The Role of Vitamin D in Inflammatory Bowel Disease Pathogenesis and Severity

THE ROLE OF VITAMIN D IN INFLAMMATORY BOWEL DISEASE PATHOGENESIS AND SEVERITY

by Berkeley Limketkai

A dissertation submitted to Johns Hopkins University in conformity with the requirements for the degree of Doctor of Philosophy

Baltimore, Maryland

March, 2017

ABSTRACT

Vitamin D has traditionally been known for its role in bone metabolism, but more

recently been implicated in immune function. Emerging evidence has further suggested

that vitamin D may be involved in inflammatory bowel disease (IBD) pathogenesis and

activity. Studies found lower vitamin D levels among IBD patients, particularly in the

setting of increased disease activity, implying that vitamin D influenced IBD. However,

interpretation of these findings is challenged by the concept of reverse causation, where

intestinal inflammation is already known to reduce vitamin D levels. These studies have

therefore not established vitamin D deficiency as a cause or effect of IBD in humans. The

principal objective of this thesis was to clarify the causal role of vitamin D in IBD

pathogenesis and severity, while employing diverse methodologic approaches to overcome the issue of reverse causation.

The first study was a case-control comparison of vitamin D levels from sera that

were prospectively collected before and around the time of diagnosis of Crohn’s disease

(CD). This longitudinal analysis permitted the evaluation of vitamin D levels as

precursors to CD pathogenesis, while reducing interference by reverse causation. Vitamin

D levels were similar between cases and controls prior to diagnosis, but significantly

lower among cases around the time of diagnosis. The second study was an ecologic study

that evaluated the association of ultraviolet (UV) exposure – a surrogate marker of

vitamin D that was not influenced by IBD – and IBD hospitalization severity. Low UV

exposures were associated with increased rates of hospitalization, prolonged

hospitalization, and need for bowel surgeries. The third study was a systematic review of

randomized trials that evaluated the effect of vitamin D supplementation on IBD activity.

Randomization in these trials would have theoretically permitted assessment of the effect of vitamin D on IBD independent of disease activity. The meta-analysis of 4 studies with

12-month follow-up showed a trend toward fewer clinical relapses, but this was not statistically significant. Nonetheless, most studies had small sample sizes, significant risk of bias, and substantial methodologic heterogeneity.

In conclusion, this thesis investigation did not find vitamin D to significantly influence CD pathogenesis. There was nonetheless some evidence to suggest that vitamin

D may influence IBD activity. Further investigation with high-quality and larger randomized trials of vitamin D intervention in IBD is warranted.

Committee Chair James A. Tonascia, PhD, Professor of Biostatistics and Epidemiology

Committee Chair, Emeritus Theodore M. Bayless, MD, Professor Emeritus of Medicine

Research Advisor Steven R. Brant, MD, Professor of Medicine

Academic Advisor Craig W. Hendrix, MD, Professor of Medicine

Thesis Readers N. Franklin Adkinson, MD, Professor of Medicine David M. Levine, MD, Professor of Medicine Eliseo Guallar, MD, DrPH, Professor of Epidemiology Susan M. Hutfless, PhD, Assistant Professor of Medicine

iii

ACKNOWLEDGEMENTS

This dissertation is the culmination of the selfless contributions of the whole “village” that has helped me along my path of clinical and research training.

I express my deepest gratitude to Dr. Theodore Bayless who has shown immense altruism in mentorship since my early internship days through my junior faculty years. He has served as a preeminent role model whose life and passion for inflammatory bowel disease (IBD) inspired me to eventually subspecialize in the field. As he has played a strongly defining role in my budding medical career, the opportunity to pursue the Theodore M. Bayless Fellowship in Inflammatory Bowel Diseases was an incredible honor that I will cherish throughout my career.

Dr. Francis Giardiello was formally my gastroenterology fellowship director, although he also served as a career mentor during my residency and fellowship training. I truly appreciate his interest in my personal and professional development over the years. My early conversations with him while still an Osler resident solidified my desire to someday pursue further studies through the Graduate Training Programs in Clinical Investigation (GTPCI). Subsequently, as a gastroenterology fellow, Dr. Giardiello guided me through the intricacies of balancing gastroenterology fellowship, my Ph.D. studies, and in my last year at Johns Hopkins, the fellowship in IBD.

Drs. N. Franklin Adkinson and Craig Hendrix are pillars of my GTPCI studies. I am indebted to them for their great generosity in time and advice, particularly as I navigated through my Ph.D. studies and thesis research. They have also provided invaluable advice on a diversity of academic matters, such as grant funding, promotions, politics, and building a research career.

The foundations of this dissertation and my early IBD research career are credited to Drs. Steven Brant and Susan Hutfless. They served as my primary research mentors on this and other research endeavors throughout fellowship. Despite their myriad of responsibilities and very busy schedules, they have invested innumerable hours, days, and weeks over the past six years in teaching me how to conduct research.

While serving as the chair of my Ph.D. thesis committee, Dr. James Tonascia has demonstrated admirable leadership, humor, and kindness. He shepherded me along the path leading up to the final defense with a genuine interest in my success. Despite our nascent interactions over the past few months, I feel as though he had already mentored me for years. I hope to continue learning from him well beyond my graduate education.

Finally, the strongest drivers behind the person who I am …

To my parents, there is no measure for how I can repay you. Your unconditional love and sacrifice throughout my childhood have molded me into the person I have become. I took risks and steadfastly pursued my passions when you helped me overcome internal doubts. I learned to rise from failure when you pulled me out of the pit of discouragement or defeat. I aimed for the skies when you reminded me that I could dream bigger than my limited imagination. As such, I dedicate all my successes to you.

To my two sons, Aidan and Ethan, you are still too young to appreciate the inexpressible joy (and stress) you have brought your mother and me. When the days buried in work may feel endless at times, you remind me how rich my life truly is. Thank you for your patience during the many moments I could not play with you, while I focused on my thesis research and this dissertation. I hope to someday inspire you to far exceed what I have done.

To my wife, DanDan, just as I have repeated over our 16 years of marriage, I feel greatly blessed to have you in my life. You never cease to amaze me with your incredible love, patience, and selflessness. Throughout my impossible balancing act of life as a husband, father, son, physician, and scientific explorer, you have helped me pick up the pieces and have encouraged me to persevere. While I worked on this dissertation, you minimized any distractions, entertained our sons when I was not available, kept the house in order, and repaired any disorder. This would not have been possible without you. Thank you.

TABLE OF CONTENTS

ABSTRACT ...... II ACKNOWLEDGEMENTS ...... IV TABLE OF CONTENTS ...... VI LIST OF TABLES ...... VIII LIST OF FIGURES ...... IX LIST OF ABBREVIATIONS ...... X CHAPTER I. INTRODUCTION ...... 1 VITAMIN D SYNTHESIS AND ACTION...... 2 VITAMIN D AND THE IMMUNE SYSTEM ...... 3 VITAMIN D AND AUTOIMMUNE DISEASE ...... 5 VITAMIN D AND INFLAMMATORY BOWEL DISEASE ...... 7 Animal Models ...... 7 Genetic Evidence ...... 8 Epidemiologic Evidence ...... 10 Intervention Trials ...... 11 CAUSE OR EFFECT? ...... 11 SUMMARY ...... 13 FIGURES ...... 15 CHAPTER II. VITAMIN D AND CROHN’S DISEASE PATHOGENESIS ...... 17 INTRODUCTION ...... 17 METHODS ...... 18 Study Population ...... 18 Data Abstraction ...... 20 Vitamin D Measurements ...... 20 Genotyping ...... 21 Statistical Analyses ...... 22 Ethics and Study Subject Anonymity ...... 23 RESULTS ...... 24 Vitamin D Levels and Incident CD ...... 24 Gene Polymorphism Distributions ...... 25 Gene Interactions ...... 26 Sensitivity Analyses ...... 27 DISCUSSION ...... 28 TABLES ...... 34 FIGURES ...... 39 CHAPTER III. VITAMIN D AND HOSPITALIZATION SEVERITY ...... 44 INTRODUCTION ...... 44 METHODS ...... 45 UV Exposure ...... 45 Hospitalization Data ...... 46 IBD Severity ...... 47 Severity in Non-IBD Hospitalizations ...... 48 Potential Confounders ...... 48 Statistical Analysis ...... 49 RESULTS ...... 49 Rates of Hospitalizations ...... 49

Rates of Hospitalizations by UV Exposure ...... 50 Bowel Surgeries ...... 50 Lengths of Stay ...... 51 Death ...... 51 Severity of Non-IBD Hospitalizations ...... 52 Sensitivity analysis ...... 52 DISCUSSION ...... 53 TABLES ...... 57 FIGURES ...... 58 CHAPTER IV. VITAMIN D AS TREATMENT FOR INFLAMMATORY BOWEL DISEASE ...... 62 INTRODUCTION ...... 62 Description of the Condition ...... 62 Description of the Intervention ...... 63 How the Intervention Might Work ...... 64 Importance of this Review ...... 64 METHODS ...... 65 Selection Criteria ...... 65 Search Strategy ...... 66 Data Extraction ...... 67 Assessment of Risk of Bias ...... 68 Statistical Analysis ...... 68 RESULTS ...... 69 Search Results ...... 69 Excluded Studies ...... 71 Risk of Bias in Included Studies ...... 72 Effects of interventions ...... 74 DISCUSSION ...... 78 TABLES ...... 81 FIGURES ...... 110 CHAPTER V. CONCLUSION...... 114 REFERENCES ...... 118 APPENDICES ...... 130 CURRICULUM VITAE ...... 133

vii

LIST OF TABLES

Table 2-1 Patient characteristics at the time of diagnosis or index time 34 Table 2-2 Gene allele distributions 35 Table 2-3 Risk genotype distributions 36 Table 2-4 Gene polymorphisms and incident Crohn’s disease 37 Table 2-5 Akaike Information Criteria for gene interaction models 38 Table 3-1 Demographics of Crohn’s disease, ulcerative colitis, and non- 57 inflammatory bowel disease (IBD) hospitalizations Table 4-1 Characteristics of included studies 81 Table 4-2 Characteristics of excluded studies 94 Table 4-3 Disease activity in Crohn’s disease studies 95 Table 4-4 Disease activity in ulcerative colitis studies 97 Table 4-5 Inflammation in Crohn’s disease studies 98 Table 4-6 Inflammation in ulcerative colitis studies 100 Table 4-7 Vitamin D levels in Crohn’s disease studies 101 Table 4-8 Vitamin D levels in ulcerative colitis studies 103 Table 4-9 Adverse events in Crohn’s disease studies 104 Table 4-10 Adverse events in ulcerative colitis studies 106 Table 4-11 Quality of life in Crohn’s disease studies 108 Table 4-12 Quality of life in ulcerative colitis studies 109

viii

LIST OF FIGURES

Figure 1-1 Interaction between genetic, host, and environmental factors in 15 the pathogenesis of inflammatory bowel disease Figure 1-2 Vitamin D synthesis and hydroxylation pathway 16 Figure 2-1 Time points for collected serum samples 39 Figure 2-2 Vitamin D levels and relative odds of incident Crohn’s disease 40 (CD) at each time point Figure 2-3 VDR and DBP mutations and the odds of incident Crohn’s 41 disease (CD) across vitamin D tertiles Figure 2-4 Crohn’s disease (CD) risk mutations and the odds of incident CD 42 across vitamin D tertiles Figure 3-1 Directed acyclic graph of ultraviolet (UV) exposure, vitamin D, 58 and inflammatory bowel disease (IBD) severity Figure 3-2 Rates of Crohn’s disease (CD) and ulcerative colitis (UC) 59 hospitalizations according to ultraviolet (UV) exposure Figure 3-3 Rates of hospitalization severity and death according to 60 ultraviolet (UV) exposure Figure 3-4 Rates of other abdominal surgeries according to ultraviolet (UV) 61 exposure Figure 4-1 Study flow diagram 110 Figure 4-2 Risk of bias assessment 111 Figure 4-3 Forest plot comparison (random effects model) of vitamin D at 112 different doses or with placebo in studies with 12-month follow- up Figure 4-4 Forest plot comparison (fixed effects model) of vitamin D at 113 different doses or with placebo in studies with 12-month follow- up

LIST OF ABBREVIATIONS

1,25(OH)2D 1,25-dihydroxyvitamin D

7-DHC 7-dehydrocholesterol

25(OH)D 25-hydroxyvitamin D

AHRQ Agency for Healthcare Research and Quality

AIC Akaike Information Criterion

BIC Bayesian Information Criterion

CAMP Cathelicidin antimicrobial peptide

CD Crohn’s disease

CDAI Crohn’s Disease Activity Index

CI Confidence interval

CRP C-reactive protein

DBP Vitamin D binding protein dF Degrees of freedom

DSS Dextran sodium sulfate

EAE Experimental autoimmune encephalomyelitis

EC-IBD European Collaborative Study on Inflammatory Bowel Disease

ESR Erythrocyte sedimentation rate

GMCSF Granulocyte-macrophage colony-stimulating factor

HBI Harvey-Bradshaw Index

HCUP Healthcare Cost and Utilization Project

IBD Inflammatory bowel disease

ICD-9-CM International Classification of Diseases, Ninth Revision, Clinical

Modification

ICTR Institute for Clinical and Translational Research

IFN-γ Interferon-gamma

IL Interleukin

IQR Interquartile range

IU International units

LPS Lipopolysaccharide

MS Multiple sclerosis

NHANES National Health and Nutrition Examination Survey

NIS Nationwide Inpatient Sample

NOAA National Oceanic and Atmospheric Administration

OR Odds ratio

PCDAI Pediatric Crohn’s Disease Activity Index

Pre-D3 Pre-vitamin D3

RA Rheumatoid arthritis

RCT Randomized controlled trial

RNA Ribonucleic acid

RR Relative rate

RXR Retinoid-X receptor

SD Standard deviation

SEER Surveillance, Epidemiology, and End Results

SLE Systemic lupus erythematosus

SNP Single nucleotide polymorphism

Th T helper cell

TLR Toll-like receptor

TNBS 2,4,6-trinitrobenzene sulfonic acid

TNF-α Tumor necrosis factor-alpha

Treg Regulatory T cell

UC Ulcerative colitis

U.S. United States

UV Ultraviolet

UVB Ultraviolet B

VDR Vitamin D receptor

VDRE Vitamin D response element

xii

CHAPTER I. INTRODUCTION

Inflammatory bowel disease (IBD), whose major subtypes include Crohn’s disease (CD)

and ulcerative colitis (UC), is characterized by chronic relapsing and remitting

inflammation of the gastrointestinal tract. The underlying causes of IBD are yet unclear,

although the current paradigm of pathogenesis involves the interaction between genetic,

host, and environmental risk factors (Figure 1-1). Early observations had suggested a genetic determinant of disease when first-degree relatives of IBD patients were found to possess an increased risk of developing CD or UC.1-3 However, the risk conveyed by

identified genetic mutations was small and inconsistent across populations.4-7 Moreover,

discordance in monozygotic twin studies indicated that non-heritable factors strongly

contributed to IBD pathogenesis.2,3 Epidemiologic studies later suggested several

potential environmental risk factors for incident IBD, such as gastrointestinal infections,

antibiotic use, oral contraceptives, and nutrition.8-11 Similarly, vitamin D has been

proposed as a nutrient risk factor of IBD. Although traditionally known for its

involvement in calcium and phosphorus homeostasis, vitamin D has more recently been

implicated in immune regulation. Meanwhile, emerging epidemiologic evidence

demonstrated an association between vitamin D deficiency and diverse autoimmune

disorders, with further suggestions that it may likewise be an environmental factor in IBD

pathogenesis and severity.

VITAMIN D SYNTHESIS AND ACTION

Vitamin D comprises a family of fat-soluble secosteroid hormones that primarily includes two major forms: vitamin D2 (ergocalciferol) and vitamin D3 (cholecalciferol). For

humans, vitamin D can either be accumulated through dietary sources or synthesized de

novo from sterol precursors. Since ergosterol, the precursor of vitamin D2, is typically

found in yeast and fungi (ergot), humans accumulate ergocalciferol through consumption

of water plants, fish that feed on phytoplankton, fortified foods, or dietary

supplements.12,13 For cholecalciferol, ultraviolet B (UVB) exposure (280 to 315 nm) in

the epidermis leads to conversion of 7-dehydrocholesterol (7-DHC) to pre-vitamin D3

14,15 (pre-D3), which subsequently undergoes rapid thermal isomerization to vitamin D3.

Photoconversion of 7-DHC to pre-D3 can reach a maximum concentration within 15 minutes of UVB exposure.16 In the case of prolonged sunlight exposure, excess 7-DHC is

shunted to biologically inactive photoisomers: tachysterol and lumisterol. Excess pre-

vitamin D3 and vitamin D3 in the epidermis can also undergo photodegradation, so

prolonged sunlight exposure does not lead to vitamin D toxicity.17

Due to its hydrophobic nature, vitamin D circulates in the bloodstream while bound to the vitamin D binding protein (DBP) and, to a lesser degree, albumin.18,19

Vitamin D is first hydroxylated in the liver by the mitochondrial (CYP27A1) and

microsomal (CYP2R1) cytochrome P450 isoforms to form 25-hydroxyvitamin D

(25[OH]D) before being converted in the renal tubules by CYP27B1 to its active

20-22 metabolite: 1,25-dihydroxyvitamin D (1,25[OH]2D) (Figure 1-2).

Vitamin D exerts its biologic effect through the classical activation of

transcription and through more rapid-acting membrane-bound receptors. The active

23 metabolite 1,2(OH)2D initially binds the vitamin D receptor (VDR). VDR is a member

of the steroid receptor superfamily of ligand-activated transcription factors, where once

activated, it translocates into the nucleus to form a heterodimer with the retinoid-X receptor (RXR).24 The VDR-RXR complex then recruits several coactivator proteins, selectively binds the vitamin D response element (VDRE), and facilitates RNA polymerase II-mediated transcription of respective target genes.25-27 Alternatively, the

1,25(OH)2D hormone is able to generate more rapid extra-genomic effects through interaction with membrane-bound VDR and/or membrane-associated rapid response

steroid binding receptors.28,29

VITAMIN D AND THE IMMUNE SYSTEM

Although vitamin D has traditionally been known for its prominent role in calcium and

phosphorus homeostasis, emerging data have shown it to instead possess other effects.13

In particular, studies in the 1980s began to implicate vitamin D in immune function. The

discovery that 1,25(OH)2D induced the differentiation of murine myeloid leukemia cells

to macrophages prompted a search for its receptor in these cells.30 VDR was subsequently

found in a broad host of immunologic cells, with constitutive expression in diverse

hematopoietic cells, including myeloblasts, promyelocytes, monoblasts, and

macrophages.31-33 Although resting T lymphocytes have negligible VDR levels, VDR

expression is inducible by T cell activation.32,34

Evidence of 25(OH)D-1-α-hydroxylase (CYP27B1) in macrophages, dendritic

cells, and B cells,35-37 further suggested involvement of vitamin D in the regulation of the

immune system. That is, local conversion of 25(OH)D to the active 1,25(OH)2D

represents an autocrine or paracrine mechanism of immune cell signaling. Furthermore,

expression of 1-α-hydroxylase in these cells are synergistically induced by interferon-

gamma (IFN-γ) and Toll-like receptor activation,38,39 which are often associated with

immune system function. This mechanism of induction differs from the role of renal 1-α-

hydroxylase in calcium homeostasis, which is regulated by the parathyroid hormone and

1,25(OH)2D.

Of interest in IBD, vitamin D possesses inhibitory effects on immunity. In

monocytes, 1,25(OH)2D can suppress TLR2 and TLR4 expression, leading to an

impaired inflammatory response to lipopolysaccharide (LPS).40 There is additionally

dose-dependent suppression of LPS-induced tumor necrosis factor-alpha (TNF-α) production. This downregulation of TLR expression and hyporesponsiveness to bacterial cell wall antigens are reversible through VDR antagonism.

In lymphocytes, vitamin D’s modulatory effects occur through several putative mechanisms. First, vitamin D impairs lymphocyte proliferation and differentiation.37,41-43

It further inhibits differentiation and maturation of dendritic cells, 44-46 which indirectly

leads to T cell anergy by impaired activation of alloreactive T cells.46,47 Second, vitamin

D exerts an immunomodulatory effect by expansion of regulatory T (Treg) cells.48-51 In

2,4,6-trinitrobenzene sulfonic acid (TNBS)-treated mice used to produce colitis,

exogenous 1,25(OH)2D administration shifts the T helper (Th) 1 / Th17 profile to Th2

and Treg cells.52 Third, vitamin D modulates the release of inflammatory cytokines.

Vitamin D promotes synthesis of the anti-inflammatory cytokine interleukin (IL)-10,

while inhibiting synthesis of pro-inflammatory cytokines (e.g., IL-2, IL-6, IL-12, IL-17,

IL-22, IL-23, IFN-γ , TNF-α).40,42,52-56 It also directly inhibits transcription of the

granulocyte-macrophage colony-stimulating factor (GMCSF) gene.57

Vitamin D is additionally involved in innate immunity through induced

expression of pattern recognition receptors and antimicrobial peptides. In a state of

epidermal injury, 1,25(OH)2D stimulates transcription of genes for TLR2 and its co-

receptor CD14 in keratinocytes.58 Toll-like receptors (TLR) belong to a family of molecular pattern recognition receptors that sense microbial components and signal downstream expression of antimicrobial peptides. TLR activation in macrophages

39 induces expression of VDR, CYP27B1, and cathelicidin. Moreover, the 1,25(OH)2D

pathway directly activates transcription of human cathelicidin antimicrobial peptide

(CAMP) and defensin β2 (DEFB2/HBD2).59,60 Inversely, a defect or deficiency in VDR,

39,58 CYP27B1, or 1,25(OH)2D can impair induced expression of cathelicidin.

