4.88
(0.53)
<0.0001 2.28
(0.37)
<0.0001 -5.85 (0.56) <0.0001
Latinos vs Chinese Americans -8.23
(0.37)
<0.0001 -1.06
(0.52)
0.0424 3.28
(0.36)
<0.0001 -1.08 (0.54) 0.0479
Axial Length 1.4 (0.11) <0.0001 -2.04
(0.17)
<0.0001 -0.91
(0.12)
<0.0001 -2.91 (0.17) <0.0001
Disc Area 0.8 (0.31) 0.01 6.81 (0.45) <0.0001 4.71
(0.32)
<0.0001 6 (0.45) <0.0001
Disc Signal 0.14 (0.1) 0.16 2.03 (0.15) <0.0001 0.76
(0.11)
<0.0001 1.42 (0.14) <0.0001
50
Supplementary Table 2.5. Semipartial Correlation Coefficients for Effect of Overall Model and Individual Variables on Contribution
to RNFL Thickness Using Linear Mixed Models
Variable Average SR
2
Temporal SR
2
Superior SR
2
Nasal SR
2
Inferior SR
2
Model 0.169 (0.16-0.184) 0.252 (0.24-0.267) 0.139 (0.13-
0.153)
0.072 (0.063-
0.084)
0.153 (0.14-0.167)
Age 0.03 (0.023-0.038) 0.002 (0.001-0.005) 0.027 (0.021-
0.034)
0.002 (0.001-
0.004)
0.031 (0.024-
0.038)
Sex 0.001 (0-0.003) 0.009 (0.005-0.013) 0 (0-0.002) 0 (0-0.002) 0.002 (0-0.004)
Hypertension 0.001 (0-0.002) 0 (0-0.001) 0.001 (0-
0.002)
0 (0-0.001) 0.001 (0-0.002)
Diabetic 0.001 (0-0.003) 0 (0-0.001) 0 (0-0.002) 0.001 (0-0.003) 0.001 (0-0.002)
Ethnicity
African Americans vs. Chinese
Americans
0.051 (0.042-0.06) 0.189 (0.17-0.203) 0.015 (0.01-
0.02)
0.007 (0.004-
0.011)
0.02 (0.014-0.026)
Latinos vs Chinese Americans 0.006 (0.003-0.01) 0.087 (0.076-0.098) 0.001 (0-
0.002)
0.015 (0.01-0.02) 0.001 (0-0.002)
Axial Length 0.013 (0.008-0.018) 0.019 (0.014-0.025) 0.02 (0.014-
0.026)
0.008 (0.005-
0.012)
0.037 (0.03-0.045)
Disc Area 0.028 (0.021-0.035) 0.001 (0-0.002) 0.028 (0.021-
0.035)
0.027 (0.021-
0.034)
0.02 (0.015-0.027)
Disc Signal 0.01 (0.006-0.015) 0 (0-0.001) 0.013 (0.009-
0.018)
0.004 (0.002-
0.007)
0.006 (0.003-0.01)
Abbreviations: SR
2
, Semipartial correlation coefficients
51
Figure 2.1. Flow Chart of the Overall Study
Abbreviations: AFEDS, African American Eye Disease Study; CHES, Chinese-American Eye Disease Study; LALES, Los Angeles
Latino Eye Study
52
Figure 2.2. Line Series Plot of RNFL Thickness by Clock-Hour Stratified by Ethnicity
53
Abbreviations: C-A, Significant difference between CHES and AFEDS (Tukey’s P < 0.05); C-L, Significant difference between
CHES and AFEDS (Tukey’s P < 0.05); C-L, Significant difference between CHES and LALES (Tukey’s P < 0.05); A-L, Significant
difference between AFEDS and LALES (Tukey’s P < 0.05); AFEDS, African American Eye Disease Study; CHES, Chinese-
American Eye Disease Study; LALES, Los Angeles Latino Eye Study
Statistically significant differences adjusted for age and gender at clock hour using estimated marginal between studies
54
Figure 2.3. A) Linear Mixed Regression Line Plots of Average RNFL Thickness with Age Across Three Ethnicities. B) Linear Mixed
Regression Line Plots of Quadrant Specific RNFL Thickness with Age Across Three Ethnicities
55
Abbreviations: AFEDS, African American Eye Disease Study; CHES, Chinese-American Eye Disease Study; LALES, Los Angeles
Latino Eye Study
Age-related loss is significant across all quadrants and ethnicities
56
Figure 2.4. Clock Hour of RNFL Thickness Decrease for Each 10-year Increase in Age Separated by Ethnicity
57
Supplementary Figure 2.1. Line Series Plot of RNFL Thickness Stratified by Ethnicity in Participants with Suspect or Untreated
Glaucoma
Abbreviations: C-A, Significant difference between CHES and AFEDS (Tukey’s P < 0.05), C-L, - Significant difference between
CHES and AFEDS (Tukey’s P < 0.05), C-L - Significant difference between CHES and LALES (Tukey’s P < 0.05), A-L*- -
58
Significant difference between AFEDS and LALES (Tukey’s P < 0.05); AFEDS, African American Eye Disease Study; CHES,
Chinese-American Eye Disease Study; LALES, Los Angeles Latino Eye Study
59
Supplementary Figure 2.2. Clock-Hour of RNFL Thickness Decrease for Each 10-year Increase in Age Separated by Ethnicity in
Participants with Suspect or Untreated Glaucoma
60
REFERENCES
1. Huang D, Swanson EA, Lin CP, et al. Optical coherence tomography. Science.
1991;254(5035):1178–81.
2. Alamouti B, Funk J. Retinal thickness decreases with age: An OCT study. Br. J.
Ophthalmol. 2003;87(7):899–901.
3. Knight OJ, Girkin CA, Budenz DL, et al. Effect of race, age, and axial length on
optic nerve head parameters and retinal nerve fiber layer thickness measured by
cirrus HD-OCT. Arch. Ophthalmol. 2012;130(3):312–318.
4. Varma R, Skaf M, Barron E. Retinal nerve fiber layer thickness in normal human
eyes. Ophthalmology. 1996;103(12):2114–2119.
5. Realini T, Zangwill LM, Flanagan JG, et al. Normative Databases for Imaging
Instrumentation. J. Glaucoma. 2015;24(6):480–483.
6. Health C for D and R. Workshops & Conferences (Medical Devices) - Public
Workshop - FDA/American Glaucoma Society Workshop on the Validity,
Reliability, and Usability of Glaucoma Imaging Devices, October 5, 2012. Center
for Devices and Radiological Health; (Accessed October 18,
2018).(https://www.fda.gov/MedicalDevices/NewsEvents/WorkshopsConferences/
ucm318305.htm#date). (Accessed October 18, 2018)
7. Budenz DL, Anderson DR, Varma R, et al. Determinants of Normal Retinal Nerve
Fiber Layer Thickness Measured by Stratus OCT. Ophthalmology.
2007;114(6):1046–1052.
8. Girkin CA, McGwin G, Sinai MJ, et al. Variation in optic nerve and macular
structure with age and race with spectral-domain optical coherence tomography.
Ophthalmology. 2011;118(12):2403–2408.
9. Chi T, Ritch R, Tsai C, et al. Racial differences in optic nerve head parameters.
Arch. Ophthalmol. 1989;107(6):836–839.
10. Racette L, Boden C, Kleinhandler SL, et al. Differences in visual function and
optic nerve structure between healthy eyes of blacks and whites. Arch.
Ophthalmol. 2005;123(11):1547–1553.
11. Marsh BC, Cantor LB, Wudunn D, et al. Optic nerve head (ONH) topographic
analysis by stratus OCT in normal subjects: Correlation to disc size, age, and
ethnicity. J. Glaucoma. 2010;19(5):310–318.
12. Seider MI, Lee RY, Wang D, et al. Optic disk size variability between African,
Asian, white, hispanic, and filipino Americans using heidelberg retinal
tomography. J. Glaucoma. 2009;18(8):595–600.
13. Biswas S, Lin C, Leung CKS. Evaluation of a myopic normative database for
analysis of retinal nerve fiber layer thickness. JAMA Ophthalmol.
2016;134(9):1032–1039.
14. Mwanza J-C, Oakley JD, Budenz DL, et al. Ability of Cirrus
TM
HD-OCT Optic
Nerve Head Parameters to Discriminate Normal from Glaucomatous Eyes.
61
Ophthalmology. 2011;118(2):241-248.e1.
15. Varma R, Paz SH, Azen SP, et al. The Los Angeles Latino Eye Study: design,
methods, and baseline data. Ophthalmology. 2004;111(6):1121–1131.
16. Varma R, Hsu C, Wang D, et al. The Chinese American eye study: Design and
methods. Ophthalmic Epidemiol. 2013;20(6):335–347.
17. McKean-Cowdin R, Fairbrother-Crisp A, Torres M, et al. The African American
Eye Disease Study: Design and Methods. Ophthalmic Epidemiol. 2018;25(4):306–
314.
18. Garcia-Martin E, Pinilla I, Idoipe M. Intra and interoperator reproducibility of
retinal nerve fibre and macular thickness ... Acta. 2010;3768.
19. Keane PA, Mand PS, Liakopoulos S, et al. Accuracy of retinal thickness
measurements obtained with Cirrus optical coherence tomography. Br. J.
Ophthalmol. 2009;93(11):1461–1467.
20. Cheung CY, Chen D, Wong TY, et al. Determinants of quantitative optic nerve
measurements using spectral domain optical coherence tomography in a
population-based sample of non-glaucomatous subjects. Investig. Ophthalmol. Vis.
Sci. 2011;52(13):9629–9635.
21. Khawaja AP, Chan MPY, Garway-Heath DF, et al. Associations with retinal nerve
fiber layer measures in the EPIC-Norfolk Eye Study. Invest. Ophthalmol. Vis. Sci.
2013;54(7):5028–34.
22. Zhang X, Francis BA, Dastiridou A, et al. Longitudinal and Cross-Sectional
Analyses of Age Effects on Retinal Nerve Fiber Layer and Ganglion Cell Complex
Thickness by Fourier-Domain OCT for the Advanced Imaging for Glaucoma
Study Group. Trans Vis Sci Tech. 2016;5(2):1–9.
23. Bowd C, Zangwill LM, Blumenthal EZ, et al. Imaging of the optic disc and retinal
nerve fiber layer: the effects of age, optic disc area, refractive error, and gender. J.
Opt. Soc. Am. Opt. image Sci. vision. 2002;19(1):197–207.
24. Patel NB, Lim M, Gajjar A, et al. Age-associated changes in the retinal nerve fiber
layer and optic nerve head. Investig. Ophthalmol. Vis. Sci. 2014;55(8):5134–5143.
25. Parikh RS, Parikh S, Sekhar GC, et al. Diagnostic Capability of Optical Coherence
Tomography (Stratus OCT 3) in Early Glaucoma. Ophthalmology.
2007;114(12):2238–2243.
26. Park SB, Sung KR, Kang SY, et al. Comparison of Glaucoma Diagnostic
Capabilities of Cirrus HD and Stratus Optical Coherence Tomography. Arch.
Ophthalmol. [electronic article]. 2009;127(12):1603.
(http://archopht.jamanetwork.com/article.aspx?doi=10.1001/archophthalmol.2009.
296). (Accessed October 18, 2018)
27. Kashani AH, Zimmer-Galler IE, Shah SM, et al. Retinal Thickness Analysis by
Race, Gender, and Age Using Stratus OCT. Am. J. Ophthalmol. 2010;149(3):496-
502.e1.
62
28. Glaucoma, Open-angle | National Eye Institute.
(https://nei.nih.gov/eyedata/glaucoma). (Accessed June 4, 2019)
29. Leung CKS, Ye C, Weinreb RN, et al. Impact of age-related change of retinal
nerve fiber layer and macular thicknesses on evaluation of glaucoma progression.
Ophthalmology. 2013;120(12):2485–2492.
30. Varma R, Tielsch JM, Quigley HA, et al. Race-, Age-, Gender-, and Refractive
Error—Related Differences in the Normal Optic Disc. Arch. Ophthalmol.
1994;112(8):1068–1076.
31. Oliveira C, Harizman N, Girkin CA, et al. Axial length and optic disc size in
normal eyes. Br. J. Ophthalmol. 2007;91(1):37–39.
32. Zheng F, Wu Z, Leung CKS. Detection of Bruch’s Membrane Opening in Healthy
Individuals and Glaucoma Patients with and without High Myopia.
Ophthalmology. 2018;125(10):1537–1546.
33. Ho H, Tham YC, Chee ML, et al. Retinal Nerve Fiber Layer Thickness in a
Multiethnic Normal Asian Population: The Singapore Epidemiology of Eye
Diseases Study. Ophthalmology. 2019;126(5):702–711.
34. Beck RW, Messner DK, Musch DC, et al. Is There a Racial Difference in
Physiologic Cup Size? Ophthalmology. 1985;92(7):873–876.
35. Tsai CS, Zangwill L, Gonzalez C, et al. Ethnic differences in optic nerve head
topography. J. Glaucoma. 1995;4(4):248–57.
36. Fechtner R. Review of Normative Database Construction in Available OCT
Models.
(https://www.fda.gov/downloads/MedicalDevices/NewsEvents/WorkshopsConfere
nces/UCM326697.pdf). (Accessed October 18, 2018)
37. González-García AO, Vizzeri G, Bowd C, et al. Reproducibility of RTVue Retinal
Nerve Fiber Layer Thickness and Optic Disc Measurements and Agreement with
Stratus Optical Coherence Tomography Measurements. Am. J. Ophthalmol.
2009;147(6):1067-1074.e1.
38. Varma R, Ying-Lai M, Francis BA, et al. Prevalence of open-angle glaucoma and
ocular hypertension in Latinos: The Los Angeles Latino Eye Study.
Ophthalmology. 2004;111(8):1439–1448.
39. Hougaard JL, Kessel L, Sander B, et al. Evaluation of heredity as a determinant of
retinal nerve fiber layer thickness as measured by optical coherence tomography.
Investig. Ophthalmol. Vis. Sci. 2003;44(7):3011–3016.
40. Nannini DR, Kim H, Fan F, et al. Genetic Risk Score Is Associated with Vertical
Cup-to-Disc Ratio and Improves Prediction of Primary Open-Angle Glaucoma in
Latinos. Ophthalmology. 2018;125(6):815–821.
41. Nannini DR, Torres M, Chen YDI, et al. A genome-wide association study of
vertical cup-disc ratio in a latino population. Investig. Ophthalmol. Vis. Sci.
2017;58(1):87–95.
63
Chapter 3. A Genome-Wide Association and Admixture Study of Retinal Nerve
Fiber Layer Thickness in a Latino Population
ABSTRACT
Imaging of the retinal nerve fiber layer (RNFL) has helped quantify the ganglion
cell complex. Previous studies across racial/ethnic groups suggests the role of heritability
in RNFL thickness. We 1) determined heritability of RNFL thickness, 2) identified
variants associated with RNFL thickness, and 3) performed admixture mapping of RNFL
thickness in Hispanic individuals. We used the population-based Los Angeles Latino Eye
Study (LALES) consisting of 2,112 individuals. Participants were genotyped on the
Illumina OmniExpress or Illumina SOL BeadChip. We performed linear mixed
regression to assess the association between single variants and RNFL thickness. We
calculated local ancestry for participants using the 1000 Genomes and performed linear
mixed regression for each ancestral population on RNFL thickness. Genetic heritability
of RNFL thickness ranged from 10%-46%. Using standard association mapping, we
identified two genome-wide significant SNPs associated RNFL thickness in the nasal and
inferior quadrant. One SNP, rs7916697 (within ATOH7), was associated with nasal
RNFL thickness. The other variant, rs10007907, was associated with RNFL thickness in
the inferior quadrant. Using admixture mapping, we identified that in 17q21, African
haplotypes were associated with lower RNFL thickness. This study provides the largest
study on the genetic contributions to RNFL thickness in Hispanics.
