Effects of Ultraviolet Radiation Exposure on Oxidative Stress Markers on the Human Ocular Surface

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Kristina Marie Haworth, OD, MS

Graduate Program in Vision Science

The Ohio State University

2014

Dissertation Committee:

Heather Chandler, PhD, Advisor

Donald Mutti, OD, PhD

Karla Zadnik, OD, PhD

Tatiana Oberyszyn, PhD

Kristina Marie Haworth

2014

Abstract

Purpose: To evaluate feasibility, repeatability, and relationships between measures of ocular sun exposure, conjunctival ultraviolet autofluorescence (UVAF), hexanoyl-lysine

(HEL), and 8-hydroxy-2’-deoxyguanosine (8OHdG) for in vivo human tear and conjunctival cell sample types.

Methods: Fifty healthy volunteers were seen for 2 visits 14±2 days apart. Ocular sun exposure was estimated by questionnaires quantifying time outdoors and ocular protection habits. Conjunctival UVAF images were obtained using a Nikon D7000 camera system adapted with filters to photograph only UV excited visible light.

Collection of tear samples was by glass microcapillary tube; conjunctival cells by cytology brush. Image analysis was accomplished using ImageJ image analysis software.

HEL and 8OHdG ELISAs were conducted for both tear and conjunctival cell samples.

Normality of distributions for each outcome measure was evaluated (Shapiro-Wilk); data was transformed to approximate normal distributions. Dichotomous outcome variables for occupation, contact lens wear, conjunctival UVAF, and ocular sun exposure were evaluated. Furthermore, estimates of repeatability for ocular sun exposure, UVAF and the

ELISAs were evaluated using Bland-Altman plots with mean bias and 95% limits of agreement calculated. Linear or logistic regression was used to evaluate for relationships of continuous and dichotomous outcome variables respectively. Multiple linear regression models were built using a forward stepwise approach. ii

Results: Descriptive statistics were generated for outcome measures of ocular sun exposure, conjunctival UVAF, HEL expression in tears and cells, and 8OHdG expression in tears. Bland-Altman plots produced demonstrated repeatability for each outcome measure and for the ELISA assays. Non-normally distributed raw data for each of the continuous outcome measures were transformed using either log10 (ocular sun exposure,

HEL expression in tears and cells, and 8OHdG expression in tears) or square root

(UVAF) transformations to approximate normal distributions. Several univariate relationships were identified; multivariate linear regression analysis supported significant relationships for each outcome variable: ocular sun exposure with outdoor occupation; tear HEL expression with increased tear 8OHdG expression and decreased cell HEL expression; cell HEL expression with decreased tear HEL expression, decreased contact lens wear, and inversely with age; tear 8OHdG expression with increased tear HEL expression, and inversely with age; total UVAF with decreased tear 8OHdG expression.

Conclusions: We demonstrate feasibility and repeatability over a two week time period for ocular sun exposure and UVAF, and for previously unreported ELISA measurements of oxidative stress levels in human tears (8OHdG) and conjunctival cells (HEL). Our

HEL ELISA findings for tears are notably higher than previous reports on human tears, possibly due to experimental differences. Results suggest that ocular surface lipid peroxidation is accompanied by DNA damage, lipid peroxidation of tears and conjunctival cells fluctuate inversely, and cellular changes causing UVAF are inhibitory toward DNA damage evidence in tears. Other factors that may influence expression of

iii ocular surface oxidative stress biomarkers are gender, season of collection, contact lens wear, and UV exposure.

Dedication

This document is dedicated to my son, Ethan, for his inspiring

unquenchable thirst for knowledge and insightful wisdom.

Acknowledgments

I would like to thank my advisor, Heather Chandler, for her diligent assistance to coach me through laboratory analysis protocols, presentation development, and ultimately with this dissertation. Another tireless supporter, Karla Zadnik, has provided support for my research from her heart and by providing financial resources, and to her I will always be grateful. Thanks are also well deserved by Don Mutti, chief statistics consultant for my dissertation, whose enthusiasm for research is unrivaled. I am also thankful for the many wonderful friends I have made during my time as a graduate student who have provided me with untold levels of intellectual stimulation and emotional support. In summary, I will forever be grateful for each person I have had the privilege of knowing, that without my experience as a graduate student at The Ohio State

University College of Optometry, would have left an incalculable void in my life.

Vita

1964...... Born, Glendale, CA

1982...... High School Diploma, Fresno Adventist

Academy

1986...... BBA Business Administration, Loma

Linda University

1993...... OD Doctor of Optometry, Southern

California College of Optometry

1993 to 2001 ...... Clinical Optometrist, Fresno, CA

2001 to 2007 ...... Optometric Practice Owner, Kristina M.

Haworth, OD, Fresno, CA

2009...... MS Vision Science, The Ohio State

University

2007 to 2008 ...... Graduate Teaching Associate, College of

Optometry, The Ohio State University

2008 to 2011 ...... Postdoctoral Fellow, College of Optometry,

The Ohio State University

vii

2011 to present ...... Assistant Professor - Practice, College of

Optometry, The Ohio State University

Publications

Haworth KM and Chandler HL. Oxidative Stress Measures and Correlations with Lipid and DNA Damage Assays for Human Tears and Conjunctival Cells. (In preparation)

Pucker AD and Haworth KM. The Presence & Significance of Polar Meibum & Tear

Lipids. The Ocular Surface, accepted manuscript available online: 8-OCT-2014; DOI information: 10.1016/j.jtos.2014.06.002.

Haworth KM, Nichols JJ, Thangavelu M, Sinnott LT, Nichols KK. Examination of human meibum collection and extraction techniques. Optometry and Vision Science Apr

2011;88(4):525-533.

Peer-reviewed Abstracts

Haworth KM and Chandler HL. Conjunctival Ultraviolet Autofluorescence and Ocular

Sun Exposure. 2014. Accepted for presentation at the American Academy of Optometry annual meeting, Denver, CO, on Wednesday, November 12, 2014. Paper.

viii

Haworth KM and Chandler HL. Oxidative Stress Measures and Correlations with Lipid and DNA Damage Assays for Human Tears and Conjunctival Cells. 2014. Association for Research in Vision and Ophthalmology annual meeting, Orlando, FL. ARVO Meeting

Abstracts April 30, 2014 55:2769. Poster.

Mutti DO, Shiley KA, Haworth KM, Zadnik K. Dietary Vitamin D Consumption in

Young Adult Myopes. 2011. Association for Research in Vision and Ophthalmology annual meeting, Fort Lauderdale, FL. ARVO meeting abstracts April 22, 2011 52:2702.

Poster.

Haworth KM, Chen J, Green-Church KB, Nichols KK. Comparison of Various

Collection Techniques for Human Meibum with Mass Spectrometric Analysis. 2009.

American Academy of Optometry annual meeting, Orlando, FL. American Academy of

Optometry Archives #90725. Paper.

Haworth KM, Nichols JJ, Thangavelu M, Sinnott LT, Nichols KK. Examination of

Human Meibum Collection and Extraction Techniques. 2009. Association for Research in

Vision and Ophthalmology annual meeting, Fort Lauderdale, FL. ARVO meeting abstracts April 11, 2009 50:2544. Paper.

Fields of Study

Major Field: Vision Science

Table of Contents

Abstract ...... ii

Dedication ...... v

Acknowledgments...... vi

Vita ...... vii

Table of Contents ...... xi

List of Tables ...... xvi

List of Figures ...... xviii

Chapter 1: Introduction ...... 1

1.1 Public Health Relevance and Project Significance ...... 1

Chapter 2: Background ...... 5

2.1 Ultraviolet Radiation Exposure ...... 5

2.1.1 Ultraviolet Radiation Exposure and Ocular Absorption ...... 5

2.1.2 Peripheral Light Focusing Effect (Coroneo Effect) ...... 6

2.1.3 Ultraviolet Radiation Protection with Eyewear ...... 7

2.1.4 Ultraviolet Radiation Exposure Assessment Challenges ...... 10

2.1.5 Ultraviolet Radiation Protection and Assessment...... 13 xi

2.2 Oxidative Stress...... 17

2.2.1 Mechanisms of production ...... 17

2.2.2 Effects on cellular components and activity ...... 17

2.2.3 Methods of oxidative stress assessment ...... 20

2.3 Fluorescence ...... 24

2.3.1 Extrinsic (induced) fluorescence ...... 26

2.3.2 Intrinsic fluorescence (autofluorescence) ...... 27

Chapter 3: Methods ...... 33

3.1 Research Design ...... 33

3.2 Sample Size ...... 33

3.3 Subjects ...... 34

3.4 Measurement / Instrumentation ...... 34

3.4.1 Tear collection ...... 34

3.4.2 Conjunctival brush cytology ...... 35

3.4.3 Ocular surface vital staining ...... 35

3.4.4 Ocular sun exposure ...... 36

3.4.5 Quantification of Oxidative Damage by HEL ELISA ...... 37

3.4.6 Quantification of Oxidative Damage by 8OHdG ELISA ...... 39

3.4.7 Conjunctival UV Autofluorescence imaging and analysis ...... 40

xii

Chapter 4: Results ...... 45

4.1 Subject characteristics ...... 45

4.2 Descriptive statistics ...... 46

4.2.1 Feasibility study ...... 46

4.2.2 Cross-sectional study ...... 57

4.4 Regression analyses...... 76

4.4.1 Ocular sun exposure ...... 87

4.4.2 Tear HEL expression ...... 90

4.4.3 Cell HEL expression ...... 94

4.4.4 Tear 8OHdG expression ...... 97

4.4.5 Total UVAF ...... 101

Chapter 5: Discussion ...... 103

5.1 Logistic regression ...... 103

5.1.1 Occupation ...... 103

5.1.2 Contact lens wear ...... 104

5.1.3 Season of collection ...... 105

5.1.4 Dichotomous UVAF and ocular sun exposure ...... 106

5.2 Ocular sun exposure assessment ...... 107

5.2.1 Ocular sun exposure repeatability ...... 109

xiii

5.2.2 Ocular sun exposure linear regression ...... 111

5.3 Oxidative stress assessment ...... 113

5.3.1 Lipid damage detection ...... 113

5.3.1.1 Tear HEL expression repeatability ...... 114

5.3.1.2 Tear HEL expression linear regression ...... 116

5.3.2.1 Cell HEL expression repeatability ...... 119

5.3.2.2 Cell HEL expression linear regression ...... 123

5.3.2 DNA damage detection ...... 126

5.3.2.1 Tear 8OHdG expression repeatability ...... 127

5.3.2.2 Tear 8OHdG expression linear regression...... 129

5.4 Conjunctival ultraviolet autofluorescence detection ...... 134

5.4.1 Conjunctival Ultraviolet Autofluorescence repeatability ...... 136

5.4.2 Conjunctival ultraviolet autofluorescence linear regression ...... 138

5.5 Limitations ...... 140

Chapter 6: Summary and Future Directions ...... 142

6.1 Summary ...... 142

6.1.1 Feasibility study ...... 142

6.1.2 Cross-sectional study ...... 142

6.2 Future Directions ...... 143

xiv

References ...... 145

Appendix A: Questionnaires……………………………………………………………156

List of Tables

Table 1. Ground reflectance of UVB for various terrain surfaces.17 ...... 10

Table 2. Descriptive statistics and repeatability for feasibility study reported measures. 46

Table 3. Statistically significant correlations identified using Pearson's correlation analysis for feasibility study, confirmed for small sample sizes using Spearman's Rho. . 52

Table 4. Percent change storage effects for tear 8OHdG expression in ten samples stored at -80 degrees Celsius over one year...... 56

Table 5. Descriptive statistics for continuous outcome variable raw data...... 57

Table 6. Shapiro-Wilk test for normal distribution of continuous outcome variable raw data...... 58

Table 7. Descriptive statistics for continuous outcome variable transformed data...... 59

Table 8. Bland-Altman estimates of repeatability for raw and transformed data from cross-sectional study outcome measures, and for visit 1 ELISA transformed values...... 60

Table 9. Univariate logistic regression for raw and transformed outcome variables: occupation, contact lens wear and season of collection...... 78

Table 10. Univariate linear regression results for raw data continuous outcome variables ocular sun exposure and total conjunctival UV autofluorescence...... 79

Table 11. Univariate linear regression results for raw data continuous outcome variables tear HEL expression, cell HEL expression, and tear 8OHdG expression...... 80

xvi

Table 12. Univariate linear regression results for transformed continuous variables ocular sun exposure and total conjunctival UV autofluorescence...... 81

Table 13. Univariate linear regression results for transformed continuous outcome variables tear HEL expression, cell HEL expression, and tear 8OHdG expression...... 82

Table 14. Univariate linear regression for reported activities by visit for transformed continuous outcome variable total UVAF...... 83

Table 15. Univariate linear regression for reported activities by visit for transformed continuous variables tear HEL expression, cell HEL expression, and tear 8OHdG expression...... 84

Table 16. Linear regression models summary for ocular sun exposure...... 89

Table 17. Linear regression models summary for tear HEL expression...... 93

Table 18. Linear regression models summary for cell HEL expression...... 96

Table 19. Linear regression models summary for tear 8OHdG expression...... 100

Table 20. Linear regression models summary for total UVAF...... 102

xvii

List of Figures

Figure 1. A proposed model of the precorneal tear film showing the relationship and interaction of lipid-binding proteins and the outer lipid layer.7 Reproduced with permission of Association for Research in Vision and Ophthalmology in the format republish in a thesis/dissertation via Copyright Clearance Center...... 4

Figure 2. Relative absorption of UV radiation within ocular structures.19 Reproduced with permission, from Lucas RM, An epidemiological perspective of ultraviolet exposure-- public health concerns. Eye Contact Lens. Jul 2011;37(4):168-175...... 6

Figure 3. Peripheral Light Focusing effect: light incident at 120° temporally is enhanced by a factor up to 20 times at the nasal limbus...... 7

Figure 4. UV exposure with common spectacle designs.17 Reproduced with permission, from Sliney DH, How light reaches the eye and its components. International journal of toxicology. Nov-Dec 2002;21(6):501-509...... 9

Figure 5. Monthly averaged annual ambient erythemally weighted UVR, 1997-2003.1

Reproduced with permission of the publisher, from Lucas RJ, McMichael T, Smith W,

Armstrong B, Solar Ultraviolet Radiation: Global burden of disease from solar ultraviolet radiation. In: Environmental Burden of Disease Series, No. 13, Geneva,

World Health Organization, 2006...... 12

xviii

Figure 6. ROS action in cells.9 Reproduced with permission, from Avery SV, Molecular targets of oxidative stress. Biochem J. Mar 1 2011;434(2):201-210. © the Biochemical

Society...... 19

Figure 7. Jablonski diagram of absorbance, non-radiative decay, and fluorescence. By

Jacobkhed, 2012 (Creative Commons license CC0).107...... 26

Figure 8. Emission spectra from intrinsic tissue fluorophores. Reproduced with permission of Springer-Verlag New York Inc., Principles of Fluorescence Spectroscopy, third edition, Lakowicz JR, Chapter 3, 2006;106 permission conveyed through Copyright

Clearance Center, Inc...... 29

Figure 9. Conjunctival UV autofluorescence imaging system...... 41

Figure 10. Bland-Altman plot illustrating test-retest variability of ocular sun exposure in

20 subjects...... 48

Figure 11. Bland-Altman plot illustrating test-retest variability of tear HEL levels in 20 subjects...... 49

Figure 12. Bland-Altman plot illustrating test-retest variability of cell HEL levels in 7 subjects...... 50

Figure 13. Bland-Altman plot illustrating test-retest variability of tear 8OHdG levels in

20 subjects...... 51

Figure 14. Correlation between ocular sun exposure and tear HEL levels for first twenty subjects. The linear regression equation is shown in the box. Correlation coefficients and associated p values are further specified in Table 2...... 53

xix

Figure 15. Correlation between cell HEL levels and tear 8OHdG levels for first twenty subjects. The linear regression equation is shown in the box. Correlation coefficients and associated p values are further specified in Table 2...... 54

Figure 16. Correlation between tear HEL levels and tear 8OHdG levels for first twenty subjects. The linear regression equation is shown in the box. Correlation coefficients and associated p values are further specified in Table 2...... 55

Figure 17. Bland-Altman plot illustrating test-retest variability of raw data ocular sun exposure in 50 subjects...... 61

Figure 18. Bland-Altman plot illustrating variability of raw data conjunctival UVAF in

34 subjects...... 62

Figure 19. Bland-Altman plot illustrating test-retest variability of raw data tear HEL expression in 50 subjects...... 63

Figure 20. Bland-Altman plot illustrating test-retest variability of raw data cell HEL expression in 37 subjects...... 64

Figure 21. Bland-Altman plot illustrating test-retest variability of raw data tear 8OHdG expression in 50 subjects...... 65

Figure 22. Bland-Altman plot illustrating test-retest variability of log10 transformed data for ocular sun exposure in 50 subjects...... 66

Figure 23. Bland-Altman Altman plot illustrating variability of square root transformed data for total conjunctival UVAF in 34 subjects...... 67

Figure 24. Bland-Altman plot illustrating test-retest variability of log10 transformed data for tear HEL expression in 50 subjects...... 68

Figure 25. Bland-Altman plot illustrating test-retest variability of log10 transformed data for cell HEL expression in 37 subjects...... 69

Figure 26. Bland-Altman plot illustrating test-retest variability of log10 transformed data for tear 8OHdG expression in 50 subjects...... 70

Figure 27. Bland-Altman plot illustrating test-retest variability of log10 transformed data for visit 1, well 1 and well 2, for tear HEL ELISA in 50 subjects...... 71

Figure 28. Bland-Altman plot illustrating test-retest variability of log10 transformed data for visit 1, well 1 and well 2, for cell HEL ELISA in 37 subjects...... 72

Figure 29. Bland-Altman plot illustrating test-retest variability of log10 transformed data for visit 1, well 1 and well 2, for tear 8OHdG ELISA in 50 subjects...... 73

Figure 30. Bland-Altman plot illustrating repeatability for ocular sun exposure during winter season of collection (n=15)...... 74

Figure 31. Normal Q-Q plot illustrating normal distribution of ocular sun exposure for spring season of collection...... 75

Figure 32. Normal Q-Q plot illustrating non-normal distribution for ocular sun exposure during winter season of collection...... 76

Figure 33. Questionnaire estimating proportion of time spent using personal ocular protective measures and time outdoors...... 157

Figure 34. Questionnaire estimating hours spent outdoors in seven activity categories. 158

xxi

Chapter 1: Introduction

1.1 Public Health Relevance and Project Significance

Ultraviolet radiation (UV) exposure is a significant public health problem. The

World Health Organization (WHO) estimates UV exposure is responsible for significant loss of quality of life and 60,000 premature deaths annually.1 Ambient UV exposure is influenced by distal environmental factors such as cloud cover and latitude; and is further modified by proximal individual factors such as behavior and skin pigmentation.

While UV exposure is known to be beneficial for systemic vitamin D production, no ocular benefits are known. In fact, WHO identified nine diseases with enough support in the literature to be associated with UV exposure; among those nine diseases, three are specific to the eye: cortical cataract, pterygium, and squamous cell carcinoma of the cornea and conjunctiva.1 Other ocular diseases thought to be associated with UV exposure include photokeratitis, ocular melanoma,2 and age-related macular degeneration.3 On the ocular surface, corneal conditions of photokeratitis and pterygia are the most common. It is thought these UV exposure associated diseases involve oxidative stress.

Oxidative stress is known to be an important component of UV induced damage.

In general, oxidative stress occurs when the production of reactive oxygen species

(ROS) outweighs the capacity of antioxidants to detoxify the system. While ROS are a

1 normal part of healthy biological processes, extended levels of elevated oxidative stress leads to tissue injury and molecular damage.4 Levels of oxidative stress can be influenced by ultraviolet (UV) radiation exposure, toxic chemicals, and genetic responses.5

The 2012 National Eye Institute (NEI) Vision Research Needs, Gaps, and

Opportunities6 report identifies oxidative stress as playing a key role in cataract formation and other age-related degenerative diseases. Oxidative stress induced ocular oxidative damage is thus recognized as an important research need. Furthermore, the need for development of effective treatments by identifying new biomarkers to better define ocular surface disease is emphasized.6 Findings from this study will provide insight into potential biomarkers of UV induced oxidative stress and may lead to future development of methods for treatment and prevention of UV associated ocular disease.

On a cellular level, ROS are capable of interacting with lipids in cell membranes causing lipid peroxidation; formation of DNA oxidation, protein oxidation, and accumulation of deleterious products may result. The first point of contact for UV exposure on the ocular surface is the tear film. At the forefront of the tear film is the lipid bilayer, with the tear film being the primary source of oxygen for the cornea (Figure 1).7

Additionally, proteins within the lipid bilayer and the adjacent aqueous phase are continuously exposed to the elements and are therefore subject to possible UV exposure-induced oxidative damage (Figure 1).7 Furthermore, just posterior to the tear film lays the conjunctiva with its large surface area (approximately 80% of the ocular surface) composed in part, by several layers of mitochondria rich epithelial cells, with

2 each cell enclosed by a lipid bilayer membrane.8 Each of these lipid bilayer structures are subject to ROS mediated UV damage and untold downstream effects.9

Ultraviolet-induced oxidative stress has been demonstrated to cause protein, lipid, and DNA damage in animal models10,11 and cell culture;12,13 however, human studies showing UV-induced ocular surface oxidative stress are not found in the literature. This void in our knowledge base and the scientific literature is a barrier to progress in the field for development of viable effective methods to reduce or eliminate ocular disease induced by UV exposure. Based on animal models it is anticipated that markers of oxidative stress will be detected on the human ocular surface, with higher amounts of UV exposure resulting in increased oxidative stress.

Findings from this study will create the necessary groundwork for future studies to help understand the mechanisms of UV damage on human ocular tissue. Therapeutic benefits may emerge from this enhanced knowledge and aid in development of new methods for prevention of UV-induced ocular disease.

The objective of this study is to test the hypothesis that expression of damaged protein, lipid, and DNA products will be higher in human tears and conjunctival cells and that the area of conjunctival UV autofluorescence will be greater for individuals exposed to greater levels of UV radiation when compared with those exposed to less UV radiation.

Chapter 2: Background

2.1 Ultraviolet Radiation Exposure

2.1.1 Ultraviolet Radiation Exposure and Ocular Absorption

The solar spectrum is divided into three general categories: visible, UV, and infrared. The most physiologically relevant spectral category is the UV region. The UV spectrum was further divided in 1932 at the Second International Congress on Light into three regions: UVA (400-315 nm), UVB (315-280 nm), and UVC (280-100 nm).14 It is well established UVB radiation causes about 80% of the harmful effects in human skin, with UVA contributing about 20% of the damage.15 Ultraviolet B is the erythemal action spectrum that causes most skin damage from temporary acute (sunburn) or more permanent chronic exposure (nevi or cancer).15,16 Nonetheless, skin absorption varies by wavelength with increasingly deeper tissue absorption with UVA compared to UVB.16

Similar UV absorption characteristics are present in ocular components as the cornea absorbs most UVB, the aqueous and lens absorb most UVA, and only a small, yet clinically significant, portion (~1%) of UVA reaches the retina (Figure 2).17 By comparison, the cornea is approximately 1000 times more sensitive to UVB compared with UVA.18 Thus, each of the ocular components is subject to UV radiation exposure and the possible damaging effects.

2.1.2 Peripheral Light Focusing Effect (Coroneo Effect)

The most prominent ocular surface damage associated with UV exposure is explained by the Peripheral Light Focusing effect (Coroneo effect).20 The Peripheral

Light Focusing (PLF) effect estimates up to 20-fold amplification of UV at the nasal limbus (area between the conjunctiva and cornea) when light is incident on the temporal side of the eye at an angle of approximately 120 degrees (Figure 3). The PLF effect is believed responsible for the predilection of pterygia21 and cortical cataracts22,23 to the nasal region. Because the area where UV amplification is greatest contains limbal stem cells,24 damage to the limbus may hold significant implications for corneal and

6 conjunctival health and disease. Personal UV protective measures capable of protecting the limbal region from the PLF are limited to goggles, UV-blocking soft contact lenses, and to a lesser degree, wrap-around style sunglasses.17,25

Figure 3. Peripheral Light Focusing effect: light incident at 120° temporally is enhanced by a factor up to 20 times at the nasal limbus.

2.1.3 Ultraviolet Radiation Protection with Eyewear

Quality of available non-prescription sunglass lenses varies considerably by manufacturer with respect to UV transmission levels.26,27 Although most prescription lenses and many higher quality sunglass lenses are designed to block UV transmission, protection of the eyes is only from directly incident frontal UV rays, leaving superior, temporal, and inferior portions of the eye potentially exposed (Figure 4).17 Of particular importance are areas of temporal and inferior exposure. As discussed above, the PLF

7 effect describes how temporal exposure can result in concentrated UV damage to the sensitive limbal region. In fact, mannequin studies using photodetectors positioned to measure UV at the nasal limbus have shown common styles of sunglasses (non wrap- around) provide little to no protection from PLF.25 Exposure to the inferior ocular surface is also important to consider as a substantial amount of chronic UV exposure to the eyes can occur secondary to ground reflected rays from horizontal surfaces.17 Ground reflections of UV vary greatly based on the surface material: for example, snow is approximately 100 times more reflective than grass (Table 1).17 In addition, it is thought that wearing poor fitting or poor quality sunglasses may actually increase ocular UV exposure by decreasing the natural responses of pupil constriction and squinting.17

Repeated UV exposure is established as a cause of damage to cellular components;28 while ocular UV exposure varies by individual,29 the importance of its assessment is apparent.