VITAMIN D AND AUTOIMMUNE DISEASE

Vitamin D has been implicated in the pathogenesis of several autoimmune diseases. For

one, experimental autoimmune encephalomyelitis (EAE) is an inducible murine model of

multiple sclerosis (MS) that involves a Th1 response against myelin basic protein and

other autoantigens in the central nervous system. Administration of 1,25(OH)2D was found to prevent development or reverse progression of EAE.61,62 Interestingly, but not surprisingly, severe vitamin D deficiency can also interfere with the pathogenesis of EAE in mice.63 This seeming paradox may stem from the impairment of cell-mediated

immunity in a state of severe vitamin D deficiency.64

In humans, a prospective evaluation of dietary vitamin D intake in the Nurses’

Health Study and Nurses’ Health Study II revealed that women in the highest quintile of vitamin D intake had a 33% lower age-adjusted risk of MS.65 Likewise, intake of vitamin

D supplements was associated with a 41% reduction in risk of MS. These findings were later corroborated in another prospective study that showed a dose-response trend, where higher serum 25(OH)D levels were associated with decreased odds of incident MS.66

Restriction site polymorphisms on the VDRE gene have also been linked with increased risk of MS. In a study of 77 MS patients compared with 95 healthy controls, there was a greater prevalence of the b allele (92.9% vs. 84.2%; P = 0.01) and homozygote bb genotype (85.7% vs. 71.6%; P = 0.03) of the Bsm1 polymorphism among the MS patients.67 There was also increased prevalence of the AA genotype (27.3% vs. 9.5%; P <

0.01) of the Apa I polymorphism among MS patients.68

Similar to the Nurses’ Health Study for MS, a large prospective cohort of 29,520 women (ages 55-69), based on the Iowa Women’s Health Study, revealed that women with greater intake of dietary and supplementary vitamin D had a lower risk of rheumatoid arthritis (RA).69 Vitamin D supplementation was additionally shown to inhibit progression of infection- or collagen-induced arthritis in mice.70

The first reported observation of vitamin D deficiency in systemic lupus erythematosus (SLE) included 7 out of 12 prednisone-treated adolescents who were

71 found to have low 1,25(OH)2D levels. Since that report in 1979, there have been over

20 additional published studies, mostly correlating vitamin D deficiency with SLE and/or disease activity.72 The studies were all small, which may have limited the power in the few that did not find statistical significance. Nonetheless, the largest study had a multi-

center European and Israeli cohort of 378 SLE patients in whom the investigators found

an inverse relationship between 25(OH)D levels and disease activity scores.73

The evidence of vitamin D’s role in other autoimmune diseases is less robust,

largely due to a dearth of data, but still suggests a relationship between vitamin D

deficiency and disease pathogenesis. For autoimmune diabetes, 1,25(OH)2D treatment

was shown to arrest development or progression of diabetes in NOD mice, an animal

model for human autoimmune diabetes (insulin-dependent diabetes mellitus or juvenile diabetes).48,74,75 There are also emerging data for a role of vitamin D in other

immunologic conditions, such as asthma and atopy.76-78 Taken together, the accumulating

in vitro, in vivo, and epidemiologic data implicate vitamin D in immune-related disease.

VITAMIN D AND INFLAMMATORY BOWEL DISEASE

The relationship between vitamin D and IBD has been explored in animal models, in

vitro assays, and human observational studies. Although none of the individual studies

confirms that vitamin D deficiency leads to IBD in humans, the totality of evidence is

compelling.

Animal Models

In the IL-10 knockout mouse model, the mice may spontaneously develop enterocolitis.

Vitamin D deficiency in these mice accelerates the development of enterocolitis, while

also leading to more severe diarrhea, wasting disease, and death.79 However, treatment

with 1,25(OH)2D leads to amelioration of these symptoms, and treatment of calcitriol in

TNBS-induced colitis is associated with a decrease in Th1/Th17 cells in favor of Th2 and

7 regulatory T cells.52 Similarly, administration of a VDR agonist attenuates colitis in IL-10 knockout mice or TNBS-induced colitis.80,81 VDR agonism also reduces expression of pro-inflammatory cytokines and lymphocyte infiltration in the lamina propria of mice with dextran sodium sulfate (DSS)-induced colitis.82

Genetic Evidence

VDR knockouts serve as another model of vitamin D deficiency by ablating a step along the putative pathway of vitamin D’s mechanism of action. As such, VDR and IL-10 double knockout mice have been observed to develop more severe colitis than either

VDR or IL-10 single knockouts.83-85 These findings illustrate that functional vitamin D deficiency can trigger IBD in a susceptible mouse model and that vitamin D deficiency alone may not be sufficient for pathogenesis. This would be consistent with observations in humans where the widespread prevalence of vitamin D deficiency far exceeds the prevalence of IBD.

In humans, genetic studies have elucidated an association between vitamin D- related risk alleles and IBD. In particular, polymorphisms in the VDR gene have been associated with IBD, but appear to vary according to population. In a study of European

Caucasian patients (245 CD, 158 UC, 164 controls), the homozygote tt genotype of the

TaqI polymorphism on the VDR gene was significantly more common in CD (22%) patients than UC patients (12%) or controls (12%) (P = 0.02).86 In a study of Iranian patients (80 CD, 150 UC, 150 controls), the f allele of the FokI polymorphism was more common in CD (49.4%) and UC (36.3%) patients compared with controls (26.7%) (P <

0.01 and P = 0.01, respectively).87 In Jewish Ashkenazi patients (129 CD, 72 UC, 143

controls), the BB genotype of the BsmI polymorphism was associated with UC (21% vs.

11%; P = 0.04), but not CD (14% vs. 11%; P not significant), when compared with

controls.88 A larger study of 1,359 Irish participants (660 IBD patients) did not find any

significant association between VDR gene polymorphisms in FokI, BsmI, ApaI, or TaqI

and IBD.89 While this finding may contradict prior studies, the population was homogenously Irish and prior studies have demonstrated variations in significant polymorphisms based on racial backgrounds. Relevant associations differed among studies, but they were limited by small cohorts. However, in the highly-powered IBD immunochip study, there was genome wide evidence of association establishing a VDR intronic polymorphism (rs11168249) as a risk factor for IBD, although the effect size was small (OR 1.05).90 Homozygotes for the Thr420Lys mutation in the vitamin D binding

protein (DBP) have also been associated with CD and UC, although the true functional

effect of these variations remains unclear.91

A link between vitamin D and CD risk genetics has additionally been

demonstrated through in vitro experiments of monocytic and epithelial cells, where

1,25(OH)2D induced expression of NOD2/CARD15, defensin β2, and cathelicidin in the

presence of muramyl dipeptide, but not in cells with homozygous non-functional NOD2

due to coding mutations (i.e., the effect of these mutations does not arise through

alteration of NOD2 expression).92 Inadequate recognition or clearance of bacterial

antigens may contribute to the abnormal immunologic response in CD, although the

specific mechanisms for NOD2- or ATG16L1-mediated risk of CD are still unclear.

These findings nonetheless indicate a possible mechanism of gene-environment

interaction in CD pathogenesis.

Epidemiologic Evidence

On a population level, epidemiologic observations have revealed a north-south gradient for IBD, where more northern geographic regions with less sunlight exposure have a higher incidence of disease.93-97 The European Collaborative Study on Inflammatory

Bowel Disease (EC-IBD) reported 80% and 40% greater incidence of CD and UC, respectively, when comparing northern and southern European centers.98 Similar latitudinal trends have been noted in individual European countries, such as in France

95,96,99 and Scotland.93 One French study went further to compare surface UV radiation intensity and IBD incidence,96 and finding that low sunlight exposure was associated with an increased incidence of CD, but not UC. Retrospective analyses of the Caris Life

Science database and the Nurses’ Health Study, a prospective cohort of 72,719 women

(ages 40 or above) revealed similar north-south incidence trends in the United

States.100,101 A separate study of the Nurses’ Health Study additionally found an inverse correlation between predicted 25(OH)D levels (indirectly estimated in regression models from measured factors, such as dietary and supplement intake, sunlight exposure, race, and body mass index) and CD incidence.102

The temporal relationship between vitamin D and IBD activity was however unclear in these studies. On a patient level, a collaboration among several medical centers in Boston retrospectively evaluated the effects of vitamin D deficiency and normalization of deficient levels in 3217 IBD patients.103 Consistent with earlier findings, vitamin D level was associated with a two-fold risk of IBD-related surgery and hospital admissions among CD and UC patients. Vitamin D-deficient CD patients who had normalization of their 25(OH)D levels had a reduced risk of surgery, but not hospitalization. However,

vitamin D-deficient UC patients who had normalization of 25(OH)D levels did not

experience a change in their risk of surgery or hospitalization.

Intervention Trials

If vitamin D clinically influenced the natural course of IBD, one would surmise that

vitamin D supplementation could serve as an adjunctive therapy for induction and/or

maintenance of remission. A randomized placebo-controlled trial of 1200 IU of cholecalciferol daily in 94 total CD patients showed improvement in serum 25(OH) levels after 3 months.104 The relapse rate in the cholecalciferol arm was lower, but did not

reach statistical significance (13% vs. 29%; P = 0.06). The study may have been underpowered and/or the effect of vitamin D may be not as strong as hoped. Similarly, a

Hungarian trial of 1,25(OH)2D supplementation in CD patients found improvement in

disease activity indices and C-reactive protein (CRP) at week 6.105 Another study that

used adaptive methods for vitamin D3 supplementation (either 1000 IU daily with

progressive escalation or a fixed 5000 IU daily) in CD patients with mild-moderate

disease found that vitamin D repletion to 40 ng/mL was accompanied by improvement in

25(OH)D levels, disease activity indices, and quality of life scores.106 Most patients

required the maximum allotted 5000 IU of daily cholecalciferol to reach a serum

25(OH)D level of 40 ng/mL.

CAUSE OR EFFECT?

Vitamin D deficiency can be commonly found in children and adults with IBD.107-113

After adjusting for dietary intake and seasonal variation, one study reported significantly

113 lower 25(OH)D3 levels in CD patients when compared with healthy controls. While

inflammation and diarrhea can explain malabsorption of the fat-soluble vitamin D,114,115

absorption and bioavailability are still on average 30% lower than normal in patients with

quiescent disease.115 Therefore, unlike with other immune-related diseases (e.g., MS, RA,

SLE) that have been linked to vitamin D deficiency, the nature of fat-soluble vitamin malabsorption in IBD confounds the interpretation of current evidence. It is therefore unclear whether vitamin D deficiency is a cause or effect of IBD in humans.

This question of cause or effect (reverse causation) becomes particularly important for observational studies that attempt to link vitamin D deficiency with disease pathogenesis. On one hand, the large Nurses’ Health Study showed lower predicted serum 25(OH)D levels prior to CD diagnosis over a follow-up period of 22 years.102 A

major caveat is the fact that serum 25(OH)D concentrations were indirectly predicted

based on surveyed data of dietary choices, supplemental D intake, sunlight exposure,

regional UVB intensity, body mass index, and race. Direct measurement of serum

25(OH)D prior to the onset of IBD has been absent due to significant logistical and

financial challenges in prospectively identifying individuals in the general population at

risk for developing IBD and collecting pre-disease serum samples. Moreover, the

temporal causation of vitamin deficiency and actual disease onset has not yet been

established. Therefore, given the currently available evidence, the question remains as to

whether vitamin D deficiency is truly a cause or effect of IBD in humans.

SUMMARY

The serendipitous discovery of VDR in immune cells opened the doors for investigation

of the relationship between vitamin D and immunoregulation. Vitamin D deficiency has

since been shown to influence pathogenesis of autoimmune disorders. Animal studies and

epidemiologic data suggest that the same relationship exists in IBD. However, unlike

other non-gastrointestinal autoimmune disorders, IBD can itself reduce vitamin D levels

due to malabsorption. The dilemma of reverse causation (does p cause q or does q cause

p?) currently represents the greatest challenge in establishing vitamin D deficiency as an

environmental risk factor of IBD.

In this thesis dissertation, we present a series of studies that employ diverse

methodologic approaches in an attempt to overcome reverse causation when exploring

the role of vitamin D in IBD pathogenesis and severity. Chapter II discusses a case- control study that compared directly measured 25(OH)D levels in prospectively collected sera among individuals who eventually developed CD and age-, sex-, and race-matched controls. The findings of this study provide a unique longitudinal view of vitamin D levels before and after diagnosis, thus reducing interference by reverse causation.

Chapter III discusses the analysis of a national hospitalization dataset that evaluated variations in ultraviolet exposure – as a surrogate of vitamin D not influenced by disease- related malabsorption – and its association with hospitalization severity and need for bowel surgery. Finally, Chapter IV discusses a systematic review and meta-analysis of randomized controlled trials (RCTs) that compared vitamin D at different doses or with placebo, while measuring disease activity as primary or secondary outcomes. The

13 randomization process is intended to homogenize the populations among treatment arms, thus minimizing confounding and permitting attribution of observed effects to the primary intervention. In this case, we evaluated whether vitamin D has any measurable effect on intestinal inflammation and/or clinical relapse.

FIGURES

Figure 1-1. Interaction between genetic, host, and environmental factors in the pathogenesis of inflammatory bowel disease. The current paradigm of inflammatory bowel disease pathogenesis involves the complex interaction of an individual’s genetic risk, gut microbiota, environmental factors, and an aberrant host immune response.

Figure 1-2. Vitamin D synthesis and hydroxylation pathway. Vitamin D can either be synthesized in the epidermis (ultraviolet-mediated conversion of 7-dehydrocholesterol to cholecalciferol) or consumed from food sources. While bound to the vitamin D binding protein or albumin, vitamin D is then transported to the liver, where it is hydroxylated to 25-hydroxyvitamin D. The 25-hydroxyvitamin D molecule is subsequently transported to the renal tubules, where it is hydroxylated to 1,25-dihydroxyvitamin D (active form).

CHAPTER II. VITAMIN D AND CROHN’S DISEASE PATHOGENESIS

INTRODUCTION

The etiology of CD remains unclear, although suspected to involve interactions between

genetic and environmental factors (Figure 1-1). Over 225 genetic risk polymorphisms

have so far been identified, but only account for a minority of CD cases.90,116 Moreover,

monozygotic twins, who presumably have identical genetic makeup, have poor

concordance for CD, suggesting the existence of non-heritable contributors to CD

pathogenesis.117 Epidemiologic studies have subsequently identified several potential environmental risk factors, such as gastrointestinal infections, medication use, and diet.8,11,118

Vitamin D has also been considered as a modifier of CD risk.102 While vitamin D

has traditionally been associated with calcium and phosphorus metabolism, it has more

recently been found to possess immunomodulatory properties. In IL-10 knockout mice,

vitamin D deficiency has been shown to cause diarrhea, wasting disease, and mortality;

these symptoms were absent among the mice that were either vitamin D sufficient or

supplemented with vitamin D.79 Similarly, concurrent knockout of the VDR in these mice

can lead to a more severe and accelerated manifestation of colitis.84 Administration of

exogenous vitamin D or a VDR agonist have further been demonstrated to ameliorate

colitis in IBD mouse models.80,82 In humans, ecologic studies in Europe and the United

States have found a north-south gradient of CD incidence. That is, regions with less sunlight exposure and presumably less natural synthesis of vitamin D have greater CD incidence.95,98,100 An independent analysis of the prospective Nurses’ Health Study found

that higher predicted 25(OH)D levels were associated with a reduced risk of CD.102

Additionally, cohort studies have correlated low vitamin D levels with increased CD

activity,97,108,119 although the converse could also be true where increased CD activity

leads to vitamin D deficiency. The latter can occur through intestinal malabsorption,

alterations in the diet, and reduced sunlight exposure (from decreased physical activity

and sunlight avoidance while on antimetabolite therapy).

This concept of reverse causation comprises the fundamental challenge in

assessing directionality of effect in the relationship between vitamin D and CD. To

overcome this limitation, longitudinal vitamin D levels would need to be measured before

and after the onset of CD. The aforementioned Nurses’ Health Study had found that pre-

disease 25(OH)D levels predicted the risk of CD, but the levels were indirectly estimated

based on a regression model that employed other related factors, such as dietary and

supplement intake, sunlight exposure, race, and body mass index.102 As such, the principal objective of this study was to clarify the question whether vitamin D deficiency leads to CD, vice versa, or both, by directly measuring 25(OH)D levels in prospectively collected serum samples among individuals who did and did not eventually develop CD.

METHODS

Study Population

The source population included United States (U.S.) military personnel who had served in active duty for 6 or more years from January 1, 1998 to December 31, 2011 and had no diagnosis of inflammatory bowel disease (CD, UC, or non-specific intestinal inflammation) based on diagnostic codes for at least 3 years. For this case-control study,

cases included 400 individuals with an original definition of at least 2 ICD-9-CM

(International Classification of Diseases, Ninth Revision, Clinical Modification) codes for CD (555), while controls included 400 age-, sex-, race-, and military unit-matched individuals. To validate the case selection criteria, chart review was performed on 300 cases, revealing misclassification in 104 (34.7%) cases. The case definitions were therefore revised to require either (i) chart review-confirmed cases of CD based on endoscopic findings; or (ii) at least 3 ICD-9-CM codes for CD (555) from gastroenterology clinic visits. Using the new criteria, there were subsequently 240 cases and 240 matched controls available for this study. The positive predictive value of the new criteria, while using the chart review as the gold standard, was 98.0% and the negative predictive value was 100%. However, given the increased stringency of the chart review criteria, some CD cases may have been inadvertently excluded, including 24 with “possible CD”, 12 with a single episode of ileitis, 3 with a single episode of jejunitis, 3 with recurrent perianal fistulae, and 3 with a single episode of terminal ileal thickening and recurrent anal fissures.

For each participant, 3 serum specimens were obtained from the Department of

Defense Serum Repository at 3 separate time points leading up to CD diagnosis (for cases) or matched index time (for controls). The time points were 3 to 8 years before diagnosis or index time (pre-2); 3 months to 3 years before diagnosis or index time (pre-

1); or, 3 months before and up to 21 months after diagnosis or index time (pre-0) (Figure

2-1).

Data Abstraction

Data abstraction from the medical records was performed by gastroenterology fellows and supervised by two gastroenterologists who specialized in the diagnosis and management of IBD at Walter Reed National Military Medical Center (WRNMMC). All were active military personnel. Due to security limitations, the dataset provided for this study was limited to age, sex, race, year of diagnosis for cases or matched index time for controls, rank at diagnosis, highest education level achieved, and deployment history.

The time of serum collection was provided as the number of days relative to the time of diagnosis or index time. The month and geographic location of residence at the time of serum collection were not available.

Vitamin D Measurements

At each time point for cases and controls, 50 µL of serum was aliquoted and stored at -

20oC for the 25(OH)D assays. The 25(OH)D levels were measured using the DiaSorin

25-Hydroxyvitamin D 125I radioimmunoassay kit (Saluggia, Italy) and performed by the

Institute for Clinical and Translational Research (ICTR) Core Laboratory at the Johns

Hopkins Bayview Medical Center (Baltimore, Maryland). To account for potential

random errors in batch processing, cases and matched controls were placed into the same

plate by the study investigators with the position of the samples randomly placed by case

status and plated only by sample numerical ID with the case-control status unknown to

the study investigators. Cases and control assignments were blinded to all laboratory

investigators.

Genotyping

We assumed each participant’s genomic data to have remained stable throughout the

follow-up period. Serum from the pre-1 and pre-0 samples (50 µL each) were thus pooled for greater yield. DNA purification was performed using column-based purification methods with the QIAGEN QIAmp DNA Blood Mini Kit ® (Germantown,

Maryland), as outlined in the product protocol. In brief, the samples were incubated at

56oC for 10 minutes in proteinase lysis solution and mixed with DNAse/RNAse

inhibitors. DNA was isolated via several cycles of centrifugation through spin columns

with DNA-binding silica membranes. Residual contaminants were removed with buffer

washes and the DNA retrieved with an elution buffer. Primers and TaqMan® probes

were designed to target the following single nucleotide polymorphism (SNP) sites:

rs2282679 (DBP), rs731236 (TaqI site of VDR), rs10870077 (CARD9), rs2066845

(NOD2 SNP17), rs2066844 (NOD2 SNP20), rs2066847 (NOD2 C-insertion), rs2241880

(ATG16L), rs13361189 (IRGM), rs10870077 (CARD9), and rs601338 (FUT2). Primers

and probes pairs were initially validated on whole genomic DNA and then on DNA

purified from matching sera of 10 non-study samples, using real-time polymerase chain

reaction (PCR) on the Applied Biosystems 7900HT Fast Real-Time PCR System and

compared with results from direct sequencing. The validated primers and probes were

then applied to the designated serologic samples, pipetted into a 384-well plate,

accompanied by water (negative) and genomic DNA (positive) controls. PCR master mix

and corresponding primer-probe pairs were added. Samples were then amplified using

real-time PCR and genotypes determined.

Statistical Analyses

Categorical variables were compared using the χ2 and continuous variables were compared using the Student t test. Conditional logistic regression was used to estimate the relative odds of CD according to 25(OH)D levels when comparing cases and controls at each time point. Vitamin D levels were categorized as tertiles and quintiles, based on the distribution of 25(OH)D levels among controls. The 25(OH)D levels for the tertiles were 5.84 to 27.52 ng/mL, 27.57 to 35.66 ng/mL, and 35.68 to 71.96 ng/mL, respectively. The 25(OH)D levels for the quintiles were 5.84 to 23.11 ng/mL, 23.25 to

29.09 ng/mL, 29.13 to 34.17 ng/mL, 34.27 to 39.19 ng/mL, and 39.21 to 71.96 ng/mL, respectively. Additional analyses evaluated vitamin D according to pre-defined levels of deficiency (< 21 ng/mL), insufficiency (21 to 30 ng/mL), and sufficiency (≥ 30 ng/mL).