64
Optical coherence tomography (OCT) has allowed visualization and
quantification of the structure of the circumpapillary peripapillary retinal nerve fiber
layer (RNFL) (1). OCT quantifies the area around the optic nerve head (ONH), which is a
measure of the axons of the ganglion cell complex distributed over the entire retina.
Changes in RNFL thickness are associated with aging and optic neuropathies, including
glaucoma, conditions that are distinguished by degeneration of the ganglion cell axons
and changes of the neuroretinal rim (2–4). While age is suggested to be a determinant of
RNFL thickness (5), few studies have identified consistent factors associated with RNFL
thickness. Previous studies have shown that the ONH and RNFL thickness varies
significantly across racial/ethnic groups. For instance, compared to those of European
descent, Hispanics and Asians have higher RNFL thickness overall (5,6). There is
additional evidence for the difference in RNFL thickness between those of African
descent having thinner RNFL compared to those of European descent (3,7).
These studies across racial/ethnic groups also suggests the role of heritability in
RNFL thickness. There is evidence suggesting the high heritability of ocular traits (8),
and previous studies have examined the heritability using family-based design study
designs in characterizing the RNFL thickness. A twin study using monozygotic twins to
dizygotic twins estimated the heritability at 66% (9), which increased to 82% after
correction for age and other imaging factors (9). A family-based study estimated the
heritability of RNFL thickness at a range from 40%-49% (10). Due to this high
heritability discovered in family-based designs, further examining the genes and
associated function with RNFL will describe the functional processes underlying retinal
development. Other consortia have used genome-wide association mapping approaches to
65
discover functional variants within genes associated with ONH variation, including
vertical cup-disc ratio (11–13). Despite the studies that have examined the optic nerve
head, there are few that have examined the heritability in RNFL thickness, and to our
knowledge, only one genome-wide association study (GWAS) has been completed in
understanding the RNFL (14). Other groups have used targeted gene mapping approaches
by examining variants only in loci associated with glaucoma (SIX1/SIX6) (15,16).
However, these studies were completed in populations in those with predominantly
European or Asian ancestry.
There is evidence for the role of heritability and ancestry in the variation of RNFL
thickness, and leveraging the mixed genetic ancestry in Latino populations may help in
refining the mapping and improving estimation of the genes associated with various traits
(17). Those with Hispanic descent generally consist of three ancestral populations,
including European, Native American, and African ancestry; this technique identifies the
variation by ancestry groups and its association with a trait. Prior studies have examined
the relationship of global ancestry to various ophthalmologic traits, including glaucoma
and intraocular pressure (18,19). However, no studies have further subset this global
ancestry to the chromosomal level and identify chromosomal segments that are
associated with a trait. Admixture mapping approaches at the local ancestral level, rely on
the differences in the allele frequency at varying chromosomal segments across the entire
genome that differ across ancestral populations. Therefore, by performing admixture
mapping we will discover regions that differ between populations, but then further
identify individual variants that fall underneath a significant admixture peak. These
individual variants might not pass a standard genome-wide association mapping
66
approach, therefore, leveraging the differences in ancestral population limits the total
number regions that need to be mapped. These approaches may allow for identification of
variants not discovered using standard GWAS techniques.
The aims of the current study illustrated the following with regard to the genetic
features of RNFL thickness: we 1) characterized the overall heritability of RNFL
thickness, 2) identified individual single nucleotide polymorphisms (SNPs) associated
with RNFL thickness, 3) performed admixture mapping using local ancestry and
identified chromosomal regions associated with RNFL thickness, and finally 4) validated
these findings in an independent Chinese-American population. We utilized a sample of
2,112 individuals from the third wave of the population-based Los Angeles Latino Eye
Study (LALES).
MATERIALS AND METHODS
Study design and participants
For the aims of this study, we utilized the Los Angeles Latino Eye Study
(LALES), a population-based study of eye disease conducted in Los Angeles, California.
This study was conducted in the 8-year follow up (LALES III), which utilized genetics
and OCT imaging data conducted between 2010-2014. The full methods for this study
have been described previously (20–22). Participants were eligible for the study
depending on self-reported ethnicity, being 40 years and older at baseline and living
within respective census tracts of La Puente, a city within Los Angeles County, CA.
Trained interviewers completed in-home interviews, and participants were then invited to
complete clinical examinations by ophthalmic technicians and ophthalmologists. Written
informed consent was obtained for all participants. Study approval was obtained from the
67
Los Angeles County/University of Southern California Medical Center Institutional
Review Board and adhered to the tenets of the Declaration of Helsinki.
Optical coherence tomography imaging
OCT imaging of RNFL was performed using the Cirrus HD-OCT 4000 (Carl
Zeiss Meditec, Dublin, CA, USA). The trained ophthalmic technicians collected Optic
Disc Cube 200x200 scans for the right (OD) and left (OS) eye separately. This algorithm
identifies the center of the optic disc and automatically traces a 3.46 mm 360º circle
around the optic disc. To calculate RNFL thickness, the anterior and posterior boundaries
of the RNFL are delineated first. The system then calculates the RNFL thickness by
counting the number of pixels between the anterior and posterior boundaries along each
point on the circle of the A-scan.16 RNFL thickness measurements are categorized into
clock hour, quadrants, and by overall average. To assess ONH measures, the algorithm
identifies the termination of Bruch’s membrane to determine the disc edge. The software
then extracts the rim width around the optic disc by measuring the thickness of the
neuroretinal tissue in the optic nerve. Final segmentation data were extracted for all
subsequent analyses.
For quality control of OCT, a scan was retained if it had signal strength ≥7 out of
10, as recommended by the manufacturer for high quality (23,24). To exclude scans with
potential any segmentation errors, scans were excluded if any values for RNFL were
equal to zero. We assessed symmetry across eyes and excluded participants if the
differences between eyes was greater than 30µm. If participants had multiple quality
scans that passed our thresholds, only one randomly selected scan per eye was retained.
These quality control procedures ensured high-quality images used in the segmentation
68
for the RNFL thickness values. From the comprehensive in-clinic ocular examinations,
which collected ophthalmic clinical and demographic information, we additionally
extracted covariate data including gender and age.
Genotyping and imputation
LALES participants were genotyped using either the Illumina OmniExpress
BeadChip (~730k markers, Illumina, San Diego, CA) for n=4,122 samples or the
Illumina Hispanic/SOL BeadChip (~2.5 million markers, Illumina, San Diego, CA) for
n=716. Genotyping was performed at the Los Angeles Biomedical Research Institute. To
identify more variants across the entire genome and to allow for the harmonization of the
two datasets, we performed genetic imputation. In both datasets, all SNPs were filtered
on a genotype call rate of 90% and excluded SNPs with a Hardy-Weinberg equilibrium
P-values less than 10
-8
. Additionally, we removed any monomorphic SNPs, or variants
with alleles that could not be matched to the correct forward strand in the hg19 genome
assembly. For the OmniExpress and SOL arrays, this retained 697,553 and 1,608,650
markers respectively. For sample level quality control, we removed any duplicate
samples and missingness per individual at 90%. This retained 3,642 and 706 individuals
for downstream imputation.
Due to differences in arrays used and data freezes at respective time points,
participants were imputed in two batches. An initial imputation was done using the
OmniExpress array completed in 2015. Imputation was performed using SHAPEIT
(Version 2, available at
https://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html) (25) to phase
samples followed by IMPUTE2 (26) for imputation. We utilized the entire 1000
69
Genomes Project (1KGP) as the reference panel. The second batch of imputation was
performed on samples genotyped on the SOL array. Phasing and imputation were
performed using EAGLE2 (27) and positional Burrows-Wheeler transform (PBWT) (28),
respectively, using the Sanger Imputation Server (29). We utilized the same 1KGP
reference population.
Data across both imputation batches were combined. Before merging the two
batches, all variants were filtered on an Information score >0.4. The information score
was calculated variance of each genotype probability is considered for the uncertainty of
a genotype call. Duplicate samples across the two arrays were removed retaining 4,406
individuals for all subsequent analyses. After combining the OCT phenotype data with
this dataset, we retained 2,112 individuals for all analyses of RNFL thickness.
Additionally, a new imputation Information score, similar to the IMPUTE2 score, was
calculated for the final harmonized set. This final dataset resulted in 13,122,188 SNPs
and when restricted to SNPS with MAF>0.01, all downstream analyses used 4,918,941
variants.
Statistical methods
Ancestry estimation.
We used RFMix (30) (Version 2, https://github.com/slowkoni/rfmix) to estimate
the proportion of European (EUR), AmerIndian, (AMR), and African (AFR) local
ancestry across all autosomes in this Latino population using the recommended minimum
node size of 5. RFMix uses a sliding window to infer ancestry within a window using a
conditional random field parameterized by random forests. We used the 1KGP population
and extracted all individuals with a superpopulation of EUR, AMR, or AFR and matched
70
all common variants between both datasets. Additionally, we excluded the first and the
last 2 Mb from the telomeres of each chromosome due to likely inaccurate reconstruction
of local ancestry in the regions (31). This created 22,391 intervals that spanned length of
all autosomes. We then created a measure of global ancestry by taking the local ancestry
proportion and summing and averaging across the length of each region to create a global
measure.
Genome-wide association mapping.
We used linear mixed regression models as implemented in the GEMMA
software (32) to determine the association between SNPs and the RNFL measures. This
approach is an efficient exact method, where a kinship matrix between all samples is
constructed and then incorporated into all downstream association tests to account for
both cryptic relatedness and population stratification. We used a leave-one-chromosome-
out approach in construction of the kinship matrix to avoid proximal contamination (33).
After constructing the relatedness matrix, we then tested the association between each
variant and RNFL (average and at each quadrant) adjusting for age and gender. We did
not control for global ancestry, due to population stratification being controlled for using
the mixed model approach. Single variants were declared genome-wide significant at
p<5x10
-8
. Additionally, we calculated the genomic inflation value for all GWAS, to
determine to level of population stratification in residual confounding of our results.
Heritability and admixture mapping.
To assess the overall genetic heritability of traits, we used the Genome-wide
Complete Trait Analysis (GCTA) (34) software (v1.92.0) estimate variance explained by
all the SNPs. We constructed genetic relationship matrices (GRM) to determine the
71
genetic relatedness between individuals and then used the GCTA-GREML (Genome-
based restricted maximum likelihood) module to calculate the variance explained in
RNFL thickness by all variants combined.
We used admixture mapping techniques to identify additional regions and variants
associated with RNFL thickness that may not have been captured using standard
association approaches (35). We then conducted association analyses between our
measures and RNFL thickness using two measures levels: global and local ancestry. To
conduct the global association analyses, we used a linear regression model where RNFL
thickness measures were regressed on proportion of AFR and AMR ancestry, adjusting
for age and sex. Local ancestry effects were estimated using the number of counts of an
individual’s ancestry (ranging from 0-2 for European, AmerIndian, and African,
separately) across each region as the unit of analysis. Then the mapping technique was
performed similar manner as the single variant analysis, where each count of ancestry is
regressed on RNFL thickness adjusting for age and sex using the linear mixed models as
implemented in GEMMA. Local ancestry regions were declared significant at P <
5.7x10
-5
, determined based on simulation to yield a genome-wide type I error of 0.05 in a
Hispanic population (36).
We examined all single variants within the significantly associated local ancestry
regions. Significant loci discovered in local ancestry analyses were then tested at the
individual SNP level to discover potentially missed associations from standard
association mapping techniques (37,38). P-values were adjusted for the number of tests
performed underneath a local ancestry region. Analyses were performed with the R
software (version 3.5.0).
72
Functional annotation.
To annotate all discovered regions and variants, we used the Functional Mapping
and Annotation (FUMA) software to visualize the results and perform functional
annotation (39). Additionally, within the FUMA software, we were able to use gene-level
mapping to determine if SNPs that were within gene were also associated with RNFL
thickness measures. This was completed using Multi-marker Analysis of GenoMic
Annotation (MAGMA) (40), where SNPs were mapped to potential genes if the variants
were within a 10kb window of a gene. The MAGMA gene-based analysis accounts for
gene size and LD structure, and uses a multiple linear principal components regression
model, and utilizes an F-test to compute the gene p-value. Gene-based analyses were
declared statistically significant at P < 2.67x10
-6
because SNPs were mapped to 18,711
protein-coding genes and corrected to an overall false discovery level of 0.05.
Validation.
To replicate the association signals discovered in LALES, we used an
independent dataset of Chinese Americans derived from the Chinese American Eye
Study (21). This study contained 1484 samples with complete genotyping, OCT imaging
and covariate data. Genotyping of this dataset was performed on the Infinium
HumanExome BeadChip with additional custom SNPs associated with various
ophthalmologic conditions. Due to the discrepancies in arrays used across the studies, not
all variants discovered in the LALES association study were available on this array.
Therefore, we extracted all available SNPs that either in linkage disequilibrium at R
2
≥0.2
or within 100 kb of a significant locus and tested the relationship between RNFL
thickness and SNPs within that region, using the top SNP as the marker for the locus. The
73
association between RNFL thickness and SNPs were tested using a multiple linear
regression of adjusted for sex, age, and the first three principal components.
RESULTS
Study population
The LALES participants with full OCT and genotype data are described in Table
1 for 2,112 individuals. The average age of this population was 62.2 (SD=8.7) years.
Similar to the base population, most of the study population was female (60.5%).
Additionally, there appeared to be no differences between the participants with available
genotyping and imaging data compared to the participants with only OCT imaging data
available (all P > 0.05). The RNFL thickness in aggregate was 93.51 µm (SD=11.2) and
ranged from 63.12 µm in the temporal quadrant to 123.05 µm in the inferior quadrant.
RNFL thickness measured values are similar between those included in the genetic study
compared to those that were not (P > 0.05). Additionally, our sample is comprised of
Hispanics (primarily of Mexican origin), that constitute mostly AmerIndian, followed by
European, and African descent in their global ancestry.
Genome-wide association mapping results
Table 3.2 highlights all SNPs that were associated either at the genome-wide
significant level (P < 5x10
-8
) or suggestive (P < 1x10
-6
) with RNFL thickness across all
quadrants and averaged using linear mixed model regression. Figure 1a shows the
Manhattan plot P-values in the nasal quadrant. The most significant SNP, which also
reached genome-wide significance, was rs7916697. This association was observed in the
nasal quadrant in chromosome 10:69991853 (hg19 position) with P = 1.25x10
-10
. The
effect allele (G, EAF=0.622) was associated with an increase in nasal quadrant RNFL
74
thickness [β (SE)= 2.28 (0.35)]. This SNP is located within ATOH7 gene. Another SNP
in this region (rs10762199) was associated with nasal quadrant RNFL thickness and
reached genome-wide significance. However, these two SNPs were in relatively high LD
with one another (R
2
=0.59) in the 1000 Genomes AmerIndian population and were
considered as one combined locus.