Table 1. Ground reflectance of UVB for various terrain surfaces.17

Representative Diffuse reflectance terrain surfaces solar UVB Green mountain grassland 0.8% to 1.6% Dry, parched grassland 2% to 3.7% Wooden boat dock 6.4% Black asphalt 5% to 9% Concrete pavement 8% to 12% Atlantic beach sand (dry) 15% to 18% Atlantic beach sand (wet) 7% Sea foam 25% to 30% Dirty snow 50% Fresh snow 88%

2.1.4 Ultraviolet Radiation Exposure Assessment Challenges

Factors influencing UV exposure include distal environmental factors such as cloud cover, latitude, season, ozone levels, and atmospheric pollution.1 In addition, proximal individual factors further modify UV exposure; among them are behavior, genetic variation in skin pigmentation and sun sensitivity, and cultural dress.1 While environmental factors may have equal effects on UV exposure to a population at a specific time and location, significant variance is likely to occur on an individual level.

Because refinements for ocular UV exposure measurements are not yet fully developed, many studies report values for ocular UV exposure based on dermatology research which includes factors such as skin pigmentation and sun sensitivity.30-34 Sun exposure tolerance for skin has been defined in terms of Minimum Erythemal Dose (MED) and 10

Minimum Melanogenic Dose (MMD).35 However, the nature of the ocular surface is quite obviously different than that of the skin (i.e. differences in surface area and pigmentation) suggesting that estimations of cutaneous exposure may not be relatable to ocular UV exposure.36 Furthermore, UV exposure levels are frequently estimated by comparing the subjects’ lifetime history of geographic location with general ambient UV levels as determined by ground monitors or satellites.3,32-34,36-39

Geographic location influences individual UV responses on some level due to latitude and altitude changes. Latitude changes have a profound effect on the amount of ambient UV exposure possible with latitudes nearest the equator having the highest levels of ambient UV radiation (Figure 5).1 In fact, for each degree change of latitude towards the equator, erythemally effective UV radiation increases by approximately 3–4%.40

Furthermore, when considering differing altitudes at approximately equivalent latitude, a

3-4% decrease in UV radiation occurs per 1000 feet descent.40 Decreased pollution levels and stratospheric ozone are also factors that may cause increased ambient UV radiation levels.40 In addition, it cannot be ignored that individual exposure is also greatly dependent on personal activities that are known to vary throughout the day and by season of the year.40

Figure 5. Monthly averaged annual ambient erythemally weighted UVR, 1997-2003.1

Reproduced with permission of the publisher, from Lucas RJ, McMichael T, Smith W,

Armstrong B, Solar Ultraviolet Radiation: Global burden of disease from solar ultraviolet radiation. In: Environmental Burden of Disease Series, No. 13, Geneva,

World Health Organization, 2006.

Ultraviolet radiation levels also are known to vary both diurnally and seasonally.

Much of the diurnal variation is the result of the solar zenith angle.41 Because the solar zenith angle changes throughout the day, ambient UV intensity is greatest when the sun is directly overhead and lowest when the sun is near the horizon.41 Seasonal variation of ambient UV occurs due to changes in daytime lengths and geographic proximity to the sun. Seasonal variations thus cause higher ambient UV levels during summer months 12 when days are longer and the earth is closer to the sun, when compared to winter months.42,43

It is commonly thought the highest UV exposure occurs during the summer because outdoor activities are generally greater during summer months due to longer, warmer days. In fact, while personal ambient exposure is estimated at 30% of total ambient UV throughout the year, seasonal variation of ambient UV is well known, with personal UV exposure in winter approximately 40% of UV exposure during summer.43

While this may prove true, it is important to consider individual activities and surface reflectance together for a more full assessment of ocular UV exposure.43 As one example, individuals who engage in summer outdoor activities over grass (low reflectance) may require more hours of activity to reach an equivalent level of UV exposure as those who engage in winter snow (high reflectance) activities.44

2.1.5 Ultraviolet Radiation Protection and Assessment

Assessment of UV exposure in most epidemiologic studies has been through methods subject to recall bias and with disregard for individual activities outside of personal protection habits.45 Subsequently, many epidemiologic studies report associations (or lack of association) between UV exposure and disease based on subject recall for place of residence and occupation compared with recorded ambient UV levels for each reported location.45 Potential for error is significant using this method as ambient

UV levels vary up to 30 percent dependent on factors such as season and shade cover.46

Hence, for this reason and those discussed earlier, attempts to assess lifetime UV exposure may have limited relevance for assessing correlations to disease development

13 for individuals. However, estimates of UV exposure may be more reliable for recall of activities than for recall of time spent outdoors.47 Consequently, in an effort to simplify and improve accuracy of outdoor activity recall, Kwok, et al. reduced more than 4,000 activity types gleaned from activity diaries into a study-friendly number of seven; the seven item questionnaire was shown to account for approximately 94% of time spent outdoors,48 and was thus chosen for use in our study. However, because of the influence of individual protective measures on actual UV exposure, it is equally vital to assess the use of ocular sun protective measures while spending time outdoors.25

Questionnaires to assess skin UV protective measures typically include questions about exposed parts of the body, tendency to seek shade, and use of hats and sunscreen.49

Self-reported responses to these questionnaires correlate only weakly to personal UV dosimetry.49 It is clear that recall bias and ecologic fallacy from these non-validated questionnaires causes significant dilution of the contribution UV exposure may have on disease formation. Furthermore, UV exposure varies up to 10-fold by body part and posture (sitting or standing).50,51 Hence, assessment of ocular exposure presents a need for special considerations.

Ocular UV exposure is known to vary significantly from ambient UV levels and skin exposure, largely due to the position of the eyes in the human body and the common use of ocular specific UV protective measures (i.e. use of hats, clear or dark spectacles, and UV-blocking contact lenses).25,29 Ocular tissues are protected by the upper brow bone, eyebrows and lashes from some solar zenith angles, particularly the one with highest ambient UV intensity (when the sun is directly overhead).17,52 Ocular UV

14 exposure is heavily influenced by the angle of incident light. Erythemal UV damage is maximal at solar noon with careful UV skin protection recommended two hours before and four hours after solar noon.53-55 Ocular exposure to solar UV radiation peaks bimodally at solar elevation angles of 40° for UVA and 50° for UVB.52 Thus, although there is some seasonal variation,56 evidence supports that the time of day most significant to ocular tissues is bimodal, occurring about three to four hours before and after the period of most intense overhead ambient UV.57,58 This means the bimodal distribution for ocular UV radiation peaks during times of day commonly believed to be “safe” from UV exposure. This has the possibility to result in an inadvertent increase in ocular UV exposure due to diminished use of UV protective measures.58 As such, it is likely that a simple questionnaire about the duration of time spent outdoors compared with ambient

UV levels is not the most effective way to measure UV exposure. In support of this,

Threlfall and colleagues demonstrated that when compared with general ambient estimation of sun exposure, accounting for ocular specific sun exposure measures resulted in a nearly 3 fold increase in the odds ratio for development of pterygium.29

Their findings suggest that it is useful to obtain self-reported estimates of the proportion of time ocular protective measures are used while an individual is outdoors.

An additional method developed to improve UV exposure assessment is personal

UV dosimetry. Personal UV dosimetry has taken several forms, beginning with polysulphone film badges.59 Since the unintentional discovery of UV responsive properties of polysulphone approximating the action spectra for human skin in 1976,59 the polysulphone badge became widely used in research.60 Polysulfone badges are

15 composed of a UV radiation exposure degradable thermoplastic polymer that is durable, inexpensive, and widely available.59 Polysulphone badges are usually worn on the chest to gather UV exposure information from the front exposed area.60 Specifically to improve

UVB detection for the eye, spectacle mounted polysulphone badges have been developed;61 however, use of this device has not been widely implemented. In addition, polysulphone contact lenses were used in an isolated study in an effort to improve assessment of ocular specific UV exposure.62 This study demonstrated through polysulphone contact lens wear that ocular UV exposure was significantly overestimated when compared to polysulphone badges worn in other facial locations.62 Polysulphone

UV dosimetry is limited by low saturation levels, lack of temporal resolution, and large variability (low specificity).63 As an alternative to polysulphone, biologically weighted personal UV dosimeters have been developed such as VioSpor® (BioSense, Germany).

The VioSpor® UV dosimeter uses DNA of spores known to have a similar response as human skin to UV exposure.64,65 Interest in the UV exposure research community for greater accuracy and temporal resolution in assessment of personal UV dosimetry has opened the door for development of electronic UV dosimeters.

One example of an electronic personal UV dosimeter is commercially available from the National Institute of Water and Atmospheric Research (NIWA) in New Zealand in a convenient wristwatch size design.66 The NIWA personal UV dosimeters contain an

AlGaN photodiode radiation sensor calibrated to sense the erythemal UV spectrum.44

NIWA dosimeters have several advantages over other personal UV dosimeters in that

16 they are time-stamped, lightweight, water resistant, and can gather data continuously for several weeks.44

2.2 Oxidative Stress

2.2.1 Mechanisms of production

Ultraviolet radiation is one source of what is called oxidative stress, known to be an important component of UV induced damage.67 Generally, oxidative stress occurs when the production of oxidative molecules exceeds the capacity of antioxidant repair mechanisms to detoxify the system. The primary source of oxidative molecules in humans is from cellular metabolism, with the mitochondrial electron transport chain the largest contributor.68 It is estimated 0.3 - 1% of oxygen produced in each cellular electron transport chain “leaks” from the system producing superoxide and allowing for uncontrolled activity of the resultant reactive oxygen species (ROS).68 In fact, each cell’s nuclear DNA experiences approximately 1000 oxidative events per day.69 Ultraviolet radiation exposure is well established in cell culture and animal models as producing a wide variety of ROS, particularly highly reactive singlet oxygen, through the photooxidation mechanism.70 Increasing levels of UV exposure could potentially overload cellular mechanisms for oxidative detoxification, subsequently increasing the risk of oxidative stress induced disease.69,71 Despite the presence of antioxidant systems, oxidative stress is thought to be responsible for a wide variety of disease processes in virtually all body systems and oxidative molecules are known to increase with aging.72

2.2.2 Effects on cellular components and activity

Cellular components are heavily influenced by oxidative stress. Components primarily damaged are lipids, proteins and DNA. While damage of each cellular component is well established in the literature, the specific mechanisms and pathways involved are not well elucidated. The mechanism of entry into the cell is thought to often be through membrane lipid peroxidation.9 Lipid bilayer membranes surround each cell in the human body. The lipid bilayer membranes contain numerous polyunsaturated fatty acids (PUFAs).9 PUFAs are highly susceptible to lipid peroxidation because they contain multiple double bonds in methylene groups.73 Access of ROS through the cell membrane following lipid peroxidation is thus easily facilitated.

Once penetration of the cell membrane through lipid peroxidation occurs, several cytotoxic effects may follow (Figure 6). Oxidation of DNA may be directly caused by

ROS singlet oxygen and the hydroxyl radical.9 Iron catalyzed oxidation through the

Fenton reaction, a significant source of additional hydroxyl radicals, may be the most damaging to DNA as iron directly binds to the DNA backbone.9 Direct damage to DNA by chromosomal rearrangement, strand breakage, and dimer formation is cytotoxic; yet, indirect damage by loss of integrity of proteins necessary for preservation of DNA is arguably the most important.74

Figure 6. ROS action in cells.9 Reproduced with permission, from Avery SV, Molecular targets of oxidative stress. Biochem J. Mar 1 2011;434(2):201-210. © the Biochemical

Society.

Protein oxidation susceptibility is influenced by several factors and may result in inactivation, misfolding, or cross-linking.9 Metabolic disruption secondary to protein oxidation may occur early in many disease processes.9,75,76 Subsequent events following protein oxidation may include production of additional ROS, activation of matrix metalloproteinases (MMPs), cytokines, and growth factors.9,76 Changes to these proteins can have profound effects on cellular activity, through both inactivation of critical functions and initiation of cytotoxic functions.9 Unless abated, altered cellular activity can lead to apoptosis, mistranslation, and disease causing protein aggregation.9

Abatement of oxidative stress in the human body is composed of multiple overlapping systems.77 The principle counteracting force against oxidative stress is antioxidants. An antioxidant is any process that diminishes or eliminates oxidation. Many antioxidant enzymes are activated to reduce oxidation including glutathione, superoxide dismutase, catalase, and aldehyde dehydrogenase.78 Therefore, assessment of the antioxidant capacity of the body can provide insight to disease processes.

2.2.3 Methods of oxidative stress assessment

Methods to assess oxidative stress and the damage it may cause are numerous and varied. It is possible to directly detect ROS, yet more commonly deleterious products formed by ROS actions are measured.79 The Biomarkers of Oxidative Stress Study

(BOSS) sponsored by the National Institute for Environmental Health Sciences (NIEHS) suggests several possible assays of value for assessment of biologically relevant oxidative stress.80 Biologically relevant oxidative stress markers include those for damage to lipids, proteins, and DNA, as well as assessment for antioxidant activity.80

Reactive Oxygen Species

Reactive oxygen species are capable of producing significant damage to proteins, lipids, and DNA yet are transient in nature and challenging to detect. The most direct in vivo detection method is through electron paramagnetic resonance (EPR) spectroscopy.81

However, a significant disadvantage of EPR is that due to the chemical conditions required for detection of ROS signals it is nearly impossible to measure them directly in vivo.81 It is possible to measure ROS indirectly using colorimetric assays, yet these ROS

20 assays suffer from a lack of repeatability and specificity largely due to expected normal biochemical fluctuations and are not included in BOSS study recommended assays.79,80

Lipids

Lipid peroxidation initiated by the hydroxyl radical is thought to be the initial site of oxidation in many reactive species attacks.9 The lipid bilayer is replete with PUFAs.

The PUFAs contain several double bonds including reactive hydrogens (methyl groups) that are extremely sensitive to lipid peroxidation events.73 Since each of the trillions of human cell membranes contains a lipid bilayer, susceptibility to lipid peroxidation is highly relevant to human health and disease.

To assess lipid peroxidation it is necessary to consider not only direct damage lipid alterations, but also various indirect lipid-derived products and multiple protein alterations that are generated by lipid peroxidation. Direct oxidative damage to PUFAs may result in formation of lipid hydroperoxides, followed by later formation of aldehydes.82 Lipid hydroperoxides can be assayed directly, yet historically lipid peroxidation assays most commonly evaluate aldehyde formation.82 Assays for aldehyde formation include the widely published TBARS assay (with significant limitations),83 direct measurement of malondialdehyde levels (MDA), and 4-hydroxy-2-nonenal (4-

HNE). 4-HNE is an abundantly produced aldehyde formed by lipid peroxidation and is considered to be more sensitive and specific for lipid peroxidation than MDA.84 4-HNE is produced from oxidation of omega-6 PUFAs; this is of particular importance as omega-6 fatty acids comprise a considerable portion of the Western diet.85 Furthermore,

4-HNE has high chemical reactivity and is implicated in development of several

21 pathological conditions.84 On the ocular surface, 4-HNE has been studied in human conjunctival cells86 where it has been shown to be increased in human atopic ocular surface disease and in murine corneal tissues following UVB exposure.11

Indirect assessment of lipid-derived products through protein modification is possible to evaluate both lipid hydroperoxide-derived products and aldehyde-derived products. One lipid hydroperoxide- derived product of note is another oxidation product from omega-6 fatty acids termed Hexanoyl-lysine (HEL).82 Although an amide product with much lower chemical reactivity than the aldehyde 4-HNE, HEL may represent a product formed earlier in the process of omega-6 lipid peroxidation.82 Few publications assess oxidative compounds on the ocular surface; yet, significant differences in HEL levels of human tears was demonstrated following cigarette smoke exposure,87 a well- established source of ROS.

Proteins

Protein oxidation may form several types of modified products. The BOSS study suggests biologically relevant assays for protein oxidation include protein carbonyls, tyrosine products, and methionine sulfoxidation.80 Although protein carbonyls are only one of the suggested protein oxidation assays in the BOSS study, protein carbonyl assays are considered a standard biomarker for oxidative stress to proteins.76 Protein carbonyl groups are side chains added to either aldehyde or ketone groups of existing proteins by the actions of ROS causing a change in protein function.76

Carbonylation of proteins is an irreversible process, thus it is a stable product to quantify and a useful index of oxidative injury.79 Although carbonylated proteins are

22 marked for proteasomal degradation, when there is a high rate of carbonylation or more complex protein aggregates are formed the proteasome system capacity will be exceeded and many damaged proteins will escape degradation.76 Protein carbonyl assays have been used in studies of many systemic diseases.76 Specific to the eye, assessment of protein carbonyls was used to evaluate aqueous humor samples from glaucomatous88 eyes, those with pseudoexfoliation,89,90 and those with macular degeneration.91 Furthermore, increased protein carbonyl formation following UVB radiation exposure has recently been shown in murine corneal tissue.92 Protein carbonyl evaluation has also been performed in cultured corneal endothelial93 and human lens epithelial cells.94

DNA

Several forms of damage to DNA may occur as a result of UV radiation generated

ROS including direct oxidative DNA damage, strand breaks, and defective repair; many assays are available to detect such damage.

Direct genetic damage to DNA and RNA is known to occur from ROS attack. The most widely accepted assay for oxidative DNA damage is 8-Oxo-2’-deoxyguanosine (8-

OHdG).95 8-OHdG is formed from oxidative DNA damage through the interaction of hydroxyl radicals and the DNA base guanosine.96,97 It is the most abundant oxidative product formed from DNA, and is pro-mutagenic.96,97 8-OHdG is considered by many to be the gold standard for detection of oxidative DNA damage.95 Associated with the ocular surface, 8-OHdG has been detected in surgically obtained human conjunctival specimens for those with conjunctivochalasis and pterygium,98,99 cultured human corneal

23 epithelial cells following UVB radiation,100 murine corneal tissue samples following

UVB exposure11 and tears of type 2 diabetic mice.101

DNA strand break genotoxicity can be measured by the comet assay (single cell gel electrophoresis) and has been widely used. The comet assay may provide additional insight to disease processes combined with other assessments of oxidative genetic damage.102 Furthermore, UV-generated ROS may cause altered activation of tumor suppressor protein p53, a transcription factor known as the ‘guardian of the genome’ as it is responsible for cell cycle regulation, protein transcription regulation, and maintenance of genetic code integrity.103 Subsequently, quantification of UV-induced activation of p53 may be a useful indicator of DNA damage, as altered protein transcription and pathologic alteration of multiple DNA regulated functions may follow.103

Antioxidants

Antioxidants are critical for the survival and maintenance of cellular integrity through their absorbance, scavenging, enzymatic degradation, and recycling of other antioxidants (i.e., oxidized glutathione) caused by ROS and their potentially damaging effects to cells, including those present on the ocular surface.78 Several assays for antioxidants are widely used in research including total antioxidant capacity, ascorbic acid, tocopherols, glutathione, glutathione ratio (GSH/GSSG), and uric acid.80 Although antioxidants are critical to the overall picture of the oxidative environment, it is beyond the scope of this dissertation to discuss them fully. Antioxidant activity on the ocular surface is one area that will be explored further in future studies.

2.3 Fluorescence 24

Fluorescence occurs when absorbance (excitation) is followed by emission, a principle described by the Stokes shift often using a Jablonski diagram.104 The Stokes shift describes the principle necessary for fluorescence to occur, where absorption of higher energy photons at shorter wavelengths is followed by emission of photons at lower energy longer wavelengths (Figure 7).104 Fluorescence excitation and emission peaks can be affected by environmental influences such as temperature and exposure intensity and duration.105 Fluorescence is classified as either induced (extrinsic) or autofluorescence

(intrinsic).106

Figure 7. Jablonski diagram of absorbance, non-radiative decay, and fluorescence. By

Jacobkhed, 2012 (Creative Commons license CC0).107

2.3.1 Extrinsic (induced) fluorescence

Extrinsic (induced) fluorescence has many health care assessment and research applications.106 Fluorescence is induced when an extrinsic compound is used to create the effect of fluorescence.106 Fluorescein sodium is commonly used in ocular assessment of the anterior and posterior segment.108-110 As one example, with peak excitation at 490nm and peak emission at about 530nm, fluorescein sodium shows a bright green-yellow glow when used topically on ocular surface pathology.110 Fluorescein is also widely used to induce fluorescence for enhanced study of a variety of other tissues and body systems.106

While induced fluorescence has many biomedical applications, intrinsic fluorescence

(autofluorescence) is of primary concern for the present study.

2.3.2 Intrinsic fluorescence (autofluorescence)

Intrinsic fluorescence (autofluorescence) follows the same principles that describe extrinsic fluorescence including the Stokes shift and environmental effects.104,111 The key difference is that autofluorescence requires no external substance for fluorescence to occur. Several biologically relevant molecules can transiently autofluoresce when excited within the UV spectrum.106 Among them are critical redox molecules such as nicotinamide adenine dinucleotide (NADH), tryptophan, and green fluorescent protein from marine animals.106 Of these molecules, tryptophan and NADH are known UV absorbing molecules present in many proteins.106,112 Although transient in their autofluorescence, these substances can be used to evaluate disease status and may participate in forming multi-molecular substances, such as lipofuscin and collagen.112

Lipofuscin accumulation occurs in many tissues with quantitative increases over time and is associated with aging and disease.76 Autofluorescence of collagen is more apparent following UV-induced protein cross-linking112 and many biological substances form fluorescent complexes following UV-induced oxidative stress.113 Furthermore, a wide variety of aldehydes form during lipid peroxidation, many of which are known to autofluoresce.114,115 Tissue autofluorescence is therefore composed of several types of compounds, making the source of origin difficult to discern.106,112

Autofluorescence excitation and emission spectra vary widely in part due to the chemical compounds generating the signal. Within the visible spectrum a wide range of

27 wavelengths generate several different colors of fluorescence ranging from violet to blue, green, yellow and red; yet, perhaps the most impressive are those where the excitation wavelength is in the invisible spectra (such as UV) and emission is in the visible spectrum (400-700nm).106 Autofluorescent compounds generated in biomedical tissue samples can be differentiated to a degree by spectral curve profiling, although overlapping spectral curves add an element of uncertainty to analyses (Figure 8).106

Clearance Center, Inc.

While the color spectra used in Figure 8 approximate the emission spectra for each substance evaluated, it can be seen that significant overlap exists for excitation wavelengths and that emission spectra are rather broad.106 Of the substances in Figure 8, lipofuscin (lipo-pigments) exhibits the broadest excitation and emission spectra;116-118 collagen is also seen to have a broad autofluorescence spectral range.112 Collagen has been studied extensively and is known to have several autofluorescence emission 29 maximums ranging from 310 to 530 nm.112 Ocular surface pathologies have demonstrated a broad range of tissue autofluorescence, thought to originate from collagen, elastin, and the varied cell and tissue types present.119 Specific to the cornea, Chai and colleagues demonstrated autofluorescence spectral emission profiles of corneas following exposure to UVA ranging from 390 to 682 nm, with the autofluorescence attributed to collagen.120

It is known that fluorescence properties of the same compound may vary with tissue type, scattering, absorption and environmental factors.111 Consequently, although there are challenges with ascertaining the causes of tissue autofluorescence, there is promise of one day using fluorescence emission spectra as a clinical diagnostic tool.

Tissue autofluorescence in vivo holds much potential as a clinical diagnostic tool.

Challenges for detection in tissue are primarily due to scattering and absorption. Methods to reduce the signals from scattering and absorption largely center on subtraction of reflectance from fluorescence signals. Confocal fluorescence microscopy is a three- dimensional technique widely used to detect fluorescence in tissue with fluorophore excitation and achieves removal of undesired signals with a pinhole aperture in front of the detector.121 While confocal fluorescence microscopy is most commonly used to detect non-fluorescent proteins labeled with autofluorescent proteins (extrinsic fluorescence), the technique can also be applied to detect autofluorescence. Specific to the eye, confocal scanning laser ophthalmoscopy is a method based on confocal microscopy principles for imaging the human retina in vivo.116,122 In addition, other variations of confocal microscopy imaging are useful for detecting autofluorescence generated by therapeutic corneal collagen cross-linking.123, 124 While few publications reporting confocal

30 autofluorescence imaging of the conjunctiva can be found,119,125 detection of conjunctival autofluorescence through another method, conjunctival ultraviolet autofluorescence, has recently been reported in several studies.126-133

Conjunctival Ultraviolet Autofluorescence

Conjunctival ultraviolet autofluorescence (UVAF) measurements have recently been reported using methods adapted from principles used in dermatologic assessment.127,130,134 Conjunctival UVAF uses a digital single lens reflex (SLR) camera system to document UV induced fluorescence in visible light photographs. The camera system incorporates appropriate filters for the flash (UV transmission only) and camera

(visible light only).126-130,132,133,135 Of note, the flash must transmit sufficient UV light to cause excitation of conjunctival tissue; thus, it is important to avoid use of UV blocking filters on flash covers. Alternatively, one small study has demonstrated that conjunctival autofluorescence can occur with excitation outside of the UV wavelength range (488 nm) and supports the idea that multiple factors are responsible for conjunctival autofluorescence.125

Conjunctival UVAF is supported as a reliable and valid biomarker for time spent outdoors by evidence from studies performed in Australia and the Norfolk Island Eye

Study.132 Clinical support for the utility of conjunctival UVAF was demonstrated in subjects with pterygia131 and in those with myopia.129 As epidemiological evidence supports increased incidence of pterygia29 and decreased onset of myopia136 being associated with time spent outdoors, strength of conjunctival UVAF as a clinically useful biomarker is noteworthy. In fact, the association for myopia protection was stronger for

31 conjunctival UVAF than for time outdoors.129 Additionally, a dose-response increase in conjunctival UVAF was noted with increased time outdoors for those with pterygium.131

Further support is given to conjunctival UVAF as a biomarker for chronic time outdoors in schoolchildren as UVAF is more apparent in those who spend more time outdoors and with increasing age.127 Of interest, while 81% of schoolchildren ages 12 to 15 years showed evidence of conjunctival UVAF, clinical evidence of conjunctival tissue changes were not frequently found.127 However, Sherwin and colleagues also demonstrated a decrease in conjunctival UVAF with advancing age.130 Consequently, based on this evidence, it was hypothesized that conjunctival UVAF may be useful in detecting areas of preclinical tissue changes thought to be associated with UV radiation exposure, specifically as a subacute ocular UV dosimeter.