The latter thresholds for vitamin D were not used as the primary categorization method, because the gradated relationship between vitamin D levels and CD incidence were yet unknown. Genetic associations with incident CD were evaluated using a binary category

(wild-type or risk allele) for each gene. Gene-environment interactions were subsequently evaluated using the binary gene categories and vitamin D levels coded as tertiles. These interaction analyses were not performed with vitamin D levels coded as quintiles to minimize the number of discrete categories in the gene-vitamin D interaction terms. The wild-type polymorphism at the highest vitamin D tertile was used as the reference group in the conditional logistic regression models.

To test the robustness of our primary analyses, sensitivity analyses were performed using (i) only chart-reviewed cases; or (ii) all cases that were not excluded

after chart review, but did not necessarily meet the criteria of 3 ICD-9-CM codes for CD from gastroenterology clinic visits. Sensitivity analysis of the gene models was performed by first creating a global multivariable model that included vitamin D levels coded as tertiles, all genotyped polymorphisms, and the longitudinal serum number (-2, -

1, 0). Derivative models that included subsets of genes were then compared with the

original model using the Akaike Information Criterion (AIC) and Bayesian Information

Criterion (BIC). The primary analysis of vitamin D and relative odds of CD were then

performed using the multivariable model with the lowest AIC. Statistical significance

was defined as a two-tailed α of 0.05. All analyses were performed using Stata MP 13.1

(StataCorp, College Station, Texas).

Ethics and Study Subject Anonymity

This study was approved by the Institutional Review Boards of the Johns Hopkins

University School of Medicine (Baltimore, Maryland) and the WRNMMC (Bethesda,

Maryland). The study protocol was also reviewed and approved by the Armed Forces

Health Surveillance Center (AFHSC). Chart review was limited to the physicians at

WRNMMC. Only a limited dataset of medical and demographic data was approved by

the AFHSC. This dataset was linked to the serum sample identifier by the AFHSC and no

investigators (including the physicians at WRNMMC) have the code to this linkage,

which the AFHSC will destroy at the study’s conclusion.

RESULTS

The study population included 240 cases and 240 age-, sex-, race-, and military unit- matched controls (Table 2-1). The mean age was 28 years, and the participants were mostly male (83.8%) and white (81.7%). The most common military grade at diagnosis was enlisted (83.8%) and the most common education level achieved at diagnosis was a high school degree (72.3%). There were fewer cases (61.2%) than controls (72.9%) to ever be deployed overseas (P < 0.01).

Vitamin D Levels and Incident CD

When classifying 25(OH)D levels as tertiles, the relative odds of incident CD were not significantly different between cases and controls with each increasing tertile for the pre-

2 (odds ratio [OR] 1.04; 95% confidence interval [CI] 0.80 – 1.35) or pre-1 (OR 0.96;

95% CI 0.76 – 1.21) samples (Figure 2-2A). However, the relative odds of incident CD were significantly lower with each increasing tertile for the pre-0 samples (OR 0.74; 95%

CI 0.59 – 0.94). When classifying 25(OH)D levels as predefined thresholds of vitamin D deficiency, insufficiency, and sufficiency, the results were similar. The relative odds of incident CD were not significantly different between cases and controls with each increasing stratum for the pre-2 (OR 1.03; 95% CI 0.77 – 1.39) or pre-1 (OR 0.90; 95%

CI 0.69 – 1.18) samples. The relative odds of incident CD were significantly lower with each increasing stratum for the pre-0 samples (OR 0.75; 95% CI 0.57 – 0.99). When classifying 25(OH)D levels as quintiles, the relative odds of incident CD were not significantly different between cases and controls with each increasing quintile for the pre-2 (OR 1.06; 95% CI 0.91 – 1.24) or pre-1 (OR 1.01; 95% CI 0.88 – 1.16) samples

(Figure 2-2B). The relative odds of incident CD were significantly lower with each

increasing quintile for the pre-0 (OR 0.85; 95% CI 0.74 – 0.98) samples. These results do

not change after incorporating deployment history into a multivariable model, although

prior deployment was consistently associated with a lower odds of CD at all time points.

Gene Polymorphism Distributions

The TaqI region of VDR had different allele frequencies for cases and controls: T (0.60

vs. 0.65) and C (0.40 vs. 0.35) (P = 0.02) (Table 2-2). DBP had similar allele frequencies

between cases and controls: A (0.72 vs. 0.75) and C (0.28 and 0.25) (P = 0.36). For the

CD risk alleles, cases and controls differed in their allele frequencies for NOD2 SNP 20

(C: 0.92 vs. 0.96; T: 0.08 vs. 0.04; P = 0.04), ATG16L (A: 0.42 vs. 0.52; G: 0.58 vs.

0.48; P = 0.02), and IRGM (C: 0.18 vs. 0.11; T: 0.82 vs. 0.89; P = 0.01). Otherwise, cases and controls had similar allele frequencies for CARD9, NOD2 SNP17, NOD2 C- insertion, and FUT2.

When evaluating gene polymorphisms, cases had significantly more polymorphisms than controls in VDR (66.5% vs. 55.1%; P = 0.01), NOD2 (28.4% vs.

16.1%; P < 0.01), ATG16L (81.6% vs. 72.9%; P = 0.02), and IRGM (31.4% vs. 19.5%;

P < 0.01) (Table 2-3). Polymorphisms in these genes were associated with an increased relative odds of CD: VDR (OR 1.57; 95% CI 1.09 – 2.27); NOD2 (OR 2.22; 95% CI 1.36

– 3.63); ATG16L (OR 1.70; 95% CI 1.08 – 2.67); and IRGM (OR 2.21; 95% CI 1.36 –

3.58) (Table 2-4). Polymorphisms in the other genes were otherwise not associated with incident CD.

Gene Interactions

Analysis of gene-environment interactions focused on the vitamin D-related genes (VDR,

DBP) and the CD risk genes (NOD2, ATG16L, IRGM) found in this study to be associated with an increased risk of CD. For VDR, the relative odds of CD remained constant across vitamin D tertiles among VDR wild-type and variant genotypes in the pre-1 samples (Figure 2-3A). However, the odds of CD were significantly higher at the lowest vitamin D tertile for both VDR wild-type (OR 2.61; 95% CI 1.24 – 5.46) and variant (OR 2.80; 95% CI 1.42 – 5.54) genotypes in the pre-0 samples (Figure 2-3B).

Variant VDR genotypes had a significantly greater risk of CD across vitamin D tertiles.

For DBP, the relative odds of CD remained constant across vitamin D tertiles among

DBP wild-type and variant genotypes in the pre-1 samples (Figure 2-3C). Similar to

VDR, the odds of CD were significantly higher at the lowest vitamin D tertile for both

DBP wild-type (OR 1.88; 95% CI 1.02 – 3.48) and variant (OR 2.24; 95% CI 1.16 –

4.33) genotypes in the pre-0 samples.

For NOD2, the odds of CD were generally similar across vitamin D tertiles in the pre-1 samples (Figure 2-4A). The odds of CD were significantly higher at the lowest vitamin D tertile for both NOD2 wild-type (OR 1.77; 95% CI 1.02 – 3.10) and mutant

(OR 3.86; 95% CI 1.47 – 10.17) genotypes in the pre-0 samples (Figure 2-4B). For

ATG16L, the odds of CD remained constant across vitamin D tertiles among the

ATG16L wild-type and risk genotypes in the pre-1 samples (Figure 2-4C). The odds of

CD were significantly higher at the lowest vitamin D tertile for risk ATG16L genotypes

(OR 3.00; 95% CI 1.13 – 8.00) in the pre-0 samples (Figure 2-4D). For IRGM, the odds

of CD were significantly higher at the lowest vitamin D tertile for risk IRGM genotypes

(OR 2.42; 95% CI 1.10 – 5.30) in the pre-1 samples (Figure 2-4E). The odds of CD were significantly higher at the lowest vitamin D tertile for both NOD2 wild-type (OR 1.80;

95% CI 1.03 – 3.15) and mutant (OR 3.90; 95% CI 1.77 – 8.58) genotypes in the pre-0 samples (Figure 2-4F).

Sensitivity Analyses

During the course of case validation, misclassification was detected, which led to some cases being excluded from the study. To test the robustness of our primary analyses, diverse case definitions were applied to the conditional logistical regression model. When using only chart-reviewed cases, the relative odds of incident CD were not significantly different with each increasing vitamin D tertile for the pre-2 (OR 1.05; 95% CI 0.94 –

1.18) or pre-1 (OR 0.96; 95% CI 0.86 – 1.06) samples. The relative odds of incident CD were significantly lower with each increasing tertile for the pre-0 (OR 0.89; 95% CI 0.80

– 0.99) samples. When using all cases that were not excluded after chart review, but did not necessarily meet the final rigorous case definition, the relative odds of incident CD were not significantly different with each increasing tertile for the pre-2 (OR 1.04; 95%

CI 0.82 – 1.32) or pre-1 (OR 0.89; 95% CI 0.71 – 1.11) samples. The relative odds of incident CD were significantly lower with each increasing tertile for the pre-0 (OR 0.77;

95% CI 0.62 – 0.96) samples. These sensitivity analyses using variant case definitions provided consistent results as our primary analyses.

Among the tested models that included multiple genes as covariates, the global multivariable model had the lowest AIC (Table 2-5). Sensitivity analysis using this global

27 model did not reveal vitamin D to be independently associated with incident CD (OR

0.90; 95% CI 0.77 – 1.06). When the same model was separately analyzed by time points, the relative odds of incident CD were not significantly different across vitamin D tertiles for the pre-2 (OR 1.18; 95% CI 0.88 – 1.58) or pre-1 (OR 0.97; 95% CI 0.75 – 1.26) samples. The relative odds of incident CD were significantly lower with each increasing vitamin D tertile for the pre-0 (OR 0.68; 95% CI 0.52 – 0.89) samples. These findings were consistent with all prior analyses.

DISCUSSION

To our knowledge, this is the first study that directly measures vitamin D levels and evaluates their longitudinal relationship with risk of CD, and samples were collected prior to CD diagnosis. Our findings revealed that pre-diagnosis vitamin D levels were not associated with incident CD. By contrast, vitamin D levels were inversely associated with the relative odds of CD around the time of diagnosis. These findings were consistently observed even after stratifying vitamin D levels as tertiles, quintiles, and pre-defined thresholds (deficiency, insufficiency, and sufficiency). In finding that vitamin D levels were only decreased among cases in the peri/post-diagnosis specimens, the data suggest that vitamin D deficiency probably does not confer a significantly increased risk of CD.

Instead, the presence of CD may have contributed to lower vitamin D levels.

These findings support current assumptions that CD may adversely affect vitamin

D levels in patients, even among those with quiescent disease.115 However, they contradict earlier studies that have implicated vitamin D in CD pathogenesis. One potential explanation for this discrepancy is that vitamin D either does not – or only

weakly – influences CD risk, particularly given the complex multi-factorial

underpinnings of CD pathogenesis. In the regression analyses, even the tested genetic risk

polymorphisms had stronger effect sizes than vitamin D. The weakness of vitamin D’s

effect on CD may be further illustrated by intervention trials that have not found a clear

or strong benefit of vitamin D supplementation on IBD-related outcomes (Chapter IV).

The largest randomized trial of vitamin D as a treatment for CD involved 94 participants

and found no significant difference in clinical relapse rates, although there was a trend

toward statistical significance (P = 0.06).104 Alternatively, vitamin D may not lie along the causal pathway of CD pathogenesis, but exists as a confounder influenced by inflammation.120

In the genetic analyses, this study found an association between the TaqI site in

VDR with incident CD. This relationship has previously been described in a European

Caucasian cohort, but less strongly among those from other racial and ethnic

backgrounds.86-88 DBP was not associated with CD in our study, although it may have

been underpowered to detect an effect. One prior study of 636 IBD cases and 248

controls found an inverse association between the presence of DBP variants and IBD.91

As the functional effect of these polymorphisms are currently unknown, the investigators

hypothesized that a reduced affinity could facilitate vitamin D release at its target tissue.

We additionally evaluated known CD risk polymorphisms in our study population.

NOD2, ATG16, and IRGM were expectedly associated with an increased risk of CD. The

lack of an association with CARD9 or FUT2 was not surprising, given our relatively

small sample size relative to typical genome-wide association studies. In the gene-

vitamin D interaction analyses, there were no clear signals of interaction. While the

presence of VDR and known CD risk polymorphisms increased the relative odds of CD,

they did not significantly alter the relationship of vitamin D with CD.

This study had several strengths. In an attempt to overcome the challenges of

reverse causation when interpreting previous studies, this study directly measured

25(OH)D levels in samples that were collected before and around the time of diagnosis.

This longitudinal comparison in matched cases and controls allowed for a clear

assessment of trends in vitamin D levels before diagnosis (when disease presumably has

no effect on vitamin D levels) and after diagnosis (when disease has a potential effect on

vitamin D levels). An additional strength of this study was the use of validated case

definitions. We used a combination of chart review and ICD-9-CM criteria to define and

validate our case criteria, thus reducing the risk of misclassification bias in our study

population. As an illustration of the robustness of the study findings, we obtained strongly consistent results across sensitivity analyses that employed different case definitions (original and validated), different categorization schemes for vitamin D levels,

and different gene regression models.

There were nonetheless inherent limitations of the study, including sample size,

lack of granular data, selection bias, and single measurements of 25(OH)D levels. Our

initial sample size estimates for a power of 0.8 required 400 cases and 400 controls.

However, due to the revised diagnostic classification schemes, the study population was

60% of the original target. This decrease resulted in our ability to detect a difference in

relative odds of 0.60 or 1.68 with a power of 0.80. On the other hand, sensitivity analyses

using less conservative case criteria (although not necessarily much less accurate) and a

larger sample size revealed similar findings. Moreover, statistically significant

associations were detected in pre-0 samples indicating that the sample size was not

necessarily too small. We additionally performed post hoc sample size estimates that revealed the need for approximately 1408 cases for pre-2 and 3298 cases for pre-1 to

achieve a power of 0.8 to detect an association between vitamin D and CD, if present.

These large numbers again suggest that the effect of pre-diagnosis vitamin D on CD may

be so small if even at all present. Even if we were to achieve statistically significant

results at such large sample sizes, the clinical significance would be unclear. Moreover,

the required sample size would be much larger than feasibly attainable. The next largest

prospective studies with pre-diagnostic CD sera were performed using military samples

and had approximately 100 participants. An additional limitation of the study was the

lack of more granular data, such as season of serum collection, disease activity,

medication use, and vitamin supplementation at the time of each sample. We additionally

did not have data to elucidate actual dates of disease onset, which may differ from the

dates of diagnosis. On the other hand, the pre-diagnosis samples should be far enough

from the time of diagnosis to fairly assume disease-free status at the time of serum

collection. If we had significantly misclassified pre-diagnosis samples with subclinical

disease as being disease-free, we would have expected to observe an association between

CD and 25(OH)D levels among the pre-1 samples. We did not find this association

except with the pre-0 samples. Thirdly, our study population included mostly white

males, which unsurprisingly reflects a selection bias of the military population and

interferes with generalizability of the results. Nonetheless, the racial distribution was

similar to the demographics of CD in the general population. Similar studies in other

demographics would be warranted, but difficult to achieve due to the need for large

prospectively collected serum samples in individuals with unclear risk of CD. If a high-

risk population were to be sampled (e.g., relatives of CD patients), one limitation would

be the influence of potentially stronger confounders than vitamin D. Finally, due to

limited sera availability, 25(OH)D levels were measured as single measurements, which

may have been subject to undetectable random error. To assess the reliability of

thesemeasurements, we assayed 20 randomly distributed duplicates (each duplicate

placed on consecutive plates). The mean absolute difference in 25(OH)D levels between

duplicates was 2.7 ng/mL (SD 2.0; range 0.3 – 7.5) after exclusion of one clear outlier

(difference of 41.7 ng/mL), indicating good intra-batch precision. We also ensured that

all case and control pairs were assayed on the same plate to optimize the reliability of

measurements for our matched pair analyses and remove potential batch effects. When

comparing 25(OH)D levels between African-American and Caucasian controls, African-

Americans expectedly maintained a mean difference of -10.3 to -9.2 ng/mL across time

points. These findings are an example of the internal sample reliability of our

measurements. Moreover, the consistency of our results across sensitivity analyses

further suggest that random variations from individual 25(OH)D levels would likely not

have changed our inferences.

This prospective case-control study was thus far the most direct evaluation of the

relationship between longitudinal 25(OH)D level and CD incidence. We found no

association between pre-diagnosis vitamin D and CD, thus calling into question prior

assessments that vitamin D deficiency may lead to the development of CD. We

nonetheless found an association between post-diagnosis vitamin D and CD, suggesting

that prior studies that associated vitamin D deficiency with CD were affected by the

32 phenomenon of reverse causation. As this study specifically evaluated the relationship between vitamin D and CD pathogenesis, it does not necessarily answer the question about the relationship between vitamin D and CD severity.

TABLES

Table 2-1. Patient characteristics at the time of diagnosis or index time. Characteristic Cases (n = 240) Controls (n = 240) P value Age, years (SD) a 28.2 (0.4) 28.3 (0.4) 0.82 Sex (%) 1.00 Male 201 (83.8) 201 (83.8) Female 39 (16.2) 39 (16.2) Race (%) 1.00 White 196 (81.7) 196 (81.7) Black 22 (9.2) 22 (9.2) Other 22 (9.2) 22 (9.2) Rank (%) 0.80 Enlisted 202 (84.2) 200 (83.3) Officer 38 (15.8) 40 (16.7) Highest degree (%) 0.56 High school 171 (71.2) 176 (73.3) Bachelor’s 29 (12.1) 29 (12.1) Master’s 9 (3.8) 10 (4.2) Other 31 (12.9) 25 (10.4) Ever deployed? (%) < 0.01 No 93 (38.8) 65 (27.1) Yes 147 (61.2) 175 (72.9) Abbreviations: SD, standard deviation. a Age presented as mean

Table 2-2. Gene allele distributions. Gene Allele Cases Controls P value VDR T 0.60 0.65 0.02 C 0.40 0.35 DBP A 0.72 0.75 0.36 C 0.28 0.25 CARD9 C 0.59 57.2 0.69 G 0.41 42.8 NOD2 SNP 17 C 0.03 0.01 0.06 G 0.97 0.99 NOD2 SNP 20 C 0.92 0.96 0.04 T 0.08 0.04 NOD2 C- W 0.94 0.97 0.05 insertion C 0.06 0.03 ATG16L A 0.42 0.52 0.02 G 0.58 0.48 IRGM C 0.18 0.11 0.01 T 0.82 0.89 FUT2 A 0.48 0.51 0.31 G 0.52 0.49

Table 2-3. Risk genotype distributions. Gene Variant Genotype Cases (%) Controls (%) P value VDR A/G 66.5% 55.1% 0.01 G/G DBP G/T 47.3% 41.5% 0.21 G/G CARD9 C/G 62.8% 66.5% 0.39 G/G NOD2 Multiple genes a 28.4% 16.1% < 0.01 ATG16L A/G 81.6% 72.9% 0.02 G/G IRGM C/C 31.4% 19.5% < 0.01 C/T FUT2 A/A 28.9% 23.3% 0.17 a NOD2 mutants were defined based on the presence of at least one of the following mutant genotypes: SNP 17 (C/C, C/G), SNP 20 (C/T, T/T), C-insertion (*/C, C/C).

Table 2-4. Gene polymorphisms and incident Crohn’s disease. Gene Variants OR 95% CI P value VDR A/G 1.57 1.09 – 2.27 0.02 G/G DBP G/T 1.25 0.87 – 1.80 0.23 G/G CARD9 C/G 0.84 0.58 – 1.24 0.38 G/G NOD2 Multiple genes a 2.22 1.36 – 3.63 < 0.01 ATG16L A/G 1.70 1.08 – 2.67 0.02 G/G IRGM C/C 2.21 1.36 – 3.58 < 0.01 C/T FUT2 A/A 1.45 0.92 – 2.29 0.11 Abbreviations: CI, confidence interval; OR, conditional odds ratio. a NOD2 mutants were defined based on the presence of at least one of the following mutant genotypes: SNP 17 (C/C, C/G), SNP 20 (C/T, T/T), C-insertion (*/C, C/C).

Table 2-5. Akaike Information Criteria for gene interaction models.a Gene Covariates dF AIC BIC DBP, TaqI, CARD9, NOD2, ATG16L, IRGM, FUT2 9 1239 1286 TaqI, NOD2, ATG16L, IRGM, FUT2 7 1243 1279 TaqI, CARD9, NOD2, ATG16L, IRGM 7 1247 1283 TaqI, NOD2, ATG16L, IRGM 6 1249 1285 Abbreviations: AIC, Akaike Information Criterion; BIC, Bayesian Information Criterion; dF, degrees of freedom. a Multivariable models included vitamin D tertiles, longitudinal serum batch, and listed genes as covariates.

FIGURES

Figure 2-1. Time points for collected serum samples. Three serum specimens were obtained for each participant. The time points were 3 to 8 years before diagnosis or index date (pre-2), 3 months to 3 years before diagnosis or index date (pre-1), and 3 months before and up to 21 months after diagnosis or index date (pre-0).

Figure 2-2. Vitamin D levels and relative odds of incident Crohn’s disease (CD) at each time point. A. When vitamin D levels were stratified according to tertiles, vitamin D was not associated with incident CD at pre-2 and pre-1 time points. Increased vitamin D levels were associated with lower incidence of CD at pre-0. B. When vitamin D levels were stratified according to quintiles, vitamin D was not associated with incident CD at pre-2 and pre-1 time points. Increased vitamin D levels were associated with lower incidence of CD at pre-0.