The results for the single variant analyses are shown in Figures 3.2 and 3.3.
Figure 1b highlights the next genome-wide significant SNP, rs10007907, located at
chromosome 4:168444121 (4q32.3) for the inferior quadrant. The effect allele (T,
EAF=0.02) was associated with an increase in microns of inferior RNFL thickness [β
(SE)= 11.57 (3.00)]. This SNP is in an intergenic region; however, the closest genes are
the SPOCK3 and ANXA10 genes, (288 kb and 57 kb) away respectively. Additional
suggestive SNP associations are highlighted in Table 2. The suggestive SNP to highlight
is rs146732699 in chromosome 1:220815581. The effect allele (A, EAF=0.021) was
associated with a decrease in superior RNFL thickness [β (SE)= -8,78 (1.75), P =
5.46x10
-7
]. The variant is within the MARK1 (microtubule affinity-regulating kinase 1)
gene. Additionally, in chromosome 6:161520743 the SNP rs7762953, was associated
with average decrease in RNFL thickness. The effect allele (C, EAF=0.016) was
associated with decreasing superior RNFL thickness [β (SE)= -9.58 (1.93), P = 7.55x10
-
7
]. The variant is within MAP3K gene (mitogen-activated protein kinase). Lastly, the
variant rs80289734, located in in chromosome 4:21643440, was associated with inferior
RNFL thickness. The effect allele (T, EAF=0.014) was associated with higher RNFL
thickness [β (SE)= 11.89 (2.41), P = 8.37x10
-7
]. This SNP is within an intron region of
the KCNIP4 (Potassium voltage-gated channel interacting protein 4) gene. The
75
Manhattan plots for SNP for the superior quadrant, temporal quadrant, and average
RNFL thickness are shown in Supplementary Figures 1-3. Across all quadrants and in
aggregate, the genomic control values ranged from 1.017 in the superior quadrant to 1.03
in the nasal quadrant. The use of linear mixed-models and the genomic control values,
through the GEMMA software, and indicate that both cryptic relatedness and population
stratification has been accounted for adequately.
In Figure 2, we show the results of the MAGMA gene-based analyses of the
association between SNPs mapped to genes and RNFL thickness. In the nasal quadrant,
three genes (ATOH7, MYPN, and PBLD2) within chromosome 10 were associated with
RNFL thickness (gene level P = 1.85x10
-7
, P = 7.22x10
-7
, P = 2.27x10
-6
, respectively).
Additionally, NCLN was significantly associated with RNFL thickness in the temporal
quadrant (gene P = 8.38x10
-7
). Lastly, one suggestive association was observed between
ECE2 and average RNFL thickness (gene P = 3.09x10
-6
).
Heritability and Admixture Mapping
We identified the total proportion of heritability that can be due to genetic
variation using a family-based approach (by estimating two genetic related matrices
(GRM), and these results are summarized in Table 3. The genetic heritability for RNFL
thickness estimates using genome-based restricted maximum likelihood (GREML)
ranged from 10% to 46%, and all were statistically significant except for in the superior
quadrant. This indicates that there are statistically significant contributors to RNFL
thickness beside environmental components.
When we examined the association between overall global ancestry and RNFL
thickness, there are modest associations as shown in Table 3. In general, increasing
76
proportion of African ancestry, while adjusting for AmerIndian ancestry, was associated
with thinner RNFL across all quadrants, however these effects were not statistically
significant. Betas for the effect ranged from -1.7 in the nasal quadrant to -17.62 in the
inferior quadrant. With AmerIndian ancestry, we observed a statistically significant
association between increasing proportion of ancestry and increasing thickness of RNFL
in the inferior quadrant adjusting for age, sex, and African descent [β (SE)= 4.97 (2.52) P
< 0.05].
We estimated the local ancestry at all autosomes for our participants; we mapped
22,391 intervals that differed in the three ancestral populations for our participants that
spanned a mean length of 119,736 bp (ranging from 254- 2,4190,684 bp). The intervals
were constructed using between 5-24315 variants, with a mean of 560.8 variants. When
comparing the local ancestry estimates to its association with RNFL thickness, adjusting
for age and sex when using a linear mixed model. We observed 50 total regions that
passed the genome-wide local ancestry admixture mapping threshold. Three of the
regions for European descent were associated with RNFL thickness and 47 regions were
discovered for African descent. For instance, for the analysis of European descent, we
observed that in the region chromosome 18: 64251191-64503343, having European
descent in this region was associated with a decrease in superior quadrant RNFL
thickness [β (SE)=-2.00 (0.48), P = 3.65x10
-5
]. However, no single SNPs within this
region were associated with RNFL thickness after adjusting for the total number of
comparisons. Results of the admixture mapping where SNPs underneath any admixture
mapping peak were also significant after adjusting for the false discovery rate are
highlighted (Table 4). For instance, in chromosome 17:49748954-49897766, we observed
77
that African descent was associated with lower average RNFL thickness [β (SE)=-3.00
(0.71), P = 2.60x10
-5
]. Underneath this peak, the strongest association SNP was
rs28655804, where the effect allele (G, EAF=0.019) was associated with thinner RNFL
[β (SE)=-5.32 (1.23), P =1.63x10
-5
, FDR Q = 0.001]. This SNP is within the intron
region of CA10 (carbonic anhydrase 10. This variant’s allele frequency differs between
those of European, African, and AmerIndian descent in the 1000 Genomes population
(AF=0.001, 0.294, and 0.023, respectively). Another region chromosome 17:46826047-
47034177 had one SNP, rs16944762, and the effect allele (A, AF=0.076) was associated
with decreased average RNFL thickness [β (SE)=-2.37 (0.60), P = 8.48x10
-5
, FDR Q =
0.014]. The allele frequencies also differed between African and European and
AmerIndian descent (AF=0.484, 0.050, and 0.063, respectively). The SNP was within the
intronic region of the gene SNF8.
Lastly, for any variants discovered in single variant or local ancestry admixture
analyses, we attempted to validate the results in all other quadrant measures within the
same Hispanic population and with an external dataset of Chinese-Americans, from the
Chinese-American Eye Study (CHES) (21). We extracted any SNPs that were in LD at
R
2
>0.2 or within 100 kb of the target SNP and used the new SNP as the marker for the
locus. The most significant SNP, rs7916697, discovered in the nasal region did not reach
genome-wide significance in the average, inferior, or temporal region and was not
associated with superior RNFL thickness in the LALES population. However, the effect
of the SNP was in the same direction across all significant quadrants. This variant was
associated with temporal quadrant RNFL thickness in the Chinese-American cohort (P =
0.004), however the effect allele indicated a lower RNFL thickness for each additional
78
effect variant. The second genome-wide significant SNP (rs10007907) was replicated in
the average, nasal, and superior quadrant RNFL thickness, and additionally, the effect
alleles were consistently in the same direction. Lastly, the SNPs discovered using
admixture local ancestry mapping was validated at all quadrants. The effect alleles
affected RNFL thickness in the same magnitude. These two variants were not genotyped
in the Chinese-American validation dataset, and we were therefore not able to validate
these results.
DISCUSSION
This study used joint association and admixture mapping techniques to identify
the loci associated with RNFL thickness, and to our knowledge this is the first genetic
mapping study of RNFL thickness completed. Specifically, we leveraged the ethnic and
genetic diversity to perform this study in a Hispanic population. In our standard
association mapping techniques, we identified two genome-wide significant SNPs
associated with OCT measured RNFL thickness in the nasal and inferior quadrant. The
first of these SNPs, rs7916697, in within or nearby the genes ATOH7, MYPN, and PBLD.
The other variant, rs10007907, was associated with RNFL thickness in the inferior
quadrant. Additionally, we discovered five other unique variants that had suggestive
associations with RNFL thickness. Gene-based analyses in the temporal quadrant and in
average suggested the role of the genes NCLN and ECE2, respectively. By leveraging
admixture in this population due to three ancestral populations (European, African, and
AmerIndian), we discovered a region of chromosome 17q21, where an increasing number
of African ancestral haplotypes was associated with lower RNFL thickness. After
limiting the search to this region, we identified that these SNPs were within the genes
79
CA10 and SNF8W. These variants identified in this study would not have passed a
standard association mapping approach.
SNP-based analyses discovered the most significant SNP, rs7916697. This
variant and others in high linkage disequilibrium in the 10q21.3 locus have been
examined in other GWAS studies related to retinal structures, however, these studies
were in relationship to optic disc size (12,13), vertical cup-disc ratio (11), or glaucoma
(41). This variant is within the ATOH7 gene, which have been studied in its role as a
transcription factor responsible for retinal ganglion cell development (42,43).
Additionally, the other genes in the gene-based approach that were significantly
associated with RNFL thickness, MYPN has found to be expressed in rod bipolar cells
(44). The other gene, PBLD is not expressed in developing eyes (43). The linkage
disequilibrium pattern in this region may have emphasized the genes that may not have as
much of an effect. We did not adjust for optic disc size in our main results, due to the
correlation between optic disc area and RNFL thickness. Histological studies have shown
that optic nerve fiber count will increase with enlarged ONH (45,46). As a sensitivity
analysis, we did adjust for optic disc, and this did attenuate the signal in this region to no
longer being statistically significant. However, the lead SNP remained highly significant
(P = 7.37 x 10
-7
). This indicates that this region is strongly associated with retinal
ganglion cell development.
The second genome-wide significant SNP is rs10007907, in 4q32.3, which is in
an intergenic region near ANXA10 and SPOCK3. While there is no known function of
ANXA10 in the retina, deficiency in a related familial gene, ANXA2, was associated with
problems in fibroblast growth in a rat neonatal retina (47). SPOCK3 is found to be
80
expressed in a novel type of retinal bipolar cell, when measuring using single cell
transcriptomics approaches (48). The NCLN gene, which was associated in the temporal
quadrant RNFL thickness, is important in the Nodal/TGF-beta signaling pathway, which
is involved in cell fate decisions during embryonic development (49).
Using our admixture mapping approach, we identified regions and additionally
variants associated with African ancestry and RNFL thickness. The top variant within
one of the mapped loci was rs28655804, which is in the gene CA10. This gene and
corresponding protein are predominantly expressed in the CNS (50,51). This gene binds
with neurexin to form a complex that is important in signaling and neuronal development
(50). This gene was also discovered to be significantly downregulated in canine
glaucoma and suggested to have function in retinal neuroinflammation (52). Another
variant discovered through admixture mapping, rs16944762, was within the gene SNF8.
This gene encodes a vacuolar sorting-associated protein that has been shown to be
functional within retinal progenitor cells (53).
When examining the overall representation of local ancestry in our population, the
admixture mapping approaches discovered regions of African descent in our participants
that were associated with RNFL thickness. We observed that increasing regions of
African descent genome were associated with lower RNFL thickness. This is consistent
with previous findings and our own, where African-American ethnicities tended to have
thinner RNFL compared to those who self-reported as White, Hispanic, and Chinese
(3,5,54). Other studies have also leveraged local ancestry at known glaucoma genes and
identified that African descent was associated with increased glaucoma risk (18). This
may be important in determining the normative values across different populations, as the
81
clinical interpretations for certain conditions, including glaucoma, may vary depending
on what may be heritable compared to pathological deterioration (3).
One of the strengths of this study is that we performed this analysis in a
population-based study of eye disease in a Hispanic population. Other studies rely on
crude estimates of global ancestry to control for the potential role of admixture (55), but
we adequately controlled for both population stratification and cryptic relatedness in this
population using linear mixed modeling. Another strength of this study is that we used
admixture mapping to complement the standard genome-wide association mapping. By
leveraging the multiple level admixture in this Hispanic group, we identified a novel
region associated with RNFL thickness. Previous studies have examined the relationship
in ophthalmologic traits using global measures (19,56), which may mask the true
functional regions/variants. Lastly, we were able to validate some of our discoveries in an
independent dataset and in a Chinese population. However, the results were not
consistent for all quadrants.
One limitation of this study is that we used multiple regions of the RNFL for
analysis. The quadrants are highly correlated, however all quadrants and the average
RNFL thickness are important in diagnosing optic neuropathies. While the results were
generally in agreement, we did not always find the same effect at each quadrant. For
subsequent studies, we hope to develop further methods that will account for the repeated
measures and the correlation amongst these measures to better quantify the amount of
genetic variation that is responsible for RNFL thickness variability. Additionally, we
were not able to validate our findings in another Latino population, but this would be of
82
utmost importance, as glaucoma and other optic neuropathies have a high prevalence in
this population (57).
In summary, we have found that RNFL thickness is a heritable trait using novel
association and admixture mapping techniques. As this study contained the largest
sample of Hispanics, we leveraged the power to perform local ancestry admixture
mapping alongside association techniques. Specifically, we have discovered and also
validated previous the potential role of the 10q21.3 region on changes in the retina.
Additionally, we discovered novel regions using both association and admixture mapping
techniques within the 4q32.3 and 17q21 regions. Further validation and functional studies
are needed to implicate the underlying pathology of the retina.
83
Table 3.1. Analytical Cohort of the Los Angeles Latino Eye Study Compared to Participants with Imaging Data Only
Covariate LALES with Imaging + Genotyping
(n=2112)
Mean (SD)
LALES with Imaging only
(n=2650)
Mean (SD)
P-value of
differences
Age (10 years) 6.23 (0.87) 6.22 (0.87) 0.60
Female - n (%) 1271 (60.18) 1603 (60.49) 0.85
Temporal RNFL 63.12 (10.75) 63.03 (10.68) 0.79
Superior RNFL 114.79 (17.09) 114.74 (17.26) 0.91
Nasal RNFL 73.2 (11.51) 73.22 (11.43) 0.94
Inferior RNFL 123.05 (19.03) 123.05 (19.22) 0.99
Average RNFL 93.54 (11.2) 93.51 (11.27) 0.94
Global
Ancestry
African 0.05 (0.04)
AmerIndian 0.561 (0.16)
European 0.387 (0.16)
84
Table 3.2. Top Summary Results for Association Mapping of Variants and RNFL Thickness Across All Quadrants in the Los Angeles
Latino Eye Study.