The biochemical basis of conjunctival UVAF is not well understood. Suggested biochemical sources of conjunctival UVAF tissue changes include altered collagen structure, areas of cellular mitochondrial activity, presence of known fluorophores

NADH or tryptophan, and lipofuscin.125,127,128,133 In addition, glucose-6-phosphate- dehydrogenase (G6PD) deficiency (a common x-chromosome linked enzymopathy) has been associated with pterygium fibroblasts and increased autofluorescence;137 G6PD is known to be rapidly activated as an initial response to oxidative stress. While the biochemical basis of conjunctival UVAF remains to be delineated, its use as an objective measure of time spent outdoors may prove to be a valuable component in the assessment of UV-associated ocular disease.

Chapter 3: Methods

3.1 Research Design

An initial feasibility study was conducted in which human tear and conjunctival cell samples were collected and analyzed for the purpose of establishing normal oxidative stress values. Additionally, protocol development for imaging of conjunctival UV autofluorescence (UVAF) was performed. Developed protocols and outcome measures from the feasibility study were then used to estimate sample size for the cross-sectional portion of the study.

3.2 Sample Size

Sample size was not initially established as the feasibility study was for protocol development. It was estimated that completion of 20 subjects would be sufficient for the feasibility study. Following data analysis of the initial 20 subjects, sample size for the cross-sectional portion of the study was estimated based on outcome measures of ocular sun exposure, HEL and 8OHdG expression at 25 per group, for a total of 50 subjects, to determine a 25% difference between those who have low amounts of UV exposure when compared with those who have large amounts of UV exposure.

3.3 Subjects

Twenty volunteers were seen for the feasibility portion of the study; subject data collected in the feasibility study was included in the cross-sectional portion of the study.

A total of fifty volunteers completed the study. Each study participant was seen for 2 visits 14 ± 2 days apart. All adults 18 years and older were included, with exclusion criteria only for those 1) currently pregnant or lactating by self report, or 2) with sporadic use of topical ocular medications by self report. Prior to initiation of the study, this research was approved by The Ohio State University Institutional Review Board.

Following explanation of the nature of the study and possible consequences, enrollment in the study commenced for each participant when they signed an informed consent document following the tenets of the Declaration of Helsinki. Following in-person screening for ocular pathology by slit lamp examination, a brief medical and ocular history questionnaire including proportion of time outdoors and use of ocular protective measures,131 an outdoor activity questionnaire,48 and the Ocular Surface Disease Index©

(OSDI©, Allergan)138 were completed. Next, tear and conjunctival samples were collected, as described below, from each eye on both visits. For contact lens wearers, current contact lens brand information when known was collected from each subject to determine UV-absorbing status.

3.4 Measurement / Instrumentation

3.4.1 Tear collection

Glass microcapillary tubes (Drummond Microcaps, Sigma-Aldrich, St. Louis, MO,

USA) were used to collect tears from the inferior tear meniscus. Tear samples were

34 collected with care to minimize stimulation of tearing and inadvertent collection of cells.

The quantities of tears collected ranged from 0 to 12 µL per eye. After collection, tears were removed from the microcapillary tubes and stored in microcentrifuge tubes.

Individual tear samples were stored at -80°C until analysis.

3.4.2 Conjunctival brush cytology

Conjunctival cells were collected by cytology brush (CytoSoft Cytology brush or

Medical Calcium Alginate swab, Fisher Scientific, Pittsburg, PA, USA) from the interpalpebral nasal and temporal regions of each eye. While the participant’s gaze was opposite to the direction of the current collection area, the cytology brush was carefully rotated 3 to 5 times in each region. A new brush was used for each region of each eye, with a total of four brushes used per participant per visit. Each brush was then placed in a separate microcentrifuge tube. Individual brushes containing cell samples were stored at

-80°C until ready for analysis.

3.4.3 Ocular surface vital staining

Following sample collection, 5 µL of non-preserved fluorescein sodium (Leiter’s

Compounding Pharmacy, San Jose, CA, USA) was instilled into the inferior cul-de-sac of each eye and the corneas examined with a biomicroscope and cobalt blue filter for possible evidence of superficial punctuate keratitis and/or corneal abrasion during sample collection. National Eye Institute ocular surface divisions and grading scales were used to evaluate each eye: fluorescein sodium staining was recorded for each of five corneal regions with numeric values ranging from 0 to 3.139 Next, 5 µL of non-preserved

Lissamine green (Leiter’s Compounding Pharmacy) was instilled into the inferior cul-de-

35 sac of each eye and the conjunctivas examined with a biomicroscope for evidence of conjunctival cell loss and/or sample collection induced abrasions. National Eye Institute ocular surface divisions and grading scales were used to evaluate each eye: Lissamine green staining was recorded for each of six conjunctival regions on a quantitative scale from 0 to 3.139

3.4.4 Ocular sun exposure

Ocular sun exposure was estimated over a two week time period by subject responses on two questionnaires developed from the existing literature. First, a questionnaire of ocular protection habits over the most recent two week time period was administered.131 In this questionnaire, subjects were asked to categorize the proportion of time spent wearing the following ocular protective measures when outside: prescription glasses, protective hats, and sunglasses (Appendix A). To minimize recall bias, participants were asked to estimate the proportion of time each protective measure was worn using the following nominal categories: never, seldom, half time, usually, and always.131 These categories were subsequently used for quantitative analyses with the following scale assigned: never = 0, seldom = 0.25, half time = 0.50, usually = 0.75, and always = 1.00. In addition, each protective measure was assigned a number relative to their ocular exposure specific effectiveness:29 no protection = 1.00, hat alone = 0.65, clear spectacles alone = 0.23, dark spectacles alone = 0.11, hat with clear spectacles =

0.11, hat with dark spectacles = 0.07. To further clarify this component of ocular exposure calculations, “no protection” provides no ocular exposure protection, thus is given a proportion of 1.00, or 100% ocular exposure; alternatively, “hat with dark

36 spectacles” provides ocular exposure protection of about 93%, or an ocular exposure proportion of 0.07, or 7%.

A second questionnaire was used to estimate the hours each participant spent in outdoor activities. The outdoor activity questionnaire asked each participant to estimate over the past two weeks the number of hours spent doing each of the following activities: driving, gardening, home, walking/light exercise, recreation, water, and leisure.48 The outdoor activity questionnaire provides more specific examples to further explain the general categories (Appendix A). The mean total number of hours reported less those spent driving were used in subsequent calculations. Hours reported for driving were not included because most car windshields provide UV protection.140 As an example for the final calculation of ocular sun exposure, consider a participant reporting 31 hours outdoors and wearing sunglasses half the time, the ocular sun exposure would be calculated as follows: Thirty-one hours outdoors multiplied by 11% (ocular protection effectiveness for sunglasses) and by 50% (half time wear of sunglasses), then added to the thirty-one hours multiplied by 100% (full ocular sun exposure) and by 50% (for the half time when sunglasses were not worn). An equation would appear as below:

Ocular sun exposure = (31 X 0.11 X 0.50) + (31 X 1.00 X [1 – 0.50]) = 17.21

hours3.4.5 Quantification of Oxidative Damage by HEL ELISA

The first outcome variable studied to evaluate oxidative stress damage was lipid peroxidation in tears and conjunctival cells. This variable was measured using a commercially available HEL ELISA86,87 (Northwest Life Science Specialties, LLC,

Vancouver, WA, USA). Tear samples from the right and left eyes were pooled by subject

for each visit prior to conducting the assay. Tear samples were diluted to 1:20 concentration with 1X phosphate buffered saline (PBS) before ELISA analysis. During the feasibility study including twenty subjects, cell samples from seven subjects were designated to be evaluated by HEL ELISA and seven by 8OHdG ELISA (discussed in further detail below in section 3.4.6). The six remaining subject cell samples were designated to be evaluated by a protein carbonyl ELISA (Northwest Life Science

Specialties); however, a valid standard curve could not be obtained and subsequently the samples could not be quantitatively analyzed for protein carbonyl. Additionally, 8OHdG was below limits of detection in the selected ELISA for the seven designated cell samples in the feasibility portion of the study. Consequently, all 30 additional cell samples obtained for the cross-sectional portion of the study were thus designated to cell HEL

ELISA evaluation. For the HEL ELISA, cells were disrupted by sonication (Microson ultrasonic cell disruptor, Misonix, Inc., Farmingdale, NY, USA) with PBS on ice for 30 seconds, 4 times. Following cell disruption, the four samples collected for each subject were pooled into one microcentrifuge tube for each visit. The HEL ELISA was followed according to manufacturer’s recommendations. Briefly, 50 µL of each sample or standard was added to each well (performed in duplicate), followed by 50 µL of primary antibody and overnight incubation at 4°C. Subsequently, the microplate was washed 3 times with provided wash buffer and 100 µL of secondary antibody was added to each well; the plate was then sealed and incubated at room temperature for 1 hour. Following incubation, the plate was again washed 3 times with provided wash buffer, 100 µL TMB

(3,3’,5,5’ tetramethylbenzidine) substrate was then added, followed by a final incubation

38 for 15 minutes at room temperature in the dark. After adding 100 µL stop solution, absorbance was measured at 450 nm (Tecan –iControl Infinite 200 Pro, Tecan Trading

AG, Switzerland). As recommended by the HEL ELISA manufacturer, a four-parameter logistic standard curve was generated using an online software program (Readerfit.com;

Hitachi Solutions America) to quantify sample HEL levels from the absorbance data obtained.

3.4.6 Quantification of Oxidative Damage by 8OHdG ELISA

The second outcome variable studied to evaluate for oxidative stress damage was

DNA damage in tears and conjunctival cells. This variable was measured using a commercially available 8-OHdG ELISA141 (Northwest Life Science Specialties). Tear samples from the right and left eyes were pooled by subject for each visit prior to conducting the assay. Tear samples were diluted to 1:20 concentration with 1X PBS before ELISA analysis. A subset of ten tear samples were evaluated for possible changes in 8OHdG adduct formation during storage at -80°C over a one year period. Cell samples were processed for DNA extraction and digestion using the QIAamp DNA Micro Kit

(QIAGEN, Valencia, CA, USA), including use of carrier RNA. Subsequent to DNA extraction, DNA digestion was performed by incubating DNA in QIAamp kit provided buffer at 95°C for 5 minutes, followed by a rapid chill on ice for 5 minutes. Samples were then incubated for 2 hours at 37°C with nuclease P1 and 20 mM sodium acetate (pH 5.2).

3U alkaline phosphatase and 100 mM Tris-HCl (pH 7.5) was then added to all samples and incubated for 1 hour at 37°C. Finally, all samples were centrifuged at 6000 x g for 5 minutes. After completion of DNA digestion, supernatant for each subject was pooled

39 into one vial for each visit. The 8OHdG ELISA was performed according to manufacturer’s recommendations and the steps followed were identical to the above described HEL ELISA protocol. Manufacturer recommended four-parameter logistic standard curves were generated using an online software program (Readerfit.com;

Hitachi Solutions America) to quantify 8OHdG levels based on the obtained absorbance data.

3.4.7 Conjunctival UV Autofluorescence imaging and analysis

Conjunctival UVAF was also evaluated with the goal of identifying an objective biomarker of cumulative UV exposure. Conjunctival UVAF is reported as a valid measurement in those with suspected alterations in oxidative stress levels, such as pterygium patients.131 Conjunctival UVAF images were obtained using a camera with appropriate flash and filter system.129 The camera system was comprised of a Nikon

D7000 Digital SLR camera body, 16.2 megapixel (Nikon USA, Melville, NY), fit with a

Nikon AF-S VR Micro-NIKKOR 105mm f/2.8G IF-ED lens (Nikon USA, Melville,

NY). The camera lens was fit with a Hoya 62mm HMC UV-IR filter (Hoya/Kenko

Tokina Co., Ltd., Tokyo, Japan) and a Hoya 62mm polarizer (Hoya/Kenko Tokina Co.,

Ltd., Tokyo, Japan). In addition, a Nikon R1 Speedlight kit (Nikon USA, Melville, NY) was attached to the system. Flash units on the Speedlight kit were modified by removing the standard Fresnel plastic flash covers and replacing them with non-UV absorbing glass slides. Adaptation of the commercial flash kit was further required to accommodate necessary flash filters; hence, custom filter holders were machined to fit the modified flash units (Figure 9).

Figure 9. Conjunctival UV autofluorescence imaging system.

Control and UVAF images were successfully obtained for temporal and nasal regions of each eye for the final 34 subjects enrolled at visits one and two. For both control and UVAF images the following settings were used: lens aperture was f/16 in ‘A’ mode, center-weighted metering, zero exposure compensation, built-in flash turned off, and zero flash compensation. Control images were obtained first in normal room lighting with the camera set at 200 ISO and Baader 2” Polarization Filters for Herschel Wedge

(Baader Planetarium, Germany) in place over the flash units. Next, for UVAF images all room lights were turned off, the camera was set at 1600 ISO and Baader U-Filter Venus

UV ZWL 350nm filters (Baader Planetarium, Germany) were placed over the flash units.

A penlight was used to assist in positioning the subject and obtaining sharp focus on conjunctival tissues, yet was turned off prior to acquiring each image.

Areas of conjunctival UVAF were analyzed using ImageJ (National Institutes of

Health) software. First, images were converted to 8-bit format prior to analysis. Next, the measurement scale was set to millimeters (mm) following calibration using a millimeter ruler image. Following scale setting, the freehand selection tool was used to trace each area of UVAF to allow for the highest flexibility level in area estimation. Each area for the nasal and temporal regions was measured twice and the average obtained for use in statistical analyses. Subsequently, total UVAF was calculated by adding the average areas for each region of both eyes, or the sum total of four measured areas per subject.

3.5 Statistical Analysis

Data was analyzed using IBM SPSS Statistics 19 or 21 (IBM Corporation,

Armonk, New York, USA) software. Means, standard deviations, percentages, and

42 frequencies were used to summarize data collected. The initial feasibility study data

(n=20) served to establish normal values. Normal values were subsequently used to calculate sample size estimates for the cross-sectional study to detect differences between groups for two continuous or dichotomous variables (n=25 per group, total of

50 subjects). In addition, repeatability based on test-retest variability over the two week period between visits for ocular sun exposure, total UVAF and ELISAs was evaluated using Bland-Altman plots. Furthermore, Pearson’s correlation analysis was conducted to evaluate relationships between measures; Spearman’s Rho was used to confirm correlations for very small sample sizes. Finally, storage effects for a subset of ten tear

8OHdG samples were evaluated by calculating percent changes in tear 8OHdG expression over one year of storage.

During the cross-sectional portion of the study, in addition to the feasibility study data analyses described above, regression was performed to evaluate predictors of the outcome variables. Continuous variables found to not follow a normal distribution by Shapiro-Wilk test of normality and Q-Q plots were transformed using either square root or log10 transformations. After transformation, linear regression was initially done for univariate analyses, followed by forward stepwise model building. Linear regression was used to model relationships for continuous outcome variables of tear HEL levels, cell HEL levels, tear 8OHdG levels, ocular sun exposure, and total UVAF. Independent variables considered for each model included age, gender, occupation (indoor or outdoor), season (winter or spring), contact lens wear, tear HEL levels, cell HEL levels, tear 8OHdG levels, ocular sun exposure, and total UVAF. In addition, logistic

43 regression was used to evaluate season of collection, occupation, contact lens wear, total

UVAF and ocular sun exposure as dichotomous outcome variables. Statistical significance was accepted at probability level less than 5%.

Chapter 4: Results

4.1 Subject characteristics

The feasibility study was comprised of 40 eyes (from 20 subjects), 75% female,

100% Caucasian, with a mean age of 38 ± 14 (range 22-64) years from The Ohio State

University, College of Optometry. Subject data and samples were collected during spring

2013. Two subjects were classified as having mild dry eye for both visits based on their responses to the OSDI©. Corrective lenses were worn by 12 subjects: seven wore spectacles, eight wore contact lenses. Of the eight subjects who wore contact lenses, two reported wearing known UV absorbing lenses, three reported wearing known non-UV absorbing lenses, and the remaining three were not aware of the specific brand of contact lenses worn. All subjects completed the study and there were no adverse events.

The cross-sectional study was comprised of 100 eyes (from 50 subjects total; average values from both eyes used in calculations), 58% female, with a mean age of 38

± 13 (range 21-64) years from The Ohio State University. Subject data and samples were collected during spring 2013, winter 2014, and spring 2014. Nine subjects reported having outdoor occupations, with the remaining 41 reporting indoor occupations. Four subjects were classified as having mild dry eye at visit one based on their responses to the

OSDI©. Corrective lenses were worn by 30 subjects: 20 wore spectacles, 10 wore contact lenses. Of the 10 subjects who wore contact lenses, only three reported wearing known

UV absorbing lenses, three reported wearing known non-UV absorbing lenses, and the remaining four were not aware of the specific brand of contact lenses worn. All subjects completed the study and there were no adverse events.

4.2 Descriptive statistics

4.2.1 Feasibility study

Descriptive statistics were generated for measures of ocular sun exposure, tear and cell HEL expression, and tear 8OHdG expression (Table 2). Although cell samples were collected with two different types of devices, comparable amounts of DNA were collected from each (data not shown), thus both devices were treated equally in data analysis. As discussed in section 3.4.5, while DNA was detected in the conjunctival cell samples, 8OHdG levels were below limits of detection, and the remaining samples were not evaluated for protein carbonyl expression as a valid standard curve was not attained.

As such, the only reportable data from cell samples was obtained from samples designated to the HEL ELISA.

Table 2. Descriptive statistics and repeatability for feasibility study reported measures.

Measure n Mean (±SD) Bias 95% Limits of Agreement Ocular Sun Exposure 20 4.38 (±3.02) hours -0.04 -6.66 +6.59 Tear HEL expression 20 24794.86 (±9256.71) nmol/L +533.19 -25799.39 +26865.76 Cell HEL expression 7 19.07 (±8.20) nmol/L -3.70 -21.32 +13.92 Tear 8OHdG expression 20 66.72 (±26.95) ng/mL -3.40 -82.78 +75.98

Bland-Altman scatter plots were generated for ocular sun exposure (Figure 10), tear HEL levels (Figure 11), cell HEL levels (Figure 12), and tear 8OHdG levels (Figure 13) to obtain estimates of repeatability based on test-retest variability for each measure, with mean differences (bias) and 95% limits of agreement calculated (Table 2). A range of agreement was defined as mean bias ± 1.96 standard deviations. For each Bland Altman plot, the solid line shows bias (mean differences); dotted lines show 95% limits of agreement. Thus it was demonstrated that ocular sun exposure, tear HEL expression, cell

HEL expression, and 8OHdG expression were detectable and repeatable over a two week time period with methods used in the present study.

Figure 10. Bland-Altman plot illustrating test-retest variability of ocular sun exposure in

20 subjects.

Figure 11. Bland-Altman plot illustrating test-retest variability of tear HEL levels in 20 subjects.

Figure 12. Bland-Altman plot illustrating test-retest variability of cell HEL levels in 7 subjects.

Figure 13. Bland-Altman plot illustrating test-retest variability of tear 8OHdG levels in

20 subjects.

4.2.1.1 Correlations of feasibility study outcome measures

Pearson’s correlation analysis revealed several statistically significant relationships between outcome variables (Table 3). Positive linear correlations at α=0.05 level (two-tailed) of significance were found for ocular sun exposure with tear HEL levels (r=0.454, p=0.044; Figure 14); for tear 8OHdG levels with both cell HEL levels

(r=0.903, p=0.005; Figure 15) and tear HEL levels (r=0.619, p=0.004; Figure 16); and for tear HEL levels with cell HEL levels (r=0.817, p=0.025). Recognizing the small sample size specifically for cell HEL levels (n=7), nonparametric correlation analysis using

Spearman’s Rho was also performed (Table 3). Statistically significant correlations remained for tear 8OHdG levels with cell HEL levels (r=0.893, p=0.007); while the correlation coefficient was high for tear and cell HEL, it did not remain significant

(r=0.714, p=0.071).

Table 3. Statistically significant correlations identified using Pearson's correlation analysis for feasibility study, confirmed for small sample sizes using Spearman's Rho.

Pearson's Correlation r r² p Ocular sun exposure and tear HEL expression 0.454 0.206 0.044 Tear 8OHdG and cell HEL expression 0.903 0.815 0.005 Tear 8OHdG and tear HEL expression 0.619 0.383 0.004 Tear HEL and cell HEL expression 0.817 0.667 0.025 Spearman's Rho r r² p Tear 8OHdG and cell HEL expression 0.893 0.797 0.007 Tear HEL and cell HEL expression 0.714 0.510 0.071

As test-retest repeatability of ocular sun exposure, tear and cell HEL expression, and tear 8OHdG expression was demonstrated with the initial twenty subject feasibility study, subject recruitment was continued using the same protocols, with the exception of focusing on recruitment of those who spend a significant amount of time outdoors.

4.2.1.2 Storage effects for tear 8OHdG samples

A subset of tear samples collected during the feasibility study were evaluated for possible formation of additional 8OHdG adduct during storage at -80°C over one year.

Percent changes showed an increase in tear 8OHdG expression for nine of ten samples, and an overall increase in samples tested of nearly 16 (±14.63) percent (Table 4). The storage effect identified may have had some influence on levels of 8OHdG expression detected in the present study as tear samples were stored up to six months following collection.

Table 4. Percent change storage effects for tear 8OHdG expression in ten samples stored at -80 degrees Celsius over one year.

Storage effects for tear 8OHdG (ng/mL) 2013 2014 ID Mean 1 Mean 2 Difference % Change 11 61.77 75.64 13.87 22.45 12 66.30 77.23 10.93 16.49 13 62.86 91.84 28.98 46.11 14 58.73 79.57 20.84 35.49 15 53.13 67.61 14.48 27.25 16 70.14 77.51 7.38 10.52 17 108.49 122.80 14.30 13.18 18 124.25 139.38 15.13 12.18 19 46.45 51.76 5.31 11.44 20 125.06 117.60 -7.47 -5.97 Total group 11-20 77.72 90.10 12.38 15.93 56

4.2.2 Cross-sectional study

Descriptive statistics were generated for measures of ocular sun exposure, conjunctival UVAF, tear and cell HEL ELISA, and tear 8OHdG ELISA (Table 5). As described in section 3.4.5, although DNA was detected in the conjunctival cell samples,

8OHdG levels were below limits of detection, and the remaining cell samples designated to protein carbonyl ELISA were not evaluated due to inability to obtain a valid standard curve. Thus, the only reportable data from cell samples was obtained from the HEL

ELISA. Conjunctival UVAF was present in 85% of subjects tested (29 of 34 subjects).

Table 5. Descriptive statistics for continuous outcome variable raw data.

Continuous variable n Min Max Mean (±SD) Median (Raw data) Ocular sun exposure (hours) 50 0.55 60.24 8.86 (±11.97) 5.57 Tear HEL levels (nmol/L) 50 3705.23 51424.78 17368.02 (±9878.42) 14791.33 Cell HEL levels (nmol/L) 37 10.38 284.37 53.94 (±55.53) 34.27 Tear 8OHdG levels (ng/mL) 50 29.04 125.06 66.13 (±19.99) 65.37 Total UVAF (mm²) 34 0 30.64 9.15 (±9.47) 6.31

Violations of assumptions for normal data distribution were evident for each outcome variable by Shapiro-Wilk test for normality (Table 6) and Q-Q plots. Each variable was positively skewed to the right, toward lower values. The outcome variable of ocular sun exposure demonstrated the highest degrees of both skewness (2.90) and kurtosis (9.01); cell HEL levels showed similar values (skewness=2.50, kurtosis=7.93) for non-normally distributed data. Data was transformed to better approximate normal

57 distributions by either log10 or square root transformation (Table 7). Following transformation, each variable passed the Shapiro-Wilk test for normality with the exception of total UVAF. Square root transformation of total UVAF most closely approximated a normal distribution and so was used in further analyses. As the number of observations is relatively small for total UVAF (n=34) and thus may not have sufficient power to pass the Shapiro-Wilk test, comparison of the mean and median was done to confirm approximation of a normal distribution (Table 7).

Table 6. Shapiro-Wilk test for normal distribution of continuous outcome variable raw data.

Continuous variable n Shapiro-Wilk Skewness Kurtosis (Raw data) Ocular sun exposure (hours) 50 0.63 (p<0.0001) 2.90 9.01 Tear HEL levels (nmol/L) 50 0.93 (p=0.005) 1.07 1.54 Cell HEL levels (nmol/L) 37 0.73 (p<0.0001) 2.50 7.93 Tear 8OHdG levels (ng/mL) 50 0.92 (p=0.02) 1.12 1.85 Total UVAF (mm²) 34 0.87 (p=0.001) 0.79 -0.57

Table 7. Descriptive statistics for continuous outcome variable transformed data.

Continuous variable ₁₀ n Min Max Mean (±SD) Median (Transformed data) Log transformed Ocular sun exposure (hours) 50 -0.26 1.78 0.69 (±0.47) 0.75 Tear HEL levels (nmol/L) 50 3.57 4.71 4.17 (±0.27) 4.17 Tear 8OHdG levels (ng/mL) 50 1.46 2.10 1.80 (±0.13) 1.82 Cell HEL levels (nmol/L) 37 1.02 2.45 1.56 (±0.39) 1.53 Square Root transformed Total UVAF (mm²) 34 0 5.54 2.43 (±1.82) 2.51

Bland-Altman142 scatter plots were produced using raw data for ocular sun exposure (Figure 17), total UVAF (Figure 18), tear HEL levels (Figure 19), cell HEL levels (Figure 19), and tear 8OHdG levels (Figure 21) to obtain estimates of repeatability based on test-retest variability for each measure. Next, Bland-Altman scatter plots were produced using transformed data for ocular sun exposure (Figure 22), total UVAF

(Figure 23), tear HEL levels (Figure 24), cell HEL levels (Figure 25), and tear 8OHdG levels (Figure 26) with some evidence for mitigation of non-normal distribution; however, outliers were still evident. In addition, Bland-Altman scatter plots were produced for visit 1 ELISA for well 1 and 2 for each sample to estimate repeatability of the tear HEL ELISA (Figure 27), cell HEL ELISA (Figure 28), and tear 8OHdG ELISA

(Figure 29) for sample types specific to the present study. Finally, to further evaluate the seasonal influence on ocular sun exposure a Bland-Altman plot of repeatability for ocular sun exposure during winter (Figure 30) and normal Q-Q plots for ocular sun exposure during spring (Figure 31) and winter (Figure 32) season were produced. Mean differences

(bias) and 95% limits of agreement calculated for each variable using each type of data

(Table 8). A range of agreement was defined as mean bias ± 1.96 standard deviations.