Figure 2-3. VDR and DBP polymorphisms and the odds of incident Crohn’s disease (CD) across vitamin D tertiles. The relative odds of CD were compared between cases and controls at different vitamin D tertiles using the wild-type at the highest vitamin D tertile as the reference. The 25(OH)D levels for the tertiles were 5.84 to 27.52 ng/mL, 27.57 to 35.66 ng/mL, and 35.68 to 71.96 ng/mL, respectively. A. At the pre-1 period, the odds of CD remained constant across vitamin D tertiles for both VDR wild-type and variant genotypes. B. At the pre-0 period, the odds of CD were significantly higher at the lowest vitamin D tertile for both VDR wild-type (P = 0.01) and variant (P < 0.01) genotypes. Variant VDR genotypes had a significantly greater risk of CD across vitamin D tertiles. C. At the pre-1 period, the odds of CD remained constant across vitamin D tertiles for both DBP wild-type and variant genotypes. D. At the pre-0 period, the odds of CD were highest at the lowest vitamin D tertile for both DBP wild-type (P = 0.04) and variant (P = 0.02) genotypes.

Figure 2-4. Crohn’s disease (CD) risk mutations and the odds of incident CD across vitamin D tertiles. The relative odds of CD were compared between cases and controls at different vitamin D tertiles using the wild-type at the highest vitamin D tertile as the reference. The 25(OH)D levels for the tertiles were 5.84 to 27.52 ng/mL, 27.57 to 35.66 ng/mL, and 35.68 to 71.96 ng/mL, respectively. A. At the pre-1 period, the odds of CD were generally similar across vitamin D tertiles. B. At the pre-0 period, the odds of CD were significantly higher at the lowest vitamin D tertile for both NOD2 wild-type (P =

0.04) and mutant (P < 0.01) genotypes. C. At the pre-1 period, the odds of CD remained constant across vitamin D tertiles for both ATG16L wild-type and risk genotypes. D. At the pre-0 period, the odds of CD were significantly higher at the lowest vitamin D tertile for risk ATG16L genotypes (P = 0.03). E. At the pre-1 period, the odds of CD were significantly higher at the lowest vitamin D tertile for risk IRGM genotypes (P = 0.03). F. At the pre-0 period, the odds of CD were significantly higher at the lowest vitamin D tertiles for both IRGM wild-type (P = 0.04) and risk genotypes (P < 0.01).

CHAPTER III. VITAMIN D AND HOSPITALIZATION SEVERITY

INTRODUCTION

Prior studies have found low vitamin D levels to be associated with increased IBD

severity. IBD murine models with VDR-knockout experienced more severe colitis than wild-type mice,121,122 while exogenous administration of vitamin D or an analogous VDR

agonist ameliorated symptoms.79,82 In humans, a large retrospective study of 403 CD and

101 UC patients found that vitamin D deficiency was associated with increased disease

activity and decreased quality of life in CD, but not UC.108 A smaller cross-sectional study of 34 UC patients found that vitamin D-deficient patients were significantly more likely than vitamin D-sufficient patients to have active disease (68% vs. 33%; P = 0.04) and be on corticosteroid therapy (47% vs. 7%; P = 0.02).119 A case-control study

involving 34 CD patients and 34 age- and sex-matched controls similarly found lower vitamin D levels among CD patients and these levels were inversely correlated with

disease activity.97

While these findings demonstrate an inverse association between vitamin D and

IBD severity, they do not clarify directionality of effect. Notably, IBD – particularly

active disease – can lead to vitamin D malabsorption and subsequently reduced vitamin D

levels. To address the challenges of reverse causation while investigating the relationship

between vitamin D and IBD severity, we needed a surrogate marker that strongly

correlated with vitamin D levels, but was not significantly influenced by IBD. Ultraviolet

(UV) exposure has previously been shown to be a stronger determinant of vitamin D

status than dietary intake or oral supplementation.123,124 Moreover, in our directed acyclic

graph, UV exposure does not have a bidirectional relationship with IBD and is

technically not considered a confounder (Figure 3-1).

We therefore hypothesized that low UV exposure would be associated with increased rates of hospitalizations and disease severity among IBD patients, but not necessarily among non-IBD patients. For this study, we used UV surface data from the

National Oceanic and Atmospheric Administration (NOAA) and hospitalization information from the Nationwide Inpatient Sample (NIS) from 1998 through 2010. We further compared the relationship between UV exposure and severity among non-IBD patients to confirm whether the associations were specific to IBD and not general trends in hospitalization outcomes.

METHODS

UV Exposure

Daily UV radiation data were collected from the NOAA for 57 cities throughout the continental U.S., Alaska, and Hawaii between 1998 through 2010. Ground-based UV exposure was calculated by NOAA using a validated radiative transfer model that accounts for regional and temporal ozone, sun-earth distance, solar zenith angle, and altitude;125,126 cloud cover was estimated from satellite-based ozone data, while surface

albedo was kept constant at 5%. Tropospheric pollution and haze were not considered in

the UV index. We quantified UV exposure for each hospitalization as the UV index (one

unit equals a noontime erythemal dose of 25 mW/m2) for the corresponding month

(averaged UV over all days of the month), year, and hospital location (state level). UV exposures were then stratified according to rounded UV index thresholds defined by the

World Health Organization: low (0-2), moderate (3-5), high (6-7), and very high (8 and above).127 Admission month was not available for Florida, so UV exposure could not be assigned to hospitalizations in this state.

Hospitalization Data

The NIS is the largest database of hospitalizations in the U.S. The NIS is generated annually as part of the Healthcare Cost and Utilization Project (HCUP), sponsored by the

Agency for Healthcare Research and Quality (AHRQ). The database samples approximately 1,000 non-federal, short-term, acute care hospitals, while excluding psychiatric, substance abuse, and short-term rehabilitation facilities.128 The sampled

hospitals comprise approximately 8 million annual inpatient records and represent over

39 million hospitalizations per year. The authors obtained permission from AHRQ to

access the NIS dataset and used it according to HCUP guidelines (www.hcup-

us.ahrq.gov/team/NationwideDUA.jsp). This study protocol (NA83700) was approved by

the Institutional Review Board of the Johns Hopkins University School of Medicine.

The NIS was analyzed for all hospitalizations between 1998 through 2010. Diagnoses

and types of procedures were based on ICD-9-CM codes in primary or secondary

diagnosis or procedure positions. IBD hospitalizations were defined as those with a

corresponding discharge code for CD (ICD-9-CM code 555) or UC (ICD-9-CM code

556) in the primary and a secondary diagnosis positions, while non-IBD hospitalizations

included those without an ICD-9-CM code 555 or 556 in any diagnosis position. Records

without an available admission month (n = 54,835 records) and states with ≥10% missing

race information for a particular year (n = 156,718 records from 19 states) were excluded.

Information of interest included demographic data (age, sex, and race), types of

abdominal surgery, inpatient or peri-hospitalization mortality, hospitalization characteristics (admission month, year, length of stay), and hospital-specific features

(state, bed size). The Charlson-Deyo comorbidity index is an aggregate severity score of comorbid diagnoses, designed for use with administrative data and based on the presence of 17 medical conditions, such as cardiovascular disease, chronic pulmonary disease, liver disease, diabetes, and malignancy. The comorbidity index was calculated for each hospitalization according to published diagnostic criteria.129,130

IBD Severity

Hospitalization severity outcomes were assessed among IBD hospitalizations. These outcomes of interest included rates of CD or UC hospitalization, prolonged hospitalization, need for bowel surgery, and death. A prolonged hospitalization was defined as the top quartile length of hospital stay among all hospitalizations, which is equivalent to > 7 days. Bowel surgery comprised incisions, excisions, anastomoses, and other therapeutic interventions of the intestinal tract (ICD-9-CM procedural codes 45.0,

45.3 – 46.2, 46.4 – 46.5, 46.7, 48.3 – 48.6). CD-related surgeries included bowel resections and proctectomies (45.5 – 45.9, 48.4, 48.61 – 48.65). UC-related surgeries included partial or total colectomies, and proctectomies (45.7 – 45.8, 45.92 – 45.95, 48.4,

48.61 – 48.65). Only procedures associated with IBD were included in these categories.

Severity in Non-IBD Hospitalizations

To compare hospitalization trends for IBD and some other disorders, hospitalizations

rates were also calculated for cirrhosis (ICD-9-CM code 571.5), cardiovascular disease

(430 – 438), chronic obstructive pulmonary disease (490 – 505, 506.4), and diabetes

mellitus (250.0 – 250.7). To assess whether UV-associated trends observed with IBD-

related surgeries among IBD hospitalizations are unique or suggestive of general surgical

practice patterns, we evaluated the rates of bowel surgery and common surgeries

unrelated to IBD, including cholecystectomies (ICD-9-CM 51.2) and appendectomies

(ICD-9-CM 47.0 – 47.1), in non-IBD hospitalizations.

Potential Confounders

Age, sex, race, Charlson-Deyo comorbidity index, admission type, and hospital size were included as potential confounders of hospitalization severity. Race was categorized as white, black, or other. Admission types included emergent or urgent, elective, and other admissions. Hospital size is thought to influence the volume and complexity of IBD cases encountered in a particular hospital.131 Hospital size was pre-assigned in NIS based on

the number of beds as “small”, “medium”, or “large” according to approximately one-

third cutoff points for each size category within a given region, urban or rural location,

and teaching status combination. We considered urban location and teaching status as

potential confounders, but did not ultimately include them in the model because of their

strong associations with hospital size.

Statistical Analysis

All aggregate data were analyzed using survey procedures that accounted for the complex

sampling design of the NIS. The sampling weights provided with NIS generally permit

calculation of national estimates, and require that at least 1 observation per sampled

hospital be included in the analysis for correct variance estimation.132 Disease severity was measured among IBD and non-IBD hospitalizations. Rates of CD or UC hospitalizations were calculated over all-cause hospitalizations within each UV exposure group. Poisson regression was used to estimate relative rates of prolonged hospitalization, abdominal surgeries, and death according to UV exposure. Multivariable models adjusted for age, female gender, race, Charlson-Deyo comorbidity index, and hospital size.

Variance estimations were done with Taylor series linearization. P for trend was calculated to assess for a dose-response pattern of severity outcomes across UV exposures. We also performed a sensitivity analysis of our case definition to understand the robustness of our findings by defining IBD hospitalizations using all diagnostic positions rather than the first two diagnostic positions. Statistical significance was defined as a two-sided α of less than 0.05. Statistical analyses were performed using SAS

9.3 (Cary, NC) and Stata SE 12.1 (College Station, TX).

RESULTS

Rates of Hospitalizations

The study population included 289,928,496 hospitalizations in the U.S. between 1998 through 2010. Of these hospitalizations, there were 649,932 for CD and 384,267 for UC

(Table 3-1). The median age at hospitalization was 40.7 years (interquartile range [IQR]

28.1 – 55.1) for CD and 46.1 years (IQR 30.2 – 64.1) for UC. For both CD and UC, there

were slightly more females and the majority of patients were white, which was

comparable with non-IBD hospitalizations.

Rates of Hospitalizations by UV Exposure

The rates of CD and UC hospitalizations were greatest in the lowest UV index group and had a statistically significant decline with increasing UV exposure (Figure 3-2). For CD, the low and moderate UV index groups had 217.8 (95% CI 216.8 – 218.8) and 203.8

(95% CI 202.9 – 204.7) hospitalizations per 100,000 overall hospitalizations, compared with the very high UV group rate of 182.5 (95% CI 181.4 – 183.6). The P for trend was <

0.001. For UC, the low and moderate UV groups had 123.2 (95% CI 122.5 – 124.0) and

119.3 (95% CI 118.6 – 119.9) hospitalizations per 100,000 residents, compared with the very high UV group rate of 113.8 (95% CI 112.9 – 114.7; P for trend = 0.033).

Bowel Surgeries

Overall, bowel surgeries in CD were significantly more common in the low (relative rate

[RR] 1.24; 95% CI 1.16 – 1.32), moderate (RR 1.18; 95% CI 1.11 – 1.26), and high (RR

1.17; 95% CI 1.09 – 1.24) UV index groups, compared with the very high UV group.

Lower UV exposures had a significant trend for more bowel surgeries (P for trend <

0.001) (Figure 3-3A). For CD-related bowel resection, a similar trend was noted across

UV groups: low (RR 1.29; 95% CI 1.20 – 1.38), moderate (RR 1.22; 95% CI 1.13 –

1.31); high (RR 1.20; 95% CI 1.12 – 1.29) (P for trend < 0.001) (Figure 3-3B). Similar to

CD patients, overall bowel surgeries for UC were also more common in the low (RR

1.21; 95% CI 1.09 – 1.33), moderate (RR 1.15; 95% CI 1.06 – 1.24), and high (RR 1.14;

95% 1.05 – 1.25) UV index groups, with a significant trend across UV exposures (P for trend < 0.001). This observation was consistent with relative colectomy rates in the low

(RR 1.35; 95% CI 1.15 – 1.58), moderate (RR 1.24; 95% CI 1.09 – 1.42), and high (RR

1.22; 95% CI 1.06 – 1.40) UV groups. There was a significant trend of increasing colectomy rates with lower UV exposures (P for trend < 0.001). UV exposure did not influence rates of other abdominal surgeries unrelated to IBD, including cholecystectomies (CD: P for trend = 0.184; UC: P for trend = 0.746) or appendectomies

(CD: P for trend = 0.919; UC: P for trend = 0.435) (Figure 3-4).

Lengths of Stay

IBD patients in the low UV index group were more likely to have prolonged hospitalizations (> 7 days) than those in the very high UV index group (Figures 3-3C).

The trends were statistically significant for both CD (P for trend < 0.001) and UC (P for trend < 0.001).

Death

Inpatient mortality was greater in the low UV indices compared with the very high UV indices for CD (RR 1.76; 95% CI 1.14 – 2.71; P for trend = 0.046; Figure 3-3D).

Although inpatient mortality rates for UC were statistically similar in the low (RR 1.24;

95% CI 0.92 – 1.66; P for trend = 0.110), moderate (RR 1.16; 95% CI 0.87 – 1.55), and high (RR 1.08; 95% CI 0.79 – 1.48) UV index groups, the point estimates steadily increased across UV exposures.

Severity of Non-IBD Hospitalizations

When analyzing rates of non-IBD hospitalizations, there was no significant trend across

UV exposures for cardiovascular disease (P trend = 0.098) and diabetes mellitus (P trend

= 0.500). Hospitalization rates for cirrhosis had a significant trend in hospitalization rates

(P trend = 0.007), but with decreasing hospitalizations at lower UV exposures; this is the

inverse of the trend observed with IBD. Chronic obstructive pulmonary disease shared a

similar directional trend of hospitalization rates as IBD along UV exposures (P trend <

0.001). For the other surrogates of disease severity, the low UV indices compared with

the high UV indices had higher rates of prolonged hospitalizations (RR 1.09; 95% CI

1.07 – 1.11) and deaths (RR 1.13; 95% CI 1.09 – 1.17). Trends were also significant for

increasing rates of prolonged hospitalizations (P for trend < 0.001) and deaths (P for trend < 0.001) with lower UV exposures. In contrast to IBD hospitalizations, lower UV indices among non-IBD hospitalizations did not increase rates of overall bowel surgeries

(P trend = 0.931) and bowel resections (P trend = 0.441). There was a significant decline in rates of cholecystectomies (P trend < 0.001) and appendectomies (P trend < 0.001) with lower lowest UV exposures for non-IBD hospitalizations (Figure 3-4).

Sensitivity analysis

When examining hospitalizations for CD or UC using all diagnosis positions, similar trends were observed as in the primary analyses. Among CD hospitalizations, low UV exposure was associated with more prolonged hospitalizations (P for trend = 0.004), overall bowel surgeries (P for trend < 0.001), bowel resections (P for trend < 0.001), and deaths (P for trend = 0.001). For UC hospitalizations, low UV exposure was also

associated with more prolonged hospitalizations (P for trend = 0.015), overall bowel

surgeries (P for trend = 0.038), colectomies (P for trend = 0.007), but not deaths (P for trend = 0.068).

DISCUSSION

We found that low UV exposure was associated with increased IBD hospitalization rates and severity. IBD hospitalization rates, prolonged hospitalization, and bowel surgery were consistently greater in the lower UV indices for both CD and UC. Mortality rates were low with small differences across UV exposures, but there was a significant increase in mortality with lower UV exposure for CD and a trend toward significance for

UC.

These findings were consistent with some, but not all, prior epidemiologic data that showed a north-south gradient relationship with IBD incidence.93,96,98,100,101 Studies

in France and Scotland found associations with CD, but not UC.93,96 On the other hand,

analogous epidemiologic studies in all of Europe and the U.S. detected a north-south

gradient for CD and UC, albeit a weaker association for UC.98,100 A possible explanation

for this discrepancy is the shorter latitudinal breadth in the French and Scottish studies

than their U.S. and pan-European counterparts, which would affect their ability to detect

a weaker association for UC.

Large administrative databases have been used to examine the relationship

between UV exposure and other conditions expected to be associated with UV exposure,

latitude and vitamin D. For instance, studies that linked UV data to the Surveillance,

Epidemiology, and End Results (SEER) database and state cancer registries (New York,

New Jersey, Illinois, California, Texas, and Florida) found higher UV indices to be

associated with an increased risk of melanoma.133,134 In a separate study using the Model

Reporting Area for Blindness Statistics and the National Health and Nutrition

Examination Survey (NHANES) datasets, regions with greater sunlight exposure were

found to have higher cataract-to-control ratios for persons aged 65 years or older.135

Similarly, an analysis of multiple sclerosis prevalence in the U.S. and Canada showed a

strong inverse correlation between UV index and multiple sclerosis distribution.136 A

large case-control study also found that high vitamin D levels were associated with lower

risk of multiple sclerosis.66

There are several strengths in this study, including the method of measuring UV,

measure of disease severity, and use of control outcomes to help rule out practice

variation unrelated to IBD. First, in measuring UV exposure, we employed data that were

derived using a validated model to estimate ground-level UV exposure, thus providing an additional layer of specificity than satellite-based sensors. The model incorporates several factors, including regional and temporal sun-earth distance, solar zenith angle, ozone, cloud cover, and altitude. In addition to the use of ground-based UV estimations, our methods combined regional and temporal measures to account for seasonal and weather- pattern variations in UV exposure; this contrasts from other studies that only evaluated geographic latitude to indirectly infer UV exposure. Geographic latitude is a less specific indicator of UV exposure due to seasonal variations in UV exposure that do not strictly follow a latitudinal gradient.137 Second, in contrast with other studies which measured the

relationship between UV or latitude and IBD incidence or prevalence, this study was able

to evaluate relative UV exposure and IBD severity during hospitalization, given its larger

sample size. Third, we explored severity outcomes in non-IBD hospitalizations to assess

whether geographic or surgical practice patterns contributed to our observations of UV

exposure and IBD-related outcomes. Prolonged hospitalizations and deaths were

increased with low UV exposures in the non-IBD hospitalizations, so it is unclear how

much of the similar observations among IBD hospitalizations is due to increased IBD

severity across UV exposures or due to general hospitalization trends. Nonetheless, in contrast to IBD-specific hospitalizations, UV exposures in non-IBD hospitalizations did not influence rates of bowel surgeries, bowel resections, and colectomies. Moreover, rates of other abdominal surgeries unrelated to IBD, including cholecystectomies and appendectomies, were not influenced by UV exposure among IBD hospitalizations.

These findings further reinforce our observation that low UV and higher bowel surgery rates is a unique, and not random, phenomenon in IBD hospitalizations.

There are several limitations of this study. First, our analyses were constrained by data available in a large administrative database that includes information about hospitalizations (and does not identify when multiple hospitalizations occurred within a single individual) and does not have complete information on the potential confounder smoking nor other important disease outcomes such as endoscopic and radiologic findings, medication use, or readmissions. However, for confounding to occur there needs to be a relationship between the confounder and disease. Although smoking is strongly associated with CD and UC, there is no evidence that smoking is associated with day-to-day variability in UV, despite potential regional differences in smoking rates.

Similarly, we were unable to identify a known confounder strongly associated with both

UV and IBD or IBD outcomes that was likely to change our inference. Second, an

inherent limitation in administrative databases is the completeness and accuracy of ICD-9

coding for IBD, comorbidities, or procedures. Although the codes used in this study have

not been validated with the NIS, we selected codes that were either validated in other

databases or based on reasonable codes used in clinical practice.130,138 Third, some

records had to be excluded due to missing admission time or race data, which did not

permit us to calculate national estimates using weighted data. We were nonetheless still

able to use weighted regression analyses to make technically valid comparisons among

UV groups, provided we include at least one observation from each sampled hospital for

correct variance estimation.132

In summary, our findings demonstrate at the population level an association

between low UV exposure and increased IBD hospitalization rates and severity. Although

the precise mechanism for UV effect on IBD is currently unknown, the proposed mediator for this relationship is vitamin D. Further investigation using vitamin D intervention at the individual level would help clarify the causal relationship of vitamin D with disease activity and outcomes.