rsID CHR BP EA EAF Beta SE P nSNPs Loci start Loci end Quadrant
rs7916697 10 69991853 G 0.622 2.28 0.35 1.25x10
-10
568 69909562 70254610 Nasal
rs10007907 4 168444121 T 0.02 11.57 2.00 8.00 x10
-9
22 168403745 168470026 Inferior
rs6857031 4 64906630 C 0.01 8.33 1.62 3.10 x10
-7
62 64587322 65000141 Average
rs17812521 8 298274 A 0.107 2.85 0.56 3.94 x10
-7
54 272839 431501 Nasal
rs146732699 1 220815581 A 0.021 -8.78 1.75 5.46 x10
-7
182 220672171 220902172 Superior
rs7762953 6 161520743 C 0.016 -9.58 1.93 7.55 x10
-7
12 161440366 161527784 Superior
rs80289734 4 21643440 T 0.014 11.89 2.41 8.37 x10
-7
3 21643440 21671728 Inferior
Abbreviations: CHR, chromosome; BP, base pair; RA, reference allele; EA, effect allele; nSNPs, total SNPs within a loci
85
Table 3.3. Overall Genetic Heritability Estimates of RNFL Thickness and Effect of Global Ancestry on RNFL Thickness in the Los
Angeles Latino Eye Study
Heritability Global Ancestry
a
Region R2 (SE) AMR – β (SE) AFR – β (SE)
Temporal RNFL 0.193 (0.118)* 2.26 (1.5) -9.93 (6.19)
Superior RNFL 0.095 (0.116) -1.48 (2.29) -3.15 (9.44)
Nasal RNFL 0.46 (0.12)* -2.9 (1.59) -1.7 (6.56)
Inferior RNFL 0.123 (0.099)* 4.97 (2.52)* -17.62 (10.39)
Average RNFL 0.218 (0.115)* 0.71 (1.49) -8.03 (6.14)
Abbreviations: AMR: AmerIndian, AFR: African
*-p<0.05
a
European as referent population
86
Table 3.4. Summary Results of Local African Ancestry Associated with RNFL thickness in the Los Angeles Latino Eye Study
Loci
AFR
Ancestry
β (SE)
Ancestry
P
Top SNP
Top SNP
BP
EA EAF β (SE) P FDR Quadrant
EUR
Freq
AFR
Freq
AMR
Freq
17:46826047-
47034177
-3.68 (0.72)
3.81x10
-7
rs16944762 47007497 A 0.076 -2.37 (0.6) 8.48 x10
-5
0.014 A 0.05 0.48 0.06
17:46554688-
46826047
-3.34 (0.69)
1.29 x10
-6
rs9906920 46577130 G 0.981 4.76 (1.21) 8.21 x10
-5
0.008 A 1.00 0.78 0.97
17:49748954-
49897766
-3 (0.71)
2.60 x10
-5
rs28655804 49764963 G 0.019 -5.32 (1.23) 1.63 x10
-5
0.001 A 0.00 0.29 0.02
17:49694262-
49748954
-3 (0.71)
2.60 x10
-5
rs113335834 49732063 A 0.094 1.93 (0.56) 6.23 x10
-4
0.019 A 0.16 0.01 0.11
17:47034177-
47211560
-4.64 (1.11)
3.10 x10
-5
rs872539 47196993 G 0.726 1.73 (0.57) 2.41 x10
-3
0.036 S 0.79 0.68 0.76
17:49748954-
49897766
-4.91 (1.21)
4.76 x10
-5
rs9904278 49763099 T 0.018 -8.63 (2.11) 4.61 x10
-5
0.002 I 0.00 0.29 0.02
17:46826047-
47034177
-5 (1.23)
4.86 x10
-5
rs16944762 47007497 A 0.076 -3.79 (1.02) 2.07 x10
-4
0.031 I 0.05 0.48 0.06
Abbreviations: CHR, chromosome; BP, base pair; EA, effect allele; EAF, effect allele frequency; A, average quadrant; S, superior
quadrant; I, inferior quadrant; Freq, frequency; EUR, European; AFR, African; AMR, AmerIndian
87
Table 3.5. Validation of Top SNPs P-values and the Association with RNFL Thickness in the Los Angeles Latino Eye Study and the
Chinese American Eye Study
Latinos Temporal
RNFL
Superior
RNFL
Nasal RNFL Inferior
RNFL
Average
RNFL
rs7916697 2.37 x 10
-4
* 2.44 x 10
-1
1.25 x 10
-10
* 1.95 x 10
-4
* 1.26 x 10
-5
*
rs10007907 2.69 x 10
-1
8.33 x 10
-3
* 8.93 x 10
-6
* 8.00 x 10
-9
* 1.06 x 10
-6
*
rs28655804 4.79 x 10
-4
* 2.77 x 10
-3
* 3.84 x 10
-4
* 4.61 x 10
-5
* 1.63 x 10
-5
*
Chinese Americans
rs7916697 4.17 x 10
-3
* 1.90 x 10
-1
6.27 x 10
-2
1.94 x 10
-1
9.21 x 10
-2
*p<0.05
88
Figure 3.1. A) Global Ancestry for All Participants in the Los Angeles Latino Eye Study. B) Local Ancestry for Three Randomly
Selected Individuals
AFR
EUR
AMR
UNK
89
Global and local ancestry was estimated using RFMix in the Los Angeles Latino Eye Study population. Abbreviations: AFR, African
descent; EUR, European descent; AMR, AmerIndian descent; UNK, Unknown
90
Figure 3.2. Manhattan Genome-wide Association Plot for A) Nasal Quadrant B) Inferior
Quadrant RNFL Thickness
The genome-wide association study was completed in the Los Angeles Latino Eye Study
population after imputation to approximately four million variants. The red line indicates
genome-wide significant associations (P = 5 x10
-8
)
91
Figure 3.3. Manhattan Gene-Based Plot for A) Nasal Quadrant and B) Temporal
Quadrant RNFL Thickness
A gene-based association was completed in the Los Angeles Latino Eye Study population
after imputation to approximately four million variants and assigned to 18.711 protein
coding genes. The red line indicates genome-wide significant associations (P = 2.67 x10
-
6
)
92
Figure 3.4. Manhattan Local Ancestry Plot for Inferior Quadrant RNFL Thickness and
Local African Ancestry
Genome-wide local ancestry association was completed in the Los Angeles Latino Eye
Study population using RFMix. The red line indicates genome-wide significant
associations (P = 5.6 x10
-5
)
93
Supplementary Figure 3.1. Manhattan A) Genome-Wide Association Plot for Average
RNFL Thickness and B) Gene-Based Plot for Average RNFL Thickness
Both the genome-wide and gene-based association studies were completed in the Los
Angeles Latino Eye Study population after imputation to approximately four million
variants and assigned to 18.711 protein coding genes. The red line indicates genome-wide
significant associations (P = 2.67 x10
-6
)
94
Supplementary Figure 3.2. Manhattan A) Genome-Wide Association Plot for Superior
RNFL Thickness and B) Gene-Based Plot for Superior RNFL Thickness
Both the genome-wide and gene-based association studies were completed in the Los
Angeles Latino Eye Study population after imputation to approximately four million
variants and assigned to 18.711 protein coding genes. The red line indicates genome-wide
significant associations (P = 2.67 x10
-6
)
95
Supplementary Figure 3.3. Manhattan Genome-wide Association Plot for Temporal
Quadrant RNFL Thickness
The genome-wide association study was completed in the Los Angeles Latino Eye Study
population after imputation to approximately four million variants. The red line indicates
genome-wide significant associations (P = 5 x10
-8
)
96
Supplementary Figure 3.4. Manhattan Gene-Based Plot for Inferior Quadrant RNFL
Thickness
A gene-based association was completed in the Los Angeles Latino Eye Study population
after imputation to approximately four million variants and assigned to 18.711 protein
coding genes. The red line indicates genome-wide significant associations (P = 2.67 x10
-
6
)
97
REFERENCES
1. Huang D, Swanson EA, Lin CP, et al. Optical coherence tomography. Science.
1991;254(5035):1178–81.
2. Alamouti B, Funk J. Retinal thickness decreases with age: An OCT study. Br. J.
Ophthalmol. 2003;87(7):899–901.
3. Knight OJ, Girkin CA, Budenz DL, et al. Effect of race, age, and axial length on
optic nerve head parameters and retinal nerve fiber layer thickness measured by
cirrus HD-OCT. Arch. Ophthalmol. 2012;130(3):312–318.
4. Varma R, Skaf M, Barron E. Retinal nerve fiber layer thickness in normal human
eyes. Ophthalmology. 1996;103(12):2114–2119.
5. Budenz DL, Anderson DR, Varma R, et al. Determinants of Normal Retinal Nerve
Fiber Layer Thickness Measured by Stratus OCT. Ophthalmology.
2007;114(6):1046–1052.
6. Girkin CA, McGwin G, Sinai MJ, et al. Variation in optic nerve and macular
structure with age and race with spectral-domain optical coherence tomography.
Ophthalmology. 2011;118(12):2403–2408.
7. Girkin CA, Sample PA, Liebmann JM, et al. African Descent and Glaucoma
Evaluation Study (ADAGES): II. Ancestry differences in optic disc, retinal nerve
fiber layer, and macular structure in healthy subjects. Arch. Ophthalmol.
2010;128(5):541–550.
8. Sanfilippo PG, Hewitt AW, Hammond CJ, et al. The Heritability of Ocular Traits.
Surv. Ophthalmol. 2010;55(6):561–583.
9. Hougaard JL, Kessel L, Sander B, et al. Evaluation of heredity as a determinant of
retinal nerve fiber layer thickness as measured by optical coherence tomography.
Investig. Ophthalmol. Vis. Sci. 2003;44(7):3011–3016.
10. Van Koolwijk LME, Despriet DDG, Van Duijn CM, et al. Genetic contributions to
glaucoma: Heritability of intraocular pressure, retinal nerve fiber layer thickness,
and optic disc morphology. Investig. Ophthalmol. Vis. Sci. 2007;48(8):3669–3676.
11. Nannini DR, Torres M, Chen YDI, et al. A genome-wide association study of
vertical cup-disc ratio in a latino population. Investig. Ophthalmol. Vis. Sci.
2017;58(1):87–95.
12. Ramdas WD, van Koolwijk LME, Ikram MK, et al. A genome-wide association
study of optic disc parameters. PLoS Genet. 2010;6(6):1–12.
13. Springelkamp H, Mishra A, Hysi PG, et al. Meta-analysis of Genome-Wide
Association Studies Identifies Novel Loci Associated With Optic Disc
Morphology. Genet. Epidemiol. 2015;39(3):207–216.
14. Venturini C, Hysi P, Nag A, et al. Genetic factors and pathways affecting retinal
nerve fiber layer thickness. Invest. Ophthalmol. Vis. Sci. 2013;54(15):4500–4500.
15. Yoshikawa M, Yamashiro K, Nakanishi H, et al. Association of SIX1/SIX6 locus
98
polymorphisms with regional circumpapillary retinal nerve fibre layer thickness:
The Nagahama study. Sci. Rep. 2017;7(1):4393.
16. Kuo JZ, Zangwill LM, Medeiros FA, et al. Quantitative trait locus analysis of
SIX1-SIX6 with retinal nerve fiber layer thickness in individuals of European
descent. Am. J. Ophthalmol. 2015;160(1):123-130.e1.
17. Brown LA, Sofer T, Stilp AM, et al. Admixture Mapping Identifies an Amerindian
Ancestry Locus Associated with Albuminuria in Hispanics in the United States. J.
Am. Soc. Nephrol. 2017;28(7):2211–2220.
18. Restrepo NA, Laper SM, Farber-Eger E, et al. Local genetic ancestry in CDKN2B-
AS1 is associated with primary open-angle glaucoma in an African American
cohort extracted from de-identified electronic health records. BMC Med.
Genomics. 2018;11(S3):70.
19. Nannini D, Torres M, Chen YDI, et al. African Ancestry Is Associated with
Higher Intraocular Pressure in Latinos. Ophthalmology. 2016;123(1):102–108.
20. Varma R, Paz SH, Azen SP, et al. The Los Angeles Latino Eye Study: design,
methods, and baseline data. Ophthalmology. 2004;111(6):1121–1131.
21. Varma R, Hsu C, Wang D, et al. The Chinese American eye study: Design and
methods. Ophthalmic Epidemiol. 2013;20(6):335–347.
22. McKean-Cowdin R, Fairbrother-Crisp A, Torres M, et al. The African American
Eye Disease Study: Design and Methods. Ophthalmic Epidemiol. 2018;25(4):306–
314.
23. Garcia-Martin E, Pinilla I, Idoipe M. Intra and interoperator reproducibility of
retinal nerve fibre and macular thickness ... Acta. 2010;3768.
24. Keane PA, Mand PS, Liakopoulos S, et al. Accuracy of retinal thickness
measurements obtained with Cirrus optical coherence tomography. Br. J.
Ophthalmol. 2009;93(11):1461–1467.
25. Delaneau O, Marchini J, Consortium T 1000 GP, et al. Integrating sequence and
array data to create an improved 1000 Genomes Project haplotype reference panel.
Nat. Commun. 2014;5(1):3934.
26. Howie BN, Donnelly P, Marchini J. A flexible and accurate genotype imputation
method for the next generation of genome-wide association studies. PLoS Genet.
2009;5(6):e1000529.
27. Loh PR, Danecek P, Palamara PF, et al. Reference-based phasing using the
Haplotype Reference Consortium panel. Nat. Genet. 2016;48(11):1443–1448.
28. Durbin R. Efficient haplotype matching and storage using the positional Burrows-
Wheeler transform (PBWT). Bioinformatics. 2014;30(9):1266–1272.
29. McCarthy S, Das S, Kretzschmar W, et al. A reference panel of 64,976 haplotypes
for genotype imputation. Nat. Genet. 2016;48(10):1279–1283.
30. Maples BK, Gravel S, Kenny EE, et al. RFMix: A discriminative modeling
approach for rapid and robust local-ancestry inference. Am. J. Hum. Genet.
99
2013;93(2):278–288.
31. Bhatia G, Tandon A, Patterson N, et al. Genome-wide scan of 29,141 African
Americans finds no evidence of directional selection since admixture. Am. J. Hum.
Genet. 2014;95(4):437–444.
32. Zhou X, Stephens M. Genome-wide efficient mixed-model analysis for association
studies. Nat. Genet. 2012;44(7):821–4.
33. Yang J, Zaitlen NA, Goddard ME, et al. Advantages and pitfalls in the application
of mixed-model association methods. Nat. Genet. 2014;46(2):100–106.
34. Yang J, Benyamin B, McEvoy BP, et al. Common SNPs explain a large proportion
of the heritability for human height. Nat. Genet. 2010;42(7):565–9.
35. Zhang J, Stram DO. The Role of Local Ancestry Adjustment in Association
Studies Using Admixed Populations. Genet. Epidemiol. 2014;38(6):502–515.
36. Schick UM, Jain D, Hodonsky CJ, et al. Genome-wide Association Study of
Platelet Count Identifies Ancestry-Specific Loci in Hispanic/Latino Americans.
Am. J. Hum. Genet. 2016;98(2):229–242.
37. Yorgov D, Edwards KL, Santorico SA. Use of admixture and association for
detection of quantitative trait loci in the Type 2 Diabetes Genetic Exploration by
Next-Generation Sequencing in Ethnic Samples (T2D-GENES) study. In: BMC
Proceedings. BioMed Central; 2014:S6.
38. Galanter JM, Gignoux CR, Torgerson DG, et al. Genome-wide association study
and admixture mapping identify different asthma-associated loci in Latinos: The
Genes-environments & Admixture in Latino Americans study. J. Allergy Clin.
Immunol. 2014;134(2):295–305.
39. Watanabe K, Taskesen E, van Bochoven A, et al. Functional mapping and
annotation of genetic associations with FUMA. Nat. Commun. 2017;8(1):1826.
40. de Leeuw CA, Mooij JM, Heskes T, et al. MAGMA: Generalized Gene-Set
Analysis of GWAS Data. PLoS Comput. Biol. 2015;11(4):e1004219.
41. Cao D, Jiao X, Liu X, et al. CDKN2B polymorphism is associated with primary
open-angle glaucoma (POAG) in the Afro-Caribbean population of Barbados,
West Indies. PLoS One. 2012;7(6):e39278.
42. Gao Z, Mao CA, Pan P, et al. Transcriptome of Atoh7 retinal progenitor cells
identifies new Atoh7-dependent regulatory genes for retinal ganglion cell
formation. Dev. Neurobiol. 2014;74(11):1123–1140.