Table 8. Bland-Altman estimates of repeatability for raw and transformed data from cross-sectional study outcome measures, and for visit 1 ELISA transformed values.

Relative to the 95% limits of agreement for each variable, mean bias values are not notably different from zero with the exception of raw data for cell HEL levels (9.7%).

Measure n Bias 95% Limits of Agreement Raw data Ocular sun exposure (hours) 50 +0.57 -14.52 +15.66 Tear HEL levels (nmol/L) 50 +529.27 -17873.70 +18932.24 Cell HEL levels (nmol/L) 37 -19.24 -218.20 +179.72 Tear 8OHdG levels (ng/mL) 50 +3.53 -57.64 +64.70 Total UVAF (mm²) 34 +0.19 -1.75 +2.12 Transformed data Ocular sun exposure (log10 hours) 50 -0.15 +0.98 -1.28 Tear HEL levels (log10 nmol/L) 50 +0.02 +0.47 -0.43 Cell HEL levels (log10 nmol/L) 37 -0.16 +0.98 -1.30 Tear 8OHdG levels (log10 ng/mL) 50 +0.05 +0.72 -0.62 Total UVAF (sqrt mm²) 34 +0.01 +0.48 -0.46 Visit 1 ELISA transformed data Tear HEL ELISA (log10 nmol/L) 50 +0.02 +0.44 -0.39 Cell HEL ELISA (log10 nmol/L) 37 -0.04 +0.25 -0.33 Tear 8OHdG ELISA (log10 ng/mL) 50 -0.04 +0.06 -0.15

Figure 17. Bland-Altman plot illustrating test-retest variability of raw data ocular sun exposure in 50 subjects.

Figure 18. Bland-Altman plot illustrating variability of raw data conjunctival UVAF in

34 subjects.

Figure 19. Bland-Altman plot illustrating test-retest variability of raw data tear HEL expression in 50 subjects.

Figure 20. Bland-Altman plot illustrating test-retest variability of raw data cell HEL expression in 37 subjects.

Figure 21. Bland-Altman plot illustrating test-retest variability of raw data tear 8OHdG expression in 50 subjects.

Figure 22. Bland-Altman plot illustrating test-retest variability of log10 transformed data for ocular sun exposure in 50 subjects.

Figure 23. Bland-Altman Altman plot illustrating variability of square root transformed data for total conjunctival UVAF in 34 subjects.

Figure 24. Bland-Altman plot illustrating test-retest variability of log10 transformed data for tear HEL expression in 50 subjects.

Figure 25. Bland-Altman plot illustrating test-retest variability of log10 transformed data for cell HEL expression in 37 subjects.

Figure 26. Bland-Altman plot illustrating test-retest variability of log10 transformed data for tear 8OHdG expression in 50 subjects.

Figure 27. Bland-Altman plot illustrating test-retest variability of log10 transformed data for visit 1, well 1 and well 2, for tear HEL ELISA in 50 subjects.

Figure 28. Bland-Altman plot illustrating test-retest variability of log10 transformed data for visit 1, well 1 and well 2, for cell HEL ELISA in 37 subjects.

Figure 29. Bland-Altman plot illustrating test-retest variability of log10 transformed data for visit 1, well 1 and well 2, for tear 8OHdG ELISA in 50 subjects.

Figure 30. Bland-Altman plot illustrating repeatability for ocular sun exposure during winter season of collection (n=15).

Figure 31. Normal Q-Q plot illustrating normal distribution of ocular sun exposure for spring season of collection.

Figure 32. Normal Q-Q plot illustrating non-normal distribution for ocular sun exposure during winter season of collection.

4.4 Regression analyses

Univariate regression was carried out prior to modeling. Logistic regression showed a number of significant relationships for dichotomous outcome variables occupation and season of collection; no significant predictors were identified for contact lens wear (Table 9). In linear regression, data was evaluated in both raw form and after transformation to normal distributions for comparison. Univariate linear regression

76 revealed several significant relationships for the raw data (Tables 10 and 11) and transformed data (Tables 12 and 13) continuous outcome variables ocular sun exposure, total UVAF, tear HEL expression, cell HEL expression, and tear 8OHdG expression.

Transformed data is described before raw data as the transformed data is presumed to be the more robust of the two data sets. In addition, univariate linear regression of subject reported activities by visit (Appendix A) was used to identify the possible influences on continuous outcome variables for ocular sun exposure, conjunctival UVAF, tear HEL expression, cell HEL expression, and tear 8OHdG expression (Tables 14 and 15).

Table 9. Univariate logistic regression for raw and transformed outcome variables: occupation, contact lens wear and season of

collection.

Occ (IN=1, OUT=2) C L wear (Y=1, N=0) Season (W=1, S=2) Regr Coeff p OR 95% CI Regr Coeff p OR 95% CI Regr Coeff p OR 95% CI

Age 0.02 0.425 1.02 0.97-1.09 -0.04 0.180 0.96 0.90-1.02 <0.0001 0.996 1.00 0.95-1.05 Gender (F=1, M=2) 1.91 0.028 6.75 1.24-36.91 -1.29 0.131 0.28 0.05-1.47 -1.93 0.005 0.15 0.04-0.57 Occupation (IN=1, OUT=2) -20.07 0.999 * -22.97 0.999 * C L wear (Y=1, N=0) -19.97 0.999 * 1.58 0.153 4.85 0.56-42.26 Season collected (W=1, S=2) -21.61 0.997 * 1.58 0.153 4.85 0.56-42.26

Transformed data

8 Ocular sun exposure 1.53 0.074 4.62 0.86-24.79 -0.53 0.501 0.59 0.12-2.78 -0.28 0.671 0.75 0.20-2.79 Total UVAF 0.52 0.043 1.69 1.02-2.81 -0.44 0.279 0.64 0.29-1.43 -0.46 0.036 0.63 0.41-0.97 Tear HEL expression -4.35 0.009 0.013 <0.0001-0.34 3.38 0.052 29.23 0.98-876.20 3.81 0.008 45.02 2.72-744.40 Tear 8OHdG expression -5.62 0.078 0.004 <0.0001-1.86 1.74 0.537 5.69 0.02-1416.78 3.91 0.133 49.92 0.30-8195.10 Cell HEL expression 1.72 0.110 5.60 0.68-46.40 -4.60 0.087 0.01 0.000-1.93 -1.90 0.054 0.149 0.02-1.03

Raw data Ocular sun exposure 0.07 0.024 1.08 1.01-1.14 -0.07 0.302 0.93 0.82-1.06 -0.04 0.147 0.96 0.92-1.01 Total UVAF 0.10 0.031 1.10 1.01-1.20 -0.07 0.442 0.94 0.79-1.11 -0.10 0.028 0.91 0.83-0.99 Tear HEL expression** -1.89 0.015 0.15 0.03-0.70 0.61 0.086 1.84 0.92-3.68 1.46 0.008 4.30 1.46-12.70 Tear 8OHdG expression -0.05 0.079 0.96 0.91-1.01 0.01 0.552 1.01 0.98-1.05 0.03 0.100 1.03 0.99-1.07 Cell HEL expression 0.006 0.314 1.01 0.99-1.02 -0.10 0.166 0.91 0.79-1.04 -0.02 0.067 0.98 0.97-1.00 p: two-tailed test of significance *Model failure due to insufficient data **Raw data for tear HEL expression is per 10,000 unit change

Table 10. Univariate linear regression results for raw data continuous outcome variables ocular sun exposure and total

conjunctival UV autofluorescence.

Ocular Sun Exposure Total UVAF Regr Coeff p 95% CI Regr Coeff p 95% CI

Ocular sun exposure 0.03 0.821 -0.22 to 0.28 Total UVAF 0.06 0.821 -0.47 to 0.58 Tear HEL expression* -1.72 0.327 -5.20 to 1.77 -0.91 0.580 -4.22 to 2.40 Tear 8OHdG expression 0.06 0.511 -0.12 to 0.23 -0.16 0.041 -0.31 to -0.007

0 7 Cell HEL expression 0.05 0.262 -0.04 to 0.13 0.01 0.637 -0.05 to 0.08

9 Age -0.13 0.356 -0.40 to 0.15 0.03 0.855 -0.26 to 0.31 Gender (F=1, M=2) -1.42 0.683 -8.37 to 5.54 5.36 0.099 -1.07 to 11.80 Occupation (IN=1, OUT=2) 12.85 0.003 4.71 to 20.98 8.46 0.019 1.49 to 15.44 C L wear (Y=1, N=0) -4.72 0.269 -13.21 to 3.77 -4.50 0.440 -16.23 to 7.23 Season collected (W=1, S=2) -5.78 0.119 -13.10 to 1.53 -7.56 0.018 -13.75 to -1.36 p: two-tailed test of significance *Raw data for tear HEL expression is per 10,000 unit change

Table 11. Univariate linear regression results for raw data continuous outcome variables tear HEL expression, cell HEL expression, and tear 8OHdG expression.

Tear HEL expression Cell HEL expression Tear 8OHdG expression Regr Coeff p 95% CI Regr Coeff p 95% CI Regr Coeff p 95% CI

Ocular sun exposure -116.84 0.327 -353.96 to 120.28 0.79 0.262 -0.62 to 2.19 0.16 0.511 -0.32 to 0.64 Total UVAF -106.31 0.580 -493.92 to 281.30 0.53 0.637 -1.73 to 2.79 -0.77 0.041 -1.52 to -0.03 Tear HEL expression* -14.46 0.124 -33.07 to 4.15 0.47 0.001 0.21 to 0.73 Tear 8OHdG expression 228.17 0.001 101.00 to 355.33 0.23 0.630 -0.733 to 1.20 Cell HEL expression -45.91 0.124 -104.99 to 13.17 0.03 0.630 -0.09 to 0.15 Age -178.05 0.115 -400.88 to 44.77 -1.15 0.141 -2.70 to 0.40 -0.43 0.059 -0.88 to 0.02

80 Gender (F=1, M=2) -7075.98 0.011 -12446.88 to -1705.08 7.08 0.704 -30.44 to 44.61 -16.77 0.003 -27.34 to -6.20

Occupation (IN=1, OUT=2) -9374.02 0.009 -16241.88 to -2506.16 22.23 0.303 -20.90 to 65.37 -13.04 0.076 -27.51 to 1.43 C L wear (Y=1, N=0) 6299.90 0.071 -555.50 to 13155.31 -42.64 0.149 -101.37 to 16.09 4.19 0.559 -10.13 to 18.50 Season collected (W=1, S=2) 8658.94 0.003 2998.64 to 14319.23 -39.14 0.033 -74.98 to -3.29 10.37 0.093 -1.80 to 22.54 p: two-tailed test of significance *Raw data for tear HEL expression is per 10,000 unit change

Table 12. Univariate linear regression results for transformed continuous variables ocular sun exposure and total conjunctival UV autofluorescence.

Ocular Sun Exposure Total UVAF Regr Coeff p 95% CI Regr Coeff p 95% CI

Ocular sun exposure 0.03 0.969 -1.37 to 1.43 Total UVAF 0.002 0.969 -0.09 to 0.10 Tear HEL expression -0.14 0.591 -0.65 to 0.37 -0.71 0.545 -3.08 to 1.66 Tear 8OHdG expression 0.51 0.327 -0.53 to 1.55 -6.12 0.015 -10.99 to -1.25 Cell HEL expression 0.11 0.600 -0.30 to 0.51 -0.19 0.828 -1.98 to 1.60

Age -0.01 0.091 -0.02 to 0.001 0.007 0.801 -0.05 to 0.06 Gender (F=1, M=2) -0.09 0.500 -0.36 to 0.17 0.95 0.133 -0.30 to 2.20 Occupation (IN=1, OUT=2) 0.32 0.064 -0.02 to 0.65 1.51 0.031 0.15 to 2.87 C L wear (Y=1, N=0) -0.11 0.509 -0.44 to 0.22 -1.25 0.265 -3.48 to 0.99 Season collected (W=1, S=2) -0.06 0.678 -0.35 to 0.23 -1.36 0.028 -2.57 to -0.16 p: two-tailed test of significance

Table 13. Univariate linear regression results for transformed continuous outcome variables tear HEL expression, cell HEL

expression, and tear 8OHdG expression.

Tear HEL expression Cell HEL expression Tear 8OHdG expression Regr Coeff p 95% CI Regr Coeff p 95% CI Regr Coeff p 95% CI

Ocular sun exposure -0.04 0.591 -0.21 to 0.12 0.08 0.600 -0.21 to 0.36 0.04 0.327 -0.04 to 0.12 Total UVAF -0.02 0.545 -0.07 to 0.04 -0.008 0.828 -0.09 to 0.07 -0.03 0.015 -0.05 to -0.006 Tear HEL expression -0.52 0.028 -0.99 to -0.06 0.16 0.017 0.03 to 0.30 Tear 8OHdG expression 0.69 0.017 0.13 to 1.25 0.53 0.284 -0.46 to 1.52 Cell HEL expression -0.25 0.028 -0.47 to -0.03 0.06 0.284 -0.05 to 0.18

2 Age -0.004 0.225 -0.01 to 0.002 -0.004 0.489 -0.02 to 0.007 -0.003 0.079 -0.005 to 0

Gender (F=1, M=2) -0.17 0.024 -0.31 to -0.02 0.11 0.384 -0.15 to 0.37 -0.11 0.002 -0.18 to -0.04 Occupation (IN=1, OUT=2) -0.28 0.003 -0.46 to -0.10 0.24 0.102 -0.50 to 0.54 -0.09 0.069 -0.18 to 0.007 C L wear (Y=1, N=0) 0.19 0.042 0.007 to 0.37 -0.41 0.042 -0.81 to -0.02 0.03 0.545 -0.06 to 0.12 Season collected (W=1, S=2) 0.23 0.003 0.08 to 0.39 -0.26 0.045 -0.51 to -0.006 0.06 0.128 -0.02 to 0.14 p: two-tailed test of significance

Table 14. Univariate linear regression for reported activities by visit for transformed continuous outcome variable total UVAF.

Activities Total UVAF Regr Coeff p 95% CI Visit 1 Wear of prescription eyeglasses when outside -0.03 0.876 -0.41 to 0.34 Wear of protective hat when outside 0.37 0.202 -0.21 to 0.94 Wear of sunglasses when outside 0.23 0.348 -0.26 to 0.70 Portion of day spent outdoors while working 0.72 0.034 0.06 to 1.38

8 Portion of day spent outdoors while not at work -0.07 0.876 -0.95 to 0.82

Visit 2 Wear of prescription eyeglasses when outside -0.12 0.522 -0.48 to 0.25 Wear of protective hat when outside 0.22 0.445 -0.36 to 0.81 Wear of sunglasses when outside 0.26 0.291 -0.23 to 0.75 Portion of day spent outdoors while working 0.87 0.020 0.16 to 1.59 Portion of day spent outdoors while not at work -0.57 0.310 -1.68 to 0.55 p: two-tailed test of significance

Table 15. Univariate linear regression for reported activities by visit for transformed continuous variables tear HEL expression, cell HEL expression, and tear 8OHdG expression.

Activities Tear HEL expression Cell HEL expression Tear 8OHdG expression Regr Coeff p 95% CI Regr Coeff p 95% CI Regr Coeff p 95% CI Visit 1 Wear of prescription eyeglasses when outside -0.03 0.256 -0.08 to 0.02 0.004 0.923 -0.08 to 0.09 -0.002 0.857 -0.03 to 0.02 Wear of protective hat when outside -0.06 0.135 -0.15 to 0.02 0.04 0.597 -0.11 to 0.18 -0.01 0.601 -0.05 to 0.03 Wear of sunglasses when outside 0.07 0.033 0.006 to 0.14 -0.004 0.948 -0.12 to 0.12 0.01 0.506 -0.02 to 0.04 Portion of day spent outdoors while working -0.10 0.061 -0.20 to 0.005 0.06 0.505 -0.12 to 0.23 -0.02 0.482 -0.07 to 0.03 Portion of day spent outdoors while not at work 0.03 0.645 -0.10 to 0.15 0.06 0.546 -0.15 to 0.28 0.02 0.497 -0.04 to 0.08

Visit 2

8 Wear of prescription eyeglasses when outside -0.02 0.394 -0.06 to 0.03 -0.05 0.202 -0.13 to 0.03 -0.01 0.567 -0.03 to 0.02

Wear of protective hat when outside -0.12 0.002 -0.20 to -0.05 0.06 0.394 -0.08 to 0.19 -0.02 0.499 -0.06 to 0.03 Wear of sunglasses when outside 0.01 0.779 -0.05 to 0.70 -0.06 0.309 -0.16 to 0.05 -0.01 0.678 -0.04 to 0.03 Portion of day spent outdoors while working -0.12 0.018 -0.22 to -0.02 0.14 0.092 -0.02 to 0.31 -0.01 0.814 -0.06 to 0.05 Portion of day spent outdoors while not at work -0.11 0.101 -0.25 to 0.02 -0.14 0.214 -0.36 to 0.08 -0.02 0.683 -0.09 to 0.06 p: two-tailed test of significance

Logistic regression

Logistic regression was conducted for dichotomous outcome variables of occupation, contact lens wear, and season of collection (Table 9). Total UVAF and ocular sun exposure were also tested as dichotomous variables relative to the median; neither variable was found to be a significant predictor of the other as an outcome variable

(p=0.293).

Based on sample size calculations, the present study is powered to detect only about a 40% effect size for outdoor occupation (n=9) and contact lens wear (n=10); therefore, each was tested as a dichotomous outcome variable with logistic regression to determine possible covariates. Outdoor occupation showed two significant predictors using the transformed data: significant predictors of outdoor occupation were identified for higher total UVAF (p=0.043) and lower tear HEL expression (p=0.009) (Table 9).

When compared with raw data, four significant predictors of outdoor occupation were identified: male gender (p=0.028), higher ocular sun exposure (p=0.024), higher total

UVAF (p=0.031) and lower tear HEL levels (p=0.015) (Table 9). Contact lens wear was not found to have significant predictive relationships with any of the logistic regression predictors tested using either the raw or transformed data; however, using transformed data higher tear HEL expression nearly reached significance as a predictor for contact lens wear (p=0.052) (Table 9).

Using a similar approach, season of collection was evaluated as a dichotomous variable to identify possible covariates. Transformed data showed higher levels of total

UVAF (p=0.036) predicted winter season of collection; higher tear HEL expression

(p=0.008) predicted spring season of collection. In addition, using transformed data higher cell HEL expression was a nearly significant predictor of winter season of collection (p=0.054) (Table 9). Using the raw data, three variables were identified as significant: higher levels of total UVAF (p=0.028) and male gender (p=0.005) were again found to be predictors of winter season of collection; higher tear HEL expression was again found to predict spring season of collection (p=0.008) (Table 9).

Linear regression models

Multivariate linear regression forward stepwise models were initially built using the best fit transformed data. Independent variables considered for each model included age, gender, occupation (indoor or outdoor), season (winter or spring), contact lens wear, tear HEL levels, cell HEL levels, tear 8OHdG levels, ocular sun exposure, and total

UVAF. Sensitivity testing was subsequently done for several specific concerns. First, given apparent cell HEL expression bias toward increased variability with increased expression levels in the Bland-Altman plot (Figure 19), each model was also tested using log10 transformed data only from baseline (visit one) cell HEL expression. Effects of this variable change on each model are discussed below. Second, sensitivity was tested for the single outlier for ocular sun exposure (Figure 17) by elimination; no significantly different results were found for any of the models. Third, the four subjects determined to have mild dry eye based on their responses to the OSDI© at visit one were eliminated from the data set; again, no significantly different results were found for any of the models. Finally, multiple linear regression models were built and tested using the best fit raw data; sensitivity to elimination of cell HEL expression was also evaluated for the raw

86 data. For ease of interpretation, regression coefficients are discussed here only for the raw data linear regression models.

4.4.1 Ocular sun exposure

Univariate linear regression for ocular sun exposure as a continuous outcome variable was conducted to identify possible covariates using raw (Table 10) and transformed (Table 12) data. Possible predictors tested were total UVAF, tear HEL expression, tear 8OHdG expression, cell HEL expression, age, gender, occupation, contact lens wear, and season of collection. No significant covariate relationships were identified for any of these possible predictors of ocular sun exposure using the transformed data. For the raw data, outdoor occupation was found to be a predictor of higher ocular sun exposure (p=0.003), with an increase of nearly 13 hours of ocular sun exposure expected for those with outdoor occupations when compared with those having indoor occupations.

The best fit multiple linear regression model using transformed data for ocular sun exposure considered as a continuous outcome variable was found to include occupation and age; gender was also included in this model (Table 16). This model provided a weak description of ocular sun exposure (adjusted R2=0.12; F=3.27; p=0.030). Individual coefficient analysis showed a significant contribution for occupation (p=0.016). Age

(p=0.060) and gender (p=0.192) were not found to be significant in this model. There were no changes to the best fit model when using cell HEL expression data only from visit one as a possible predictor of ocular sun exposure.

Using the raw data, the best fit multiple linear regression model for ocular sun exposure included occupation and tear 8OHdG expression (Table 16). Age and gender were added to this model. The full model was again only a weak description of ocular sun exposure (adjusted R2=0.19; F=3.89; p=0.009). However, after adjusting for age and gender, occupation remained highly significant (p<0.0001), with the expectation that on average, those with outdoor occupations will have approximately 16 additional hours of ocular sun exposure when compared with those who have indoor occupations. Age

(p=0.297), gender (p=0.252) and tear 8OHdG expression (p=0.427) were not significant predictors of ocular sun exposure in this model. There were no changes to this model when cell HEL expression was removed as a possible predictor.

Table 16. Linear regression models summary for ocular sun exposure.

Ocular sun exposure adj R² F test p value coeff effect Transformed data Full best fit model 0.12 3.27 0.030 Occupation (IN=1, OUT=2) 0.016 0.43 + OUT Age 0.060 -0.01

Gender (F=1, M=2) 0.192 -0.18

Cell HEL visit one only No change Raw data Full best fit model 0.19 3.89 0.009 Occupation (IN=1, OUT=2) <0.0001 16.04 + OUT Tear 8OHdG expression 0.427 0.07

Age 0.297 -0.14 Gender (F=1, M=2) 0.252 -4.09

Cell HEL data removed No change

Ocular sun exposure for activities reported by visit for wear of prescription eyeglasses, wear of sunglasses, and for wear of a protective hat when outside were used to calculate estimated hours of ocular sun exposure; thus, these are not considered possible predictors. However, as expected, increasing the proportion of the day spent outdoors while working was associated with higher levels of ocular sun exposure (visit 1: p=0.014; visit 2: p=0.034), while increasing the proportion of the day spent outdoors while not at work was not found to be significant (visit 1: p=0.552; visit 2: p=0.380).

4.4.2 Tear HEL expression

Univariate predictors identified to be statistically significantly associated with tear

HEL expression were tear 8OHdG expression, cell HEL expression, gender, occupation, contact lens wear, and season of collection. For the transformed data (Table 13), tear

8OHdG expression (p=0.017) was identified as a predictor of higher levels of tear HEL expression. While tear 8OHdG expression was found to have a positive relationship with tear HEL expression, cell HEL expression (p=0.028) was inversely related and thus predicted lower tear HEL expression. In addition, other predictors for tear HEL expression demonstrate female gender (p=0.024), indoor occupation (p=0.003), spring season of collection (p=0.003), and contact lens wear (p=0.042) predict higher expression. Using the raw data (Table 11), significant univariate predictors of tear HEL expression were slightly different, with higher tear 8OHdG expression (p=0.001) remaining associated with higher tear HEL expression, while cell HEL expression was not significant (p=0.124). Furthermore, female gender (p=0.011), indoor occupation

(0.009), and spring season of collection (p=0.003) continued to predict higher tear HEL expression; contact lens wear was not significant (p=0.071) using the raw data.

The best fit multiple linear regression model for describing tear HEL expression using transformed data was found to include season, tear 8OHdG expression, and cell

HEL expression (Table 17). Age and gender were included in the regression analysis to control for potential effects of both variables. Overall, the model explained a moderate proportion of variance in tear HEL expression (adjusted R2=0.25; F=3.44; p=0.014).

Analysis of each predictor variable showed significance only for increase with tear

8OHdG expression (p=0.041) and decrease with cell HEL expression (p=0.011). Age

(p=0.270), gender (p=0.963), and season (p=0.798) were not found to be significant in the model. As described previously, the model was also tested using cell HEL expression data only from visit one, with no evidence of significant differing results from the model above (Table 17).

Using the raw data, the best fit multiple linear regression model for tear HEL expression included tear 8OHdG expression, season, cell HEL expression, and age (Table

17). In addition, gender was added to this model. The full model provided a moderate description of tear HEL expression (adjusted R2=0.31; F=4.23; p=0.005). After adjusting for age and gender, tear 8OHdG expression (p=0.011) and cell HEL expression (p=0.045) remained significant. Thus, based on the raw data we can predict that for one unit of increase in tear 8OHdG expression, an increase on average of 224 units of tear HEL expression. Additionally, for each unit of increase in cell HEL expression, we predict on average about a 58 unit decrease in tear HEL expression. Age (p=0.261), gender

(p=0.750), and season (p=0.844) were not significant predictors of tear HEL expression in this model.