TABLES

Table 3-1. Demographics of Crohn’s disease, ulcerative colitis, and non- inflammatory bowel disease (IBD) hospitalizations Characteristic Crohn’s Disease Ulcerative Colitis Non-IBD Hospitalizations (n) 649,932 384,267 288,894,297 Median age (IQR) 40.7 (28.1 – 55.1) 46.1 (30.2 – 64.1) 53.9 (31.3 – 72.7) Female (%) 376,637 (58.1) 205,906 (53.8) 172,271,721 (59.8) Race (%) White 517,788 (81.9) 292,569 (78.8) 189,651,855 (67.8) Black 69,069 (10.9) 32,946 (8.9) 38,296,620 (13.7) Other 45,005 (7.1) 45,963 (12.4) 51,963,642 (18.6) Charlson-Deyo Index (%) 0 526,474 (81.0) 296,425 (77.1) 165,811,790 (57.4) 1-2 111,281 (17.1) 75,339 (19.6) 91,533,838 (31.7) ≥ 3 12,178 (1.9) 12,503 (3.3) 31,548,668 (10.9) Admission Type (%) Emergent/Urgent 460,630 (79.2) 243,725 (75.2) 168,803,647 (72.0) Elective 120,728 (20.8) 80,285 (24.8) 65,351,789 (27.9) Other 101 (0.0) 33 (0.0) 217,712 (0.1) Hospital Size (%) Small (1 – 249 beds) 85,131 (13.2) 46,673 (12.2) 38,736,810 (13.5) Medium (25 – 449 beds) 165,985 (25.6) 98,189 (25.6) 78,112,071 (27.1) Large (45 – 450+ beds) 396,040 (61.2) 238,187 (62.2) 171,125,893 (59.4) Abbreviations: IBD, inflammatory bowel disease; IQR, interquartile range.

FIGURES

Figure 3-1. Directed acyclic graph of ultraviolet (UV) exposure, vitamin D, and inflammatory bowel disease (IBD) severity. In the proposed directed acyclic graph to test the assumption that vitamin D influences IBD severity, UV exposure directly affects vitamin D levels, which subsequently affects IBD severity. UV exposure does not directly influence IBD severity and is therefore not technically considered a confounder.

Figure 3-2. Rates of Crohn’s disease (CD) and ulcerative colitis (UC) hospitalizations according to ultraviolet (UV) exposure. Low UV exposures were associated with higher rates of CD and UC hospitalizations. Rates were calculated as the number of CD or UC hospitalizations over all-cause hospitalizations.

Figure 3-3. Rates of hospitalization severity and death according to ultraviolet (UV) exposure. Low UV exposures were associated with increased rates of Crohn’s disease (CD) and ulcerative colitis (UC) severity outcomes: (A) all bowel surgeries; (B) bowel resection; (C); prolonged hospitalization (length > 7 days); and (D) death. For non-IBD hospitalizations, low UV exposures were not associated with rates of all bowel surgeries and bowel resection, but were associated with prolonged hospitalization and death.

Figure 3-4. Rates of other abdominal surgeries according to ultraviolet (UV) exposure. Low UV exposures were associated with decreased rates of cholecystectomies and appendectomies for non-inflammatory bowel disease (IBD) hospitalizations, but not for Crohn’s disease (CD) or ulcerative colitis (UC).

CHAPTER IV. VITAMIN D AS TREATMENT FOR INFLAMMATORY BOWEL

DISEASE

INTRODUCTION

While observational studies may provide supportive evidence of causality between vitamin D and IBD, they are potentially limited by unmeasured confounders. A classic example has been the previous association of lifestyle with the “English disease”

(rickets). In 1822, the Polish physician Jȩdrzej Sniadecki had written in his book “On the

Physical Education of Children” that exposure to “dry, open and pure air … especially in the sun” would cure rickets.139 While this may be true as a whole, we now know that UV exposure – and not the “dry, open and pure air – improves vitamin D levels and, in turn, bone health. To test causality between vitamin D and IBD, while minimizing unmeasured confounders, we thus performed a systematic review of randomized trials that evaluated the effect of vitamin D on IBD activity.

Description of the Condition

IBD is a chronic inflammatory disorder of the gastrointestinal tract. Clinical manifestations may include abdominal pain, cramping, diarrhea, and blood in stools. CD patients may also manifest strictures, abscesses, and/or fistulae. The incidence and prevalence of IBD have been increasing worldwide with the highest rates in Europe and

North America.140 The highest reported annual incidence of CD and UC are found in

Europe at 12.7 and 24.3 new diagnoses per 100,000 person-years, respectively. The

highest reported prevalence of CD and UC are 322 and 505 per 100,000 persons,

respectively. By contrast, the lowest overall incidence of CD (0.04 to 5.0 new diagnoses

per 100,000 person-years) and UC (0.1 to 6.3 new diagnoses per 100,000 person-years)

are found in Asia and the Middle East. The lowest prevalence for CD (0.9 to 67.9 per

100,000 persons) and UC (4.9 to 168.3 per 100,000 persons) are also found in these

regions. The mechanisms for the temporal increase in IBD incidence are unclear,

although some hypothesized reasons include lifestyle changes, urbanization, medication

exposure, and nutrition.

Description of the Intervention

Vitamin D is a fat-soluble hormone that is derived through sunlight exposure or oral consumption. The amount of vitamin D synthesis from sunlight (or UVB) exposure depends on several factors, such as duration of exposure, percentage of body surface area exposed, skin tone, latitude, season, and cloud cover. For the body to synthesize the equivalent of 1000 IU oral vitamin D, Caucasians with lighter skin tones and 25% body exposure would need 20 to 60 minutes under direct sunlight at noon in Boston

(Massachusetts, United States) in March, while dark-skinned African-Americans would need approximately 83 minutes to achieve the same dose of vitamin D synthesis.17

Oral consumption of vitamin D may include dietary sources or pharmacologic

supplementation. Foods rich in vitamin D include fish, beef liver, eggs, and vitamin D-

fortified food products. In the United States, the average daily intake from the diet is 200

to 240 IU; by comparison, the Institute of Medicine recommends a daily intake of 600 IU

for all persons ≤ 70 years old and 800 IU for persons above 70 years old.141 Vitamin D

supplementation is thus often needed to maintain normal vitamin D levels (> 30 ng/mL),

which approximately 37% of the population in the United States endorses taking.142

Vitamin D is often included in multivitamins, ranging from 50 to 1000 IU per pill.

Typical over-the-counter and prescription formulations may range from 400 IU per day to

50,000 IU per week. In IBD where intestinal malabsorption, dietary restrictions, and lifestyle changes may occur, the need for vitamin D supplementation may be even greater, although not clearly defined.143

How the Intervention Might Work

Vitamin D has traditionally been known for its prominent role in calcium and phosphorus

homeostasis, although it has been more recently implicated in immune function. At the

molecular level, vitamin D participates in regulating immune cell differentiation and

proliferation.37,42,43,49 In turn, vitamin D deficiency has been associated with the

pathogenesis of several autoimmune diseases, including EAE, RA, and MS.66,69 Similarly

in IBD, mice with a vitamin D receptor knockout have been shown to develop severe

gastrointestinal inflammation,83,84 while administration of exogenous vitamin D or an

analog reduces expression of pro-inflammatory cytokines and lymphocyte infiltration in

the lamina propria of a DSS-induced colitis mouse model.82 Epidemiologic studies have

additionally associated vitamin D deficiency with increased risk of incident disease,

hospitalization and surgery.102,103,144

Importance of this Review

The currently available immunosuppressive drugs do not induce and maintain remission

in all patients, while additionally possessing significant adverse effect profiles, such as

increased risk of infections and malignancy. Vitamin D possesses immunomodulatory

properties that may play a role in IBD pathogenesis and disease severity. This systematic

review will therefore evaluate the effect of vitamin D supplementation on IBD activity.

Results from this review will help determine whether existing data support the use of

vitamin D as a potential economical, low-risk, adjunctive therapy for the treatment of

IBD.

METHODS

Selection Criteria

We reviewed RCTs of vitamin D administration, as a primary or secondary comparator, in the treatment of CD and UC patients. Abstracts and peer-reviewed journal articles published in all languages were eligible. Studies that did not include disease activity outcomes were excluded. We included children and adults with a clinical diagnosis of

IBD. We were primarily interested in evaluating trials for CD and UC separately, but did not exclude trials that aggregated CD and UC patients in their results. We did not encounter trials that mixed IBD and non-IBD participants.

The interventions included all forms of vitamin D supplementation, including vitamin D-specific and combination formulations. One study included a vitamin D analog: alfacalcidol. The studies generally compared vitamin D supplementation with placebo, but some included vitamin D at different doses. Studies that used equivalent doses of vitamin D across treatment arms were excluded from this systematic review, as they would be unable to demonstrate any comparable effect – even if present – of vitamin

D supplementation on IBD-related outcomes.

The primary outcomes of interest were all reported measures of disease activity and inflammation. Disease activity was reported as changes in severity index (i.e.,

Harvey-Bradshaw Index [HBI], Crohn’s Disease Activity Index [CDAI], Pediatric

Crohn’s Disease Activity Index [PCDAI], Mayo UC score) and clinical relapse.

Inflammation was reported through serologic markers (i.e., erythrocyte sedimentation rate [ESR] or CRP), endoscopic findings, or histology. Secondary outcomes were vitamin

D levels, newly reported adverse events, and changes in quality of life. Vitamin D levels were variably reported as relative changes in levels, measured levels at various intervals, or differences of levels between treatment arms. Adverse events were reported as individual symptoms experienced within each treatment arm, although not necessarily distinguishing the effects specifically from vitamin D or other drugs used in the treatment arm. Quality of life measures included changes in the standard IBD Questionnaire score or Short IBD Questionnaire score.

Search Strategy

We searched the following databases since their inception through 14 October 2014:

MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials. The electronic search strategy used is outlined in Appendix 1. All languages were eligible for inclusion. We checked reference lists of all primary studies and review articles for additional references. We also searched ClinicalTrials.gov (clinicaltrials.gov) and the

World Health Organization International Clinical Trials Registry Platform

(apps.who.int/trialsearch) for relevant trials.

All titles and abstracts were screened for potential inclusion. Abstracts that were

not published in peer-review form were included in this systematic review. Potentially

eligible abstracts that were published in peer-review journals underwent full-text review for determination of inclusion or exclusion.

Data Extraction

Data on the study design, patient characteristics, interventions, and outcomes were abstracted using a standardized extraction form. Data collected included the following:

• Study design – type of trial, method of allocation concealment, method of

randomization, method of blinding patients, providers and outcome assessors,

months and years of study recruitment and date of trial completion;

• Intervention and comparator details – number of people randomized and analyzed

for each group, type of vitamin D, brand of product, dose, administration method,

frequency of use, and adherence;

• Participants – country, total number of participants, age at randomization, age at

diagnosis, sex, disease state at baseline (active, quiescent), concomitant

medications at baseline and during the trial, prior medications, presence of fistula,

baseline vitamin D level, baseline disease activity level, inclusion and exclusion

criteria;

• Study-specified primary outcome and details on the power calculation for that

outcome and time point;

• Method used to handle missing data;

• Primary and secondary outcomes for each time point of interest;

• Duration of treatment and length of follow-up; and

• Funding for the trial and industry support for study authors.

Due to the heterogeneity of methods used to report outcomes, outcome measures were

recorded as free text in the following groups: (i) disease activity; (ii) inflammation; (iii)

vitamin D levels; (iv) adverse events; and (v) quality of life.

Assessment of Risk of Bias

Each included study was assessed for risk of bias using criteria outlined in the Cochrane

Handbook for Systematic Reviews of Interventions.145 Items that were assessed included the following: (i) random sequence generation; (ii) allocation concealment; (iii) blinding of participants and personnel; (iv) blinding of outcome assessment; (v) incomplete outcome data; (vi) selective outcome reporting; and (vii) other biases, including study size. We graded each potential source of bias as good (low risk of bias), fair, poor (high risk of bias), or unclear. Justification was provided for each graded risk of bias assessment. For items with incomplete reporting to reliably assess risk of bias, these were designated as having an unclear risk of bias.

Statistical Analysis

The unit of analysis was each randomized participant. Missing data and imputation methods were reported in the assessment of risk of bias. The I2 statistic was used to

quantitatively measure heterogeneity among trials in the meta-analysis. We also performed a qualitative assessment of clinical and methodologic heterogeneity. There

were 4 studies included in the meta-analysis. We used reported results from each study

68 based on intention to treat and evaluated the relative odds of clinical relapse over a 12- month follow-up period. Clinical relapse was assessed based on the study’s reported outcome of relapse, absence of remission, or abnormal CRP throughout the follow-up period. Due to significant methodologic heterogeneity among studies, we used a random- effects model for the primary meta-analysis and a fixed-effects model for sensitivity analysis. Two of the 4 studies aggregated CD and UC patients, which limited the ability to perform subgroup analyses according to IBD type. Results were tabulated separately for CD and UC, where possible. If we had at least 10 trials to compare, we would have created a funnel plot to assess for reporting bias. Statistical significance was defined as a two-sided α of less than 0.05. Statistical analyses were performed using RevMan 5.3.5

(Cochrane Collaboration, United Kingdom).

RESULTS

Search Results

The electronic search strategy generated 235 abstracts and journal articles for screening

(Figure 4-1). Of these, 80 duplicates were removed and 132 were excluded after screening the titles and abstracts. There were 3 abstracts that met criteria for inclusion and 20 full-text journal articles to be reviewed. Eleven of the full-text articles were excluded due to lack of appropriate comparison arms or outcomes of interest, leaving 3 abstracts and 9 journal articles for inclusion: Baars 2010;146 Bernstein 1996;147 Boothe

2011;148 Dendrinos 2008;149 Jorgensen 2010;104 Kitazaki 2009;150 Mathur 2014;151 Pappa

2012;152 Pappa 2014;153 Raftery 2013;154 Vogelsang 1995;155 and Wingate 2014.156 Seven of the 12 studies were performed in the United States (Bernstein 1996, Boothe 2011,

Dendrinos 2008, Mathur 2014, Pappa 2012, Pappa 2014) and Canada (Wingate 2014).

Four studies were performed in Europe (Baars 2010, Jorgensen 2010, Raftery 2013,

Vogelsang 1995) and one was performed in Japan (Kitazaki 2009). Table 4-1 details the characteristics of the included studies.

Participants

The twelve studies involved a total of 583 participants. Six studies included 315 CD-only participants (Boothe 2011, Dendrinos 2008, Jorgensen 2010, Raftery 2013, Vogelsang

1995, Wingate 2014). Two studies included 57 UC-only participants (Kitazaki 2009,

Mathur 2014). Four studies included 211 IBD participants (Baars 2010, Bernstein 1996,

Pappa 2012, Pappa 2014) who were not differentiated between CD or UC. Of the 9 journal articles, 3 studies were performed in children with a mean age ranging from 14.3 to 16. 3 years (Pappa 2012, Pappa 2014, Wingate 2014); the other studies were performed in adults. There were 243 (46.8%) males and 276 (53.2%) females in the journal articles. The 3 abstracts included did not report data on age or gender demographics. When reported, the study population were predominantly white.

Intervention

All studies compared vitamin D at different doses or vitamin D with a non-vitamin D control. There was significant heterogeneity in the vitamin D doses used, ranging from

250 IU/day to 50,000 IU/week. One study compared a vitamin D analog, alfacalcidol, at

1 µg/day with alendronate 5 mg/day (Kitazaki 2009). The duration of treatment was similarly heterogeneous, ranging from 3 to 12 months.

Outcomes

For disease activity, the reported outcomes were either measured as the presence of active

disease (relapse) or change in disease activity score (i.e., HBI, CDAI, PCDAI, Mayo UC

score) during the study. For inflammation, serologic markers (i.e., ESR, CRP) were

reported in a few studies (Dendrinos 2008, Mathur 2014, Pappa 2012, Pappa 2014,

Raftery 2013, Wingate 2014). Endoscopic and histologic inflammation was noted in one study (Baars 2010). Vitamin D levels were reported as the mean levels at time of follow- up or the change from baseline. Adverse events were generally reported qualitatively with a list of symptoms. Laboratory changes in the setting of possible vitamin D toxicity, such as hypercalcemia or hyperphosphatemia, were also noted. A few studies compared the number of adverse events between intervention arms (Pappa 2012, Pappa 2014,

Wingate 2014). Quality of life was assessed via the standard or short IBD Questionnaire scoring system. Change in quality of life was reported in two studies (Mathur 2014,

Raftery 2013).

Excluded Studies

Eleven journal articles were excluded after full-text review: Abitbol 2002;157 Abitbol

2007;158 Bendix-Struve 2010;159 Bentley 2011;160 Klaus 2011;161 Pappa 2011;162

Siffledeen 2005;163 Soo 2012;164 van Bodegraven 2014;165 von Tirpitz 2000;166 and von

Tirpitz 2003.167 Table 4-2 lists the excluded studies and their respective reasons for

exclusion. Eight had no difference in vitamin D dose among intervention arms, thus

leading to an inability to assess any effect of vitamin D supplementation (Abitbol 2002,

Abitbol 2007, Pappa 2011, Siffledeen 2005, Soo 2012, van Bodegraven 2014, von Tirpitz

2000, von Tirpitz 2003). One study’s methods were vague, so we were unable to

determine whether the vitamin D dose differed across intervention arms (Klaus 2011).

One study was not an intervention trial (Bentley 2011) and one study was a subset of the

included study by Jorgensen et al (Bendix-Struve 2010).

Risk of Bias in Included Studies

Figure 4-2 outlines the risk of bias assessments for each of the included studies.

Randomization

All studies were randomized, but utilized different methods to generate the random sequence or conceal the random allocation. Block randomization was used in 6 studies

(Baars 2010, Bernstein 1996, Jorgensen 2010, Pappa 2012, Pappa 2014, Raftery 2013).

Insufficient details were provided in 3 journal articles (Kitazaki 2009, Vogelsang 1995,

Wingate 2014), while all 3 abstracts had no information on the randomization method

(Boothe 2011, Dendrinos 2008, Mathur 2014).

Allocation concealment

Allocation concealment was done by central allocation (Bernstein 1996, Jorgensen 2010,

Pappa 2012, Pappa 2014, Raftery 2013, Wingate 2014) or numbered, opaque, and sealed envelopes (Baars 2010). Insufficient details were provided in 2 journal articles (Kitazaki

2009, Vogelsang 1995) and no details were available in all 3 abstracts (Boothe 2011,

Dendrinos 2008, Mathur 2014).

Blinding

There were 5 studies that were double-blind (Bernstein 1996, Jorgensen 2010, Mathur

2014, Raftery 2013, Wingate 2014). Two studies blinded the outcomes assessors (Baars

2010, Boothe 2011), but had no details on the blinding of participants and personnel. The participants were likely not blinded. Two studies by the same investigators (Pappa 2012,

Pappa 2014) intentionally did not blind the outcomes for perceived safety reasons. Two journal articles (Kitazaki 2009, Vogelsang 1995) and one abstract (Dendrinos 2008) had no details on blinding.

Incomplete outcome data

All participants completed the study in 3 studies (Baars 2010, Mathur 2014, Raftery

2013). Where reported or able to be estimated, the attrition rate was otherwise high:

• 14% (Pappa 2012)

• 17% (Jorgensen 2010, Wingate 2014)

• 18% (Kitazaki 2009)

• 20% (Vogelsang 1995)

• 24% (Pappa 2014)

• 29% (Bernstein 1996)

For the other studies where the attrition rate was not reported or unable to be estimated, the risk of bias was noted to be unclear (Boothe 2011, Dendrinos 2008). One study reported a 24% attrition rate, but complete primary outcomes data were available in only

54% of participants (Pappa 2014). Missing data in the other studies were not clearly reported.

Selective reporting

There was no clear evidence of the presence or absence of selective reporting. The

studies listed their reported outcomes, but we did not have access to the original protocol

data.

Other potential sources of bias

All studies had small sample sizes. None had a treatment arm with at least 50 participants

and the largest trial had 94 participants (Jorgensen 2010). The relapse rate of CD in this

study was numerically different in both arms (13% vs. 29%) and almost reached

statistical significance (P = 0.056), suggesting that there may have been a limitation of

power or sample size.

Effects of interventions

Disease activity

There were 5 studies that reported disease activity outcomes in CD-only participants

(Boothe 2011, Jorgensen 2010, Raftery 2013, Vogelsang 1995, Wingate 2014) (Table 4-

3). Two studies found significantly lower disease activity scores among the vitamin D treatment arms (Boothe 2011, Vogelsang 1995) and the largest trial of 94 participants showed a trend in favor of vitamin D treatment (Jorgensen 2010). The other two studies otherwise found no difference (Raftery 2013, Wingate 2014). These aforementioned studies tested a wide range of vitamin D doses (range: 400 IU/day to 10,000 IU/day) at

different durations of follow-up (range: 3 to 12 months). There were 2 studies that

reported disease activity outcomes in UC-only participants (Kitazaki 2009, Mathur 2014)

(Table 4-4). One study reported a non-significant decrease in Mayo UC partial score

(without endoscopic subscore) when comparing vitamin D at 2000 IU/day and 4000

IU/day (Mathur 2014). There were however only 18 participants in this study. The other study reported no difference in disease activity, but did not provide data (Kitazaki 2009).

There was 1 study that reported disease activity outcomes, but did not disaggregate CD and UC participants (Bernstein 1996). After 12 months of observation, there was no difference in clinical relapse between the use of vitamin D at 250 IU/day and placebo.

This study was nonetheless small (17 participants) and used a very low dose of vitamin

Given the methodologic heterogeneity among studies, we were only able to pool the analyses of 4 studies that had included 12-month follow-up periods (Bernstein 1996,

Jorgensen 2010, Pappa 2014, Vogelsang 1995). The outcome of interest was clinical relapse within the 12-month follow-up. One study did not formally state the incidence of clinical relapse, but indicated whether the CRP level ever exceeded 1 mg/dL during the trial (Pappa 2014). For the purposes of our meta-analysis, this was assumed to signify a serologic recurrence of inflammation. The pooled analysis using a random-effects model showed a small non-significant trend toward lower risk of clinical relapse (OR 0.56; 95%

CI 0.24 – 1.29; P = 0.17) when comparing vitamin D across doses or with controls

(Figure 4-3). There were 249 participants included in this analysis, 169 of whom were

CD participants and 80 were undifferentiated IBD participants. Sensitivity analysis using a fixed-effects model provided similar results (OR 0.54; 95% CI 0.29 – 1.01; P = 0.05).