43. Ghiasvand NM, Rudolph DD, Mashayekhi M, et al. Deletion of a remote enhancer
near ATOH7 disrupts retinal neurogenesis, causing NCRNA disease. Nat.
Neurosci. 2011;14(5):578–588.
44. Woods SM, Mountjoy E, Muir D, et al. A comparative analysis of rod bipolar cell
transcriptomes identifies novel genes implicated in night vision. Sci. Rep.
2018;8(1):5506.
45. Quigley HA, Coleman AL, Dorman Pease ME. Larger Optic Nerve Heads Have
100
More Nerve Fibers in Normal Monkey Eyes. Arch. Ophthalmol.
1991;109(10):1441–1443.
46. Jonas JB, Schmidt AM, Muller-Bergh JA, et al. Human optic nerve fiber count and
optic disc size. Investig. Ophthalmol. Vis. Sci. 1992;33(6):2012–2018.
47. Ling Q, Jacovina AT, Deora A, et al. Annexin II regulates fibrin homeostasis and
neoangiogenesis in vivo. J. Clin. Invest. 2004;113(1):38–48.
48. Shekhar K, Lapan SW, Whitney IE, et al. Comprehensive Classification of Retinal
Bipolar Neurons by Single-Cell Transcriptomics. Cell. 2016;166(5):1308-
1323.e30.
49. Haffner C, Haass C. The biochemical and genetic odyssey to the function of a
nicastrin-like protein. Neurodegener. Dis. 2004;1(4–5):192–195.
50. Sterky FH, Trotter JH, Lee S-J, et al. Carbonic anhydrase-related protein CA10 is
an evolutionarily conserved pan-neurexin ligand. Proc. Natl. Acad. Sci.
2017;114(7):E1253–E1262.
51. Fagerberg L, Hallström BM, Oksvold P, et al. Analysis of the human tissue-
specific expression by genome-wide integration of transcriptomics and antibody-
based proteomics. Mol. Cell. Proteomics. 2014;13(2):397–406.
52. Jiang B, Harper MM, Kecova H, et al. Neuroinflammation in advanced canine
glaucoma. Mol. Vis. 2010;16:2092–108.
53. Zhou J, Benito-Martin A, Mighty J, et al. Retinal progenitor cells release
extracellular vesicles containing developmental transcription factors, microRNA
and membrane proteins. Sci. Rep. 2018;8(1):2823.
54. Kashani AH, Zimmer-Galler IE, Shah SM, et al. Retinal Thickness Analysis by
Race, Gender, and Age Using Stratus OCT. Am. J. Ophthalmol. 2010;149(3):496-
502.e1.
55. Manichaikul A, Palmas W, Rodriguez CJ, et al. Population structure of hispanics
in the United States: The multi-Ethnic study of Atherosclerosis. PLoS Genet.
2012;8(4):e1002640.
56. Shtir CJ, Marjoram P, Azen S, et al. Variation in genetic admixture and population
structure among Latinos: The Los Angeles Latino eye study (LALES). BMC
Genet. 2009;10(1):71.
57. Varma R, Ying-Lai M, Francis BA, et al. Prevalence of open-angle glaucoma and
ocular hypertension in Latinos: The Los Angeles Latino Eye Study.
Ophthalmology. 2004;111(8):1439–1448.
101
Chapter 4. Evaluation of the Main and Joint Associations between Ambient Air
Pollution and Retinal Vein Occlusion
ABSTRACT
Retinal vein occlusion (RVO) is a retinal vascular disease that can cause sudden,
unilateral loss of vision. We assessed the effects of ambient air pollution, including NO 2,
O 3, PM 2.5, PM 10, and traffic density, on the retinal vasculature, characterized by having RVO
using of two population-based cohort studies, the Los Angeles Latino Eye Study
(LALES) and the Chinese American Eye Study (CHES), based in the Los Angeles county
area. Air pollution measures were assessed at residence at time of recruitment. We
performed multivariable logistic regression to examine the associations between air
pollutants and RVO, adjusting for demographic characteristics. Additionally, within the
Latino study population, we tested the joint effects of known thrombosis genetic variants
and the same air pollutants on RVO by adding an interaction term to the model. While we
did not discover a direct effect between air pollution and RVO, we identified a suggestive
main association between a variant, rs2066865, within the FGG gene, and RVO.
Additionally, this variant was modified by traffic density on risk of RVO (P = 0.02). In
summary, this is the first study to examine the risk of RVO in a multiethnic population in
the United States.
102
Retinal vein occlusion (RVO) is a retinal vascular disease that may cause sudden,
unilateral loss of vision. Like thrombosis in other regions, the pathogenesis of RVO may
be due to three host systemic changes known as Virchow’s triad: venous stasis, vessel
wall damage, and hypercoagulability (1). There are two types of RVO, branch retinal
vein occlusion (BRVO) and central retinal vein occlusion (CRVO). While BRVO is more
common, CRVO tends to be more symptomatic, which can present with the sudden loss
of vision. BRVO may be asymptomatic, however, patients may have blurred vision if the
occlusion occurs near a visual field. Additionally, visual loss can result from the
associated macular edema and macular ischemia. Neovascularization, or growth of new
blood vessels, of the retina may occur and leading to further complications (2).
The prevalence of any RVO is relatively common with estimates from a pooled
international consortium at approximately 5.20 per 1000 for any RVO, 4.42 per 1000 for
BRVO and 0.80 per 1000 for CRVO (3). The highest prevalence of RVO has been found
among Hispanics (sex and age adjusted incidence rate = 6.9 per 1000), followed by
Asians, Blacks, and then whites (3.7 per 1000) (3) with an estimated 16 million adults
affected globally. The factors driving the occurrence of RVO are similar to other
vascular diseases, including being hypertensive, being diabetic, higher body mass index,
and smoking (4,5).
The presence of RVO is associated with an increased risk of cardiovascular
mortality and stroke (6) and thrombosis is an underlying pathology for both RVO and
cardiovascular events. Recent studies have suggested the role of ambient air pollution and
its association with cardiovascular disease, a leading cause of death in the United Sates
(7), and thrombosis (8–11). The components of ambient air pollution, including
103
particulate matter (PM), ozone, nitrogen oxides, and others, may induce pollution-
mediated thrombosis and associated pollution-specific epigenetic changes (10).
While the effects of ambient air pollutants have been discussed on the overall
vascular system, the role of air pollutants of ocular pathophysiology and RVO have not
been widely investigated. The surface of the eye is continuously exposed to air pollutants
and is particularly susceptible to direct contamination from the air due to the moist,
vascular nature of the sclera (12). Given the underlying role of air pollution on
cardiovascular disease and thrombosis, there is a need to evaluate the association between
environmental air pollutants and RVO. Previous studies have identified suggestive
associations between retinal vasculature events and air pollutants (13–15). However,
these studies were not done in the United States, which has a different distribution and
amalgam of air pollutants and characteristics of the exposed base population.
Genetic susceptibility may modify the impact of exposure to ambient air
pollutants on RVO. Both linkage (family-based) and genome-wide association studies
(GWAS) have identified regions of the genome that are associated with overall
thrombotic risk (16,17). A recent GWAS has identified genetic variants within coding
regions of genes that are associated with thromboembolism (17). Nine variants reached
genome-wide statistical significance in this study, and six of these variants are functional
mutations in the coagulation cascade. For example, one polymorphism is upstream FGG
gene, which codes for the fibrinogen γ’ protein. Because these variants are relatively
common within a population, it suggests the need to assess whether these genes modify
the effect of air pollution on the risk of RVO. Gene-by-environment (GxE) studies that
examine the joint effects of genetic susceptibility and environmental exposures may
104
identify a subset of the population at higher risk of RVO and may provide insight into the
function of loci in response to external exposures (18,19).
We proposed to evaluate the effects of air pollution on the retinal vasculature and
its potential interaction with candidate susceptibility loci. The analysis includes two
population-based cohort studies that were designed by the same investigative team and
shared similar study protocols, the Los Angeles Latino Eye Study (LALES) and the
Chinese American Eye Study (CHES). We hypothesize that participants living in areas
with higher pollutant concentrations of NO 2, O 3, PM 2.5, PM 10, and higher traffic density will
be associated with increased risk of RVO and test the joint effects of known thrombosis
genetic variants with ambient air pollutants on RVO, specifically in the LALES
population.
MATERIAL AND METHODS
Study design and participants
We utilized a pooled dataset from two similarly-designed population-based
studies of eye disease conducted in Los Angeles, California by the same investigative
team. We used the demographic, ocular, and fundus imaging data from the Los Angeles
Latino Eye Study (LALES III) conducted between 2010-2014, and the Chinese American
Eye Study (CHES) conducted between 2009-2013. The methods for each study have
been described previously (20–22). Participants were eligible for each respective study
depending on self-reported ethnicity, being 40 years and older for LALES and 50 years
and older for CHES, living within respective census tracts of the cities. Trained
interviewers completed in-home interviews, and participants were then invited to
complete clinical examinations by ophthalmic technicians and ophthalmologists. Written
105
informed consent was obtained for all participants. Study approval was obtained from the
Los Angeles County/University of Southern California Medical Center Institutional
Review Board and adhered to the tenets of the Declaration of Helsinki.
Vein occlusion measures
Central and branch vein occlusion measures were determined from fundus
photographs obtained during the clinical examinations. A series of detailed photographs
of the fundus was obtained in all consenting patients who underwent the clinical exam.
Each study collected seven standard fields of the fundus for each eye using the Topcon
TRC-50EX Retinal Camera (Topcon Corporation of America, Paramus, NJ) using
Ektachrome 100 film (Kodak, Rochester, NY). All standard fields from each eye were
then examined by the graders for retinal vein occlusions, and the fundus photographs
were graded using the Wisconsin Age-Related Maculopathy grading scheme (23). The
Ocular Epidemiology Grading Center at the University of Wisconsin, Madison, graded
the stereoscopic fundus photographs. The grading was completed in a masked manner.
Central vein occlusions were characterized by either occluded and sheathed retinal veins
or by retinal edema, optic disc hyperemia or edema, scattered superficial and deep retinal
hemorrhages, and venous dilation (24). To quantify branch retinal vein occlusions, a
localized area of the venule was examined for scattered superficial and deep retinal
hemorrhages, venous dilation, intraretinal microvascular abnormalities, and occluded and
sheathed retinal venules (24).
Air pollution estimation
106
Geocoding.
Participants across the two studies had their residential address collected at the
baseline home visit or clinic examination. The home addresses were then assigned a
latitude and longitude using automated geocoding provided by Texas A&M Geocoder
(http://geoservices.tamu.edu). This geocoder is an online, publicly accessible tool that
sources geodata from TIGER (Topologically Integrate Geographic Encoding and
Referencing census data), parcel, zip code tabulation areas, city, and county level entities.
In cases of ambiguous entries, the geocoder will attempt to locate the centroid of any
available reference data. For instance, if a street number is missing, the centroid of thee
street will be returned, or if a street name is specified incorrectly, the centroid of the zip
code will be returned (25). The mean matching rate of this geocoder has been shown to
be >95% (26).
Air pollution and traffic density estimation.
We utilized air pollution and traffic density estimates prepared for the University
of Southern California NIEHS by Sonoma Technology, Inc (Prepared November 2016).
We downloaded raster files for air pollution estimated of annual average NO 2, O 3, PM 2.5,
and PM 10. NO 2 and O 3 were expressed in ppb. PM 2.5 and PM 10 are represented in ug/m3.
These values were estimated from the air quality data measured at 4 closest monitoring
locations within 50 km of the grid point by inverse distance squared interpolation. The
domain extends from 32.5 to 35.5 degrees latitude and from -121.05 to -114.1 degrees
longitude, which covered all of Los Angeles and Southern California. The grid point
spacing is 0.01 degrees latitude and 0.01 degrees longitude. This corresponds to
approximately a resolving power of at least a one km square area.
107
Traffic density data was calculated from a Southern California multi-year traffic
volume dataset from Sonoma Technology Inc. Traffic density is represented as the annual
average daily traffic volume in both directions using data all roads (FCC1-FCC4,
corresponding to freeways, arterial roads, collector distributor roads, and local roads,
respectively) within a circular buffer. The density was assumed to be symmetric on both
sides of each roadway. The density values were computed using the ArcMap v10.3.1
(ESRI, Redlands, CA) density function with a 300m search radius and 10m grid. The
traffic density estimates assumed a 300m falloff to represent distance decay. We assigned
all air pollution and traffic density data for all participants using the average annual
estimates from 2010 measures. All estimates were analyzed either as continuous
measures or analyzed after categorizing into the lowest three quartiles compared to the
highest quartile.
Genotyping and gene-environment interaction
To test for the joint effects of genetics and air pollution, we utilized the LALES
population for which genotyping was available. LALES participants were genotyped
using either the Illumina OmniExpress BeadChip (~730k markers, Illumina, San Diego,
CA) for n=4,122 samples or the Illumina Hispanic/SOL BeadChip (~2.5 million markers,
Illumina, San Diego, CA) for n=716. Genotyping was performed at the Los Angeles
Biomedical Research Institute. Six of the nine variants discovered in a genome-wide
association study of thrombosis were available on both arrays and passed quality control
for all subsequent analyses. All variants were filtered on a genotype call rate of 90% and
we excluded any variants with a Hardy-Weinberg equilibrium log P-values<10
-8
.
Additionally, we removed any monomorphic variants, or variants with alleles that could
108
not be matched to the correct forward strand in the hg19 genome assembly. Variants were
coded from 0, 1, or 2 to indicate the number of alternative alleles corresponding to the
alternative alleles in the hg19 genome build. There were 2631 participants with both
genotyping and fundus imaging available (including 40 cases of RVO). We adjusted for
the number of comparisons using a Bonferroni adjustment (where six tests adjusted at an
a=0.05 corresponded to a corrected significance value of 0.008).
Statistical methods
Counts, proportions, means, and standard deviations (SD) were used to
summarize the demographic and clinical characteristics. We assessed the relationships
between average annual NO 2, O 3, PM 2.5, PM 10, and traffic density and presence of any
branch or central retinal vein occlusion using logistic regression. We adjusted all models
for age, gender, and/or study. Additionally, hypertension, diabetes mellitus (DM), and
smoking status were considered as potential covariates as they have been reported to be
associated with retinal vein occlusion (13). The studies were evaluated independently as
their measures of air pollution were different across the studies. To test for the joint
genetic and air pollution effects, we additionally added an interaction term including each
of six genetic variants on the RVO outcome, while adjusting for the same covariates. We
adjusted all tests for the Statistical analyses were performed using R (Version 3.5.2).
RESULTS
Study population
Across the two eye cohorts, there were 7,960 individuals that were eligible for
this this study with address information available (n=4,442 Chinese-Americans and
n=3,518 Latinos) (Supplementary Table 4.1). Due to concerns of geocoding quality and
109
potential for low variability across participants, we included only participants where the
geocoding software assigned to a parcel or street segment (other assignments like zip
code and city would produce the same estimates as participants were recruited within one
city). This retained a total of 4,827 participants, with 1,432 Chinese-American and 3,395
Latino participants (Supplementary Table 4.1). Participants who were able to be matched
to parcel or street segments were more likely to be younger and less diabetic compared to
those with matches to larger regions (Supplementary Table 4.2). We performed all
subsequent analyses on 4,586 participants for those with complete covariate data.