When cell HEL expression was removed from the raw data as a possible predictor, the best fit model included tear 8OHdG expression and season (Table 17). This model provided a weakly moderate description, yet with higher significance, of tear HEL expression (adjusted R2=0.27; F=5.47; p=0.001). After adjusting for age and gender, tear

8OHdG expression (p=0.026) and season of collection (p=0.036) remained significant.

Based on this model we can predict that for one unit of increase in tear 8OHdG

91 expression, on average an increase of 158 units of tear HEL expression. For spring season of collection, we predict on average a 6247 unit increase in tear HEL expression. Age

(p=0.315) and gender (p=0.552) were not significant predictors of tear HEL expression in this model.

Table 17. Linear regression models summary for tear HEL expression.

Tear HEL expression adj R² F test p value coeff effect Transformed data Full best fit model 0.25 3.44 0.014 Season (W=1, S=2) 0.798 0.02 Tear 8OHdG expression 0.041 0.75 + Cell HEL expression 0.011 -0.30 -

Age 0.270 -0.004 Gender (F=1, M=2) 0.963 0.004

Cell HEL visit one only 0.23 3.12 0.022 Season (W=1, S=2) 0.308 0.09 Tear 8OHdG expression 0.040 0.78 + Cell HEL expression 0.019 -0.23 -

Age 0.470 -0.003 Gender (F=1, M=2) 0.968 0.004 Raw data Full best fit model 0.31 4.23 0.005 Tear 8OHdG expression 0.011 223.99 + Season (W=1, S=2) 0.844 657.69 Cell HEL expression 0.045 -57.87 - Age 0.261 -153.73

Gender (F=1, M=2) 0.75 -1020.41

Cell HEL data removed 0.27 5.47 0.001 Tear 8OHdG expression 0.026 158.36 + Season (W=1, S=2) 0.036 6247.47 + SPR

Age 0.315 -102.03 Gender (F=1, M=2) 0.552 -1714.14

Univariate predictors of tear HEL expression for activities reported by visit were different for each visit (Table 15). For visit 1, increased wear of sunglasses was associated with increased tear HEL expression. Alternatively, for visit 2, both increased wear of a protective hat and increased portion of the day spent outdoors while working was supported as inhibitory toward increased tear HEL expression (Table 15).

4.4.3 Cell HEL expression

Univariate predictors found to be statistically significantly associated with cell

HEL expression using transformed data were tear HEL expression, contact lens wear, and season of collection (Table 13). Increased tear HEL expression (p=0.028), contact lens wear (p=0.042), and spring season of collection (p=0.045) predicted lower levels of cell

HEL expression. For the raw data, only season of collection was significant (p=0.033), with spring season again predicting lower cell HEL expression (Table 11).

The initial best fit multiple linear regression model using transformed data for cell

HEL expression included tear HEL expression, contact lens wear, and age; gender was also included in the model (Table 18). This model provided a weakly moderate description of cell HEL expression (adjusted R2=0.24; F=3.82; p=0.012). Individual coefficient analysis showed significant contributions for inverse relationships with tear

HEL expression (p=0.019), contact lens wear (p=0.020), and age (p=0.042). Gender

(p=0.336) was not significant in the model. The model was then tested using visit one cell

HEL data as the outcome variable, with tear HEL expression and tear 8OHdG expression found to be significant predictors (Table 18). When adjusted for age and gender in this model, an inverse relationship with tear HEL expression (p=0.016) and a positive

94 relationship with tear 8OHdG expression (p=0.012) were significant. However, age

(p=0.931) and gender (p=0.468) were not found to be significant using visit one cell HEL expression data as the outcome variable.

Using raw data, the best fit model included season, age, tear HEL expression and tear 8OHdG expression (Table 18). This model provided a weak explanation for cell HEL expression (adjusted R²=0.19; F=2.72; p=0.037). After adjusting for age and gender, tear

HEL expression remained as a significant predictor (p=0.045), with an expectation that for each unit increase in tear HEL expression we anticipate on average a decrease of

0.002 units of cell HEL expression. Although the small coefficient is somewhat challenging to interpret, when tear HEL expression was divided by 10,000, significant changes to both the model and each predictor occurred; hence, this modified data is not shown. However, for comparison, when using this approach for each 10,000 unit increase in tear HEL expression we anticipate on average a decrease of approximately 22 units of cell HEL expression. Age (p=0.107), gender (p=0.976), season (p=0.059), and tear

8OHdG expression (p=0.194) were not significant in the raw data model; however, the data suggests spring season of suggestion may be found to be associated with lower cell

HEL expression. Removal of cell HEL data sensitivity testing was not applicable to this model.

Table 18. Linear regression models summary for cell HEL expression.

Cell HEL expression adj R² F test p value coeff effect Transformed data Full best fit model 0.24 3.82 0.012 Tear HEL expression 0.019 -0.56 - Contact lens wear (Y=1, N=0) 0.020 -0.45 - W CL Age 0.042 -0.01 -

Gender (F=1, M=2) 0.336 0.12

Cell HEL visit one only 0.16 2.77 0.044 Tear HEL expression 0.016 -0.73 - Tear 8OHdG expression 0.012 1.64 +

Age 0.931 0.001 Gender (F=1, M=2) 0.468 0.11 Raw data Full best fit model 0.19 2.72 0.037 Season (W=1, S=2) 0.059 -37.18 (-SPR) Age 0.107 -1.32 Tear HEL expression 0.045 -0.002 - Tear 8OHdG expression 0.194 0.72

Gender (F=1, M=2) 0.976 -0.59

Cell HEL data removed Not applicable

There were no significant activities reported by visit that were identified as univariate predictors for cell HEL expression (Table 15).

4.4.4 Tear 8OHdG expression

Significant univariate regression relationships using transformed variables for tear

8OHdG expression were found for total UVAF, tear HEL expression, and gender (Table

13). Increased total UVAF predicted lower tear 8OHdG expression (p=0.015). On the other hand, increased tear HEL expression (p=0.017) and female gender (p=0.002) predicted higher tear 8OHdG expression. Using the raw data the same predictors of tear

8OHdG were identified with the same directions of influence as found using the transformed data (Table 11).

Tear 8OHdG expression using the initial transformed data was best described using a multiple linear regression model including gender, total UVAF, tear HEL expression, ocular sun exposure, and cell HEL expression (Table 19); age was also added to this model. A moderate relationship of the predictors to tear 8OHdG expression was found for the initial model (adjusted R2=0.52; F=6.63; p<0.0001). Individual analysis of the coefficients showed significant contributions with higher expression for tear HEL expression (p=0.003) and an inverse relationship for age (p=0.024).

The model was then tested with visit one cell HEL expression data (Table 19).

Using visit one cell HEL data the best fit multiple linear regression model included gender, total UVAF, tear HEL expression, ocular sun exposure, and cell HEL expression; age was again added to the model. A moderately strong relationship of the predictors to tear 8OHdG expression was found for this model (adjusted R2=0.55; F=7.29; p<0.0001).

In this model, tear HEL expression (p=0.001) also showed a positive association with increased tear 8OHdG expression; age (p=0.02) was again inversely associated. Gender

(p=0.143), total UVAF (p=0.091), ocular sun exposure (p=0.651), and cell HEL expression (p=0.220) were not found to be significant using the visit one cell HEL data.

When using the raw data tear 8OHdG expression was best described using a multiple linear regression model including tear HEL expression, gender, total UVAF, and cell HEL expression; age was added to this model (Table 19). A strong relationship of the predictors for tear 8OHdG expression was found for this model (adjusted R2=0.59;

F=9.81; p<0.0001). However, individual analysis of the coefficients showed significant contributions only positively for tear HEL expression (p<0.0001) and inversely for age

(p=0.030). Coefficients for tear HEL expression in the raw data models were difficult to interpret (0.001). As such, because there were no other changes to these models, coefficients for raw data tear HEL expression are displayed per 10,000 unit change

(Table 19).With a 10,000 unit increase in tear HEL expression we expect on average an increase of 10.09 units of tear 8OHdG expression. In addition, regression coefficients for this model show that for an increase in age of one year we expect on average a decrease of 0.47 units of tear 8OHdG expression.

Analysis of the raw data following removal of cell HEL expression as a possible predictor of tear 8OHdG generated a model including tear HEL expression, gender, and total UVAF (Table 19). This model was equally robust when compared to the previous one that included cell HEL expression (adjusted R2=0.58; F=12.53; p<0.0001). After adjusting for age and gender, tear HEL expression remained a significant predictor

(p<0.0001) and total UVAF nearly achieved significance as a predictor (p=0.052). Based on coefficients in this model, we predict that for a 10,000 unit increase in tear HEL

98 expression we expect on average an increase of 10.75 units of tear 8OHdG expression. In addition, there is a strong suggestion that for a one unit increase in total UVAF on average we expect tear 8OHdG to decrease by 0.53 units.

Table 19. Linear regression models summary for tear 8OHdG expression.

Tear 8OHdG expression adj R² F test p value coeff effect Transformed data Full best fit model 0.52 6.63 <0.0001 Gender (F=1, M=2) 0.155 -0.05 Total UVAF 0.095 -0.01 Tear HEL expression 0.003 0.20 + Ocular sun exposure 0.589 -0.02 Cell HEL expression 0.806 0.01

Age 0.024 -0.003 -

Cell HEL visit one only 0.55 7.29 <0.0001 Gender (F=1, M=2) 0.143 -0.05 Total UVAF 0.091 -0.01 Tear HEL expression 0.001 0.21 + Ocular sun exposure 0.651 -0.01 Cell HEL expression 0.220 0.04

Age 0.020 -0.003 - Raw data Full best fit model 0.59 9.81 <0.0001 Tear HEL expression* <0.0001 10.09 + Gender (F=1, M=2) 0.276 -5.22 Total UVAF 0.124 -0.38 Cell HEL expression 0.961 0.002

Age 0.030 -0.47 -

Cell HEL data removed 0.58 12.53 <0.0001 Tear HEL expression* <0.0001 10.75 + Gender (F=1, M=2) 0.145 -7.71 Total UVAF 0.052 -0.53 (-)

Age 0.068 -0.40 *Raw data for tear HEL expression is per 10,000 unit change 100

There were no significant activities reported by visit that were identified as univariate predictors for tear 8OHdG expression (Table 15).

4.4.5 Total UVAF

Univariate predictors for transformed total UVAF were tear 8OHdG expression, occupation, and season of collection (Table 12). Higher levels of tear 8OHdG expression were found to predict lower levels of total UVAF (p=0.015). Additionally, outdoor occupation (p=0.031) and winter season of collection (p=0.028) predicted higher total

UVAF. When using the raw data the same predictors of total UVAF were identified with the same directions of influence as found using the transformed data (Table 10).

The initial best fit multiple linear regression model using transformed data for total UVAF included only tear 8OHdG expression (Table 20). The model provided a weak description of total UVAF (adjusted R2=0.12; F=2.46; p=0.082) and was not statistically significant when controlling for non-significant age (p=0.371) and gender

(p=0.668). Interestingly, tear 8OHdG expression (p=0.037) was still found to be a significant contributor with an inverse relationship to total UVAF in this model. When the model was tested using only visit one cell HEL expression data, there was no change from the model above.

When using the raw data total UVAF was best described using a multiple linear regression model including only season (Table 20). However, when adjusted for age and gender, this model was also not significant (adjusted R2=0.10; F=2.19; p=0.110).

Individual analysis of the coefficients showed season (p=0.070), age (0.901), and gender

(p=0.423) did not make significant contributions. Although not statistically significant,

101 the regression coefficient for season suggested that for spring season of collection we can expect on average about a 6.41 unit decrease in total UVAF. There were no notable changes to this model when cell HEL expression was removed as a possible predictor.

Table 20. Linear regression models summary for total UVAF.

Total UVAF adj R² F test p value coeff effect Transformed data Full best fit model 0.12 2.46 0.082 Tear 8OHdG expression 0.037 -6.60 -

Age 0.371 -0.03 Gender (F=1, M=2) 0.668 0.29

Cell HEL visit one only No change Raw data Full best fit model 0.10 2.19 0.110 Season (W=1, S=2) 0.070 -6.41 (- SPR)

Age 0.901 -0.02 Gender (F=1, M=2) 0.423 2.81

Cell HEL data removed No change

Univariate linear regression of activities reported by visit demonstrated an increased portion of the day spent outdoors while working was associated with increased area of conjunctival UVAF for both visit 1 (p=0.034) and visit 2 (p=0.020) (Table 14).

102

Chapter 5: Discussion

5.1 Logistic regression

Logistic regression findings are discussed here for dichotomous variables occupation, contact lens wear, and season. Given that very few subjects identified outdoor occupation (n=9), were contact lens wearers (n=10), or completed the study during winter season (n=15) logistic regression was performed to identify possible covariates. However, significant relationships and possible implications for each dichotomous variable with continuous outcome variables identified following linear regression analyses are discussed further in each appropriate continuous outcome variable section.

5.1.1 Occupation

Occupation was predicted by total UVAF with similar levels of significance when using either raw (p=0.031) or transformed (p=0.043) data (Table 9). In both cases, higher levels of UVAF were associated with outdoor occupation. This outcome is expected from the potentially chronic levels of UV exposure incurred by those who work outdoors routinely. In addition, higher levels of ocular sun exposure predicted outdoor occupation using raw data (p=0.024) (Table 9). Other possible covariates of occupation include lower tear HEL expression (transformed p=0.009; raw p=0.015) and male gender

(p=0.028), both associated with outdoor occupation.

103

Although only nine subjects with outdoor occupation completed the study, it is important to consider these factors as covariates in the present study. Of note, all subjects reporting outdoor occupations were enrolled during winter. In addition, seven of the nine subjects with outdoor occupations were male. The possible influence of outdoor occupation, winter season of collection, and gender are discussed in further detail in section 5.2.2 describing results of linear regression where occupation was identified as a significant predictor of ocular sun exposure. Collectively, evidence here suggests comparisons of those with indoor occupations and those with outdoor occupations may be influenced by season of collection and male gender.

5.1.2 Contact lens wear

Logistic regression did not identify any statistically significant covariates for contact lens wear using either raw or transformed data (Table 9). This finding supports our inclusion of contact lens wearers in data analysis without reason for concern of statistical influence on our outcome measures; this is not surprising considering the small sample of contact lens wearers (n=10) in the present study data. However, univariate linear regression using transformed data demonstrated contact lens wear was a significant predictor of higher tear HEL expression (p=0.042) and lower cell HEL expression

(p=0.042) (Table 13); multivariate linear regression models supported the association with lower cell HEL expression (p=0.020) (Table 18), but not with higher tear HEL expression (Table 17). Linear regression findings are discussed further below for tear

HEL expression and cell HEL expression in sections 5.3.1.2 and 5.3.2.2, respectively.

104

Contact lens wear presents a foreign object on the ocular surface with both mechanical abrasive properties and potentially protective capabilities. Movement of a contact lens is considered necessary for maintenance of ocular surface health yet may potentially increase cellular shedding from both the conjunctival and corneal epithelium and may cause increased levels of oxidative stress on the ocular surface.143 Moreover, contact lens wear may cause squamous metaplasia with altered conjunctival epithelial cells properties.144 Alternatively, corneal protection may be conferred by contact lens wear, and in fact, contact lenses are used therapeutically to reduce the recurrence rates of recurrent corneal erosions.145 Furthermore, animal models and in vitro studies have demonstrated contact lenses with UV-blocking capabilities provide protection against

UV-induced ocular surface component damage.10,12,71,146 Moreover, contact lenses with

UV-blocking capabilities may reduce the overall UV-induced damage to the ocular surface by attenuation of the PLF effect.20,25 A well-controlled study evaluating a larger group of subjects with similar levels of UV exposure including those wearing no contact lenses and those wearing contact lenses with or without UV blocking capabilities would more accurately determine the potential relevance of contact lens wear to outcomes in the present study.

5.1.3 Season of collection

Covariates for season of collection identified using either raw or transformed data were total UVAF, male gender, and tear HEL expression (Table 9). Higher levels of total

UVAF were associated with winter season of collection with similar significance levels for both raw (p=0.028) and transformed (p=0.036) data; male gender was also associated

105 with winter season of collection (p=0.005). As discussed above in section 5.1.1, it is important to consider that all subjects reporting outdoor occupations were enrolled during winter; thus presenting a confounding factor for measuring higher levels of total UVAF in winter. Higher levels of UV exposure during the Ohio winter of 2014 were not expected as there was little snowfall and temperatures were near 0°F for several weeks; thus few people spent much time outdoors. The possible influence of outdoor occupation, winter and gender are discussed in further detail in sections 5.2.2 and 5.4.2 describing linear regression outcomes for ocular sun exposure and total UVAF, respectively.

Furthermore, higher levels of tear HEL expression consistently predicted spring season of collection with similar significance levels when comparing raw (p=0.008) and transformed (p=0.008) data (Table 9). This finding is likely confounded by the influence of all subjects with outdoor occupations completing the study during winter. In addition, it is probable the data is confounded by the influence of male gender as independently associated with lower levels of tear HEL expression (Tables 11 and 13). However, this finding does support the hypothesis that higher tear HEL expression is found during spring season of collection when higher levels of ambient UV are expected and that tear

HEL expression is a transient measure of ocular surface oxidative stress.

5.1.4 Dichotomous UVAF and ocular sun exposure

Total UVAF and ocular sun exposure were considered as dichotomous variables relative to the median. Each was tested as the outcome variable while using the other as the possible predictor; neither was found to be a significant predictor of the other

(p=0.293). This outcome is unexpected as previous reports have demonstrated a strongly

106 supported positive correlation between these outcomes.132 The negative outcome found in the present study is likely confounded by factors discussed above in sections 5.1.1 and

5.1.3, specifically occupation and season of collection. Although total UVAF was identified earlier as a predictor of outdoor occupation, the outcome here was likely influenced by the compounding factor that all subjects reporting outdoor occupations enrolled in the study during winter season. Subsequently, even those with outdoor occupations resulted in having relatively low hours of ocular sun exposure; this outcome is further discussed in section 5.2.

5.2 Ocular sun exposure assessment

While environmental factors may have equal effects on UV exposure of a population at a specific time and location, significant variance is likely to occur on an individual level due to factors including behavior, cultural dress, and variation in skin pigmentation.1 General UV exposure levels can be estimated by comparing the subjects’ lifetime history of geographic location with ambient UV levels as determined by ground monitors or satellites; however, ocular UV exposure is more challenging to estimate. 3,30-

32,34-37 Assessment of UV exposure in most epidemiologic studies has been through methods subject to recall bias and often with disregard for individual activities.45

Simplification of outdoor activity recall by Kwok and colleagues was shown to account for approximately 94% of time spent outdoors,48 and was thus chosen for use in this study. Additionally, Threlfall and colleagues showed that accounting for individual variations in ocular protective measures, as done in the current study with questionnaires,

107 can significantly improve ocular UV exposure estimates as they relate to ocular surface disease over those calculated from ambient estimates of UV exposure.29

Descriptive statistics are reported here for the first time using the present study method of ocular sun exposure estimation based on information combined from two questionnaires (Appendix A) and adjusted for the relative protective values of ocular protective measures.14,15,22 Although all questionnaires based on subject recall inescapably suffer from recall bias, a reduced timeframe required for recall of activities and time outdoors to a two week time period, as opposed to recall over several years or a lifetime.45 For the feasibility portion of the study all of our subjects worked indoors, resulting in ocular sun exposure hours detected being few, and ranged from approximately 30 minutes to 10 hours with a mean of 4.38 (±3.02) hours over a two week time period. When compared with the cross-sectional full data set of 50 subjects, including the 9 subjects reporting outdoor occupations, mean ocular sun exposure approximately doubled to 8.86 (±11.97) hours. Although the full cross-sectional data set shows a significant increase in ocular sun exposure compared with the initial feasibility study group, the resulting mean ocular sun exposure was not strikingly greater. This is most likely a reflection of subjects reporting outdoor occupation enrolling in the present study during winter season. As mentioned in section 5.1.3, winter of 2014 in Ohio was cold and harsh, with even those having outdoor occupations spending little time outdoors.

Regardless of this seasonal confounder, it is important to note that in the present study, self-reported increased portion of day spent outdoors while working predicted higher levels of ocular sun exposure (visit 1: p=0.014; visit 2: p=0.034). It would be of interest

108 to obtain a larger sample size, particularly recruiting those with markedly increased amounts of ocular sun exposure, to confirm the usefulness of the study surveys in the assessment of ocular UV exposure. Possible approaches to recruit those with increased ocular sun exposure could include a multi-center study using study surveys in locations with varying levels of ambient UV, or by following a cohort of subjects over one year.

Additionally, comparison with alternative methods to gather information about time outdoors more specific to UV exposure, such as personal UV dosimeters, would further establish the usefulness of ocular sun exposure assessment methods used in this study.63

5.2.1 Ocular sun exposure repeatability

There is good agreement for test-retest repeatability of the two measurements of ocular sun exposure over two weeks using the raw data as the mean difference (bias) is very close to zero and the scatter plot shows all of the data points, with the exception of a single outlier, are within two standard deviations of the mean difference (Figure 17).

Further inspection of the outlier revealed the subject to have reported a large number of outdoor hours for each of the two visits, although the use of sunglasses was reported 75% of the time for visit one, and only 25% of the time for visit two. This outlier highlights the challenges associated with accurately assessing ocular sun exposure. Visual inspection of the scatter plot suggests significant skewness of the data toward lower levels of ocular sun exposure and slightly increased variability with increased ocular sun exposure may be present. The variability may be explained by disparities in weather or schedules between the two visits, affecting subject motivation or ability to spend time outdoors. However, there is no definite systematic variation apparent within the data for

109 this measure. The plot shows that 95% of the time ocular sun exposure falls within approximately ± 15 hours over two weeks of time; those with greater than 15 hours may be influenced by other individual behavior factors (Figure 17). Although clinically acceptable variation for ocular sun exposure remains to be established, the test-retest variability found appears acceptable for use of this measure in future studies.

Further analysis of the repeatability using raw data for ocular sun exposure was conducted to further describe the skewness and variability of the complete data set.

Comparison of ocular sun exposure in the winter and spring shows distinct seasonal disparity for data distribution. Bland-Altman scatter plot shows spring ocular sun exposure data collected during the feasibility study (Figure 10) to be fairly evenly distributed with a narrow spread. Furthermore, normal Q-Q plot for ocular sun exposure in the spring (Figure 31) illustrates the normal distribution of this data subset. In contrast, the Bland-Altman scatter plot for winter ocular sun exposure (Figure 30) shows definite skewness toward lower values and much wider 95% limits of agreement than are present for the spring ocular sun exposure data. In support of this, the normal Q-Q plot (Figure

32) demonstrates the skewed values for the winter ocular sun exposure subset of our data

(skewness = 3.46; kurtosis = 12.94). These findings demonstrate differences in season for ocular sun exposure as a reliable outcome variable that are contradictory to the findings of Sherwin and colleagues;132 however, support is provided for future studies investigating the influence of season on ocular sun exposure and how it may relate to other outcome variables in the present study.

110

After log10 transformation of ocular sun exposure for the full cross-sectional data set (Figure 22), lower levels of ocular sun exposure with excess variability are evident.

Variability with 95% limits of agreement is approximately ±1 log unit (approximately 10 hours), with the scatter plot showing those with greater than 100 hours (2 log units) of ocular sun exposure will fall well outside the 95% limits of agreement. Although log10 transformation of ocular sun exposure allowed for analysis of the data as a normally distributed variable, it is somewhat challenging to draw meaningful conclusions from the

Bland-Altman plot, particularly as contradictory outliers remain; outliers in the transformed data fall in the negative direction for log10 values (Figure 22), while the one outlier in the raw data was in a strongly positive direction (Figure 17).

5.2.2 Ocular sun exposure linear regression

Ocular sun exposure was found to have few predictors when tested with either univariate (Tables 10 and 12) or multivariate (Table 16) regression. Multivariate models demonstrated a higher level of significance using raw data (p=0.009) when compared with transformed data (p=0.030). While no continuous variables were found to predict ocular sun exposure, occupation remained significant in both the transformed (p=0.016) and raw data (p<0.0001) multivariate models. In the multivariate model using transformed data, outdoor occupation predicted increased ocular sun exposure by a factor of approximately 2.7; using raw data, outdoor occupation predicted on average about 16 additional hours of ocular sun exposure when compared with those having indoor occupations.

111

Given the outcome from the feasibility study demonstrating a modest positive linear correlation (r = 0.454, p = 0.044) between ocular sun exposure and tear HEL levels

(Figure 14), it was anticipated a larger sample size would strengthen this relationship.

However, based on the full cross-sectional sample in the present study, this relationship has not proven substantial. As only a modest correlation was found, support for such a relationship would likely be more evident if recruitment of those with markedly increased amounts of ocular sun exposure were successful.

Collectively these results highlight the presence of known confounding variables in this study. First, all nine subjects who reported outdoor occupations were seen during the winter season. Of note, the winter of 2014 in Ohio was particularly cold, partially explaining the low ocular sun exposure. Further inspection of the small group (n=9) of outdoor workers shows the present study contains additionally confounded data. While mean ocular sun exposure for all with outdoor occupations (mean=19.40 hours) was higher than the entire study group (mean=8.86 hours), when two outliers with approximately 50 and 60 hours of ocular sun exposure were removed, the resulting mean ocular sun exposure for the remaining seven outdoor workers was only slightly higher than for the entire study group (mean=9.16 hours). For those with outdoor occupations it was unequivocally expected that an increase in ocular sun exposure would be found when compared with those who have indoor occupations. Combined with the fact that seven of nine outdoor workers were male, future studies are indicated by these results that would include more subjects who are outdoor workers with data collected over each season, and who are age and gender matched with those who are indoor workers.