The largest CD study (94 participants), which also had the least risk of bias, reported a non-significant trend toward vitamin D protecting against relapse (hazard ratio 0.42; 95%

CI 0.16 – 1.10; P = 0.056) (Jorgensen). This observation suggests that the non-significant

findings may have been due to small sample sizes and lack of power.

Inflammation

There were 3 studies that reported inflammatory marker levels in CD-only participants

(Dendrinos 2008, Raftery 2013, Wingate 2014) (Table 4-5). All 3 studies found no difference in ESR or CRP levels among treatment groups. There was only one study that reported inflammatory marker levels in UC-only participants (Mathur 2014) (Table 4-6).

The study found a difference in the change of CRP comparing vitamin D at 4000 IU/day with 2000 IU/day (-10.8 vs. -3.0), but this did not reach statistical significance (P = 0.36).

There were 3 studies that reported inflammatory marker levels, but did not disaggregate

CD and UC participants (Baars 2010, Pappa 2012, Pappa 2014). One study found that the higher vitamin D doses (1000 or 2000 IU/day) compared with the lower vitamin D dose

(400 IU/day) was associated with fewer participants with a CRP ≥ 1 mg/dL (10% vs.

31%; P = 0.04) (Pappa 2014). There were also fewer participants with an ESR ≥ 30 mm/hr (3% vs. 20%; P = 0.10), but this did not reach statistical significance, which may have reflected an issue with sample size. The other studies did not find a difference in the change of CRP levels (Pappa 2012) or endoscopic/histologic inflammation (Baars 2010) across treatment arms.

Vitamin D levels

Comparison of vitamin D level changes across studies was challenged by the significant heterogeneity in drug dosing, duration of follow-up, and reporting methods. For CD, 4

CD-specific studies associated higher vitamin D dosing with a greater increase in

25(OH)D levels (Boothe 2011, Jorgensen 2010, Raftery 2013, Wingate 2014), while the other 2 studies showed no difference in 25(OH)D levels between treatment arms

(Dendrinos 2008, Vogelsang 1995) (Table 4-7). For UC, the only UC-specific study showed a significantly greater increase in 25(OH)D levels when comparing vitamin D3 supplementation at 4000 IU/day with 2000 IU/day (Mathur 2014) (Table 4-8). In the aggregated IBD studies, one showed a significantly greater increase with vitamin D2 at

50,000 IU/week compared with vitamin D3 at 2000 IU/day or vitamin D2 at 2000 IU/day

(Pappa 2012). The other two studies showed no difference in 25(OH)D levels between

treatment arms (Bernstein 1996, Pappa 2014). One of these latter two studies showed no

difference in the 25(OH)D levels in the vitamin D group, but noted a decline in 25(OH)D

levels for the placebo group (Bernstein 1996). Nonetheless, the vitamin D dose in this

trial was the lowest of all included studies at only 250 IU/day, much lower than the

Institute of Medicine recommendations of 600 IU daily intake for persons ≤ 70 years old.

Adverse events

Adverse events were variably reported and generally not statistically compared. For CD,

there were 3 CD-specific studies that reported potential adverse effects (Dendrinos 2008,

Jorgensen 2010, Wingate 2014) (Table 4-9). One study noted an increase in diarrhea among those receiving vitamin D2 at 50,000 IU/week compared with 400 IU/week

(Dendrinos 2008), while the other 2 studies found no adverse effects attributable to

vitamin D intake. For UC, the one UC-specific study that reported potential adverse effects noted muscle pain in 1 of 20 participants receiving alfacalcidol (Kitazaki 2009)

(Table 4-10). In the aggregated IBD studies, there were no significant increases in

adverse events when receiving vitamin D (Baars 2010, Bernstein 1996, Papa 2012, Pappa

2014). One study included prednisone, which may explain the reported mood swings (4),

swollen face (3), increased appetite (2), and edema (1) (Baars 2010).

Quality of life

For CD, there was no improvement of IBD Questionnaire scores when comparing

vitamin D3 at 2000 IU/day with placebo (Raftery 2013) (Table 4-11). For UC, there was a ten-fold difference in the change in short IBD Questionnaire scores when comparing vitamin D3 at 4000 IU/day with 2000 IU/day, although this was not significant (Mathur

2014) (Table 4-12). The sample sizes for both studies were nonetheless very small.

DISCUSSION

This systematic review included 12 randomized trials, involving 583 participants, where

administration of vitamin D was compared at different doses or with placebo and where

effects on IBD activity were reported. A pooled analysis of 12-month studies showed a

trend toward vitamin D reducing the risk of clinical relapse, but the estimates were not

statistically significant. For the secondary outcomes, vitamin D supplementation was

shown to improve vitamin D levels. There were no significant adverse effects reported

that were attributable to vitamin D, suggesting that vitamin D supplementation is safe.

The data on changes in quality of life were sparse for both CD and UC.

At present, there is insufficient evidence to indicate that vitamin D improves CD or UC activity. Nonetheless, this statement does not necessarily reflect a lack of clinical

benefit from vitamin D supplementation on IBD activity, but the methodologic

differences and challenges among the included studies that have precluded clarification

of our hypothesis. For one, there was substantial heterogeneity in methodologies, such as drug dosing, follow-up duration, and reported outcomes. As an illustration of these marked differences, there was a 28-fold difference between the highest (50,000 IU/week) and lowest (250 IU/day) doses of vitamin D used in the intervention arms. The follow-up duration also widely ranged from 2 to 52 weeks. These characteristics may explain the inconsistency across studies where vitamin D supplementation was found to either have some or no benefit; none of the studies showed vitamin D to worsen measured outcomes.

Secondly, all the studies had small sample sizes, which could lead to a Type II error and/or bias effect estimates in individual studies. Half of the studies had < 20 participants in at least one treatment arm and all studies had < 50 participants per treatment arm. The largest trial with less risk of bias involved 94 participants and showed a trend toward lower clinical relapse when comparing vitamin D at 1200 IU/day versus placebo.104

While meta-analyses are intended to address these limitations in sample size, our pooled

analysis had modest numbers of 123 and 126 participants in the treatment arms. In turn,

the summary estimates showed a non-significant trend of benefit with vitamin D

supplementation. Thirdly, although randomized controlled trials are intended to reduce

the risk of confounding and bias, they cannot eliminate bias introduced through the study

design, execution, setting, or other circumstantial factors. The included studies generally

had low risk of bias in randomization, allocation, and blinding. There was nonetheless a

high risk of bias from the lack of blinding in some studies, as they deemed blinding to be

unethical. There was also a high risk of bias from attrition. Missing data were also poorly

reported and thus unclear. The only study to report completeness of primary outcomes

data collection had admitted to having them in only 54% of participants.153

In summary, the lack of consensus among the randomized trials may largely stem from the significant heterogeneity among studies and their small sample sizes. We additionally acknowledge the overall poor quality of available evidence for our primary outcomes of interest. Given the emerging signal that vitamin D could potentially reduce clinical relapse, further investigation using more consistent methodologies (dose, duration, disease activity score) and larger sample sizes are warranted to more precisely test the hypothesis that vitamin D supplementation improves IBD activity. Independent of this relationship, correction of vitamin D deficiency would still be helpful for other non- immune benefits, such as bone health. Vitamin D is also generally regarded as a safe and cost-effective supplement, which would tip the risk-benefit balance toward supplementation, especially in the presence of vitamin D deficiency that can often occur in IBD.

TABLES

Table 4-1. Characteristics of included studies.

Baars 2010 Methods Randomized controlled trial. Duration: 2 weeks. Participants IBD patients scheduled for surveillance colonoscopy between July 1, 2008 and December 31, 2009 in the Erasmus Medical Center (Rotterdam, Netherlands) or the Tweesteden Hospital (Tilburg, Netherlands). Interventions Prednisone 20 mg/day, Calcium, Vitamin D (n = 31) No treatment (n = 29) Outcomes Primary outcome: Active inflammation on histopathology biopsies Secondary outcomes: Safety Notes Median age: 46.4 Male/female distribution: 31/29 Study conducted in the Netherlands

Authors' Bias Support for judgement judgement Random sequence generation Low risk Block randomization (selection bias) Allocation concealment Low risk Allocation by numbered, opaque, sealed (selection bias) envelopes Blinding of participants and High risk No details on blinding of participants and personnel (performance bias) personnel. Participants were likely not blinded. Blinding of outcome Low risk Pathologist blinded to treatment assignment assessment (detection bias) Incomplete outcome data Low risk All randomized participants completed the (attrition bias) study. Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Bernstein 1996 Methods Randomized placebo-controlled study. Duration: 12 months

Participants Patients with Crohn's disease and ulcerative colitis seen at the UCLA Inflammatory Bowel Disease Center between March 1992 and August 1993 Inclusion criteria: Patients on corticosteroids for ≥ 3 of the 6 prior months Exclusion criteria: 1. Age < 18 years old 2. Postmenopausal women (natural or surgical) 3. Chronic liver disease (known diagnosis or unexplained liver enzyme elevations) 4. Chronic renal insufficiency (serum creatinine > 1.5 mg/dL) 5. Primary hyper- or hypoparathyroidism 6. Untreated thryoid disease 7. Deforming arthritis 8. Paget's disease of bone 9. On phenytoin or sex hormone therapy 10. Bed- or wheelchair-bound Interventions Calcium carbonate 1000 mg/day and Vitamin D 250 IU/day (n = 9) Placebo (n = 8) Outcomes Primary outcome: 1. Bone density Secondary outcomes: 1. Hemoglobin 2. Calcium 3. Phosphate 4. Magnesium 5. Albumin 6. Protein 7. Alkaline phosphatase 8. 1,25(OH)2D 9. 25(OH)D 10. Parathyroid hormone 11. Osteocalcin 12. Urea nitrogen 13. Creatinine 14. Aspartate aminotransferase 15. Alanine aminotransferase 16. Bilirubin 17. Cholesterol 18. Urinary calcium and creatinine 19. Patient compliance 20. Side effects Notes Mean ages: 35.3 years (treatment group), 36.4 (control group) Male/female distribution: 9/0 (treatment group), 5/3 (control group)

Study conducted in California (USA)

Authors' Bias Support for judgement judgement Random sequence generation Low risk Block randomization in groups of four (selection bias) Allocation concealment Low risk Central allocation of assignments by (selection bias) pharmacist Blinding of participants and Low risk Double-blinded trial personnel (performance bias) Blinding of outcome Low risk Double-blinded trial assessment (detection bias) Incomplete outcome data High risk 7/24 participants dropped out of the study. (attrition bias) Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Boothe 2011 Methods Randomized study. Duration: 26 weeks Participants Patients with Crohn's disease with 25(OH)D < 30 ng/mL (n = 15) Interventions Vitamin D 1000 IU/day Vitamin D 10,000 IU/day Outcomes Primary outcome: Harvey-Bradshaw Index Secondary outcome: Vitamin D level Notes Limited demographic data available Study conducted in New York (USA)

Incomplete outcome data Unclear risk Details not provided (attrition bias) Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Dendrinos 2008 Methods Randomized study. Duration: 7 weeks Participants Adults with Crohn's disease presenting for routine office visits to Boston Medical Center from April 2007 to November 2007 Interventions Vitamin D2 at 400 IU/week (n = 7) Vitamin D2 at 50,000 IU/week (n = 14) Outcomes Primary outcome: 1. Serum 25(OH)D Secondary outcomes: 1. Calcium 2. Hematocrit 3. ESR 4. CRP 5. Serum cytokines (GMCSF, IFN-gamma, IL-10, IL-1beta, IL-s, IL-4, IL-5, IL-6, IL-8, MCP-1, MIP-1alpha, TNF-alpha) Notes No data on age, gender, or race distribution. Study conducted in Massachussetts (USA).

Authors' Bias Support for judgement judgement Random sequence generation Unclear risk No details provided (selection bias) Allocation concealment Unclear risk No details provided (selection bias) Blinding of participants and Unclear risk No details provided personnel (performance bias) Blinding of outcome Unclear risk No details provided assessment (detection bias) Incomplete outcome data Unclear risk No details provided (attrition bias) Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Jorgensen 2010 Methods Randomized, double-blind, placebo-controlled study. Duration: 12 months Participants Adult patients (ages 18 or older) with Crohn's disease in remission seen at one of four hospitals in Denmark between September 2005 to February 2008. Remission was defined as a Crohn's Disease Activity Index < 150 with normal CRP and albumin. Remission should have been present without corticosteroids for at least 4 weeks. Patients should have normal serum calcium. Exclusion criteria: 1. Pregnancy 2. Short bowel syndrome Interventions Vitamin D3 1200 IU/day and Calcium 1200 mg/day (n = 46) Calcium 1200 mg/day (n = 48) Outcomes Primary outcome: Relapse (Crohn's Disease Activity Index ≥ 150 or an increase > 70 compared with baseline) Secondary outcome: 25(OH)D level Notes Mean ages: 36 years (treatment group), 38 (control group) Gender distribution: 33/46 female (treatment group), 29/48 (control group) Study conducted in Denmark

Authors' Bias Support for judgement judgement Random sequence generation Low risk Block randomization in groups of 10 (selection bias) Allocation concealment Low risk Central allocation of assignments by coded (selection bias) medication containers Blinding of participants and Low risk Double-blinded trial personnel (performance bias) Blinding of outcome Low risk Double-blinded trial assessment (detection bias) Incomplete outcome data High risk 16/94 participants dropped out of the study. (attrition bias) Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Kitazaki 2009 Methods Randomized study. Duration: 12 months Participants Patients with ulcerative colitis seen at Kurume University Hospital (Kurume, Japan) who were either starting or had started glucocorticoid therapy at a daily dose of at least 5 mg of prednisone or its equivalent Exclusion criteria: 1. Abnormal calcium 2. Abnormal bone-relate biochemistry (e.g., vitamin D defciency) 3. Concomitant liver disease 4. Concomitant renal disease 5. Pregnancy or breastfeeding 6. Osteoporotic fractures 7. On hormone replacement therapy 8. Previous bisphosphonate 9. Active vitamin D3 therapy Interventions Alendronate 5 mg/day (without calcium) (n = 19) Alfacalcidol 1 μg/day (without calcium) (n = 20) Outcomes Primary outcome: Bone mineral density at baseline and at 6 and 12 months Secondary outcomes: 1. Bone metabolism markers 2. Bone alkaline phosphatase 3. Urinary type I collagen cross-linked N-telopeptides 4. Serum calcium 5. Disease activity (Truelove and Witts) Notes Mean ages: 38.1 years (alfacalcido group), 41.2 (alendronate group) Male/female distribution: 12/8 (alfacalcidol group), 10/6 (alendronate group) Study conducted in Japan

Authors' Bias Support for judgement judgement Random sequence generation Unclear risk Insufficient details on random sequence (selection bias) generation Allocation concealment Unclear risk Insufficient details on allocation (selection bias) concealment using envelope method Blinding of participants and Unclear risk No details provided personnel (performance bias) Blinding of outcome Unclear risk No details provided assessment (detection bias) Incomplete outcome data High risk 7/39 randomized participants dropped out (attrition bias) of the study.

Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Mathur 2014 Methods Randomized study. Duration: 90 days Participants Patients with ulcerative colitis and a serum 25(OH)D level < 30 ng/mL at the Community Regional Medical Center (Fresno, California) Exclusion criteria: 1. Age < 18 years 2. Pregnancy 3. On vitamin D supplementation Interventions Vitamin D3 2000 IU/day (n = 8) Vitamin D3 4000 IU/day (n = 10) Outcomes 25(OH)D CRP Mayo UC disease activity (excluding endoscopy subscore) Short IBD Questionnaire Notes Age and gender distribution unknown. Study conducted in California (USA).

Authors' Bias Support for judgement judgement Random sequence generation Unclear risk Details not provided (selection bias) Allocation concealment Unclear risk Details not provided (selection bias) Blinding of participants and Low risk Double-blinded trial personnel (performance bias) Blinding of outcome Low risk Double-blinded trial assessment (detection bias) Incomplete outcome data Low risk All patients completed the trial. (attrition bias) Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Pappa 2012 Methods Randomized unblinded study. Duration: 6 weeks Participants Children (ages 5 - 21 years) with IBD and 25(OH)D ≤ 20 ng/mL within 8 weeks of enrollment Exclusion criteria: 1. Liver failure 2. Kidney failure 3. On anticonvulsant metabolized through cytochrome P450 4. Pregnancy 5. Inability to take oral medications 6. Attendance at tanning salons once weekly or more 7. On treatment for hypovitaminosis D Interventions A: Vitamin D2 2000 IU/day (n = 24) B: Vitamin D3 2000 IU/day (n = 24) C: Vitamin D2 50000 IU/week (n = 23) Outcomes Primary outcome: 25(OH)D level Secondary outcomes: 1. Parathyroid hormone 2. Hypercalciuria (urine calcium to creatinine ratio ≥ 0.20) 2. Hyperphosphatemia (serum phosphate > 5.7 mg/mL) 3. Hypercalcemia (serum calcium > 10.5 mg/dL) 4. Serum 25(OH)D > 88 ng/mL Notes Mean ages: 15.9 years (A), 14.7 years (B), 16.3 years (C) Male distribution: 58% (A), 42% (B), 61% (C) Non-Hispanic Caucasion distribution: 58% (A), 42% (B), 61% (C) Study conducted in Massachussetts (USA)

Authors' Bias Support for judgement judgement Random sequence generation Low risk Block randomization (selection bias) Allocation concealment Low risk Central allocation of assignments (selection bias) Blinding of participants and High risk No blinding personnel (performance bias) Blinding of outcome High risk No blinding assessment (detection bias) Incomplete outcome data High risk 10/71 participants dropped out of the study. (attrition bias) Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Pappa 2014 Methods Randomized, unblinded study. Duration: 12 months Participants Children (ages 5 - 21 years) with IBD and 25-hydroxyvitamin D levels > 20 ng/mL within 8 weeks of enrollment. Exclusion criteria: 1. Inability to take oral medications 2. Pregnancy 3. Liver or kidney failure 4. Use of antiepileptic medications metabolized through cytochrome P450 Interventions Vitamin D2 400 IU/day (n = 32) Vitamin D2 1000 IU/day (between May 1 and October 31) and 2000 IU/day (between November 1 and April 30) (n = 31) Outcomes Primary outcome: Probability of maintaining serum 25(OH)D ≥ 32 ng/mL at all 4 follow-up visits at 3-month intervals Secondary outcomes: 1. Hypercalciuria (urine calcium to creatinine ratio > 0.20) 2. Hyperphosphatemia (serum phosphate > 5.7 mg/mL) 3. Hypercalcemia (serum calcium > 10.5 mg/dL) 4. Serum 25(OH)D > 88 ng/mL 5. ESR 6. CRP 7. Serum IL-6 Notes Mean age: 14.8 years (standard deviation 3.1) Gender distribution: 57% female Race: 94% Non-Hispanic Caucasian Study conducted in Massachussetts (USA)

Authors' Bias Support for judgement judgement Random sequence Low risk Block randomization generation (selection bias) Allocation concealment Low risk Central allocation of assignments by (selection bias) pharmacist Blinding of participants High risk and personnel No blinding (performance bias) Blinding of outcome High risk No blinding assessment (detection bias) Incomplete outcome data High risk 15/63 participants dropped out of the study. (attrition bias) Complete primary outcomes data were only available in 34/63 participants.