The mean age of the population was 62.3 (SD=8.9). Most of the study sample was female
(61.2%) and 25% of the population were diabetic, 30% were smokers and 48% were
hypertensive (Table 4.1). There were 63 cases of BRVO and eight cases of CRVO. All
analyses were combined as any RVO due to low numbers of CRVO cases. The
participants differed on age and smoking status, while there appeared to be no differences
between RVO cases and non-RVO by sex, DM, hypertension, and ethnicity (Table 4.1).
Air pollution and traffic density exposures
Supplementary Table 4.3 shows the air pollutions estimates when combined
across studies and when stratified by study. All air pollutants and traffic density other
than NO 2 were at a higher level in the Latino population than the Chinese-American
population. The combined values for quartile categorization were driven mostly by the
Latino population as there were more participants in the LALES study population. Figure
1 highlights the four air pollutants in 2010 for the Los Angeles area colored by study. The
Chinese-American population were clustered together while the Latino population was
more spread out, as this was the later wave of the LALES data.
110
Effects of air pollutants and RVO
We found no statistically significant associations between air pollutants, traffic
density, and RVO after adjusting only for age and sex (Table 4.2). These associations
were consistently null when air pollutants and traffic density were measured as a
continuous measure or when grouped into two categories, comparing the lowest three
quartiles compared to the highest.
There were no statistically significant associations between air pollutants and
traffic density with RVO after performing a multivariable logistic regression model
additionally adjusting for hypertensive status, DM, and smoking (Table 4.3). We
additionally performed a sensitivity analysis to include all participants regardless of
geocoding quality, but results were the same as results for the higher quality geocoding
data (Supplementary Table 4.4). We examined the GxE interaction between known the
known thromboembolism functional variants and air pollution/traffic density on RVO
risk. Of the six variants tested, the main effect of the rs2066865 was associated with
RVO. Each additional A allele was associated a 16% decreased risk of RVO (β=-2.0, P =
0.02). There was a statistically significant interaction between traffic density and the
rs2066865 on risk of RVO (P = 0.02) (Table 4.4). The GxE interaction effect estimate for
this variant was 1.48, indicating that increasing numbers of the alternative allele
positively modified the effect of traffic density on the risk of BRVO (interaction P =
0.02).
DISCUSSION
This is the first study to examine the role of air pollution and traffic density on the
risk of any RVO. We used two multiethnic population-based studies of Latinos and
111
Chinese-Americans based in the Los Angeles region. While there were no direct
associations of the four air pollutants or traffic density, and association between a variant
within the FGG gene modified the effects of traffic density on the risk of RVO. We
replicated findings that hypertension and age were determinants of RVO and smoking
was nominally significantly associated with RVO. When we examined the role of genetic
variants involved in thrombosis, we identified the traffic density-FGG gene variant
(rs2066865) interaction and the association with RVO. We discovered that subjects with
each alternative allele of rs2066865 was associated with a two-fold decrease in the risk of
BRVO (P-value for interaction=0.02), however our study was based on a small number
of cases and underpowered to test for interactions. These findings are exploratory and
will require further validation in larger datasets.
PM within air pollution is defined by its size (where PM 2.5 are particles smaller
than 2.5 μm, and PM 10 are particles smaller than 10 μm but larger than 2.5 μm). In the
United States, an ~29 million people are exposed to PM 10, and ~88 million people are
exposed to PM 2.5 (27). These compounds come from both human-made and natural
sources. Acute exposure to pollutants, specifically PM, is associated with changes in
fibrinogen and other coagulation proteins, seen both in controlled experimental and
observational studies (28–30). Long term epidemiological studies have shown similar
changes in the process of clotting, however additional studies have suggested long-term
systemic inflammation caused by air pollution on changes in vasculature and fibrinogen
(31). Another mechanism of air pollution on the cardiovascular system is the role of
oxidative stress (32). Air pollutants induce oxidative stress when inhaled by triggering an
initial inflammatory response and then causing secondary systemic inflammatory effect
112
(33), which disrupts vasodilatation and inhibition of platelets (32). Overall, these
pollutants are may disrupt haemostasis, the process of maintaining and preserving blood
circulation, and systemically shifts the vasculature towards a thrombotic nature (10).
There have been previous studies that have identified positive associations
between air pollutants and RVO. For example, one large retrospective population-based
cohort study in Taiwan determined that was retinal artery occlusion was associated with
NO 2 but not PM 2.5, PM 10, or O 3 (15). This study was completed using a case-crossover
design which measured the exposure window within a five-day window. Another
similarly designed study suggested within a Polish population that retinal artery occlusion
was positively associated with short-term with preceding daily changes in NO 2, PM 10, O 3,
sulfur dioxide, and carbon monoxide concentration (14). In contrast to these findings, our
study suggests a null association between these same air pollutants and RVO. Potential
differences between the design of this study and the previous may lie with the exposure
assignment of air pollution.
First, both the current and the previously mentioned studies used residential
history as proxies to assess the ambient concentrations of air pollutants, but the previous
findings utilized the exposure assignment in a narrow period preceding the time of the
diagnosis of the retinal thrombotic event. Our study assigned the average annual exposure
in 2010. While the exposure assignment preceded the date of the outcome, we used a
window of exposure that was much longer and more distal from the time of the outcome.
Our Latino population ended data collection in 2014 while the Chinese-American
population study ended in 2013. These results could be biased with respect to our
findings; however, they were likely misclassified non-differentially and may explain the
113
null findings. While the participants could have moved between 2010 and the potential
time for onset of RVO, again this would likely be misclassified without respect to case
status.
Second, this study was not designed in a fashion that were able to accurately
capture the exact dates of incident RVO. Therefore, there is some potential for the role of
reverse causation, were cases of RVO could have moved to less polluted areas after
having symptoms or have become less exposed to outdoor air pollutants since they were
unable to perform daily life functions. This could potentially then have biased the
association toward the null if their residential address was measured at this time after they
moved.
Another important consideration in this study is the amount of variation observed
in the air pollutants. The previous studies of retinal thrombosis used larger countries to
study, Poland and Taiwan, as their source population. Poland is 25 times the size of Los
Angeles, and Taiwan is 3 times as large as Los Angeles county. This may provide more
variation in the exposure assignment compared to this study in the Los Angeles area.
There are approximately 30 air monitoring stations in the Los Angeles basin that are used
for the inverse distance interpolation as used from the California Air Resources Board
(34), which may not be able to provide the most accurate exposure information within
smaller communities. For example, we saw lower variability of most measures within the
Chinese-American population. We attempted to higher weight the extreme values, by
using the top quartile compared to the lowest three quartiles, which allowed for much
more extreme changes in the air pollution measures. For instance, with PM 2.5 the lowest
three quartiles were 11.4 µg/m
3
, 11.5 µg/m
3
and 11.5 µg/m
3
, while the highest quartile was
114
16 µg/m
3
in our Latino study population area. Traffic density, on the other hand was much
more variable since these were raster values with a decay function around the type of
roadway near a subject’s residence.
Lastly, one major consideration in the difference between the current study and
the previous findings is the actual amount of air pollutants that is the biological dose that
may be important in causing retinal vascular occlusions. While we describe the actual
concentration measures for air pollution within the Los Angeles basin in this study, the
others do not adequately describe the measured concentrations, so we were not able to
compare values across studies/regions. Global interpolations show that that the PM 2.5 is
lowest in western North American, where estimated PM 2.5 is 11, 44, and 15 µg/m
3
in
western North America, eastern Asia, and eastern Europe, respectively (35). However,
the Los Angeles basin and most cities in LA county have consistently had the highest
levels of PM across the United States (36). Therefore, the concentrations of the air
pollutants may be similar across all regions and would be appropriate to see the effects
across all regions.
We may gain a better understanding of retinal vein occlusion by also describing
the etiology of general thrombosis and its relationship with air pollution, the
manifestation of retinal occlusion in other locations. One recent meta-analysis of venous
thrombosis and three air pollutants (PM 2.5, PM 10, NO 2) found while effect estimates were
slightly elevated in risk of thrombosis there was not a statistically significant association
between either short-term or long-term risk of venous thrombosis across eight studies
(37). However, multiple experimental and randomized studies have suggested that acute
exposure to air pollutants, specifically PM 10, is associated with changes in coagulation
115
function (28–30). Long term exposure to air pollution is has been shown to reduce
clotting times with increasing levels of PM10 average over one year among healthy
controls and patients with deep vein thrombosis (30). There are approximately 1 per
1,000 cases of thromboembolism per year (38), which is lower than the rate observed of
RVO, but no studies have compared the underlying etiology of both retinal thrombotic
events and other thrombotic events. Both phenotypes share similar risk factors (4,39),
including smoking, DM, and hypertension. Because of the relatively lower burden of
fundus photography in a routine procedure compared to ultrasound for thrombotic events,
the manifestation as RVO may be an earlier predictor of overall cardiovascular health and
future studies may benefit in a systems biology approach by understanding both.
Another phenotype related to RVO is not only the occlusion event, but also the
microvasculature that plays a part in the development of RVO. Microcirculation
especially in the eye plays an important role in the overall physiology of cardiovascular
health. These minute microvascular changes could be an early marker for cardiovascular
disease. For instance, narrowed retinal vessel diameter was a marker of hypertension,
coronary heart disease, and other vasculature-related diseases in prospective cohort
studies (40,41). A population-based study completed in the Multi-Ethnic Study of
Atherosclerosis suggests that both acute and long-term exposure to ambient air pollutants
were each associated with narrower retinal arteriolar diameters in older individuals (42).
Retinal arteriole equivalents were narrower among participants residing in regions with
increased long- and short-term levels of PM 2.5 (42). Other studies found that both short-
term responses in retinal arteriole and venular vascular diameters were associated with
PM 10 and black carbon, averaged over the 24 hours before their retinal examinations (43).
116
While we did not measure microvasculature diameter in the current study, air pollution is
likely important to thrombotic events due its changes in the endothelium as part of
Virchow’s triad. Further longitudinal studies may consider the changes in
microvasculature in the natural history of RVO.
While we considered the biological role air pollution on the RVO, we also must
acknowledge the societal factors that lead to how communities are arranged for these
exposures to occur. Race and ethnicity are important predictors of neighborhood-level
measures of poverty and collective efficacy (44), and minorities may be
disproportionately living in areas of higher ambient air pollutants. In the population-
based study of RVO, blacks were found to be at increased risk of developing incident
RVO, even after controlling for various confounders including HT, components of the
metabolic syndrome, and sociodemographic factors (13). An unmeasured variable that
was not considered was neighborhood-level measures, including air pollution, which
could account for these discrepancies. The CalEnviroScreen, a social justice screen tool,
suggests that environmental health hazards disproportionately burden communities of
color in California (45). Their methodology identified that the odds of living in one of
the 10% most affected zip codes were 6.2, 5.8, and 1.8, times greater for Hispanics,
African Americans, Asian/Pacific Islanders, than for non-Hispanic Whites (45). Even
within our study, we show discrepancies in levels in air pollution across the two
ethnicities. All pollutants, except for O 3, were highest in the Latino population than the
Chinese-American population. We must consider the role that race/ethnicity plays on
environmental justice alongside the biological role of air pollutants on the retinal
vasculature.
117
We identified suggestive associations on the role of genetics and their interaction
with air pollution on risk of RVO. There have been no GWAS to identify disease-
susceptibility genes of RVO, but previous studies have identified the role of genetic
variation on the thrombosis and retinal vessel diameter. One of the first GWAS suggests
that variation within FV and ABO genes are associated with thrombosis (46). Variation
within these genes are associated with coagulation defects. Later studies have identified
genomic regions associated with thrombosis but have no identified function (17,47),
leaving room to further elucidate the overall role of genetics on thrombosis. With retinal
vessel diameter over 8 loci have been discovered across multiple loci, but again the
function of these regions remains to be identified (40,48,49). We found that variation of
FGG was suggestively significantly associated with RVO when we examined its main
effect on the risk of RVO. We have mentioned the role of fibrinogen on changes in
vasculature, and this variant is possibly responsible for reducing fibrinogen by interfering
with pre-mRNA cleaving (50,51), which suggests that further functional studies will need
to be completed.
While the specific functions of these genes may not have been discovered using
standard association techniques, using a GxE interaction approach allows to both identify
a subset of participants that may be at higher risk of an outcome even within a region of
high air pollutants and potentially better describe the functional mechanism of a gene. A
previous GxE study suggests that PM 10-fibrinogen relationship is modified by the
genotype of another fibrinogen gene, FGB, where subjects homozygous for the minor
allele had a fibrinogen response eightfold higher compared to subjects homozygous for
the major allele (52). We identified a similar response in this study with FGG, and this
118
was observed with the traffic density-gene interaction. However, our exposure window
was a multiyear window compared to a five-day period.
The limitations of this study are based on the reduced variation in air pollutants
for our participants in the Los Angeles basin. This was likely due to the fewer number of
air pollution monitoring sites, which limited the interpolation, and the clustering of the
study participants. However, future studies could, with relative ease, track the current
residence of these participants to better understand lifetime pollution. Additionally, a
proportion of the Chinese-American population did not give accurate residential address,
which may have limited interpretation of this population. This could have been due to
reporting bias or interviewer bias(53). However, we also did not identify an association
within the Latino population and therefore, could conclude that there might be few
differences between address ascertainment.
The strengths of this study include the population-based design of our studies and
the high quality of the covariates and outcome data collection. Follow up studies are
imperative in these populations as we have collected rich baseline date of a multiethnic
study in Los Angeles. In summary, this is the first study to examine the risk of RVO in a
multiethnic population in the United States. While we find no significant associations
between air pollution and RVO, we identified a suggestive association between
fibrinogen genes and RVO and an effect modification between traffic density and this
gene on risk of RVO.