112

5.3 Oxidative stress assessment

Methods to assess oxidative stress and the damage it may cause are numerous. It is challenging to directly detect ROS due to their high reactivity and short half-lives;81 therefore, it is more common to measure the deleterious products formed by ROS actions.79 The present study focused on lipid and DNA damage oxidative stress markers in human tear and conjunctival cell samples using ELISAs. Although ELISAs have been used successfully to detect oxidative stress markers in samples similar to those in the present study,86,89,141 only the HEL ELISA has previously been used for human in vivo tear samples.86,87,98,147 Evaluation of conjunctival cell HEL by ELISA has not been reported in humans, but was recently reported in a murine dry eye model.148 Additionally, the 8OHdG ELISA has not previously been used for human in vivo tear samples; however, the assay has been performed using a human corneal epithelial cell line following UVB exposure100 and surgically obtained conjunctival specimens of those with pterygia.99

5.3.1 Lipid damage detection

The first point of contact for UV exposure on the ocular surface is the tear film. At the forefront of the tear film is the lipid bilayer.7 Furthermore, just posterior to the tear film lays the conjunctiva with its large surface area composed, in part, by several layers of epithelial cells, with each cell enclosed by a lipid bilayer membrane.8 These lipid bilayer structures may be subject to ROS-mediated UV damage, and lipid peroxidation may be the first evidence of ocular surface oxidative damage.9 HEL is a lipid hydroperoxide- derived oxidation product from omega-6 fatty acids.82 As HEL represents only one step in

113 the lipid peroxidation pathway, future studies evaluating tears and ocular surface samples for different types of lipid peroxidation products would help to clarify the potential significance of other lipid peroxidation effects.

The present study has identified tear HEL values that are considerably higher than those previously reported for human tears.86,87,98,147 Reasons for this could be experimental differences in tear sample preparation and analysis, environmental induced variation (i.e. previous reports are all from Japan), dietary differences, or genetic variation in HEL production. In support of diet-induced variation, it is well established that consumption of omega-6 fatty acids is much higher for those who choose a Western diet compared with those who consume a traditional Japanese diet.85 Furthermore, increased consumption of omega-6 fatty acids may lead to altered lipogenic mRNA transcription,149 possibly facilitating increased production of HEL. Interestingly, mean tear HEL expression was higher in the feasibility study than in the full cross-sectional data. As all participants in the feasibility study were Caucasian, our findings may indeed reflect genetic variation in levels of HEL production.

In addition to the tear HEL findings, it is reported here for the first time in vivo

HEL expression by ELISA in human conjunctival cells. HEL was significantly lower in the conjunctiva compared to the tears, suggesting the majority of ocular surface HEL may be produced by damage to the tear film lipid layer. Alternatively, lipid damage may originate at the conjunctival or corneal cell level with HEL accumulation in the tear film.

Further research is needed to clarify the source of HEL on the ocular surface.

5.3.1.1 Tear HEL expression repeatability

114

While the mean difference using raw data for tear HEL expression for the full cross-sectional data set may appear high (+529.27), it should be stressed that this only represents approximately 2% of the limits of agreement values, and is thus comparatively near zero (Figure 19). Although the data point spread is somewhat wide with a range between approximately ±18,000 nmol/L HEL, there is no apparent systematic variation of the means. To date, clinically acceptable variation in tear HEL expression has not been well defined, yet the test-retest variability found here is acceptable for use of this measure in future studies.

Following log10 transformation of tear HEL expression data (Figure 24), the

Bland-Altman plot shows a small mean difference (+0.02 log units) with 95% limits of agreement representing approximately half of a log unit, or variability by approximately a factor of 3. It can be seen (Figure 24) that log10 transformation of the data provided a normal distribution of the data points. Although a few marginal outliers remain, no systematic variability is apparent for log10 transformed tear HEL expression data.

Tear HEL ELISA repeatability was evaluated for visit one, comparing the two wells used for each sample from a single visit following log10 transformation (Figure 27).

Compared with the log10 transformed tear HEL expression data (Figure 24), an equivalent mean bias is present with comparable 95% limits of agreement, yet with visual inspection a tighter distribution about the mean is evident. Furthermore, there is no evidence of systematic variation in the tear HEL ELISA for visit one samples, wells one and two.

This combination of factors provides confidence for repeatability of the ELISA kit when used with human tear samples. Consequently, as the distribution for the visit one ELISA

115 is more tightly distributed about the mean (Figure 27), one may also conclude the tear

HEL expression variability (Figure 24) between visit one and visit two reflects individual measurement variability between visits. Therefore, tear HEL expression may reflect a transient value, and further supports the need for seasonal studies.

5.3.1.2 Tear HEL expression linear regression

Multivariate models using either transformed or raw data, with or without inclusion of cell HEL sensitivity testing, were significant with raw data significance somewhat stronger when compared with transformed data models (Table 17). Increased tear 8OHdG expression was consistently identified as a predictor of higher levels of tear

HEL expression using univariate and multivariate models, and when using either transformed or raw data (Tables 11, 13, and 17). This positive predictive relationship strongly suggests that lipid modification is accompanied by DNA modification on the ocular surface. While tear 8OHdG expression was found to have a positive relationship with tear HEL expression, cell HEL expression consistently predicted lower tear HEL expression; the significance of the relationship was stronger for the transformed data. The inverse relationship between HEL expression of tear and cell samples is counterintuitive.

As previously noted, cell HEL expression demonstrates notable bias toward increased variability with increased expression using the raw data (Figure 20) and is modified but still apparent upon visible inspection of the log10 transformed data (Figure 25). Overall, this bias reduces the confidence in cell HEL expression as a valid predictor. Furthermore, the full cross-sectional study data outcome is contradictory to the feasibility study of the initial 20 subjects, where data showed a strong positive correlation relationship between

116 tear HEL and cell HEL expression (Table 3). Hence, data from the feasibility study provided support for consistently elevated HEL on the ocular surface of each subject regardless of sample type.

Given that the full cross-sectional data demonstrated an inverse relationship between tear and cell HEL expression, it is possible the correlation relationship found in the feasibility study is due to cell sample collection variability. Alternatively, it is also possible that the variability of ocular surface HEL expression overall is partly accounted for by individual differences in constitutive HEL production or another step in the lipid peroxidation pathway contributing to the formation of HEL on the ocular surface.

Furthermore, it is possible that HEL expression on the ocular surface may be affected differently in tears and conjunctival cells by other factors discussed in this study that may cause lipid peroxidation to be higher in either tears or cells such as season of collection, gender, and ocular sun exposure hours. Specifically, that increased levels of ambient UV that may be accompanied by increased ocular sun exposure in spring as compared with winter may cause increased lipid peroxidation of tears or increased cellular expression with rapid release of HEL into the tears. Possible explanations for variability in cell HEL expression are further discussed below in section 5.3.2.1.

Additional univariate predictors for tear HEL expression demonstrate female gender, indoor occupation, spring season of collection, and contact lens wear may predict higher expression. Collectively these predictors support increased tear lipid peroxidation may be present in these subject groups when compared with male gender, outdoor occupation, winter season of collection, and no contact lens wear (Table 13). However,

117 when considering them together with the fact that male gender predicted winter season of collection in the present study as discussed in section 5.1.3, it is strongly suggested that the present study data are confounded by the fact that most samples collected from males and those with outdoor occupations were collected during winter season. In addition, in multivariate models only tear 8OHdG expression and cell HEL expression were consistently significant to predict tear HEL expression, with spring season only significant when cell HEL expression was removed as a possible predictor (Table 17).

Thus, support is provided for oxidative stress biomarkers on the ocular surface being influenced by similar factors; specifically, that lipid peroxidation is consistently accompanied by DNA damage in tears, and that cell HEL expression consistently has an inverse relationship with tear HEL expression. Furthermore, spring season of collection may have an influence on tear HEL expression directly through increased tear HEL expression and indirectly by possible increased release of cellular HEL into the tears.

Subsequently, there is limited evidence in the present study to support the significance of gender, occupation, season of collection, and contact lens wear to tear HEL expression.

It was expected that increased levels of UV exposure would result in increased tear HEL expression; however, this hypothesis was not supported by the present data.

Specifically, it was expected that those with outdoor occupations would have higher levels of tear HEL when compared to those with indoor occupations. Consequently, increased exposure to UV by those with outdoor occupations was expected to cause increased levels of lipid peroxidation oxidative stress biomarkers on the ocular surface.

However, as discussed previously in section 5.2.2, study subjects with outdoor

118 occupations had only marginally higher hours of ocular sun exposure as they completed the study during the cold 2014 Ohio winter season; this may explain why tear HEL expression was not found to be increased in those with outdoor occupations.

Alternatively, it was not surprising that sample collection in the spring season was associated with increased tear HEL expression. As mentioned previously, when compared with winter, ambient UV levels increase42,43 along with outdoor activities during warmer seasons, potentially resulting in increased UV exposure. Additionally, inconclusive evidence is present for prediction of tear HEL expression by activities reported, as predictors varied comparing visit one with visit two; sunglass wear is shown to increase tear HEL expression at visit one, while use of hats and increased portion of day outdoors while working appear to be inhibitory toward tear HEL expression at visit two (Table 15). Confounding factors may be present in this data as increased portion of day outdoors while working may serve to increase use of both hats and sunglasses.

Collectively these confounding factors likely influence a bias of the data toward the null hypothesis and support analysis of spring and winter season of collection subsets of the present study data as well as future studies collecting additional data during different seasons for each of these variables. One possible study design to further evaluate the effect of season on ocular surface oxidative stress biomarkers would be a longitudinal study design, following a cohort of age and gender matched subjects with wide variation in levels of UV exposure over one year of time, evaluating ocular surface samples from subjects during each season.

5.3.2.1 Cell HEL expression repeatability

119

Of the five measures, cell HEL expression shows the largest variability. The scatter plot using raw data for the full cross-sectional data set shows a distinct funnel shaped distribution, clearly demonstrating bias toward increased variability with higher cell HEL expression (Figure 20). One probable influence on the bias is differing amounts of cells were likely collected from subjects by the study method, largely due to variation in subject tolerance of the collection process. In addition, sample processing using cell membrane disruption by sonication prior to HEL ELISA measurement was necessary to release cellular lipid peroxidation products, yet could have resulted in samples containing sheared DNA and proteins that prevented obtaining accurate absorbance readings to estimate the concentration of DNA per sample. Given the significant bias demonstrated here, alternative methods of collection and quantification of cells collected and subsequent HEL expression measurement should be explored to improve repeatability of this measure.

Feasibility data showed reasonable repeatability for cell HEL expression in the small sample number evaluated (n=7), with only a possible downward trend with increased cell HEL expression (Figure 12). Clarification of repeatability for cell HEL expression was anticipated by collection of additional data points; however, the complete cross-sectional data set served to increase bias for this measure. As mentioned previously in this section, the Bland-Altman plot using raw data (Figure 20) illustrated bias toward increased variability with increased mean levels of cell HEL expression strongly suggests measurement error, possibly from sample collection, the extraction process, or the ELISA kit. Several possible sources of cell HEL expression bias should be considered. First, cell

120 collection was done without the use of anesthetic. The reason for avoiding anesthetic was to minimize the potential influence of anesthetic on reactive species formation.150 There is some evidence to show that anesthetic alters oxidative stress adduct production;150 however, no evidence in the literature demonstrates that topical anesthetics can effect oxidative stress outcomes for ocular surface samples. The lack of anesthetic influenced the examiner’s inability to collect equal quantities of conjunctival cells from each subject’s ocular surface. By direct observation of the examiner, distinct differences in subject tolerance for the cell collection process were present. Although tolerance for collection was reasonable in that all subjects returned for visit two, the examiner was not able to complete the goal of rotating the swab equally over each conjunctival surface area of interest for all subjects. In addition, some subjects displayed clear differences in conjunctival sensitivity between nasal and temporal regions of an eye, as well as differences in conjunctival sensitivity between eyes. Evaluation of subject tolerance of sample collection and quantities of cells collected would help to clarify this outcome measure. A tolerance assessment scale could be used from the examiner and/or subject perspective. In addition, conjunctival images following sample collection with quantified area of conjunctival tissue disruption would provide additional insight into quantity of cells collected; vital staining of the conjunctiva was only used to confirm sample cells were collected and was not found statistically relevant. Finally, method development for analyzing cell HEL expression in humans using the swab collection method has not been previously reported; therefore, limited data is available for comparison with the study methods or findings for cell HEL expression. Given the challenge of accurate DNA

121 concentration estimation described earlier in this section, exploration of an assay suitable to estimate cells and/or DNA collected following sonication of samples would provide useful insight into estimating equivalent sample quantities assayed. The method of sample processing to achieve cell disruption by sonication created a challenge to subsequently evaluate the samples for DNA content or cell counting. Hence, a method to equalize the concentration of cells or DNA used prior to ELISA evaluation has been elusive to date. Possible methods to pursue to achieve this goal include Qiagen kits, gel electrophoresis, and mass spectrometry.

Following log10 transformation of cell HEL expression data (Figure 25), the mean difference is small (-0.16 log units) with variability based on 95% limits of agreement approximately ± 1.00 log unit, or approximately a factor of ±10. Some moderation of the funnel distribution is evident when the log10 transformed plot is compared with the raw data (Figure 20); yet, increased variability with increased mean values is still apparent. Thus, although log10 transformation of cell HEL expression resulted in normal distribution of the data points, a lack of confidence in study data for cell HEL expression remains.

To obtain an estimation of variability present with the ELISA kit for cell samples collected and processed using our methods, a Bland-Altman plot was produced using log10 transformed visit one samples for cell HEL values comparing wells one and two for each subject (Figure 28). The mean difference is very small (-0.04 log units), and approximately one quarter of the mean difference compared with the cell HEL expression plot for the log10 transformed data (Figure 25). In addition, 95% limits of agreement show

122 variability of about ± 0.30 log units (factor of 2), reduced to about one-fifth of variability in the cell HEL expression plot (factor of 10). Furthermore, although a suggestion of increased variability with lower values is present, the majority of data points cluster very near the mean bias. Together these outcomes support repeatability for the selected ELISA kit using our cell sample types and processing method. It is likely that the variability of the HEL ELISA kit is minor in comparison with the cell HEL expression data; therefore, one assumption that could be made is that variability is present in the subjects, with increasing cell HEL expression occurring with increased mean quantity collected.

5.3.2.2 Cell HEL expression linear regression

Higher tear HEL expression was found to consistently predict lower levels of cell

HEL expression in univariate and multivariate model analysis with similar levels of significance, with the one exception of univariate analysis of the raw data (Tables 11, 13, and 18). In addition, contact lens wear predicted lower cell HEL expression when using transformed data in both univariate analysis and the full cross-sectional transformed data multivariate model. Furthermore, spring season of collection predicted lower cell HEL expression in univariate analysis when using either transformed or raw data. However, in the multivariate models there was only a tendency toward spring season of collection

(p=0.059) predicting lower cell HEL levels when using the raw data, and failed to reach statistical significance.

The inverse relationship of tear HEL expression with cell HEL expression is counterintuitive. As discussed above in section 5.3.2.1, the validity of cell HEL expression data in the current study is questionable as clear evidence of bias toward

123 increased variability with increased expression is present (Figure 20). Additionally, a positive relationship between cell HEL and tear HEL expression was apparent during the feasibility portion of the present study (Table 3).

Contact lens wear was associated with lower cell HEL expression using transformed data in both univariate regression (Table 13) and the full multivariate model

(Table 18). This inverse relationship with contact lens wear was stronger in the multivariate model (p=0.020) than with univariate analysis (p=0.042). This outcome suggests that contact lens wear may be inhibitory toward damage to cellular lipids.

However, the outcome that contact lens wear may be associated with increased tear HEL expression while inhibitory toward conjunctival cell HEL expression is again counterintuitive as discussed earlier in section 5.1.2. While it has not yet been shown conclusively, it is possible there are factors associated with contact lens wear that affect ocular surface lipids in tears and cells differently.143 Specifically, mechanical effects from contact lens wear may cause increased cellular shedding, thus releasing cellular oxidative damage products into the tear film. Furthermore, contact lens wear may increase squamous metaplasia of conjunctival cells, possibly reducing cellular HEL production.144

Moreover, corneal and limbal protection by contact lenses with UV-blocking capabilities may reduce the levels of cell HEL expression by an overall reduction in UV-induced ocular surface damage.10,12,25,146 These are only tentative suggestions based on the findings given the small number of contact lens wearers and confounding of the present study data, particularly with cell HEL expression bias toward increased variability with

124 increased expression. Additional studies using well controlled groups of contact lens wearers and non-contact lens wearers would provide useful enlightenment on this topic.

Comparison of multivariate linear regression models showed similar significance using either transformed (p=0.012) or raw data (p=0.037), yet the strength of the relationship for each model lacked robustness (transformed data: adjusted R²=0.24; raw data: adjusted R²=0.19) (Table 18). Hence, for the transformed data there is a small improvement in the model relevance as a predictor of cell HEL expression. Possible predictors of cell HEL expression considered in the various models included tear HEL expression, contact lens wear, season, age, and tear 8OHdG expression. When examining the coefficient for tear HEL expression as a predictor in the raw data, it is of interest to note that although tear HEL expression remained significant (p=0.045) in the model, the anticipated decrease in cell HEL expression is a very small incremental change.

However, the small incremental change is likely a reflection of the fact that tear samples were diluted by a factor of 500 prior to conducting the HEL ELISA; conjunctival cell samples were not adjusted due to sample processing methods discussed in section 5.3.2.1.

Although this outcome is unanticipated and counterintuitive, the present study data provides support for the inverse relationship between tear and cell HEL expression.

Alternatively, it is possible that HEL expression on the ocular surface is influenced differently in cells and tears. For example, conjunctival cells may release HEL generated by cellular lipid damage into the tears, or tears may simply produce more HEL following increased ocular sun exposure with conjunctival cells affected later. As there are no previous reports of in vivo conjunctival cell HEL expression in humans using our

125 methods, future studies are indicated to further explain the relationship between cell and tear HEL expression.

5.3.2 DNA damage detection

Several forms of damage to DNA may occur as a result of UV radiation-generated

ROS including direct oxidative DNA damage, strand breaks, and defective repair.74,100,146

Increased oxidative DNA damage has been detected in several types of pathological conditions including cancer, autoimmune diseases, and nervous system disorders.74

Following its discovery in 1984, 8-OHdG has become the most widely accepted assay for oxidative DNA damage.95,97 While many oxidative mechanisms may contribute to its production, 8-OHdG is formed primarily from oxidative DNA damage by the interaction of hydroxyl radicals with the DNA base guanosine.96 Relevant to the ocular surface, 8-

OHdG has been detected in surgically obtained human conjunctival specimens of those with conjunctivochalasis and pterygium,98,99 cultured human corneal epithelial cells following UVB exposure,100 murine corneal tissue samples11 with significant increase following UVB exposure, and recently in type 2 diabetic murine tears.101

For the first time, the present study reports detection of human tear 8OHdG expression. One expected origin of tear 8OHdG is from damage to DNA in the conjunctival cells with subsequent accumulation in the tears. However, using present study methods 8OHdG was not detected in conjunctival cells. This may be a result of low levels of DNA collection using non-anesthetized in vivo conjunctival cytology.

Alternatively, tear 8OHdG may originate from corneal cells rather than conjunctival cells, or cellular shedding from both conjunctiva and corneal surfaces may result in free

126 ocular surface cells in the tear film that are subject to DNA damage and result in generation of tear 8OHdG. Another potential source of DNA in tears is extracellular

DNA, recently reported to be present in human tears and elevated in those with dry eye.151 These findings may stimulate future research in the area of DNA and DNA damage products present in human tears and their potential relevance to ocular disease.

5.3.2.1 Tear 8OHdG expression repeatability

Test-retest repeatability using raw data for tear 8OHdG expression over two weeks was also good, with the mean difference close to zero, as +3.53 is approximately

6% of the limits of agreement (Figure 21). However, a possible slight increase in variability with increased levels of tear 8OHdG is apparent. One possible source of this variability may be explained by factors affecting ocular sun exposure such as occupation as discussed in section 5.2. Additionally, storage effects were demonstrated in the present study with increased tear 8OHdG expression (~16%) over one year of storage at -80° C.

This storage effect was not expected as 8OHdG expression stability for more than two years has been demonstrated in urine when stored at -80° C.152 However, the present study finding suggests tear samples should be analyzed at approximately equal intervals following collection to facilitate equivalent comparison. Variability of the 8OHdG

ELISA may also play a role as the ELISA manufacturer does not provide data on the assay’s intra- or inter-assay precision; however, limited variability of the ELISA was detected in the present study and is discussed further later in this section. As this is the first time human tear 8OHdG levels have been reported, and while clinically acceptable

127 variation is yet to be determined, the test-retest variability found in the present study using raw data was acceptable.

Following log10 transformation of tear 8OHdG expression (Figure 26) a very low mean bias is present (+0.05 log units), with 95% of data points falling within approximately ± 0.70 log units (factor of 5). In addition, visual inspection shows most of the points cluster very near the mean bias. However, visual inspection also suggests a range effect with increasing tear 8OHdG expression detected in samples collected at visit two compared with samples collected during visit one. In fact, linear regression of the difference and the mean tear 8OHdG reveals a significant positive relationship between the two variables is present (R=0.42; p=0.003). Together these outcomes support the possible transient nature of tear 8OHdG expression and highlight the importance of conducting longitudinal studies comparing seasonal outcomes for this measure.

Additionally, repeatability for the 8OHdG ELISA kit when using human tear samples processed by our methods was estimated (Figure 29). Using log10 transformed data from visit one, wells one and two for each subject, the 8OHdG ELISA kit demonstrated excellent repeatability with a very low mean bias (-0.04 log units) and 95% limits of agreement of approximately ± 0.10 log units (factor of 1.25). The narrow range of variability present provides confidence in the ELISA kit when used with our sample types. Consequently, while a small portion of the variability detected in tear 8OHdG expression is due to variability within the ELISA kit, 8OHdG expression within the tear samples also contains variability. Furthermore, support is given for the presence of true variability detected between visit one and two, and for the range effect showing increased

128

8OHdG expression for visit two compared with visit one (Figure 26). The illustrated range effect for tear 8OHdG expression may be due to subject activity variation between visits one and two, and may also be influenced by storage time effects found for tear

8OHdG expression.

5.3.2.2 Tear 8OHdG expression linear regression

Increased tear HEL expression consistently predicted higher tear 8OHdG expression. This positive predictive relationship was evident for univariate analysis using transformed or raw data, and remained highly significant for multivariate models using transformed or raw data, including models for sensitivity testing of cell HEL expression

(Tables 11, 13, and 19). Furthermore, this positive relationship between tear HEL and

8OHdG expression holds for both directions of analysis, as increased tear 8OHdG expression was also found to consistently predict higher tear HEL expression as discussed in section 5.3.1.2. These results provide strong support that DNA damage occurs concurrently with lipid damage on the ocular surface.11,98 While the significance of tear HEL expression as a predictor of tear 8OHdG expression is quite strong in the transformed (p=0.003) and raw (p<0.0001) data based models, the coefficient of determination is quite small at 0.20 log unit (factor of 1.6) increase for the transformed data and a 10.09 unit increase in the raw data model per 10,000 unit increase of tear HEL expression (Table 19). The small coefficient is likely a reflection of the large dilution factor required for tears (500 X) to quantify HEL expression resulting in notably higher relative abundance for tear HEL expression when compared with tear 8OHdG expression

(20 X dilution factor).

129

The relationship between tear 8OHdG expression and tear HEL expression is further supported by the feasibility study data where tear 8OHdG levels were found to have a positive linear correlation with both cell and tear HEL levels. The strongest relationship was found between tear 8OHdG levels and cell HEL levels (r = 0.903, p =

0.005), with more than 80% (r2 = 0.815) of the total variability in cell HEL levels explained by tear 8OHdG levels (Figure 15). It was recognized the small sample size analyzed for cell HEL (n = 7) may have led to an unrealistic correlation result; nonparametric correlation analysis using Spearman’s Rho was also performed with essentially the same results (r = 0.893, p = 0.007; r2 = 0.797). However, the relationship between cell HEL and tear 8OHdG was not supported in the full cross-sectional data, possibly for factors discussed above in section 5.3.2.1 such as variability in cell sample collection and DNA estimation challenges. The relationship found between tear 8OHdG expression and tear HEL expression in the feasibility study was only modest (r = 0.619, p

= 0.004) and explained near 40% (r2 = 0.383) of the variability in tear 8OHdG expression

(Figure 16). It was anticipated a stronger association of 8OHdG expression with cell

HEL expression would be found, given that the 8OHdG ELISA reflects cellular DNA repair levels. Supporting this hypothesis, Pauloin and colleagues demonstrated increased expression of 8OHdG in human corneal epithelial cells following UVB exposure in a dose-response manner.100 Furthermore, it has been shown increased 8OHdG expression occurs in pterygium tissue when compared with adjacent normal conjunctival tissue.99

Detection of 8OHdG in tears in the present study suggests that shedding from the conjunctival and corneal epithelial cells may accumulate in the tear film, and combined

130 with the feasibility study correlation findings, suggests that DNA repair may accompany lipid peroxidative damage. Hence, it has been shown for the first time on the ocular surface using human in vivo samples that lipid damage is associated with DNA damage.

Future studies will help delineate the clinical relevance of this relationship to ocular surface disease.