Selective reporting Unclear risk Listed outcomes were reported. However, we (reporting bias) did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Raftery 2013 Methods Randomized, double-blind, placebo-controlled study. Duration: 3 months Participants Adult (ages 18 years or older) patients with Crohn's disease in remission at Tallaght Hospital (Dublin, Ireland). Remission was defined as a Crohn's Disease Activity Index < 150, CRP < 5 mg/L, and stable Crohn's disease therapy for at least 3 months. Exclusion criteria: 1. Extensive bowel resection 2. Hypersensitivity to vitamin D 3. History of hypercalcemia (corrected serum calcium > 2.66 mmol/L) 4. Supplemental vitamin D intake > 1000 IU/day 5. Antibiotic use within 4 weeks prior to enrollment 6. Renal impairment 7. Diabetes mellitus 8. Alcohol dependency 9. Urinary tract infection 10. Pregnancy 11. Short bowel syndrome 12. On bisphosphonates Interventions Vitamin D3 2000 IU/day (n = 13) Placebo (n = 14) Outcomes 25(OH)D Intestinal permeability LL-37 CRP Fecal calprotectin Crohn's Disease Activity Index Quality of life (Inflammatory Bowel Disease Questionnaire) Notes Mean ages: 36.5 years (treatment group), 36.7 (control group) Male/female distribution: 7/8 (treatment group), 6/8 (control group) Study conducted in Ireland

Authors' Bias Support for judgement judgement

Random sequence generation Low risk Block randomization in groups of 10 (selection bias) Allocation concealment Low risk Central allocation of assignments (selection bias) Blinding of participants and Low risk Double-blinded trial personnel (performance bias) Blinding of outcome Low risk Double-blinded trial assessment (detection bias) Incomplete outcome data Low risk All randomized patients completed the (attrition bias) study. Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Vogelsang 1995 Methods Randomized placebo-controlled study. Duration: 12 months Participants Adult patients with Crohn's disease who were unlikely to need hospitalization or surgery within the subsequent months Exclusion criteria: 1. Receiving estrogens 2. Receiving cholestyramine 3. Receiving calcitonin 4. Receiving fluorides Interventions Vitamin D3 1000 IU/day (n = 37) No supplementation (n = 38) Outcomes Outpatient follow-up visits every 3 months Primary outcome: Bone mineral content of the radius was measured before treatment and after 1 year Secondary outcomes: Serum alkaline phosphatase, calcium, phosphorus, 25(OH)D were measured every 6 months Notes Median ages: 39 years (treatment group), 31 years (control group) Male/female distribution: 17/20 male (treatment group), 14/24 (control group) Study conducted in Austria

Authors' Bias Support for judgement judgement Random sequence generation Unclear risk No details provided (selection bias)

Allocation concealment Unclear risk No details provided (selection bias) Blinding of participants and Unclear risk No details provided personnel (performance bias) Blinding of outcome Unclear risk No details provided assessment (detection bias) Incomplete outcome data High risk 15/75 participants dropped out of the (attrition bias) study. Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Wingate 2014 Methods Randomized study. Duration: 6 months Participants Children and adolescent patients (ages 8 - 18 years old) with quiescent Crohn's disease at McMaster's Children Hospital (Canada) and the British Columbia Children's Hospital (Canada). Exclusion criteria: 1. On corticosteroids within the prior 6 months 2. On vitamin D supplementation > 1000 IU/day Interventions Vitamin D3 400 IU/day (n = 40) Vitamin D3 2000 IU/day (n = 43) Outcomes Primary outcome: 25(OH)D level Secondary outcomes: 1. Prevalence of vitamin D inadequacy (< 16 ng/mL) 2. Proportion achieving cutoffs of 20 and 30 ng/mL 3. Pediatric Crohn's Disease Activity Index 4. Serum calcium 5. Serum phosphate 6. Urinary calcium 7. Urinary creatinine 8. CRP 9. ESR Notes Mean ages: 14.3 (standard deviation 2.3) Female distribution: 38/83 Race: 78% Caucasian Study conducted in Canada

Authors' Bias Support for judgement judgement

Random sequence generation Unclear risk Insufficient details on randomization (selection bias) sequence generation Allocation concealment Low risk Central allocation of assignments by coded (selection bias) lot numbers Blinding of participants and Low risk Double-blinded trial personnel (performance bias) Blinding of outcome Low risk Double-blinded trial assessment (detection bias) Incomplete outcome data High risk 14/83 participants dropped out of the study. (attrition bias) Selective reporting (reporting Unclear risk Listed outcomes were reported. However, bias) we did not have access to the original study protocol. Other bias High risk Small sample size (< 50 participants per treatment arm)

Table 4-2. Characteristics of excluded studies. Study Reason for exclusion Abitbol 2002 Intervention groups received the same dose of vitamin D Abitbol 2007 Intervention groups received the same dose of vitamin D Bendix-Struve 2010 Subset of one of the included studies Bentley 2011 No vitamin D intervention Klaus 2011 Unclear whether vitamin D dose differed across intervention arms Pappa 2011 Intervention groups received the same dose of vitamin D Siffledeen 2005 Intervention groups received the same dose of vitamin D Soo 2012 Intervention groups received the same dose of vitamin D van Bodegraven 2014 Intervention groups received the same dose of vitamin D von Tirpitz 2000 Intervention groups received the same dose of vitamin D von Tirpitz 2003 Intervention groups received the same dose of vitamin D

Table 4-3. Disease activity in Crohn’s disease studies. Study Comparisons Duration Effects Comments Bernstein 1. Placebo (n = 8) 12 Active disease during Participants were 1996 months trial: 4/8 vs. 4/9 (P = IBD patients (unable 2. Calcium 1000 NS) to differentiate results mg/day and Vitamin for Crohn's disease D 250 IU/day (n = 9) Remission during trial: vs. ulcerative colitis). 4/8 vs. 5/9 (P = NS) Boothe 1. Vitamin D 1000 26 weeks Mean Harvey- 2011 IU/day Bradshaw Index at 8 weeks: 6.0 ± 2.5 vs. 4.8 2. Vitamin D 10,000 ± 1.7 (P = 0.29) IU/day Mean Harvey- Bradshaw Index at 26 weeks: 7.6 ± 3.9 vs. 3.25 ± 2.2 (P < 0.05)

At 26 weeks, low-dose group had no significant change in Harvey-Bradshaw Index from baseline, but high-dose group did (P < 0.04) Jorgensen 1. Calcium 1200 12 Relapse rate (increase 2010 mg/day (n = 48) months in CDAI > 70 or CDAI ≥ 150): 14/48 vs. 6/46 2. Vitamin D3 1200 (P = 0.056) IU/day and Calcium 1200 mg/day (n = 46) Raftery 1. Placebo (n = 14) 3 months Change in CDAI from 2013 baseline (placebo 2. Vitamin D3 2000 group): P = 0.258 IU/day (n = 13) Change in CDAI from baseline (vitamin D group): P = 0.568 Change in CDAI (placebo vs. vitamin D): P = 0.119 Vogelsang 1. No 12 Flare (CDAI > 150) or 1995 supplementation (n = months need for surgery: 6/38 38) vs. 8/37

2. Vitamin D3 1000 Change in CDAI from IU/day (n = 37) baseline: -2 (-36 to 22) vs. -43 (-70 to 23) (P < 0.05) Wingate 1. Vitamin D3 400 6 months PCDAI ≤ 10 at 6 2014 IU/day (n = 40) months: 29 (73%) vs. 32 (74%) (P = 0.843) 2. Vitamin D3 2000 IU/day (n = 43)

Table 4-4. Disease activity in ulcerative colitis studies. Study Comparisons Duration Effects Comments Bernstein 1. Placebo (n = 8) 12 Active disease during Participants were IBD 1996 months trial: 4 vs. 4 (P = NS) patients (unable to 2. Calcium 1000 differentiate results for mg/day and Remission during trial: Crohn's disease vs. Vitamin D 250 4 vs. 5 (P = NS) ulcerative colitis). IU/day (n = 9) Kitazaki 1. Alendronate 5 12 No difference (data not 2009 mg/day (without months available) calcium) (n = 19)

2. Alfacalcidol 1 μg/day (without calcium) (n = 20) Mathur 1. Vitamin D3 2000 3 months Change in Mayo UC 2014 IU/day (n = 8) partial score (without endoscopic subscore): - 2. Vitamin D3 4000 0.5 ± 1.5 (P = 0.38) vs. IU/day (n = 10) -1.3 ± 2.9 (P = 0.19)

Table 4-5. Inflammation in Crohn’s disease studies. Study Comparisons Duration Effects Comments Baars 1. No treatment (n = 2 weeks Endoscopic Participants were 2010 29) inflammation: 14 IBD patients (48%) vs. 10 (32%) (P (unable to 2. Prednisone 20 = 0.21) differentiate results mg/day, Calcium, Histologic for Crohn's disease Vitamin D (n = 31) inflammation: 17 vs. ulcerative (59%) vs. 16 (52%) (P colitis). = 0.59) Dendrinos 1. Vitamin D2 at 400 7 weeks No significant increase 2008 IU/week (n = 7) in ESR or CRP

2. Vitamin D2 at 50,000 IU/week (n = 14) Pappa 1. Vitamin D2 2000 6 weeks Median change in Participants were 2012 IU/day (n = 24) CRP: 0.0 (IQR 0.0, IBD patients 0.5) mg/dL vs. 0.0 (unable to 2. Vitamin D3 2000 (IQR 0.0, 0.2) mg/dL differentiate results IU/day (n = 24) vs. 0.0 (IQR -0.2, 0.2) for Crohn's disease mg/dL (P = 0.28) vs. ulcerative 3. Vitamin D2 50,000 colitis). IU/week (n = 23) Pappa 1. Vitamin D2 400 12 CRP ≥ 1 mg/dL during Participants were 2014 IU/day (n = 32) months trial: 31% vs. 10% (P IBD patients = 0.04) (unable to 2. Vitamin D2 1000 ESR ≥ 30 mm/hr differentiate results IU/day (between May 1 during trial: 20% vs. for Crohn's disease and October 31) and 3% (P = 0.10) vs. ulcerative 2000 IU/day (between colitis). November 1 and April 30) (n = 31) Raftery 1. Placebo (n = 14) 3 months Intra-group 2013 comparisons (placebo): 2. Vitamin D3 2000 CRP (P = 0.331), IU/day (n = 13) calprotectin (P = 0.964) Intra-group comparisons (vitamin D group): CRP (P = 0.304), calprotectin (P = 0.706)

Between-group comparisons: CRP (P

= 0.624), calprotectin (P = 0.140) Wingate 1. Vitamin D3 400 6 months CRP > 5 mg/L at 6 2014 IU/day (n = 40) months: 13 (33%) vs. 19 (48%) (P = 0.171) 2. Vitamin D3 2000 ESR > 20 mm/hr at 6 IU/day (n = 43) months: 15 (40%) vs. 17 (41%) (P = 0.927)

Table 4-6. Inflammation in ulcerative colitis studies. Study Comparisons Duration Effects Comments Baars 1. No treatment (n = 29) 2 weeks Endoscopic Participants were 2010 inflammation: 14 (48%) IBD patients (unable 2. Prednisone 20 vs. 10 (32%) (P = 0.21) to differentiate mg/day, Calcium, Histologic results for Crohn's Vitamin D (n = 31) inflammation: 17 (59%) disease vs. vs. 16 (52%) (P = 0.59) ulcerative colitis). Mathur 1. Vitamin D3 2000 3 months Change in CRP: -3.0 ± 2014 IU/day (n = 8) 9.4 mg/L (P = 0.40) vs. -10.8 ± 35.0 mg/L (P = 2. Vitamin D3 4000 0.36) IU/day (n = 10) Pappa 1. Vitamin D2 2000 6 weeks Median change in CRP: Participants were 2012 IU/day (n = 24) 0.0 (IQR 0.0, 0.5) IBD patients (unable mg/dL vs. 0.0 (IQR 0.0, to differentiate 2. Vitamin D3 2000 0.2) mg/dL vs. 0.0 (IQR results for Crohn's IU/day (n = 24) -0.2, 0.2) mg/dL (P = disease vs. 0.28) ulcerative colitis). 3. Vitamin D2 50,000 IU/week (n = 23) Pappa 1. Vitamin D2 400 12 CRP ≥ 1 mg/dL during Participants were 2014 IU/day (n = 32) months trial: 31% vs. 10% (P = IBD patients (unable 0.04) to differentiate 2. Vitamin D2 1000 ESR ≥ 30 mm/hr during results for Crohn's IU/day (between May 1 trial: 20% vs. 3% (P = disease vs. and October 31) and 0.10) ulcerative colitis). 2000 IU/day (between November 1 and April 30) (n = 31)

100

Table 4-7. Vitamin D levels in Crohn’s disease studies. Study Comparisons Duration Effects Comments Bernstein 1. Placebo (n = 8) 12 No difference in Participants were 1996 months 25(OH)D levels, IBD patients 2. Calcium 1000 although placebo group (unable to mg/day and Vitamin D had decline from 33 ± 5 differentiate 250 IU/day (n = 9) ng/mL to 20 ± 3 ng/mL results for (P = 0.04) Crohn's disease vs. ulcerative colitis). Boothe 1. Vitamin D 1000 26 weeks Mean 25(OH)D levels at 2011 IU/day 8 weeks: 26 ± 8.5 ng/mL vs. 54.4 ± 22.8 (P < 2. Vitamin D 10,000 0.01) IU/day Mean 25(OH)D levels at 26 weeks: 26 ± 5.6 ng/mL vs. 64.2 ± 29.8 (P < 0.05) Dendrinos 1. Vitamin D2 at 400 7 weeks Mean change in 2008 IU/week (n = 7) 25(OH)D levels: 128% (21.3 to 41.6 ng/mL) (P 2. Vitamin D2 at = 0.007) vs. 61% (26.1 50,000 IU/week (n = to 37 ng/mL) (P = 14) 0.035). No difference in 25(OH)D levels between groups Jorgensen 1. Calcium 1200 12 Vitamin D levels 2010 mg/day (n = 48) months unchanged at 3 months for control group 2. Vitamin D3 1200 compared with increase IU/day and Calcium from mean 69 nmol/L 1200 mg/day (n = 46) (SD 31) to 96 nmol/L (SD 27) for vitamin D group (P < 0.001) Pappa 1. Vitamin D2 2000 6 weeks Increase in 25(OH)D Participants were 2012 IU/day (n = 24) levels: 9.3 ± 1.8 vs. 16.4 IBD patients ± 2.0 vs. 25.4 ± 2.5 (P < (unable to 2. Vitamin D3 2000 0.0001) differentiate IU/day (n = 24) results for Crohn's disease 3. Vitamin D2 50,000 vs. ulcerative IU/week (n = 23) colitis). Pappa 1. Vitamin D2 400 12 Change in 25(OH)D Participants were 2014 IU/day (n = 32) months levels at 3 months: -3.4 ± IBD patients 9.7 ng/mL vs. -1.5 ± 16 (unable to

101

2. Vitamin D2 1000 ng/mL (P = 0.60) differentiate IU/day (between May Change at 6 months: -1.2 results for 1 and October 31) and ± 9.2 ng/mL vs. -4.7 ± Crohn's disease 2000 IU/day (between 19 ng/mL (P = 0.40) vs. ulcerative November 1 and April Change at 12 months: - colitis). 30) (n = 31) 3.7 ± 10.8 ng/mL vs. - 5.8 ± 15.2 ng/mL (P = 0.57) Raftery 1. Placebo (n = 14) 3 months Median 25(OH)D levels 2013 at 3 months: 40.4 (IQR 2. Vitamin D3 2000 30.4, 50.4) nmol/L vs. IU/day (n = 13) 91.6 (IQR 75.5, 107.6) nmol/L (P < 0.001) Mean difference of 25(OH)D levels between groups at 3 months: 51.2 nmol/L (33.0 to 69.4) (P < 0.001) Vogelsang 1. No supplementation 12 Median change in 1995 (n = 38) months 25(OH)D levels at 12 months (control): -2.7 2. Vitamin D3 1000 (IQR -10.1, 5.5) ng/mL IU/day (n = 37) (P = 0.24) Median change in 25(OH)D levels at 12 months (treatment): 2.0 (IQR -6.7, 13) ng/mL (P = 0.35)

Between-group P = 0.19 Wingate 1. Vitamin D3 400 6 months Difference between 2014 IU/day (n = 40) groups (adjusted for baseline concentrations 2. Vitamin D3 2000 relative to 400 IU at 6 IU/day (n = 43) months; general linear model was used): -9.6 (95% CI -13.2 to -6.0) ng/mL (P < 0.001)

102

Table 4-8. Vitamin D levels in ulcerative colitis studies. Study Comparisons Duration Effects Comments Bernstein 1. Placebo (n = 8) 12 No difference in Participants were 1996 months 25(OH)D levels, IBD patients (unable 2. Calcium 1000 mg/day although placebo to differentiate and Vitamin D 250 group had decline results for Crohn's IU/day (n = 9) from 33 ± 5 ng/mL disease vs. ulcerative to 20 ± 3 ng/mL (P = colitis). 0.04) Mathur 1. Vitamin D3 2000 3 months Change in 25(OH)D 2014 IU/day (n = 8) levels: 5.00 ± 3.82 (P = 0.008) vs. 16.80 2. Vitamin D3 4000 ± 9.15 (P < 0.001) IU/day (n = 10) Pappa 1. Vitamin D2 2000 6 weeks Increase in 25(OH)D Participants were 2012 IU/day (n = 24) levels: 9.3 ± 1.8 vs. IBD patients (unable 16.4 ± 2.0 vs. 25.4 ± to differentiate 2. Vitamin D3 2000 2.5 (P < 0.0001) results for Crohn's IU/day (n = 24) disease vs. ulcerative colitis). 3. Vitamin D2 50,000 IU/week (n = 23) Pappa 1. Vitamin D2 400 12 Change in 25(OH)D Participants were 2014 IU/day (n = 32) months levels at 3 months: - IBD patients (unable 3.4 ± 9.7 ng/mL vs. - to differentiate 2. Vitamin D2 1000 1.5 ± 16 ng/mL (P = results for Crohn's IU/day (between May 1 0.60) disease vs. ulcerative and October 31) and Change at 6 months: colitis). 2000 IU/day (between -1.2 ± 9.2 ng/mL vs. November 1 and April -4.7 ± 19 ng/mL (P 30) (n = 31) = 0.40) Change at 12 months: -3.7 ± 10.8 ng/mL vs. -5.8 ± 15.2 ng/mL (P = 0.57)

103

Table 4-9. Adverse events in Crohn’s disease studies. Study Comparisons Duration Effects Comments Baars 1. No treatment (n = 2 weeks Control: headache (3), Participants were 2010 29) sleepless nights (2), vague IBD patients complaints of not feeling (unable to 2. Prednisone 20 well (1) differentiate mg/day, Calcium, results for Crohn's Vitamin D (n = 31) Treatment: mood swings disease vs. (4), headache (1), swollen ulcerative colitis). face (3), increased appetite (2), fatigue (1), edema (1) Bernstein 1. Placebo (n = 8) 12 None thought to be related Participants were 1996 months to medications IBD patients 2. Calcium 1000 (unable to mg/day and Vitamin differentiate D 250 IU/day (n = 9) results for Crohn's disease vs. ulcerative colitis). Dendrinos 1. Vitamin D2 at 400 7 weeks High-dose group: increased 2008 IU/week (n = 7) diarrhea

2. Vitamin D2 at 50,000 IU/week (n = 14) Jorgensen 1. Calcium 1200 12 No adverse effects directly 2010 mg/day (n = 48) months attributable to vitamin D intake. 2. Vitamin D3 1200 IU/day and Calcium 5 cases of mild 1200 mg/day (n = constipation (4 in no- 46) vitamin D group, 1 in vitamin D group) Pappa 1. Vitamin D2 2000 6 weeks Subjects with reported Participants were 2012 IU/day (n = 24) adverse effects: 8 (33%) vs. IBD patients 9 (38%) vs. 6 (26%) (P = (unable to 2. Vitamin D3 2000 0.70) differentiate IU/day (n = 24) results for Crohn's Nausea (6), increased thirst disease vs. 3. Vitamin D2 50,000 (5), loss of appetite (5), ulcerative colitis). IU/week (n = 23) pruritis (4), drowsiness (3), increased urination (3), rash (1), sensitive eyes (1), vomiting (1), irregular heartbeat (1), dry mouth (1), muscle pain (1). All

104

events were considered mild-moderate.

Urinary Ca:Cr ≥ 0.2: 2 (12%) vs. 4 (17%) vs. 2 (15%) (P = 1.00) Pappa 1. Vitamin D2 400 12 Subjects with clinical Participants were 2014 IU/day (n = 32) months adverse effects: 19 (59%) IBD patients vs. 15 (48%) (P = 0.38) (unable to 2. Vitamin D2 1000 Clinical adverse effects per differentiate IU/day (between subject: 3.2 ± 3.1 vs. 3.5 ± results for Crohn's May 1 and October 2.1 (P = 0.69) disease vs. 31) and 2000 IU/day Serum calcium > 10.5 ulcerative colitis). (between November mg/dL at any visit: 0 1 and April 30) (n = Serum phosphate > 5.7 31) mg/dL at any visit: 1 (3%) vs. 1 (3%) (P = 1.00) Urinary Ca:Cr ≥ 0.2 at any visit: 11 (34%) vs. 9 (29%) (P = 0.65) Wingate 1. Vitamin D3 400 6 months Serum Ca > age- 2014 IU/day (n = 40) appropriate cut-off at 6 months: 1 (3%) vs. 0 (0%) 2. Vitamin D3 2000 (P = 0.488) IU/day (n = 43) Serum phosphate: 4 (10%) vs. 6 (14%) (P = 0.739) Urinary Ca:Cr ≥ 0.6: 4 (11%) vs. 3 (7%) (P = 0.708)

105

Table 4-10. Adverse events in ulcerative colitis studies. Study Comparisons Duration Effects Comments Baars 1. No treatment (n = 2 weeks Control: headache (3), Participants were 2010 29) sleepless nights (2), vague IBD patients complaints of not feeling (unable to 2. Prednisone 20 well (1) differentiate mg/day, Calcium, results for Crohn's Vitamin D (n = 31) Treatment: mood swings disease vs. (4), headache (1), swollen ulcerative colitis). face (3), increased appetite (2), fatigue (1), edema (1) Bernstein 1. Placebo (n = 8) 12 None thought to be related Participants were 1996 months to medications IBD patients 2. Calcium 1000 (unable to mg/day and Vitamin differentiate D 250 IU/day (n = 9) results for Crohn's disease vs. ulcerative colitis). Kitazaki 1. Alendronate 5 12 Alendronate group: pruritis 2009 mg/day (without months (1), abdominal pain (1), calcium) (n = 19) diarrhea (1) Alfacalcidol group: muscle 2. Alfacalcidol 1 pain (1) μg/day (without calcium) (n = 20) Pappa 1. Vitamin D2 2000 6 weeks Subjects with reported Participants were 2012 IU/day (n = 24) adverse effects: 8 (33%) vs. IBD patients 9 (38%) vs. 6 (26%) (P = (unable to 2. Vitamin D3 2000 0.70) differentiate IU/day (n = 24) results for Crohn's Nausea (6), increased thirst disease vs. 3. Vitamin D2 50,000 (5), loss of appetite (5), ulcerative colitis). IU/week (n = 23) pruritis (4), drowsiness (3), increased urination (3), rash (1), sensitive eyes (1), vomiting (1), irregular heartbeat (1), dry mouth (1), muscle pain (1). All events were considered mild-moderate.