119
Table 4.1. Demographic characteristics of Any RVO and non-RVO in the Combined
Multiethnic LALES and CHES
Covariate Controls (n=4581) RVO (n=68) P-value
n (%) n (%)
Female 2838 (61.9) 37 (54.4) 0.25
Age
<0.001
50-59 2043 (44.6) 13 (19.1)
60-69 1605 (35.0) 23 (33.8)
70-79 696 (15.2) 25 (36.8)
80+ 237 (5.2) 7 (10.3)
Diabetic 1173 (25.6) 17 (25.0) 1.00
Ever Smokers 1371 (29.9) 28 (41.2) 0.06
Hypertensive 1713 (37.4) 34 (50.0) 0.13
Ethnicity
0.40
Chinese-Americans 1326 (29.0) 16 (23.5)
Latinos 3255 (71.1) 52 (76.5)
Abbreviations: RVO, retinal vein occlusion; LALES, Los Angeles Latino Eye Study;
CHES, Chinese-American Eye Study
120
Table 4.2. Associations Between Air Pollutants and RVO, Adjusted for Age and Sex
Only
Chinese Americans Latinos
Pollutant OR (95% CI) OR (95% CI)
Continuous
NO 2
a
2.81 (0.3-27.2) 0.92 (0.8-1.1)
O 3
a
0.41 (0.1-2.3) 1.01 (0.9-1.2)
PM 2.5
b
4.26 (0.3-74.4) 0.88 (0.7-1.3)
PM 10
b
0.69 (0.3-1.5) 1.02 (0.9-1.1)
Traffic Density
c
0.86 (0.5-1.4) 0.97 (0.8-1.2)
Quartile Q4 vs Q1-
Q3
NO 2
a
2.41 (0.7-15.4) 0.52 (0-2.4)
O 3
a
0.72 (0.3-2.1) 1.54 (0.5-3.9)
PM 2.5
b
1.65 (0.6-5.3) 0.81 (0.2-2)
PM 10
b
0.01 (0-10.1) 1.51 (0.9-2.6)
Traffic Density
c
0.52 (0.1-1.9) 1.04 (0.6-1.9)
Abbreviations: RVO, retinal vein occlusion; NO 2, Nitrogen dioxide; O 3, ozone; PM 2.5,
particulate matter >2.5µm & <10µm, PM 10, particulate matter ≥10µm; Q4, highest fourth
quartile; Q1-Q3, lowest three quartiles.
a
NO 2 and O 3 are expressed in parts per billion.
b
PM 2.5 and PM 10 are expressed in µg/m3.
c
Traffic density is a raster intensity with higher intensity indicating higher amounts of
traffic density and is log transformed
121
Table 4.3. Multivariable Associations Between Air Pollutants and RVO Adjusted for
Age, Sex, Hypertension, Diabetes, and Smoking Status
Chinese Americans Latinos
Pollutant OR (95% CI) OR (95% CI)
Continuous
NO 2
a
2.84 (0.3-27.2) 0.92 (0.8-1.1)
O 3
a
0.48 (0.1-2.7) 1.01 (0.9-1.2)
PM 2.5
b
4.37 (0.3-77.7) 0.87 (0.7-1.3)
PM 10
b
0.67 (0.3-1.4) 1.02 (0.9-1.1)
Traffic Density
c
0.82 (0.5-1.3) 0.97 (0.8-1.2)
Quartile Q4 vs Q1-
Q3
NO 2
a
2.5 (0.7-16.1) 0.52 (0-2.4)
O 3
a
0.82 (0.3-2.5) 1.55 (0.5-3.9)
PM 2.5
b
1.8 (0.6-5.8) 0.82 (0.2-2)
PM 10
b
0.01 (0-8.4) 1.49 (0.8-2.6)
Traffic Density
c
0.53 (0.1-1.9) 1.03 (0.5-1.9)
Abbreviations: RVO, retinal vein occlusion, NO 2, Nitrogen dioxide; O 3, ozone; PM 2.5,
particulate matter >2.5µm & <10µm, PM 10, particulate matter ≥10µm; Q4, highest fourth
quartile; Q1-Q3, lowest three quartiles.
a
NO 2 and O 3 are expressed in parts per billion.
b
PM 2.5 and PM 10 are expressed in µg/m3.
c
Traffic density is a raster intensity with higher intensity indicating higher amounts of
traffic density and is log transformed
122
Table 4.4. Joint Air Pollutant and Thrombosis Genetic Variants and Risk of RVO
Loci EA RSID Gene EAF PM 10 PM 2.5 Traffic Density
Interaction
β (SE) P
Interaction
β (SE) P
Interaction
β (SE), P
1:169511755 C rs6025 F5 0.40 0.54 (0.5) 0.28 0.09 (0.82) 0.91 0.9 (0.57) 0.11
4:155525276 A rs2066865 FGG 0.23 -0.93 (0.61) 0.12 -0.64 (1.12) 0.56 1.48 (0.63) 0.02
8:106590706 G rs4602861 ZFPM2 0.28 0.19 (0.52) 0.71 -1.04 (1.09) 0.34 -0.89 (0.71) 0.21
9:136149500 C rs529565 ABO 0.25 0.32 (0.53) 0.54 -0.09 (0.87) 0.92 -0.14 (0.63) 0.83
10:71245276 C rs78707713 TSPAN15 0.09 -0.38 (0.82) 0.65 -1106.16 (78494.6) 0.99 0.23 (0.91) 0.8
19:10742170 G rs2288904 SLC44A2 0.87 -0.33 (0.71) 0.64 -0.87 (1.02) 0.39 1.34 (1.12) 0.23
Abbreviations: RVO, retinal vein occlusion; EA, Effect allele; EAF, Effect allele frequency; PM 2.5, particulate matter >2.5µm &
<10µm; PM 10, particulate matter ≥10µm
123
Figure 4.1. A) NO2 and B) PM2.5 Concentration in the Combined Multiethnic Los
Angeles Latino Eye Study and Chinese American Eye Study Population
Abbreviations: LALES, Los Angeles Latino Eye Study; CHES, Chinese American Eye
Study
124
Supplementary Table 4.1. Geocoding Match Stratified by Ethnicity
Geocode Match Chinese Americans Latinos
n (%) n (%)
Parcel 1187 (26.7) 2852 (81.1)
Street Segment 245 (5.5) 543 (15.4)
USPS zip code 35 (0.8) 86 (2.4)
City 2975 (67.0) 12 (0.3)
State
25 (0.7)
Abbreviations: USPS, United States Postal Service
125
Supplementary Table 4.2. Demographic Characteristics of Geocoding Match Type in the
CHES Population
Variable City/Zip Only (n=3135) Street/Parcel (n=4287)
n (%) n (%)
Female 1894 (63.0) 923 (64.5)
Age
50-59 1393 (46.3) 748 (52.2)
60-69 991 (32.9) 523 (36.5)
70-79 403 (13.4) 123 (8.6)
80+ 224 (7.4) 38 (2.7)
Diabetic 496 (16.5) 205 (14.3)
Ever Smokers 415 (13.8) 194 (13.6)
Hypertensive 629 (20.9) 287 (20.0)
Abbreviations: CHES, Chinese American Eye Study
126
Supplementary Table 4.3. Distribution of Air Pollutants Combined and Stratified by
Ethnicity
Pollutant Quartile Chinese Americans Latinos Combined
NO 2
a
Q1 23.0 19.9 20.3
Q2 23.1 20.2 22.9
Q3 23.1 20.5 23.1
Q4 23.5 25.9 25.9
O 3
a
Q1 30.6 26.9 27.3
Q2 30.6 27.2 30.4
Q3 30.6 27.6 30.6
Q4 31.4 41.1 41.1
PM 2.5
b
Q1 11.7 11.4 11.5
Q2 11.9 11.5 11.7
Q3 11.9 11.5 11.9
Q4 12.2 16.9 16.9
PM 10
b
Q1 36.0 39.0 36.0
Q2 36.0 39.3 36.9
Q3 36.4 39.8 39.3
Q4 39.8 58.1 58.1
Traffic Density
c
Q1 2570.4 5118.8 2570.4
Q2 2570.4 9097.7 4097.9
Q3 2896.0 23340.4 13775.0
Q4 87704.8 233195.3 233195.3
Abbreviations: NO 2, Nitrogen dioxide; O 3, ozone; PM 2.5, particulate matter >2.5µm &
<10µm, PM 10, particulate matter ≥10µm;Q1,first quartile; Q2, second quartile, Q3, third
quartile; Q4, highest fourth quartile.
a
NO 2 and O 3 are expressed in parts per billion.
b
PM 2.5 and PM 10 are expressed in µg/m3.
c
Traffic density is a raster intensity with higher intensity indicating higher amounts of
traffic density
127
Supplementary Table 4.4. Sensitivity Analysis of Univariable Model of RVO and Air
Pollutants Including all Geocoding Qualities
Chinese Americans Latinos
Pollutant OR (95% CI) OR (95% CI)
Continuous
NO 2
a
1.57 (0.3-12) 0.93 (0.8-1.1)
O 3
a
0.49 (0.1-2.2) 1 (0.9-1.1)
PM 2.5
b
1.85 (0.2-22.1) 0.88 (0.7-1.3)
PM 10
b
0.86 (0.4-1.5) 1.02 (0.9-1.1)
Traffic
Density
c
1 (1-1) 1 (1-1)
Abbreviations: RVO, retinal vein occlusion; NO 2, Nitrogen dioxide; O 3, ozone; PM 2.5,
particulate matter >2.5µm & <10µm, PM 10, particulate matter ≥10µm
a
NO 2 and O 3 are expressed in parts per billion.
b
PM 2.5 and PM 10 are expressed in µg/m3.
c
Traffic density is a raster intensity with higher intensity indicating higher amounts of
traffic density and is log transformed
128
REFERENCES
1. Yau JWY, Lee P, Wong TY, et al. Retinal vein occlusion: An approach to
diagnosis, systemic risk factors and management. Intern. Med. J.
2008;38(12):904–910.
2. Clarhson JC. Natural history and clinical management of central retinal vein
occlusion. Arch. Ophthalmol. 1997;115(4):486–491.
3. Rogers S, McIntosh RL, Cheung N, et al. The Prevalence of Retinal Vein
Occlusion: Pooled Data from Population Studies from the United States, Europe,
Asia, and Australia. Ophthalmology. 2010;117(2):313–9.e1.
4. Rehak M, Wiedemann P. Retinal vein thrombosis: Pathogenesis and management.
J. Thromb. Haemost. 2010;8(9):1886–1894.
5. Risk factors for central retinal vein occlusion. The Eye Disease Case-Control
Study Group. Arch. Ophthalmol. (Chicago, Ill. 1960). 1996;114(5):545–54.
6. Hayreh SS, Zimmerman B, McCarthy MJ, et al. Systemic diseases associated with
various types of retinal vein occlusion. Am. J. Ophthalmol. 2001;131(1):61–77.
7. Roger VL, Go AS, Lloyd-Jones DM, et al. Heart Disease and Stroke Statistics—
2012 Update. Circulation. 2011;125(1):e2–e220.
8. Brook RD, Rajagopalan S, Pope CA, et al. Particulate Matter Air Pollution and
Cardiovascular Disease. Circulation. 2010;121(21):2331–2378.
9. Brook RD, Franklin B, Cascio W, et al. Air pollution and cardiovascular disease:
A statement for healthcare professionals from the expert panel on population and
prevention science of the American Heart Association. Circulation.
2004;109(21):2655–2671.
10. Robertson S, Miller MR. Ambient air pollution and thrombosis. Part. Fibre
Toxicol. 2018;15(1):1.
11. Lelieveld J, Evans JS, Fnais M, et al. The contribution of outdoor air pollution
sources to premature mortality on a global scale. Nature. 2015;525(7569):367–
371.
12. Wolkoff P. Ocular discomfort by environmental and personal risk factors altering
the precorneal tear film. Toxicol. Lett. 2010;199(3):203–212.
13. Newman-Casey PA, Stem M, Talwar N, et al. Risk factors associated with
developing branch retinal vein occlusion among enrollees in a United States
managed care plan. Ophthalmology. 2014;121(10):1939–48.
14. Grzybowski AE, Mimier MK. Evaluation of the Association between the Risk of
Central Retinal Artery Occlusion and the Concentration of Environmental Air
Pollutants. J. Clin. Med. 2019;8(2):206.
15. Cheng HC, Pan RH, Yeh HJ, et al. Ambient Air Pollution and the Risk of Central
Retinal Artery Occlusion. Ophthalmology. 2016;123(12):2603–2609.
16. Souto JC, Almasy L, Borrell M, et al. Genetic Susceptibility to Thrombosis and Its
129
Relationship to Physiological Risk Factors: The GAIT Study. Am. J. Hum. Genet.
2002;67(6):1452–1459.
17. Germain M, Chasman DI, De Haan H, et al. Meta-analysis of 65,734 individuals
identifies TSPAN15 and SLC44A2 as two susceptibility loci for venous
thromboembolism. Am. J. Hum. Genet. 2015;96(4):532–542.
18. Thomas DC, Lewinger JP, Murcray CE, et al. Invited commentary: GE-Whiz!
Ratcheting gene-environment studies up to the whole genome and the whole
exposome. Am. J. Epidemiol. 2012;175(3):203–7; discussion 208-9.
19. Gauderman WJ, Zhang P, Morrison JL, et al. Finding novel genes by testing G × E
interactions in a genome-wide association study. Genet. Epidemiol.
2013;37(6):603–613.
20. Varma R, Paz SH, Azen SP, et al. The Los Angeles Latino Eye Study: design,
methods, and baseline data. Ophthalmology. 2004;111(6):1121–1131.
21. Varma R, Hsu C, Wang D, et al. The Chinese American eye study: Design and
methods. Ophthalmic Epidemiol. 2013;20(6):335–347.
22. McKean-Cowdin R, Fairbrother-Crisp A, Torres M, et al. The African American
Eye Disease Study: Design and Methods. Ophthalmic Epidemiol. 2018;25(4):306–
314.
23. Klein R, Davis MD, Magli YL, et al. The Wisconsin Age-related Maculopathy
Grading System. Ophthalmology. 1991;98(7):1128–1134.
24. Klein R, Klein BE, Moss SE, et al. The epidemiology of retinal vein occlusion: the
Beaver Dam Eye Study. Trans. Am. Ophthalmol. Soc. 2000;98:133–41; discussion
141-3.
25. Goldberg DW. Improving geocoding match rates with spatially-varying block
metrics. Trans. GIS. 2011;15(6):829–850.
26. Chow TE, Dede-Bamfo N, Dahal KR. Geographic disparity of positional errors
and matching rate of residential addresses among geocoding solutions. Ann. GIS.
2016;22(1):29–42.
27. Seagrave JC, McDonald JD, Bedrick E, et al. Lung toxicity of ambient particulate
matter from southeastern U.S. sites with different contributing sources:
Relationships between composition and effects. Environ. Health Perspect.
2006;114(9):1387–1393.
28. Mills NL, Törnqvist H, Robinson SD, et al. Diesel exhaust inhalation causes
vascular dysfunction and impaired endogenous fibrinolysis. Circulation.
2005;112(25):3930–3936.
29. Hildebrandt K, Rückerl R, Koenig W, et al. Short-term effects of air pollution: A
panel study of blood markers in patients with chronic pulmonary disease. Part.
Fibre Toxicol. 2009;6:25.
30. Baccarelli A, Zanobetti A, Martinelli I, et al. Effects of exposure to air pollution on
blood coagulation. J. Thromb. Haemost. 2007;5(2):252–260.
130
31. Viehmann A, Hertel S, Fuks K, et al. Long-term residential exposure to urban air
pollution, and repeated measures of systemic blood markers of inflammation and
coagulation. Occup. Environ. Med. 2015;72(9):656–663.
32. Miller MR, Shaw CA, Langrish JP. From particles to patients: Oxidative stress and
the cardiovascular effects of air pollution. Future Cardiol. 2012;8(4):577–602.
33. Seaton A, Godden D, MacNee W, et al. Particulate air pollution and acute health
effects. Lancet. 1995;345(8943):176–178.
34. Air Monitoring Sites - Interactive Map.
(https://ww3.arb.ca.gov/qaweb/mapdemo/map_module.php). (Accessed June 1,
2019)
35. van Donkelaar A, Martin R V, Brauer M, et al. Global estimates of ambient fine
particulate matter concentrations from satellite-based aerosol optical depth:
Development and application. Environ. Health Perspect. 2010;118(6):847–855.
36. American Lung Association State of the Air 2013 - Most Polluted Cities. 2015 1–5
p.
37. Tang L, Wang QY, Cheng ZP, et al. Air pollution and venous thrombosis: A meta-
analysis. Sci. Rep. 2016;6:32794.
38. White RH. The epidemiology of venous thromboembolism. Circulation.
2003;107(SUPPL. 23):4I--8.
39. Gregson J, Kaptoge S, Bolton T, et al. Cardiovascular Risk Factors Associated
with Venous Thromboembolism. JAMA Cardiol. 2019;4(2):163–173.