While increased tear HEL expression predicted higher tear 8OHdG expression, increased UVAF predicted lower tear 8OHdG expression. Increased total UVAF predicted lower tear 8OHdG expression in univariate analysis using either transformed or raw data, yet failed to reach significance in the multivariate models (Tables 11, 13, and

19). Given the small sample size for total UVAF (n=34), it is possible the present study is underpowered to detect a significant relationship. This finding is likely confounded by the findings that with univariate analyses female gender predicted higher tear 8OHdG expression (Tables 11 and 13), and that total UVAF was found to be higher for those with outdoor occupation (Table 9). As discussed above in section 5.2, data was collected for those reporting outdoor occupation during winter season, and those with outdoor occupations were of primarily (7/9) male gender. Additionally, although not specific to

UVAF, this finding is contrary to outcomes predicted from previous ocular surface studies based on human corneal epithelial cells,100 pterygium tissue,99 and in murine models exposed to UV radiation, where increased corneal 8OHdG expression11 has been demonstrated following UV radiation exposure. Importantly, previous studies are of ocular surface cells and tissues, while the present study detected 8OHdG in tears.

131

It is possible conjunctival UVAF inhibits higher levels of tear 8OHdG expression as a result of altered collagen, as occurs with collagen cross-linking, shown to cause autofluorescence following UV exposure.120 Another possible source of inhibition is squamous metaplasia and its associated changes; it has been demonstrated that UVB initiates and perpetuates squamous metaplasia of conjunctival tissue153 and may result in altered cellular functions promoting downstream tissue changes. Indeed, squamous metaplasia in pingueculae has been shown to be accompanied by altered expression of keratin and proliferative biomarkers.154 Further supporting this relationship, it has been shown that by histological examination squamous metaplasia, pterygium, and squamous cell carcinoma are components of the same continuum.155 Moreover, surgically excised pterygia tissue was demonstrated to have increased 8OHdG expression when compared with adjacent normal tissue.99 Taken together, it is suggested that UV exposure- associated altered collagen and cells may have reduced function with a diminished ability for the ocular surface to produce and release 8OHdG or shed cells containing 8OHdG into the tear film. Consequently, it is suggested altered collagen combined with squamous metaplasia and related tissue changes may occur concurrently with conjunctival UVAF and subsequently UV-induced tissue damage inhibits increased tear 8OHdG expression.

Another possible explanation is that increased UV exposure, with potentially subsequent increased conjunctival UVAF, may cause increased apoptosis on the ocular surface that with chronic UV exposure may reduce the total number of conjunctival cells present to participate in DNA damage and repair activities.100 Supporting this concept is that in contrast with skin photoaging, known to be associated with epithelial thickening,156

132 thinning of the overlying epithelium has been demonstrated in pingueculae157. Future studies may serve to further describe the significance of altered tear 8OHdG expression to ocular surface disease.

Age was identified with similar levels of significance as an inverse predictor of tear 8OHdG expression in multivariate models using either transformed (p=0.024) or raw data (p=0.030) (Table 19). The inverse relationship of tear 8OHdG expression with age suggests that less DNA damage occurs with increased age or that DNA damage repair and apoptotic mechanisms may not be as efficient with increased age. It is more likely the latter is true, given that appropriate DNA damage repair and apoptosis are important functions to maintain healthy tissue and eliminate mutations; thus, increased tear 8OHdG expression in younger individuals may be a reflection of properly functioning cellular repair and maintenance. The clinical significance of a decrease of 0.003 log unit (factor of 1) in tear 8OHdG expression for each increasing year of age remains to be determined.

An inverse relationship of DNA damage with age is an unexpected finding, as the general expectation is that oxidative stress increases with age. Furthermore, a recent report has shown an increase in lipid peroxidation biomarker malondialdehyde in human tears with increased age.158 As mentioned above in section 5.3.2, potential sources of tear 8OHdG include cellular shedding of corneal and conjunctival cells and extracellular DNA.151

Alternatively, as discussed earlier in this section, it is possible altered collagen, squamous metaplasia, or cumulative apoptosis with increased age inhibits tear 8OHdG expression.99,100,120,128,135,153,159 As the present study reports tear 8OHdG in human

133 samples for the first time, additional studies are indicated to further investigate the source of tear 8OHdG and its potential relevance to ocular surface disease.

5.4 Conjunctival ultraviolet autofluorescence detection

Conjunctival UVAF is a recently identified and validated possible biomarker of ocular sun exposure developed based on principles for skin evaluation.127,130,134 Several reports from Australia and New Zealand have shown this measure to be useful for specific groups and climates.126-132 For those with pterygia, a disease known to be associated with ocular sun exposure, it was demonstrated that increased conjunctival

UVAF was associated with increased risk of prevalent pterygia.135 An increase in conjunctival UVAF with age was demonstrated with schoolchildren (ages 3 to 15 years old),127 while a decrease in conjunctival UVAF with increasing age was shown for adults

(ages 15 to 89 years old).130 Interestingly, increased area of conjunctival UVAF was demonstrated to be protective against prevalent myopia (ages 15 years old or greater).126,129

Several potential sources of conjunctival UVAF have been considered including altered collagen, the presence of NADH, altered cellular mitochondrial activity, and lipofuscin.125,127,128,133 One plausible source of conjunctival UVAF is from collagen cross-linking, demonstrated to autofluoresce following UVA exposure.120 The majority of

UVB and approximately 35% of UVA is absorbed by ocular surface tissues.19 In fact, both UVB and UVA exposure are known to be associated with UV induced ocular surface damage in those with pterygia.32 While UVB is thought to be responsible for direct DNA damage, UVA may be more responsible for initiation of oxidative stress

134 related UV damage including production of ROS, MMPs, indirect DNA base damage and collagen cross-linking.160 While potential downstream influences of UVB-induced DNA damage and mutations should not be minimized, UVA is established as a cause of collagen cross-linking,120,161,162 with altered collagen structure often permanently changing the functionality of affected tissue.112 Collagen cross-linking occurs in the stroma of corneal tissue,162 and is expected to occur in the stroma of conjunctival tissue as well. In fact, the excitation wavelengths for conjunctival UVAF in the present study were in the UVA range (320 to 380nm), previously found to be associated with collagen cross-linking autofluorescence in human skin.163 While autofluorescence of cross-linked collagen has been shown, the duration of this effect is unknown. One possible explanation for decreased conjunctival UVAF with increased age is that chronic exposure to UV may result in chronic stimulation of MMPs,153 potentially causing collagen degradation with eventual reduction in collagen autofluorescence and an altered autofluorescence profile of conjunctival tissue.164 Furthermore, down-regulation of collagen genes in adult eyes has been demonstrated.165 Lipofuscin is also known to autofluoresce,112 yet little support in the literature is found to support lipofuscin autofluorescence in conjunctival tissue. Additionally, the transient nature of autofluorescence from NADH or cellular mitochondrial activity is not as likely to have sustainable autofluorescence repeatable over two weeks as detected in the present study.112

Median total UVAF in the present study was relatively low (6.31 mm2) when compared with previous findings reported by Sherwin, et al (28.2mm2).130 However, a

135 recent report by Wolffson and colleagues of conjunctival UVAF in European eye care providers reports comparable estimates to the present study, although differences in study design prevent conclusive comparison.133 Several factors may have influenced the more than four-fold reduction found in total UVAF for our study compared with Sherwin and colleagues. First, although time spent outdoors is not reported in hours by Sherwin132 it is likely total hours spent outdoors is notably higher for the population of Norfolk Island when compared with our study subjects living in Ohio; this is expected due to the sub- tropical climate in Norfolk Island, while Ohio has a seasonally variable climate.

Secondly, Sherwin, et al129-132 and Ooi, et al127,128 report the use of colored light emitting diode (LED) fixation lights while acquiring images of conjunctival UVAF during their studies; the specific colors of LED lights used are not reported. Although the small amount of light emitted from the fixation lights may have had a limited influence, it is possible the lights increased the area of conjunctival autofluorescence by addition of unknown wavelengths to UV excitation with subsequent increased emission of visible light. This hypothesis is supported by a report from Utine and colleagues demonstrating pingueculae autofluorescence following excitation in the visible blue spectrum (488 nm) using confocal scanning laser microscopy.125 In this study, the area of autofluorescence in all evaluated pingueculae was greater than the clinically evident lesion.125 Therefore, consideration of the full spectrum of fluorescence is anticipated to more thoroughly elucidate the biological significance of conjunctival autofluorescence.

5.4.1 Conjunctival Ultraviolet Autofluorescence repeatability

136

Conjunctival UVAF (n=34) using the raw data also demonstrated good intra- observer test-retest repeatability over two weeks, with a mean difference of nearly zero

(0.19 mm²; 95% limits of agreement -1.75 to 2.12 mm²) (Figure 18). Although initially the point spread may appear somewhat wide, the limits of agreement are relatively narrow. There is no evidence of systematic bias of conjunctival UVAF in our data and the test-retest variability found here is acceptable. Clinically acceptable variability is yet to be determined for conjunctival UVAF. However, this study’s findings are in agreement with Sherwin and colleagues132 who have demonstrated reliability and validity for this measure in a genetically isolated group dwelling in a region with a fairly homogenous climate.132 In fact, intra-observer repeatability was notably better in the present study compared with the level demonstrated previously by Sherwin and colleagues (mean difference 0.36 mm²; 95% limits of agreement -5.54 to 6.25 mm²).135

Comparison with the square root transformed data for total UVAF was also done using Bland-Altman analysis (Figure 23). The mean difference here is very near zero

(+0.01 units) with a 95% limits of agreement of approximately 0.50 units. In addition, visual inspection shows that a few marginal outliers remain and again there is no evidence of systematic bias present. A disadvantage of the square root transformation is that values are challenging to interpret in meaningful units; however, there are few differences or additional information to be gleaned by comparison of transformed with the raw data Bland-Altman plot (Figure 18). The similarities between the raw and transformed data Bland-Altman plots support the excellent repeatability of total UVAF regardless of the small sample size (n=34) in this study.

137

5.4.2 Conjunctival ultraviolet autofluorescence linear regression

Multivariate linear regression models for conjunctival UVAF using either transformed or raw data were not statistically significant (Table 20). Several factors may have influenced this outcome. First, the number of subjects for which we obtained useful conjunctival UVAF images is small (n=34) and may not be sufficient to attain significance in linear regression models. Secondly, several confounding factors are present to bias our findings toward the null hypothesis as discussed in section 5.2.

Additional support for confounding by season of collection is given in the multivariate model using raw data. Although season did not achieve significance (p=0.070) in this model, there was a trend suggesting spring season of collection showed a decrease on average of about 6.41mm² total UVAF (Table 20) and is further discussed later in this section.

Higher levels of tear 8OHdG expression were found to predict lower levels of total UVAF with univariate analysis using either transformed (Table 12) or raw (Table

10) data, and remained significant (p=0.037) using transformed data in the multivariate model (Table 20). Potential explanations for the inverse relationship of tear 8OHdG expression and total UVAF are discussed earlier in section 5.3.2.2. In addition, univariate analysis of transformed and raw data showed outdoor occupation and winter season of collection predicted higher total UVAF. As discussed above in section 5.2, these findings are expected yet likely confounded by collection of data from those with outdoor occupations during winter, primarily from those of male gender. Furthermore, male

138 gender was found to independently predict lower tear 8OHdG expression in both transformed (Table 13) and raw (Table 11) data forms.

Increased UV exposure during winter of 2014 in Ohio was not anticipated given that ambient UV levels are known to be lower in winter,42,43 there was little snow fall, and temperatures were near 0°F for several weeks during the months of January and

February. Limited UV exposure during winter sample collection of the present study is supported by evidence described above in section 5.2 showing that even those having outdoor occupations had relatively few hours of ocular sun exposure. It is more likely conjunctival tissue retains UV autofluorescent properties for a minimum of several months resulting from chronic UV exposure, such as may occur with those who have outdoor occupations. Based on findings in this study, it is suggested conjunctival UVAF may be a reflection of prior UV exposure sustained for at least several months. Thus, the persistence of conjunctival UVAF is perhaps longer than suggested by Sherwin and colleagues, who suggested UVAF may serve as an indicator of acute or sub-acute UV exposure over weeks to months.129,130 As discussed above in section 5.3.2.2, the present study findings of higher levels of tear 8OHdG associated with decreased conjunctival

UVAF may be explained by UV-induced collagen cross-linking,120 with increased changes to collagen possibly causing a diminished ability of conjunctival tissue to release

8OHdG into the tears.

Univariate linear regression of portion of time spent on activities reported by visit demonstrated an increased portion of the day spent outdoors while working was associated with increased area of conjunctival UVAF for both visit one (p=0.034) and

139 visit two (p=0.020) (Table 14). This finding is expected and consistent with Sherwin and colleagues’ report of increased median total UVAF with increasing proportion of day spent outdoors.132

Identification of the possible predictors for total UVAF described above provides support for additional studies with increased sample size, recruiting age and gender matched subjects, and with a focus on potential seasonal effects for total UVAF and other study outcome variables.

5.5 Limitations

This study has several limitations, not the least of which is small sample size.

Small sample size is of particular importance for the outdoor occupation group (n=9), cell

HEL expression (n=37), and total UVAF (n=34). Each of these groups has undoubtedly influenced the outcomes for the present study, but with limited statistical power to confidently support the results. The limited statistical power is further illustrated by the limited robustness of several findings associated with the above groups. Furthermore, sensitivity testing showed some of the data is sensitive to transformation to normal distributions and to consideration of variability in cell HEL expression.

The study design is also a limitation as it was a cross-sectional feasibility study.

The cross-sectional design does not allow conclusions to be drawn about causation, but only comparison of groups for differences. Additionally, as this was in many ways overall a feasibility study, very liberal inclusion and exclusion criteria were applied.

Specifically, inquiries were not made regarding possible inducers of ocular oxidative stress including systemic diseases such as diabetes or personal habits such as smoking.

140

Although many of our subjects were from the College of Optometry where smoking prevalence is low, it is known that smoking causes ocular surface oxidative stress87 and future studies should ask subjects about smoking habits. Additionally, systemic disease type 2 diabetes mellitus is thought to cause ocular oxidative stress.101 Although it is not known if diabetes has an influence on UV associated oxidative stress, the effects could certainly be additive and future studies should inquire about diabetic diagnoses.

Another limitation is that the questionnaire based assessment of ocular sun exposure suffered from recall bias. Although the recall time was shortened to two weeks from the more common years of assessment, it was clear during subject completion of the questionnaire on hours spent outside it remained a challenge for many subjects to accurately recall their activities over the past two weeks. A more accurate method of assessing time outdoors and UV exposure would be through the use of electronic personal dosimeters.44,66 In addition, the questionnaire used in this study regarding personal ocular UV protection habits contained broad categories of classification.

Although Sherwin used a very similar questionnaire and found good association with total UVAF,132 the general categories of use of hats, spectacles, and sunglasses is clearly not specific and likely affected the calculation of ocular sun exposure. Ultimately, an ocular specific UV detector would provide greater accuracy to assess this measure.

141

Chapter 6: Summary and Future Directions

6.1 Summary

6.1.1 Feasibility study

Feasibility and repeatability was demonstrated for previously unreported ELISA measurements of oxidative stress levels in human conjunctival cells (HEL) and tears

(8OHdG). Furthermore, correlation analysis suggests that ocular sun exposure is associated with ocular surface lipid damage, and that lipid damage is associated with

DNA damage. Findings from this study establish a baseline in normal subjects for future studies comparing these outcome measures in normal groups with ocular surface disease groups.

6.1.2 Cross-sectional study

The full data set for the cross-sectional study further supports the feasibility study findings and test-retest repeatability of each outcome measure evaluated in the feasibility study with the exception of cell HEL expression. In addition to the outcome measures evaluated in the feasibility study, test-retest repeatability for conjunctival UVAF was established over a two week time period. Furthermore, repeatability for ELISA kits using the present study sample types and methods of collection was also evaluated, and each was discussed within the appropriate section. Following feasibility study estimates, cross- sectional sample size goal of 25 per group for those having indoor and outdoor

142 occupations was not achieved; despite targeted recruitment efforts, only nine subjects reporting outdoor occupations completed the study. However, based on feasibility study outcomes, subjects were recruited and samples analyzed in the same manner during the cross-sectional portion of the study.

Several findings are well supported in this study by transformed data multivariate linear regression models. First, although analysis of the full data set calls into question the reliability of our cell HEL expression analyses, the positive relationship between tear

HEL expression and tear 8OHdG expression remains strongly supported. This relationship provides support to the concept that lipid damage is associated with DNA damage on the ocular surface. Secondly, support is provided for tear and cell lipid damage on the ocular surface to occur in opposing directions; an apparent equilibrium is maintained, with either the tears or cells reflecting the majority of the damage. Thirdly, cellular changes that occur on the ocular surface may be inhibitory toward evidence for

DNA damage in tears. Fourth, contact lens wear may be inhibitory toward cellular lipid damage. Lastly, both cellular lipid damage and DNA damage products in tears may decrease with increased age. The clinical relevance of these findings remains to be elucidated through additional studies.

6.2 Future Directions

Based on our findings from the present study, BOSS recommendations, and the existing literature several directions of future studies may provide additional insight into the significance of oxidative stress to the ocular surface. Specifically, the present study supports a need for a longitudinal study with seasonal evaluation of markers. Future

143 studies with more restrictive inclusion and exclusion criteria will help to clarify and delineate the significance of our findings to ocular surface disease. It is hoped that the outcome measures reported here will provide additional insight into the importance of the tear film lipid layer for the maintenance of ocular surface equilibrium, as well as its potential role in the onset of ocular surface disease. Studies of various ocular disease groups may benefit by incorporation of outcome measures established during this study, including studies of pterygium, dry eye disease, myopia, and diabetes. Ocular surface diseases associated with UV exposure such as pterygium and conjunctival squamous cell carcinoma may benefit most directly from future studies based on the reported findings.

Furthermore, incorporation of additional oxidative stress biomarkers as suggested by the

BOSS study will provide additional insight into the potential mechanisms associated with ocular diseases, with the ultimate goal of developing individual treatment and prevention protocols.

144

References

1. Lucas RJ, McMichael, T., Smith, W., Armstrong, B. Solar Ultraviolet Radiation: Global burden of disease from solar ultraviolet radiation. In: Environment PHat, ed. Geneva2006. 2. Pane AR, Hirst LW. Ultraviolet light exposure as a risk factor for ocular melanoma in Queensland, Australia. Ophthalmic Epidemiol. Sep 2000;7(3):159- 167. 3. Cruickshanks KJ, Klein R, Klein BE. Sunlight and age-related macular degeneration. The Beaver Dam Eye Study. Arch Ophthalmol. Apr 1993;111(4):514-518. 4. Mikkelsen RB, Wardman P. Biological chemistry of reactive oxygen and nitrogen and radiation-induced signal transduction mechanisms. Oncogene. Sep 1 2003;22(37):5734-5754. 5. Williams DL. Oxidative stress and the eye. The Veterinary clinics of North America. Small animal practice. Jan 2008;38(1):179-192, vii. 6. National Eye Institute: Vision Research Needs, Gaps, and Opportunities. 2012. 7. Green-Church KB, Butovich I, Willcox M, et al. The international workshop on meibomian gland dysfunction: report of the subcommittee on tear film lipids and lipid-protein interactions in health and disease. Invest Ophthalmol Vis Sci. Mar 2011;52(4):1979-1993. 8. Pepperl JE, Ghuman, T., Gill, K.S., Zieske, J.D, and Trocme, S.D. Conjunctiva. Duane's Foundations of Clinical Ophthalmology. Vol 1: Lippincott Williams & Wilkins; 2005. 9. Avery SV. Molecular targets of oxidative stress. Biochem J. Mar 1 2011;434(2):201-210. 10. Chandler HL, Reuter KS, Sinnott LT, Nichols JJ. Prevention of UV-induced damage to the anterior segment using class I UV-absorbing hydrogel contact lenses. Invest Ophthalmol Vis Sci. Jan 2010;51(1):172-178. 11. Ibrahim OM, Kojima T, Wakamatsu TH, et al. Corneal and retinal effects of ultraviolet-B exposure in a soft contact lens mouse model. Invest Ophthalmol Vis Sci. Apr 2012;53(4):2403-2413. 12. Andley U. Protection of UVB-induced damage in cultured lens epithelial cells by UV-absorbing soft contact lenses. Invest Ophthalmol Vis Sci. 2009;50(E-Abstract 2537). 13. Youn HY, A.P. Cullen, B.R. Chou and J.G. Sivak. Phototoxicity of ultraviolet (UV) radiation: evaluation of UV-blocking efficiency of intraocular lens (IOL)

145

materials using retinal cell culture and in vitro bioassays. The Open Toxicology Journal. 2010;4:13-20. 14. Diffey BL. Sources and measurement of ultraviolet radiation. Methods. Sep 2002;28(1):4-13. 15. McLoone P, Simics E, Barton A, Norval M, Gibbs NK. An action spectrum for the production of cis-urocanic acid in human skin in vivo. The Journal of investigative dermatology. May 2005;124(5):1071-1074. 16. Laurent-Applegate LE, Schwarzkopf, S. Photooxidative stress in skin and regulation of gene expression. In: Fuchs J, Packer, L., ed. Environmental Stressors in Health and Disease. New York, NY: Marcel Dekker; 2001. 17. Sliney DH. How light reaches the eye and its components. International journal of toxicology. Nov-Dec 2002;21(6):501-509. 18. Zigman S. Environmental near-UV radiation and cataracts. Optom Vis Sci. Dec 1995;72(12):899-901. 19. Lucas RM. An epidemiological perspective of ultraviolet exposure--public health concerns. Eye Contact Lens. Jul 2011;37(4):168-175. 20. Coroneo MT. Pterygium as an early indicator of ultraviolet insolation: a hypothesis. Br J Ophthalmol. Nov 1993;77(11):734-739. 21. Kwok LS, Coroneo MT. A model for pterygium formation. Cornea. May 1994;13(3):219-224. 22. Kwok LS, Coroneo MT. Temporal and spatial growth patterns in the normal and cataractous human lens. Exp Eye Res. Sep 2000;71(3):317-322. 23. Sasaki H, Kawakami Y, Ono M, et al. Localization of cortical cataract in subjects of diverse races and latitude. Invest Ophthalmol Vis Sci. Oct 2003;44(10):4210- 4214. 24. Shanmuganathan VA, Foster T, Kulkarni BB, et al. Morphological characteristics of the limbal epithelial crypt. Br J Ophthalmol. Apr 2007;91(4):514-519. 25. Kwok LS, Kuznetsov VA, Ho A, Coroneo MT. Prevention of the adverse photic effects of peripheral light-focusing using UV-blocking contact lenses. Invest Ophthalmol Vis Sci. Apr 2003;44(4):1501-1507. 26. Leow YH, Tham SN. UV-protective sunglasses for UVA irradiation protection. International journal of dermatology. Nov 1995;34(11):808-810. 27. Moseley H, Cox NH, Mackie RM. The suitability of sunglasses used by patients following ingestion of psoralen. Br J Dermatol. Feb 1988;118(2):247-253. 28. Nishigori C. Cellular aspects of photocarcinogenesis. Photochemical & photobiological sciences : Official journal of the European Photochemistry Association and the European Society for Photobiology. Feb 2006;5(2):208-214. 29. Threlfall TJ, English DR. Sun exposure and pterygium of the eye: a dose-response curve. Am J Ophthalmol. Sep 1999;128(3):280-287. 30. Panchapakesan J, Hourihan F, Mitchell P. Prevalence of pterygium and pinguecula: the Blue Mountains Eye Study. Australian and New Zealand journal of ophthalmology. May 1998;26 Suppl 1:S2-5.

146

31. Mitchell P, Smith W, Wang JJ. Iris color, skin sun sensitivity, and age-related maculopathy. The Blue Mountains Eye Study. Ophthalmology. Aug 1998;105(8):1359-1363. 32. Taylor HR, West S, Munoz B, Rosenthal FS, Bressler SB, Bressler NM. The long-term effects of visible light on the eye. Arch Ophthalmol. Jan 1992;110(1):99-104. 33. Taylor HR, West SK, Rosenthal FS, et al. Effect of ultraviolet radiation on cataract formation. N Engl J Med. Dec 1 1988;319(22):1429-1433. 34. Taylor HR, West SK, Rosenthal FS, Munoz B, Newland HS, Emmett EA. Corneal changes associated with chronic UV irradiation. Arch Ophthalmol. Oct 1989;107(10):1481-1484. 35. Fitzpatrick TB. The validity and practicality of sun-reactive skin types I through VI. Archives of dermatology. Jun 1988;124(6):869-871. 36. West SK, Duncan DD, Munoz B, et al. Sunlight exposure and risk of lens opacities in a population-based study: the Salisbury Eye Evaluation project. JAMA. Aug 26 1998;280(8):714-718. 37. McCarty CA, Fu CL, Taylor HR. Epidemiology of pterygium in Victoria, Australia. Br J Ophthalmol. Mar 2000;84(3):289-292. 38. Cruickshanks KJ, Klein BE, Klein R. Ultraviolet light exposure and lens opacities: the Beaver Dam Eye Study. Am J Public Health. Dec 1992;82(12):1658-1662. 39. Duncan DD, Munoz B, Bandeen-Roche K, West SK. Assessment of ocular exposure to ultraviolet-B for population studies. Salisbury Eye Evaluation Project Team. Photochem Photobiol. Nov 1997;66(5):701-709. 40. Godar DE. UV doses worldwide. Photochem Photobiol. Jul-Aug 2005;81(4):736- 749. 41. McKenzie RL, Bjorn LO, Bais A, Ilyasad M. Changes in biologically active ultraviolet radiation reaching the Earth's surface. Photochemical & photobiological sciences : Official journal of the European Photochemistry Association and the European Society for Photobiology. Jan 2003;2(1):5-15. 42. Environmental Protection Agency: Monthly Average UV Index. http://www.epa.gov/sunwise/uvimonth.html. Accessed September 25, 2014. 43. Godar DE, Wengraitis SP, Shreffler J, Sliney DH. UV doses of Americans. Photochem Photobiol. Jun 2001;73(6):621-629. 44. Allen M, McKenzie R. Enhanced UV exposure on a ski-field compared with exposures at sea level. Photochemical & photobiological sciences : Official journal of the European Photochemistry Association and the European Society for Photobiology. May 2005;4(5):429-437. 45. Yam JC, Kwok AK. Ultraviolet light and ocular diseases. International ophthalmology. May 31 2013. 46. Parisi AV, Green A, Kimlin MG. Diffuse solar UV radiation and implications for preventing human eye damage. Photochem Photobiol. Feb 2001;73(2):135-139. 47. Yu CL, Li Y, Freedman DM, et al. Assessment of lifetime cumulative sun exposure using a self-administered questionnaire: reliability of two approaches. 147

Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. Feb 2009;18(2):464-471. 48. Kwok RK, Linet MS, Chodick G, et al. Simplified categorization of outdoor activities for male and female U.S. indoor workers-a feasibility study to improve assessment of ultraviolet radiation exposures in epidemiologic study questionnaires. Photochem Photobiol. Jan-Feb 2009;85(1):45-49. 49. McCarty CA. Sunlight exposure assessment: can we accurately assess vitamin D exposure from sunlight questionnaires? Am J Clin Nutr. Apr 2008;87(4):1097S- 1101S. 50. Kimlin MG, Martinez N, Green AC, Whiteman DC. Anatomical distribution of solar ultraviolet exposures among cyclists. Journal of photochemistry and photobiology. B, Biology. Oct 2 2006;85(1):23-27. 51. Parisi AV, Kimlin MG, Lester R, Turnbull D. Lower body anatomical distribution of solar ultraviolet radiation on the human form in standing and sitting postures. Journal of photochemistry and photobiology. B, Biology. Jan 2003;69(1):1-6. 52. Gao N, Hu LW, Gao Q, et al. Diurnal variation of ocular exposure to solar ultraviolet radiation based on data from a manikin head. Photochem Photobiol. May-Jun 2012;88(3):736-743. 53. Cooley JH, Quale LM. Skin cancer preventive behavior and sun protection recommendations. Seminars in oncology nursing. Aug 2013;29(3):223-226. 54. Webb AR, Slaper H, Koepke P, Schmalwieser AW. Know your standard: clarifying the CIE erythema action spectrum. Photochem Photobiol. Mar-Apr 2011;87(2):483-486. 55. Kollias N, Baqer AH, Ou-Yang H. Diurnal and seasonal variations of the UV cut- off wavelength and most erythemally effective wavelength of solar spectra. Photodermatology, photoimmunology & photomedicine. Apr 2003;19(2):89-92. 56. Hu LW, Gao Q, Xu WY, et al. Diurnal variations in solar ultraviolet radiation at typical anatomical sites. Biomedical and environmental sciences : BES. Jun 2010;23(3):234-243. 57. Hu L, Gao Q, Gao N, et al. Solar UV exposure at eye is different from environmental UV: diurnal monitoring at different rotation angles using a manikin. Journal of occupational and environmental hygiene. 2013;10(1):17-25. 58. Sasaki H, Sakamoto Y, Schnider C, et al. UV-B exposure to the eye depending on solar altitude. Eye Contact Lens. Jul 2011;37(4):191-195. 59. Davis A, Deane GH, Diffey BL. Possible dosimeter for ultraviolet radiation. Nature. May 13 1976;261(5556):169-170. 60. Schmid KL, Leyden K, Chiu YH, et al. Assessment of daily light and ultraviolet exposure in young adults. Optom Vis Sci. Feb 2013;90(2):148-155. 61. Duncan DD, Schneider W, West KJ, Kirkpatrick SJ, West SK. The development of personal dosimeters for use in the visible and ultraviolet wavelengths regions. The Salisbury Eye Evaluation Team. Photochem Photobiol. Jul 1995;62(1):94- 100.

148

62. Sydenham MM, Collins MJ, Hirst LW. Measurement of ultraviolet radiation at the surface of the eye. Invest Ophthalmol Vis Sci. Jul 1997;38(8):1485-1492. 63. Seckmeyer G, Klingebiel M, Riechelmann S, et al. A critical assessment of two types of personal UV dosimeters. Photochem Photobiol. Jan-Feb 2012;88(1):215- 222. 64. Biosense Laboratories. http://www.biosense.de/home-e.htm. Accessed September 29, 2014. 65. Quintern LE, Puskeppeleit M, Rainer P, et al. Continuous dosimetry of the biologically harmful UV-radiation in Antarctica with the biofilm technique. Journal of photochemistry and photobiology. B, Biology. Jan 1994;22(1):59-66. 66. National Institute of Water and Atmospheric Research. Personal UV Dosimeters. www.niwa.co.nz/our-services/online-services/uv-and-ozone/personal-uv- dosimeters. Accessed September 29 2014. 67. Cejkova J, Stipek S, Crkovska J, et al. UV Rays, the prooxidant/antioxidant imbalance in the cornea and oxidative eye damage. Physiol Res. 2004;53(1):1-10. 68. Turrens JF. Mitochondrial formation of reactive oxygen species. The Journal of physiology. Oct 15 2003;552(Pt 2):335-344. 69. Ames BN, Gold LS. Endogenous mutagens and the causes of aging and cancer. Mutat Res. Sep-Oct 1991;250(1-2):3-16. 70. Andley UP, Clark BA. Generation of oxidants in the near-UV photooxidation of human lens alpha-crystallin. Invest Ophthalmol Vis Sci. Apr 1989;30(4):706-713. 71. Chandler H. Ultraviolet absorption by contact lenses and the significance on the ocular anterior segment. Eye Contact Lens. Jul 2011;37(4):259-266. 72. Finkel T, Holbrook NJ. Oxidants, oxidative stress and the biology of ageing. Nature. Nov 9 2000;408(6809):239-247. 73. Lawrence GD. Chapter 5: Oxidation and Lipid Peroxidation. The Fats of Life: Essential Fatty Acids in Health and Disease. New Brunswick, NJ: Rutgers University Press; 2010. 74. Cooke MS, Evans MD, Dizdaroglu M, Lunec J. Oxidative DNA damage: mechanisms, mutation, and disease. FASEB J. Jul 2003;17(10):1195-1214. 75. Buddi R, Lin B, Atilano SR, Zorapapel NC, Kenney MC, Brown DJ. Evidence of oxidative stress in human corneal diseases. J Histochem Cytochem. Mar 2002;50(3):341-351. 76. Kastle M, Grune T. Protein oxidative modification in the aging organism and the role of the ubiquitin proteasomal system. Current pharmaceutical design. Dec 1 2011;17(36):4007-4022. 77. Umapathy A, Donaldson P, Lim J. Antioxidant delivery pathways in the anterior eye. BioMed research international. 2013;2013:207250. 78. Chen Y, Mehta G, Vasiliou V. Antioxidant defenses in the ocular surface. Ocul Surf. Oct 2009;7(4):176-185. 79. Labuschagne CF, Brenkman AB. Current methods in quantifying ROS and oxidative damage in Caenorhabditis elegans and other model organism of aging. Ageing research reviews. Sep 2013;12(4):918-930.

149

80. Biomarkers of Oxidative Stress Study. Assays. http://www.niehs.nih.gov/research/resources/databases/bosstudy/assays/index.cfm . Accessed October 8, 2014. 81. Freinbichler W, Colivicchi MA, Stefanini C, et al. Highly reactive oxygen species: detection, formation, and possible functions. Cell Mol Life Sci. Jun 2011;68(12):2067-2079. 82. Kato Y, Osawa T. Detection of lipid-lysine amide-type adduct as a marker of PUFA oxidation and its applications. Arch Biochem Biophys. Sep 15 2010;501(2):182-187. 83. Moselhy HF, Reid RG, Yousef S, Boyle SP. A specific, accurate, and sensitive measure of total plasma malondialdehyde by HPLC. J Lipid Res. Mar 2013;54(3):852-858. 84. Riahi Y, Cohen G, Shamni O, Sasson S. Signaling and cytotoxic functions of 4- hydroxyalkenals. American journal of physiology. Endocrinology and metabolism. Dec 2010;299(6):E879-886. 85. Simopoulos AP. The importance of the ratio of omega-6/omega-3 essential fatty acids. Biomedicine & pharmacotherapy = Biomedecine & pharmacotherapie. Oct 2002;56(8):365-379. 86. Wakamatsu TH, Dogru M, Ayako I, et al. Evaluation of lipid oxidative stress status and inflammation in atopic ocular surface disease. Mol Vis. 2010;16:2465- 2475. 87. Rummenie VT, Matsumoto Y, Dogru M, et al. Tear cytokine and ocular surface alterations following brief passive cigarette smoke exposure. Cytokine. Aug 2008;43(2):200-208. 88. Sorkhabi R, Ghorbanihaghjo A, Javadzadeh A, Rashtchizadeh N, Moharrery M. Oxidative DNA damage and total antioxidant status in glaucoma patients. Mol Vis. 2011;17:41-46. 89. Yagci R, Ersoz I, Erdurmus M, Gurel A, Duman S. Protein carbonyl levels in the aqueous humour and serum of patients with pseudoexfoliation syndrome. Eye (Lond). Jan 2008;22(1):128-131. 90. Alamdari DH, Kostidou E, Paletas K, et al. High sensitivity enzyme-linked immunosorbent assay (ELISA) method for measuring protein carbonyl in samples with low amounts of protein. Free Radic Biol Med. Nov 15 2005;39(10):1362- 1367. 91. Lau LI, Liu CJ, Wei YH. Increase of 8-hydroxy-2'-deoxyguanosine in aqueous humor of patients with exudative age-related macular degeneration. Invest Ophthalmol Vis Sci. Nov 2010;51(11):5486-5490. 92. Chen MH, Tsai CF, Hsu YW, Lu FJ. Epigallocatechin gallate eye drops protect against ultraviolet B-induced corneal oxidative damage in mice. Mol Vis. 2014;20:153-162. 93. Liu C, Chen Y, Kochevar IE, Jurkunas UV. Decreased DJ-1 Leads to Impaired Nrf2-Regulated Antioxidant Defense and Increased UV-A-Induced Apoptosis in Corneal Endothelial Cells. Invest Ophthalmol Vis Sci. Sep 2014;55(9):5551-5560.

150

94. Gao S, Qin T, Liu Z, et al. Lutein and zeaxanthin supplementation reduces H2O2- induced oxidative damage in human lens epithelial cells. Mol Vis. 2011;17:3180- 3190. 95. Kasai H. Chemistry-based studies on oxidative DNA damage: formation, repair, and mutagenesis. Free Radic Biol Med. Aug 15 2002;33(4):450-456. 96. Kasai H. Analysis of a form of oxidative DNA damage, 8-hydroxy-2'- deoxyguanosine, as a marker of cellular oxidative stress during carcinogenesis. Mutat Res. Dec 1997;387(3):147-163. 97. Kasai H, Hayami H, Yamaizumi Z, SaitoH, Nishimura S. Detection and identification of mutagens and carcinogens as their adducts with guanosine derivatives. Nucleic acids research. Feb 24 1984;12(4):2127-2136. 98. Ward SK, Wakamatsu TH, Dogru M, et al. The role of oxidative stress and inflammation in conjunctivochalasis. Invest Ophthalmol Vis Sci. Apr 2010;51(4):1994-2002. 99. Kau HC, Tsai CC, Lee CF, et al. Increased oxidative DNA damage, 8- hydroxydeoxy- guanosine, in human pterygium. Eye (Lond). Jul 2006;20(7):826- 831. 100. Pauloin T, Dutot M, Joly F, Warnet JM, Rat P. High molecular weight hyaluronan decreases UVB-induced apoptosis and inflammation in human epithelial corneal cells. Mol Vis. 2009;15:577-583. 101. Sano K, Kawashima M, Ito A, et al. Aerobic exercise increases tear secretion in type 2 diabetic mice. Invest Ophthalmol Vis Sci. Jul 2014;55(7):4287-4294. 102. Azqueta A, Slyskova J, Langie SA, O'Neill Gaivao I, Collins A. Comet assay to measure DNA repair: approach and applications. Frontiers in genetics. 2014;5:288. 103. Latonen L, Laiho M. Cellular UV damage responses--functions of tumor suppressor p53. Biochim Biophys Acta. Jul 25 2005;1755(2):71-89. 104. Redmond RW. Introduction to Fluorescence and Photophysics. In: Mary-Ann Mycek P, Brian W., ed. Handbook of Biomedical Fluorescence. Basel, Switzerland: Marcel Dekker, Inc.; 2003. 105. Zeng HaM, C. Fluorescence Spectroscopy and Imaging for Skin Cancer Detection and Evaluation. In: Mary-Ann Mycek P, Brian W., ed. Handbook of Biomedical Fluorescence. Basel, Switzerland: Marcel Dekker, Inc.; 2003. 106. Lakowicz JR. Principles of Fluorescence Spectroscopy. Chapter 3. third edition ed: Springer-Verlag New York Inc.; 2006. Accessed September 21, 2014. 107. Jacobkhed. Jablonski diagram of absorbance, non-radiative decay, and fluorescence. In: Jablonski_Diagram_of_Fluorescence_Only.png, ed. http://commons.wikimedia.org/wiki/File:Jablonski_Diagram_of_Fluorescence_O nly.png2012. 108. Kim YJ, Jeong CB, Hwang JM, Yang HK, Lee SH, Kim KG. New parametric imaging method with fluorescein angiograms for detecting areas of capillary nonperfusion. Healthcare informatics research. Jul 2014;20(3):191-198.

151

109. Methodologies to diagnose and monitor dry eye disease: report of the Diagnostic Methodology Subcommittee of the International Dry Eye WorkShop (2007). Ocul Surf. Apr 2007;5(2):108-152. 110. Romanchuk KG. Fluorescein. Physiochemical factors affecting its fluorescence. Surv Ophthalmol. Mar-Apr 1982;26(5):269-283. 111. Georgakoudi I, M Feld, M Muller. Intrinsic Fluorescence Spectroscopy of Biological Tissue. In: Mary-Ann Mycek P, Brian W., ed. Handbook of Biomedical Fluorescence. New York, NY: Marcel Dekker, Inc.; 2003. 112. Richards-Kortum R, Sevick-Muraca E. Quantitative optical spectroscopy for tissue diagnosis. Annual review of physical chemistry. 1996;47:555-606. 113. Yin D. Biochemical basis of lipofuscin, ceroid, and age pigment-like fluorophores. Free Radic Biol Med. 1996;21(6):871-888. 114. Esterbauer H, Schaur RJ, Zollner H. Chemistry and biochemistry of 4- hydroxynonenal, malonaldehyde and related aldehydes. Free Radic Biol Med. 1991;11(1):81-128. 115. Krohne TU, Stratmann NK, Kopitz J, Holz FG. Effects of lipid peroxidation products on lipofuscinogenesis and autophagy in human retinal pigment epithelial cells. Exp Eye Res. Mar 2010;90(3):465-471. 116. Choudhry N, Giani A, Miller JW. Fundus autofluorescence in geographic atrophy: a review. Semin Ophthalmol. Sep-Nov 2010;25(5-6):206-213. 117. Eldred GE, Katz ML. Fluorophores of the human retinal pigment epithelium: separation and spectral characterization. Exp Eye Res. Jul 1988;47(1):71-86. 118. Kennedy CJ, Rakoczy PE, Constable IJ. Lipofuscin of the retinal pigment epithelium: a review. Eye (Lond). 1995;9 ( Pt 6):763-771. 119. Steven P, Muller M, Koop N, Rose C, Huttmann G. Comparison of Cornea Module and DermaInspect for noninvasive imaging of ocular surface pathologies. J Biomed Opt. Nov-Dec 2009;14(6):064040. 120. Chai D, Gaster RN, Roizenblatt R, Juhasz T, Brown DJ, Jester JV. Quantitative assessment of UVA-riboflavin corneal cross-linking using nonlinear optical microscopy. Invest Ophthalmol Vis Sci. Jun 2011;52(7):4231-4238. 121. Rajadhyaksha MaG, S. Real-Time In Vivo Confocal Fluorescence Microscopy. In: Mary-Ann Mycek P, Brian W., ed. Handbook of Biomedical Fluorescence. New York, NY: Marcel Dekker, Inc.; 2003. 122. Delori F, Greenberg JP, Woods RL, et al. Quantitative measurements of autofluorescence with the scanning laser ophthalmoscope. Invest Ophthalmol Vis Sci. 2011;52(13):9379-9390. 123. Kruger A, Hovakimyan M, Ramirez Ojeda DF, et al. Combined nonlinear and femtosecond confocal laser-scanning microscopy of rabbit corneas after photochemical cross-linking. Invest Ophthalmol Vis Sci. Jun 2011;52(7):4247- 4255. 124. Steven P, Hovakimyan M, Guthoff RF, Huttmann G, Stachs O. Imaging corneal crosslinking by autofluorescence 2-photon microscopy, second harmonic generation, and fluorescence lifetime measurements. J Cataract Refract Surg. Dec 2010;36(12):2150-2159. 152

125. Utine CA, Tatlipinar S, Altunsoy M, Oral D, Basar D, Alimgil LM. Autofluorescence imaging of pingueculae. Br J Ophthalmol. Mar 2009;93(3):396- 399. 126. McKnight CM, Sherwin JC, Yazar S, et al. Myopia in Young Adults Is Inversely Related to an Objective Marker of Ocular Sun Exposure: The Western Australian Raine Cohort Study. Am J Ophthalmol. Jul 26 2014. 127. Ooi JL, Sharma NS, Papalkar D, et al. Ultraviolet fluorescence photography to detect early sun damage in the eyes of school-aged children. Am J Ophthalmol. Feb 2006;141(2):294-298. 128. Ooi JL, Sharma NS, Sharma S, et al. Ultraviolet fluorescence photography: patterns in established pterygia. Am J Ophthalmol. Jan 2007;143(1):97-101. 129. Sherwin JC, Hewitt AW, Coroneo MT, Kearns LS, Griffiths LR, Mackey DA. The association between time spent outdoors and myopia using a novel biomarker of outdoor light exposure. Invest Ophthalmol Vis Sci. Jul 2012;53(8):4363-4370. 130. Sherwin JC, Hewitt AW, Kearns LS, Coroneo MT, Griffiths LR, Mackey DA. Distribution of conjunctival ultraviolet autofluorescence in a population-based study: the Norfolk Island Eye Study. Eye (Lond). Jul 2011;25(7):893-900. 131. Sherwin JC, Hewitt AW, Kearns LS, Griffiths LR, Mackey DA, Coroneo MT. The association between pterygium and conjunctival ultraviolet autofluorescence: The Norfolk Island Eye Study. Acta Ophthalmol. Dec 16 2011. 132. Sherwin JC, McKnight CM, Hewitt AW, Griffiths LR, Coroneo MT, Mackey DA. Reliability and validity of conjunctival ultraviolet autofluorescence measurement. Br J Ophthalmol. Jun 2012;96(6):801-805. 133. Wolffsohn JS, Drew T, Sulley A. Conjunctival UV autofluorescence - Prevalence and risk factors. Cont Lens Anterior Eye. Aug 7 2014. 134. Asawanonda P, Taylor CR. Wood's light in dermatology. International journal of dermatology. Nov 1999;38(11):801-807. 135. Sherwin JC, Hewitt AW, Kearns LS, Griffiths LR, Mackey DA, Coroneo MT. The association between pterygium and conjunctival ultraviolet autofluorescence: the Norfolk Island Eye Study. Acta Ophthalmol. Jun 2013;91(4):363-370. 136. Jones LA, Sinnott LT, Mutti DO, Mitchell GL, Moeschberger ML, Zadnik K. Parental history of myopia, sports and outdoor activities, and future myopia. Invest Ophthalmol Vis Sci. Aug 2007;48(8):3524-3532. 137. Peiretti E, Mandas A, Cocco P, et al. Glucose-6-phosphate-dehydrogenase deficiency as a risk factor for pterygium. Invest Ophthalmol Vis Sci. Jun 2010;51(6):2928-2935. 138. Schiffman RM, Christianson MD, Jacobsen G, Hirsch JD, Reis BL. Reliability and validity of the Ocular Surface Disease Index. Arch Ophthalmol. May 2000;118(5):615-621. 139. Lemp MA. Report of the National Eye Institute/Industry workshop on Clinical Trials in Dry Eyes. Clao J. Oct 1995;21(4):221-232. 140. Moehrle M, Soballa M, Korn M. UV exposure in cars. Photodermatology, photoimmunology & photomedicine. Aug 2003;19(4):175-181.

153

141. Su H, Gornitsky M, Velly AM, Yu H, Benarroch M, Schipper HM. Salivary DNA, lipid, and protein oxidation in nonsmokers with periodontal disease. Free Radic Biol Med. Apr 1 2009;46(7):914-921. 142. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. Feb 8 1986;1(8476):307-310. 143. Mann A, Tighe B. Contact lens interactions with the tear film. Exp Eye Res. Dec 2013;117:88-98. 144. Doughty MJ. Objective assessment of squamous metaplasia of conjunctival epithelial cells as associated with soft contact lens wear versus non-lens wearers. Cornea. Oct 2014;33(10):1095-1102. 145. Fraunfelder FW, Cabezas M. Treatment of recurrent corneal erosion by extended- wear bandage contact lens. Cornea. Feb 2011;30(2):164-166. 146. Giblin FJ LL, Leverenz VR, Dang L. A class I (senofilcon A) soft contact lens prevents UVB-induced lens opacification in the rabbit in vivo. Invest Ophthalmol Vis Sci. 2009;50(E-Abstrarct 2536). 147. Wakamatsu TH, Dogru M, Matsumoto Y, et al. Evaluation of lipid oxidative stress status in Sjogren syndrome patients. Invest Ophthalmol Vis Sci. Jan 2013;54(1):201-210. 148. Li Z, Choi JH, Oh HJ, Park SH, Lee JB, Yoon KC. Effects of eye drops containing a mixture of omega-3 essential fatty acids and hyaluronic acid on the ocular surface in desiccating stress-induced murine dry eye. Curr Eye Res. Sep 2014;39(9):871-878. 149. Muhlhausler BS, Cook-Johnson R, James M, Miljkovic D, Duthoit E, Gibson R. Opposing effects of omega-3 and omega-6 long chain polyunsaturated Fatty acids on the expression of lipogenic genes in omental and retroperitoneal adipose depots in the rat. Journal of nutrition and metabolism. 2010;2010. 150. Demir MN, Demir ZA, Yalcin Tok O, et al. Oxidative stress of intracameral lidocaine and levobupivacaine on ocular tissues. Br J Ophthalmol. Aug 2010;94(8):1083-1087. 151. Tibrewal S, Sarkar J, Jassim SH, et al. Tear fluid extracellular DNA: diagnostic and therapeutic implications in dry eye disease. Invest Ophthalmol Vis Sci. Dec 2013;54(13):8051-8061. 152. Matsumoto Y, Ogawa Y, Yoshida R, Shimamori A, Kasai H, Ohta H. The stability of the oxidative stress marker, urinary 8-hydroxy-2'- deoxyguanosine (8- OHdG), when stored at room temperature. J Occup Health. 2008;50(4):366-372. 153. Ng J, Coroneo MT, Wakefield D, Di Girolamo N. Ultraviolet radiation and the role of matrix metalloproteinases in the pathogenesis of ocular surface squamous neoplasia. Invest Ophthalmol Vis Sci. Dec 2008;49(12):5295-5306. 154. Dong N, Li W, Lin H, et al. Abnormal epithelial differentiation and tear film alteration in pinguecula. Invest Ophthalmol Vis Sci. Jun 2009;50(6):2710-2715. 155. Clear AS, Chirambo MC, Hutt MS. Solar keratosis, pterygium, and squamous cell carcinoma of the conjunctiva in Malawi. Br J Ophthalmol. Feb 1979;63(2):102- 109.

154

156. Helfrich YR, Sachs DL, Voorhees JJ. Overview of skin aging and photoaging. Dermatol Nurs. Jun 2008;20(3):177-183; quiz 184. 157. Raizada IN, Bhatnagar NK. Pinguecula and pterygium (a histopathological study). Indian J Ophthalmol. Jul 1976;24(2):16-18. 158. Benlloch-Navarro S, Franco I, Sanchez-Vallejo V, Silvestre D, Romero FJ, Miranda M. Lipid peroxidation is increased in tears from the elderly. Exp Eye Res. Oct 2013;115:199-205. 159. Chan CM, Liu YP, Tan DT. Ocular surface changes in pterygium. Cornea. Jan 2002;21(1):38-42. 160. Wondrak GT, Jacobson MK, Jacobson EL. Endogenous UVA-photosensitizers: mediators of skin photodamage and novel targets for skin photoprotection. Photochemical & photobiological sciences : Official journal of the European Photochemistry Association and the European Society for Photobiology. Feb 2006;5(2):215-237. 161. Trautinger F, Mazzucco K, Knobler RM, Trenz A, Kokoschka EM. UVA- and UVB-induced changes in hairless mouse skin collagen. Archives of dermatological research. 1994;286(8):490-494. 162. Brummer G, Littlechild S, McCall S, Zhang Y, Conrad GW. The role of nonenzymatic glycation and carbonyls in collagen cross-linking for the treatment of keratoconus. Invest Ophthalmol Vis Sci. Aug 2011;52(9):6363-6369. 163. Gillies R, Zonios G, Anderson RR, Kollias N. Fluorescence excitation spectroscopy provides information about human skin in vivo. J Invest Dermatol. Oct 2000;115(4):704-707. 164. Lin SJ, Wu R, Jr., Tan HY, et al. Evaluating cutaneous photoaging by use of multiphoton fluorescence and second-harmonic generation microscopy. Opt Lett. Sep 1 2005;30(17):2275-2277. 165. Young TL, Hawthorne F, Feng S, et al. Whole genome expression profiling of normal human fetal and adult ocular tissues. Exp Eye Res. Nov 2013;116:265- 278.

155

Appendix A: Questionnaires

156

Figure 33. Questionnaire estimating proportion of time spent using personal ocular protective measures and time outdoors.

157

Figure 34. Questionnaire estimating hours spent outdoors in seven activity categories.

158