40. Ikram MK, Witteman JCM, Vingerling JR, et al. Retinal vessel diameters and risk
of hypertension: The Rotterdam Study. Hypertension. 2006;47(2):189–194.
41. Bressler NM. Retinal arteriolar narrowing and risk of coronary heart disease. Arch.
Ophthalmol. 2003;121(1):113–114.
42. Adar SD, Klein R, Klein BEK, et al. Air pollution and the microvasculature: A
cross-sectional assessment of in vivo retinal images in the population-based multi-
ethnic study of atherosclerosis (MESA). PLoS Med. 2010;7(11):e1000372.
43. Louwies T, Panis LI, Kicinski M, et al. Retinal microvascular responses to short-
term changes in particulate air pollution in healthy adults. Environ. Health
Perspect. 2013;121(9):1011–1016.
44. Kirk DS. The Neighborhood Context of Racial and Ethnic Disparities in Arrest.
Demography. 2008;45(1):55–77.
45. Cushing L, Faust J, August LM, et al. Racial/ethnic disparities in cumulative
environmental health impacts in California: Evidence from a statewide
environmental justice screening tool (CalEnviroScreen 1.1). Am. J. Public Health.
2015;105(11):2341–2348.
46. Trégouët DA, Heath S, Saut N, et al. Common susceptibility alleles are unlikely to
contribute as strongly as the FV and ABO loci to VTE risk: Results from aGWAS
approach. Blood. 2009;113(21):5298–5303.
131
47. Heit JA, Armasu SM, Asmann YW, et al. A genome-wide association study of
venous thromboembolism identifies risk variants in chromosomes 1q24.2 and 9q.
J. Thromb. Haemost. 2012;10(8):1521–1531.
48. Jensen RA, Sim X, Smith AV, et al. Novel Genetic Loci Associated With Retinal
Microvascular Diameter. Circ. Cardiovasc. Genet. 2015;9(1):45–54.
49. Sim X, Jensen RA, Ikram MK, et al. Genetic Loci for Retinal Arteriolar
Microcirculation. PLoS One. 2013;8(6):e65804.
50. Louwies T, Vuegen C, Panis LI, et al. miRNA expression profiles and retinal
blood vessel calibers are associated with short-term particulate matter air pollution
exposure. Environ. Res. 2016;147:24–31.
51. De Willige SU, Standeven KF, Philippou H, et al. The pleiotropic role of the
fibrinogen γ′ chain in hemostasis. Blood. 2009;114(19):3994–4001.
52. Peters A, Greven S, Heid IM, et al. Fibrinogen genes modify the fibrinogen
response to ambient particulate matter. Am. J. Respir. Crit. Care Med.
2009;179(6):484–491.
53. Chun C-A, Enomoto K, Sue S. Health Care Issues among Asian Americans. In:
Handbook of Diversity Issues in Health Psychology. Boston, MA: Springer US;
2007:347–365.
132
Chapter 5. Discussion
Summary of findings
The overarching goal of this dissertation was to examine the environmental and
genetic determinants of ocular central nervous system (OCNS) health by measuring
components of the retina. We assessed various aspects of the optic nerve head (ONH),
retinal nerve fiber layer (RNFL) thickness, and retinal vein occlusions (RVO) in a
multiethnic population. To our knowledge, there have been no large-scale population-
based studies to examine the environmental and genetic factors associated with these
measures of the OCNS, in a multiethnic population of Latinos, Chinese-Americans, and
African-Americans.
In Chapter 2, we pooled the largest number of racial/ethnic minorities to describe
the distribution of the ONH and RNFL thickness. We identified potential clinical and
demographic determinants of the ONH and RNFL thickness. Chinese-Americans had
higher RNFL thickness across most clock hours compared to their African-American and
Latino counterparts, after adjusting for age and sex. Chinese participants also had larger
overall disc area, cup-to-disc ratios, and vertical cup-to-disc ratios than their African-
American and Latino counterparts adjusting for these same factors. Age and ethnicity
explained the most proportion of variance in RNFL thickness. These findings suggest that
African-American and Latino eyes are aged by approximately 20 and 5 years,
respectively, with RNFL thickness compared to Chinese-Americans. Because we
controlled for a host of potential confounders, this suggests that role of genetic ancestry
and genetic heritability in our study population.
133
In Chapter 3, we utilized genetic data and 1) determined heritability of RNFL
thickness, 2) identified single genetic variants associated with RNFL thickness, and 3)
performed admixture mapping of RNFL thickness in the Latino population. Genetic
heritability of RNFL thickness ranged from 10%-46% across all quadrants and in
average. Using standard association mapping, we identified two genome-wide significant
SNPs associated RNFL thickness in the nasal and inferior quadrant. One SNP, rs7916697
(within ATOH7), was associated with nasal RNFL thickness. The other variant,
rs10007907, was associated with RNFL thickness in the inferior quadrant. Using
admixture mapping, we identified that in 17q21, African haplotypes were associated with
thinning of the RNFL.
Lastly in Chapter 4, we examined the relationship between ambient air pollution
and traffic density on the risk of RVO. We identified no statistically significantly
associations of the main effects of measures of air pollution and risk of RVO, but
replicated findings that hypertension and age were associated with RVO. When we
examined the role of genetic variants involved in thrombosis, we identified the traffic
density-fibrinogen (FGG) gene variant (rs2066865) interaction and the association with
RVO. We discovered that subjects with each alternative allele of rs2066865 was
associated with a two-fold decrease in the risk of RVO, however this variant was not
statistically significant after performing a Bonferroni adjustment for the number of
comparisons performed.
As shown in this dissertation, there are numerous potential genetic and
environmental factors associated with changes on the OCNS as measured through retinal
health. While we have considered many of the possible factors in OCNS, there are
134
opportunities to further examine additional demographic factors, including the disparities
of race/ethnicity or additional downstream functional genetics and epigenetics of the
OCNS.
Future directions
The findings from this dissertation point to the need for further work in
understanding OCNS health. First, it would be beneficial to replicate these findings in a
larger cohort, including in populations of additional racial/ethnic groups. This would
allow us to better generalize to the larger United States population. Our studies did not
have participants of European descent. While our study is unique to having a multiethnic
population based in the Los Angeles area, we were not best able to make comparisons to
other populations missing (e.g. those of European or other Asian descent). New cohorts
are actively being developed, including new European Eye Epidemiology consortium,
which will collect ophthalmological data on approximately 170,000 European
participants.(1). This cohort will harmonize data from multiple study centers across many
countries. These standardization protocols will be beneficial to validate our findings
within a larger cohort.
While validation of these results is necessary, it is also important to conduct
follow-up studies in the cohorts that have already been collected. Our study findings
describe the natural aging process in RNFL thickness and the ONH measures across
multiple populations, but the current studies used only cross-sectional data. Performing
follow-up studies with the same participants will likely give more robust results in
causality. The LALES population has been followed up at three time points, starting in
135
2004, but the two other study populations have not yet had follow-up studies conducted
to date.
Additionally, we will need further data to gain a better understanding of the role
of genetics in the retina. The first component of this is describing the role of host genetic
variation. In our study, we concluded that there were specific genes involved in changes
of the retinal structures. However, our conclusions were only generalizable to the Latino
population. The design of the genotyping data collected in our Chinese-American
population was not adequate to make comparisons across the two groups. Additionally,
we did not have any genotyping data within the African-American population. Linkage
disequilibrium, or the associations across multiple alleles, is much higher in recently
admixed populations, which include African-Americans, compared to Europeans and
Chinese populations (2,3). This will then require mapping these populations in a specific
manner to identify the causal variants responsible for the changes in the OCNS.
Functional studies of specific genes will be important in elucidating their
mechanisms in the OCNS. This presents a unique challenge because functional human
studies of the OCNS are difficult because it is difficult to gain access to the tissues of
interest. For obvious reasons, the retina and optic nerve, among others, are not readily
accessible. Currently, there are novel technologies and approaches to serve as proxies for
the OCNS. These studies have created and cultured retinal ganglion cells to be as similar
as possible to the host human environment. Once these cells are created, there are novel
technologies to isolate and characterize the key single cells and subsequently detect the
genes that are functional in the OCNS (4,5).
136
There is also a need to further study the environmental component of the OCNS.
While we did not have statistical significance for our findings on the relationship between
air pollutants and the risk RVO, validation of these null results is warranted. Better air
pollution estimates will need to be completed for this population. Our study population
was relatively restricted to a small region in the Los Angeles area. Validation on a larger
and more varied cohort, with respect to residence, might possibly reach different
conclusions. Additionally, using personal air monitors in follow-up studies would allow
for direct comparisons of estimated ambient air pollution (measured in our study through
interpolation from regional monitoring stations). Regardless, our findings, at the very
least, suggest the role that race/ethnicity play on environmental justice, where different
racial/ethnic groups are exposed differently to various air pollutants.
Overall, this dissertation suggests that there is potential for future studies to
consider not only ethnicity in future studies, but the potential functional manifestations in
the OCNS.
Public health impact
The findings of this dissertation suggest the potential roles for interventions. As
mentioned, one of the most important predictors of the OCNS structures was
race/ethnicity. We conclude that those with African ancestry have higher biological ages
with lowered RNFL thickness compared to their Chinese and Latino counterparts. While
race/ethnicity is not directly amenable to intervention, it may be important to design
screenings that incorporate race/ethnicity regarding specific optic neuropathies like
glaucoma. Glaucoma is the leading cause of global blindness, and it is estimated that
there will be 111.8 million people diagnosed with some form of glaucoma in 2040 (6).
137
Additionally, identifying the underlying genetic determinants of the OCNS, may help to
develop future targets for drugs that may be amenable to changes. Lastly, while our
findings regarding the role of environmental air pollution were not conclusive,
understanding mechanisms on air pollutants on changes in the retinal vasculature may
create recommendations on these air pollutants in those that may be at high-risk.
138
REFERENCES
1. Delcourt C, Korobelnik JF, Buitendijk GHS, et al. Ophthalmic epidemiology in
Europe: the “European Eye Epidemiology” (E3) consortium. Eur. J. Epidemiol.
2016;31(2):197–210.
2. Xu S, Huang W, Wang H, et al. Dissecting linkage disequilibrium in African-
American genomes: Roles of markers and individuals. Mol. Biol. Evol.
2007;24(9):2049–2058.
3. Shifman S, Kuypers J, Kokoris M, et al. Linkage disequilibrium patterns of the
human genome across populations. Hum. Mol. Genet. 2003;12(7):771–776.
4. Rheaume BA, Jereen A, Bolisetty M, et al. Single cell transcriptome profiling of
retinal ganglion cells identifies cellular subtypes. Nat. Commun. 2018;9(1):2759.
5. Wu J, Ding X, Hu F, et al. Single cell RNA sequencing reveals cellular diversity of
trisomy 21 retina. bioRxiv. 2019;614149.
6. Tham YC, Li X, Wong TY, et al. Global prevalence of glaucoma and projections
of glaucoma burden through 2040: A systematic review and meta-analysis.
Ophthalmology. 2014;121(11):2081–2090.
Abstract (if available)
Abstract
Population-based studies were conducted to identify demographic, clinical, and genetic determinants of ocular central nervous system (OCNS) health. We assessed various aspects of the OCNS through the retina by examining the optic nerve head (ONH), retinal nerve fiber layer (RNFL) thickness, and retinal vein occlusions (RVO) in a multiethnic population. ❧ In the first study, we pooled the largest number of racial/ethnic minorities to describe the distribution of the ONH and RNFL thickness. Chinese-Americans had higher RNFL thickness across most clock hours compared to their African-American and Latino counterparts, after adjusting for age and sex. Chinese participants also had larger overall disc area, cup-to-disc ratios, and vertical cup-to-disc ratios than their African-American and Latino counterparts. Age and ethnicity explained the greatest proportion of variance in RNFL thickness. These findings suggest that African-American and Latino eyes are aged by approximately 20 and 5 years, respectively, with RNFL thickness compared to Chinese-Americans. ❧ In the second study, we utilized whole genome genotyping genetic data and 1) determined heritability of RNFL thickness, 2) identified single genetic variants associated with RNFL thickness, and 3) performed admixture mapping of RNFL thickness in the Latino population. Genetic heritability of RNFL thickness ranged from 10%-46% across all quadrants. Using standard association mapping, two genetic variants were associated with RNFL thickness in the nasal and inferior quadrant. Using admixture mapping, we identified that in the 17q21 loci, increasing proportions of African ancestry were associated with thinning of the RNFL. ❧ Lastly, we examined the relationship between ambient air pollution and traffic density on the risk of RVO. There were no statistically significantly associations of the main effects of air pollutants and risk of RVO, but replicated the hypertension and age associations with RVO. The association between traffic density and risk of RVO was modified by a fibrinogen (FGG) genetic variant. ❧ There have been no population-based studies to examine the environmental and genetic factors associated with these measures of OCNS health. These findings address the gaps in the literature regarding clinical, demographic, and genetic factors of OCNS health as measured through the retina in a multiethnic population of Latinos, Chinese-Americans, and African-Americans.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Identifying genetic, environmental, and lifestyle determinants of ethnic variation in risk of pancreatic cancer
PDF
Genetic and environmental risk factors for childhood cancer
PDF
Chronic eye disease epidemiology in the multiethnic ophthalmology cohorts of California study
PDF
Age related macular degeneration in Latinos: risk factors and impact on quality of life
PDF
The environmental and genetic determinants of cleft lip and palate in the global setting
PDF
Genomic risk factors associated with Ewing Sarcoma susceptibility
PDF
Vision epidemiology and the impact of vision loss on vision-specific quality of life
PDF
Predictive factors of breast cancer survival: a population-based study
PDF
Ancestral/Ethnic variation in the epidemiology and genetic predisposition of early-onset hematologic cancers
PDF
The role of heritability and genetic variation in cancer and cancer survival
PDF
Genetic epidemiological approaches in the study of risk factors for hematologic malignancies
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
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
The role of genetic ancestry in estimation of the risk of age-related degeneration (AMD) in the Los Angeles Latino population
PDF
Characterization and discovery of genetic associations: multiethnic fine-mapping and incorporation of functional information
PDF
Investigating a physiological pathway for the effect of guided imagery on insulin resistance
PDF
Diet quality and pancreatic cancer incidence in the multiethnic cohort
Asset Metadata
Creator
Nousome, Darryl Reth
(author)
Core Title
Characterizing the genetic and environmental contributions to ocular and central nervous system health
School
Keck School of Medicine
Degree
Doctor of Philosophy
Degree Program
Epidemiology
Publication Date
07/26/2019
Defense Date
07/24/2019
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
Air pollution,ancestry,Epidemiology,genetics,GWAS,OAI-PMH Harvest,ocular coherence tomography,Ophthalmology,retina,retinal nerve fiber layer,retinal vein occlusion
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Mckean-Cowdin, Roberta (
committee chair
), Gauderman, Jim (
committee member
), Jiang, Xuejuan (
committee member
), Richter, Grace (
committee member
)
Creator Email
dnousome@gmail.com,drnousome@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c89-195537
Unique identifier
UC11663149
Identifier
etd-NousomeDar-7646.pdf (filename),usctheses-c89-195537 (legacy record id)
Legacy Identifier
etd-NousomeDar-7646.pdf
Dmrecord
195537
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Nousome, Darryl Reth
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
Tags
genetics
GWAS
ocular coherence tomography
retina
retinal nerve fiber layer
retinal vein occlusion