Urinary Ca:Cr ≥ 0.2: 2 (12%) vs. 4 (17%) vs. 2 (15%) (P = 1.00)

106

Pappa 1. Vitamin D2 400 12 Subjects with clinical Participants were 2014 IU/day (n = 32) months adverse effects: 19 (59%) IBD patients vs. 15 (48%) (P = 0.38) (unable to 2. Vitamin D2 1000 Clinical adverse effects per differentiate IU/day (between subject: 3.2 ± 3.1 vs. 3.5 ± results for Crohn's May 1 and October 2.1 (P = 0.69) disease vs. 31) and 2000 IU/day Serum calcium > 10.5 ulcerative colitis). (between November mg/dL at any visit: 0 1 and April 30) (n = Serum phosphate > 5.7 31) mg/dL at any visit: 1 (3%) vs. 1 (3%) (P = 1.00) Urinary Ca:Cr ≥ 0.2 at any visit: 11 (34%) vs. 9 (29%) (P = 0.65)

107

Table 4-11. Quality of life in Crohn’s disease studies. Study Comparisons Duration Effects Comments Raftery 1. Placebo (n = 14) 3 months Intra-group comparison of IBD 2013 Questionnaire scores (placebo): P 2. Vitamin D3 2000 = 0.364 IU/day (n = 13) Intra-group comparison of IBD Questionnaire scores (vitamin D group): P = 0.519 Between-group comparison of IBD Questionnaire scores: P = 0.900

108

Table 4-12. Quality of life in ulcerative colitis studies. Study Comparisons Duration Effects Comments Mathur 1. Vitamin D3 2000 3 months Change in Short IBD 2014 IU/day (n = 8) Questionnaire score: 0.1 ± 1.0 (P = 0.87) vs. 1.0 ± 1.0 (P = 0.017) 2. Vitamin D3 4000 IU/day (n = 10)

109

FIGURES

Figure 4-1. Study flow diagram.

110

Figure 4-2. Risk of bias assessment.

111

Figure 4-3. Forest plot comparison (random effects model) of vitamin D at different doses or with placebo in studies with 12-month follow-up.

112

Figure 4-4. Forest plot comparison (fixed effects model) of vitamin D at different doses or with placebo in studies with 12-month follow-up.

113

CHAPTER V. CONCLUSION

The surprising discovery of vitamin D activity in immune cells in the 1980s sparked the

cascade of investigation into its role in immune-related disorders. Bolstered by in vitro,

animal, and epidemiologic studies, vitamin D has been proposed as a mediator of IBD

pathogenesis and activity. Nonetheless, interpretation of the data from these studies was

limited by the issue of reverse causation. In this thesis dissertation, our objective was to

test the hypothesis that vitamin D deficiency is associated with IBD pathogenesis and

severity, while employing diverse methodologic approaches to overcome the limitations of reverse causation.

In Chapter II, we examined the relationship between vitamin D and CD pathogenesis. Although a post hoc analysis of the Nurses’ Health Study had previously found an inverse association between vitamin D levels and CD risk, the study was limited by the use of indirect predictors of 25(OH)D levels. This methodology is subject to unclear prediction error, particularly with increasing model complexity; although validated, the original model’s r2 was only 28%.168,169 We therefore relied on direct

measurements of 25(OH)D levels, while additionally examining them before and around

the time of diagnosis to account for reverse causation. Our data did not identify vitamin D

as an etiologic factor of CD in the primary and sensitivity analyses. A sample size issue was considered, but we had enough participants to detect a significant inverse association between vitamin D and CD around the time of diagnosis. This expected finding is consistent with biologic plausibility and a priori data, thus supporting the validity of our dataset and analysis. If we were unable to detect an association between vitamin D and

114

CD after diagnosis, as had been shown in several other studies, then the adequacy of

power would have been questioned.

Nonetheless, given the complex multi-factorial interactions that lead to CD, it is possible that the effect of vitamin D was just too weak to be detected. For one, the estimated effect sizes of the genotyped CD risk polymorphisms were notably stronger than those of vitamin D. Post hoc sample size estimations for the pre-2 and pre-1 samples also suggested that we would need very large numbers to detect a statistical change, but these would likely not have much clinical significance.

Although vitamin D levels were not found to significantly influence CD pathogenesis, there may still be a relationship between vitamin D influencing IBD severity. Factors that affect disease pathogenesis – or the development of disease – may differ from those that affect disease severity. For instance, gluten consumption does not lead to the development of celiac disease; it stems from immunogenetic factors involving aberrant HLA-DQ2 and HLA-DQ8 isoforms.170 On the other hand, gluten consumption is

strongly linked to the severity of celiac disease, whereby gluten restriction has been the

first line of treatment. Analogously, the absence of a role for vitamin D as a cause of IBD

does not negate its potential role as a modifier of IBD severity.

In Chapter III, we examined the relationship between vitamin D and IBD severity.

Prior cohort studies had already demonstrated an inverse association between the two, but they were clearly challenged by reverse causation. We therefore selected UV exposure as

a surrogate of vitamin D, as it had been noted as the strongest predictor of vitamin D

status and not directly influenced by IBD. In this study, low UV exposures were

associated with increased rates of hospitalizations and prolonged hospitalizations,

115

although these trends were not unique to the IBD population. Nonetheless, the trend of

lower UV exposure groups having increased bowel surgeries – a feature of increased IBD

severity – was unique to the IBD population. To further emphasize that UV exposure

specifically influenced IBD severity and not the general need for surgeries in the IBD

population, we examined abdominal surgeries unrelated to IBD (i.e., cholecystectomies,

appendectomies) and found no association with UV exposure. In this national hospitalization dataset, we could therefore conclude that low UV exposure, as a surrogate of vitamin D, was associated with more severe hospitalizations and the need for IBD- related surgery.

This prior study was nonetheless an ecologic study using a surrogate marker and subject to a myriad of unmeasured confounders. Instead, the randomized trial design has been heralded as the most robust methodologic approach to causal inference.

Randomization of vitamin D supplementation would most effectively isolate the effects of vitamin D on IBD severity, independent of reverse causation. In Chapter IV, we performed a systematic review and meta-analysis of RCTs that compared vitamin D at different doses or with placebo, while measuring disease activity as a primary or secondary outcome. The systematic review revealed significant methodologic heterogeneity across the included studies, particularly involving vitamin D dosing, duration of follow-up, and reported outcomes. IBD severity was often a secondary and inconsistently measured outcome in the studies. There was also significant or unclear risk of bias, high attrition, and small sample sizes. A meta-analysis of 4 studies with 12- month follow-up periods revealed an overall reduction in clinical relapse with vitamin D

supplementation, although this was not statistically significant. However, within this

116

group of studies was substantial heterogeneity that likely contributed to some

inconsistency across study findings. As such, the current level of RCT evidence is poor and unable to confirm or refute whether vitamin D supplementation improves IBD outcomes.

In conclusion, the series of investigation in this thesis dissertation was aimed at overcoming reverse causation and clarifying whether vitamin D contributes to IBD pathogenesis or severity. Vitamin D does not appear to contribute to CD pathogenesis, although it may have some effect on IBD severity. Given its known immunomodulatory properties, vitamin D could still possess beneficial effects on IBD. The strength of effect may just be too weak amidst the myriad of disease modifiers concurrently acting on intestinal inflammation. This multiplex of disease modifiers could also account for the

current scientific challenges in elucidating the cause(s) of IBD. Further investigation into

the role of vitamin D in IBD severity is nonetheless needed, particularly with high-quality

and larger randomized trials that specifically evaluate IBD activity as a primary outcome.

In any case, given the emerging signal that vitamin D could potentially benefit IBD, and

its safety, cost-effectiveness, and non-IBD health benefits, checking and repleting for

vitamin D deficiency in the clinical setting should be recommended.

117

REFERENCES

1. Orholm M, Munkholm P, Langholz E, Nielsen OH, Sorensen TI, Binder V. Familial occurrence of inflammatory bowel disease. N Engl J Med. Jan 10 1991;324(2):84-88. 2. Tysk C, Lindberg E, Jarnerot G, Floderus-Myrhed B. Ulcerative colitis and Crohn's disease in an unselected population of monozygotic and dizygotic twins. A study of heritability and the influence of smoking. Gut. Jul 1988;29(7):990-996. 3. Halfvarson J, Bresso F, D'Amato M, Jarnerot G, Pettersson S, Tysk C. CARD15/NOD2 polymorphisms do not explain concordance of Crohn's disease in Swedish monozygotic twins. Dig Liver Dis. Oct 2005;37(10):768-772. 4. Hugot JP, Chamaillard M, Zouali H, et al. Association of NOD2 leucine-rich repeat variants with susceptibility to Crohn's disease. Nature. May 31 2001;411(6837):599-603. 5. Ogura Y, Bonen DK, Inohara N, et al. A frameshift mutation in NOD2 associated with susceptibility to Crohn's disease. Nature. May 31 2001;411(6837):603-606. 6. Rioux JD, Xavier RJ, Taylor KD, et al. Genome-wide association study identifies new susceptibility loci for Crohn disease and implicates autophagy in disease pathogenesis. Nat Genet. May 2007;39(5):596-604. 7. McCarroll SA, Huett A, Kuballa P, et al. Deletion polymorphism upstream of IRGM associated with altered IRGM expression and Crohn's disease. Nat Genet. Sep 2008;40(9):1107-1112. 8. Gradel KO, Nielsen HL, Schonheyder HC, Ejlertsen T, Kristensen B, Nielsen H. Increased short- and long-term risk of inflammatory bowel disease after salmonella or campylobacter gastroenteritis. Gastroenterology. Aug 2009;137(2):495-501. 9. Garcia Rodriguez LA, Ruigomez A, Panes J. Acute gastroenteritis is followed by an increased risk of inflammatory bowel disease. Gastroenterology. May 2006;130(6):1588-1594. 10. Shaw SY, Blanchard JF, Bernstein CN. Association between the use of antibiotics in the first year of life and pediatric inflammatory bowel disease. Am J Gastroenterol. Dec 2010;105(12):2687-2692. 11. Hou JK, Abraham B, El-Serag H. Dietary intake and risk of developing inflammatory bowel disease: a systematic review of the literature. Am J Gastroenterol. Apr 2011;106(4):563-573. 12. Copping AM. Origin of vitamin D in cod-liver oil: vitamin D content of zooplankton. Biochem J. 1934;28(4):1516-1520. 13. Holick MF. Vitamin D deficiency. N Engl J Med. Jul 19 2007;357(3):266-281. 14. Holick MF, MacLaughlin JA, Clark MB, et al. Photosynthesis of previtamin D3 in human skin and the physiologic consequences. Science. Oct 10 1980;210(4466):203-205. 15. Tian XQ, Chen TC, Matsuoka LY, Wortsman J, Holick MF. Kinetic and thermodynamic studies of the conversion of previtamin D3 to vitamin D3 in human skin. J Biol Chem. Jul 15 1993;268(20):14888-14892.

118

16. Holick MF, MacLaughlin JA, Doppelt SH. Regulation of cutaneous previtamin D3 photosynthesis in man: skin pigment is not an essential regulator. Science. Feb 6 1981;211(4482):590-593. 17. Webb AR, DeCosta BR, Holick MF. Sunlight regulates the cutaneous production of vitamin D3 by causing its photodegradation. J Clin Endocrinol Metab. May 1989;68(5):882-887. 18. Bikle DD, Siiteri PK, Ryzen E, Haddad JG. Serum protein binding of 1,25- dihydroxyvitamin D: a reevaluation by direct measurement of free metabolite levels. J Clin Endocrinol Metab. Nov 1985;61(5):969-975. 19. Bikle DD, Gee E, Halloran B, Kowalski MA, Ryzen E, Haddad JG. Assessment of the free fraction of 25-hydroxyvitamin D in serum and its regulation by albumin and the vitamin D-binding protein. J Clin Endocrinol Metab. Oct 1986;63(4):954-959. 20. Ponchon G, DeLuca HF. The role of the liver in the metabolism of vitamin D. J Clin Invest. Jul 1969;48(7):1273-1279. 21. Cheng JB, Motola DL, Mangelsdorf DJ, Russell DW. De-orphanization of cytochrome P450 2R1: a microsomal vitamin D 25-hydroxilase. J Biol Chem. Sep 26 2003;278(39):38084-38093. 22. Fraser DR, Kodicek E. Unique biosynthesis by kidney of a biological active vitamin D metabolite. Nature. Nov 21 1970;228(5273):764-766. 23. Haussler MR, Norman AW. Chromosomal receptor for a vitamin D metabolite. Proc Natl Acad Sci U S A. Jan 1969;62(1):155-162. 24. Carlberg C, Bendik I, Wyss A, et al. Two nuclear signalling pathways for vitamin D. Nature. Feb 18 1993;361(6413):657-660. 25. Nishikawa J, Kitaura M, Matsumoto M, Imagawa M, Nishihara T. Difference and similarity of DNA sequence recognized by VDR homodimer and VDR/RXR heterodimer. Nucleic Acids Res. Aug 11 1994;22(15):2902-2907. 26. Sone T, Kerner S, Pike JW. Vitamin D receptor interaction with specific DNA. Association as a 1,25-dihydroxyvitamin D3-modulated heterodimer. J Biol Chem. Dec 5 1991;266(34):23296-23305. 27. Morrison NA, Shine J, Fragonas JC, Verkest V, McMenemy ML, Eisman JA. 1,25-dihydroxyvitamin D-responsive element and glucocorticoid repression in the osteocalcin gene. Science. Dec 1 1989;246(4934):1158-1161. 28. Mesbah M, Nemere I, Papagerakis P, et al. Expression of a 1,25- dihydroxyvitamin D3 membrane-associated rapid-response steroid binding protein during human tooth and bone development and biomineralization. J Bone Miner Res. Sep 2002;17(9):1588-1596. 29. Nemere I, Garbi N, Hammerling G, Hintze KJ. Role of the 1,25D(3)-MARRS receptor in the 1,25(OH)(2)D(3)-stimulated uptake of calcium and phosphate in intestinal cells. Steroids. Apr 21 2012. 30. Abe E, Miyaura C, Sakagami H, et al. Differentiation of mouse myeloid leukemia cells induced by 1 alpha, 25-dihydroxyvitamin D3. Proc Natl Acad Sci U S A. 1981;78(8):4990-4994. 31. Provvedini DM, Tsoukas CD, Deftos LJ, Manolagas SC. 1,25-dihydroxyvitamin D3 receptors in human leukocytes. Science. Sep 16 1983;221(4616):1181-1183.

119

32. Kizaki M, Norman AW, Bishop JE, Lin CW, Karmakar A, Koeffler HP. 1,25- Dihydroxyvitamin D3 receptor RNA: expression in hematopoietic cells. Blood. Mar 15 1991;77(6):1238-1247. 33. Baeke F, Korf H, Overbergh L, et al. Human T lymphocytes are direct targets of 1,25-dihydroxyvitamin D3 in the immune system. J Steroid Biochem Mol Biol. Jul 2010;121(1-2):221-227. 34. Bhalla AK, Amento EP, Clemens TL, Holick MF, Krane SM. Specific high- affinity receptors for 1,25-dihydroxyvitamin D3 in human peripheral blood mononuclear cells: presence in monocytes and induction in T lymphocytes following activation. J Clin Endocrinol Metab. Dec 1983;57(6):1308-1310. 35. Reichel H, Koeffler HP, Norman AW. Synthesis in vitro of 1,25- dihydroxyvitamin D3 and 24,25-dihydroxyvitamin D3 by interferon-gamma- stimulated normal human bone marrow and alveolar macrophages. J Biol Chem. Aug 15 1987;262(23):10931-10937. 36. Fritsche J, Mondal K, Ehrnsperger A, Andreesen R, Kreutz M. Regulation of 25- hydroxyvitamin D3-1 alpha-hydroxylase and production of 1 alpha,25- dihydroxyvitamin D3 by human dendritic cells. Blood. Nov 1 2003;102(9):3314- 3316. 37. Chen S, Sims GP, Chen XX, Gu YY, Lipsky PE. Modulatory effects of 1,25- dihydroxyvitamin D3 on human B cell differentiation. J Immunol. Aug 1 2007;179(3):1634-1647. 38. Stoffels K, Overbergh L, Giulietti A, Verlinden L, Bouillon R, Mathieu C. Immune regulation of 25-hydroxyvitamin-D3-1alpha-hydroxylase in human monocytes. J Bone Miner Res. Jan 2006;21(1):37-47. 39. Liu PT, Stenger S, Li H, et al. Toll-like receptor triggering of a vitamin D- mediated human antimicrobial response. Science. Mar 24 2006;311(5768):1770- 1773. 40. Sadeghi K, Wessner B, Laggner U, et al. Vitamin D3 down-regulates monocyte TLR expression and triggers hyporesponsiveness to pathogen-associated molecular patterns. Eur J Immunol. Feb 2006;36(2):361-370. 41. Kreutz M, Andreesen R, Krause SW, Szabo A, Ritz E, Reichel H. 1,25- dihydroxyvitamin D3 production and vitamin D3 receptor expression are developmentally regulated during differentiation of human monocytes into macrophages. Blood. Aug 15 1993;82(4):1300-1307. 42. Tsoukas CD, Provvedini DM, Manolagas SC. 1,25-dihydroxyvitamin D3: a novel immunoregulatory hormone. Science. Jun 29 1984;224(4656):1438-1440. 43. Manolagas SC, Provvedini DM, Murray EJ, Tsoukas CD, Deftos LJ. The antiproliferative effect of calcitriol on human peripheral blood mononuclear cells. J Clin Endocrinol Metab. Aug 1986;63(2):394-400. 44. Piemonti L, Monti P, Sironi M, et al. Vitamin D3 affects differentiation, maturation, and function of human monocyte-derived dendritic cells. J Immunol. May 1 2000;164(9):4443-4451. 45. Griffin MD, Lutz W, Phan VA, Bachman LA, McKean DJ, Kumar R. Dendritic cell modulation by 1alpha,25 dihydroxyvitamin D3 and its analogs: a vitamin D receptor-dependent pathway that promotes a persistent state of immaturity in vitro and in vivo. Proc Natl Acad Sci U S A. Jun 5 2001;98(12):6800-6805.

120

46. Penna G, Adorini L. 1 Alpha,25-dihydroxyvitamin D3 inhibits differentiation, maturation, activation, and survival of dendritic cells leading to impaired alloreactive T cell activation. J Immunol. Mar 1 2000;164(5):2405-2411. 47. Adorini L, Penna G, Giarratana N, et al. Dendritic cells as key targets for immunomodulation by Vitamin D receptor ligands. J Steroid Biochem Mol Biol. May 2004;89-90(1-5):437-441. 48. Gregori S, Giarratana N, Smiroldo S, Uskokovic M, Adorini L. A 1alpha,25- dihydroxyvitamin D(3) analog enhances regulatory T-cells and arrests autoimmune diabetes in NOD mice. Diabetes. May 2002;51(5):1367-1374. 49. Jeffery LE, Burke F, Mura M, et al. 1,25-Dihydroxyvitamin D3 and IL-2 combine to inhibit T cell production of inflammatory cytokines and promote development of regulatory T cells expressing CTLA-4 and FoxP3. J Immunol. Nov 1 2009;183(9):5458-5467. 50. Barrat FJ, Cua DJ, Boonstra A, et al. In vitro generation of interleukin 10- producing regulatory CD4(+) T cells is induced by immunosuppressive drugs and inhibited by T helper type 1 (Th1)- and Th2-inducing cytokines. J Exp Med. Mar 4 2002;195(5):603-616. 51. Ghoreishi M, Bach P, Obst J, Komba M, Fleet JC, Dutz JP. Expansion of antigen- specific regulatory T cells with the topical vitamin d analog calcipotriol. J Immunol. May 15 2009;182(10):6071-6078. 52. Daniel C, Sartory NA, Zahn N, Radeke HH, Stein JM. Immune modulatory treatment of trinitrobenzene sulfonic acid colitis with calcitriol is associated with a change of a T helper (Th) 1/Th17 to a Th2 and regulatory T cell profile. J Pharmacol Exp Ther. Jan 2008;324(1):23-33. 53. Palmer MT, Lee YK, Maynard CL, et al. Lineage-specific effects of 1,25- dihydroxyvitamin D(3) on the development of effector CD4 T cells. J Biol Chem. Jan 14 2011;286(2):997-1004. 54. Alroy I, Towers TL, Freedman LP. Transcriptional repression of the interleukin-2 gene by vitamin D3: direct inhibition of NFATp/AP-1 complex formation by a nuclear hormone receptor. Mol Cell Biol. Oct 1995;15(10):5789-5799. 55. Bhalla AK, Amento EP, Krane SM. Differential effects of 1,25-dihydroxyvitamin D3 on human lymphocytes and monocyte/macrophages: inhibition of interleukin- 2 and augmentation of interleukin-1 production. Cell Immunol. Apr 1 1986;98(2):311-322. 56. D'Ambrosio D, Cippitelli M, Cocciolo MG, et al. Inhibition of IL-12 production by 1,25-dihydroxyvitamin D3. Involvement of NF-kappaB downregulation in transcriptional repression of the p40 gene. J Clin Invest. Jan 1 1998;101(1):252- 262. 57. Towers TL, Freedman LP. Granulocyte-macrophage colony-stimulating factor gene transcription is directly repressed by the vitamin D3 receptor. Implications for allosteric influences on nuclear receptor structure and function by a DNA element. J Biol Chem. Apr 24 1998;273(17):10338-10348. 58. Schauber J, Dorschner RA, Coda AB, et al. Injury enhances TLR2 function and antimicrobial peptide expression through a vitamin D-dependent mechanism. J Clin Invest. Mar 2007;117(3):803-811.

121

59. Wang TT, Nestel FP, Bourdeau V, et al. Cutting edge: 1,25-dihydroxyvitamin D3 is a direct inducer of antimicrobial peptide gene expression. J Immunol. Sep 1 2004;173(5):2909-2912. 60. Gombart AF, Borregaard N, Koeffler HP. Human cathelicidin antimicrobial peptide (CAMP) gene is a direct target of the vitamin D receptor and is strongly up-regulated in myeloid cells by 1,25-dihydroxyvitamin D3. FASEB J. Jul 2005;19(9):1067-1077. 61. Lemire JM, Archer DC. 1,25-dihydroxyvitamin D3 prevents the in vivo induction of murine experimental autoimmune encephalomyelitis. J Clin Invest. Mar 1991;87(3):1103-1107. 62. Cantorna MT, Hayes CE, DeLuca HF. 1,25-Dihydroxyvitamin D3 reversibly blocks the progression of relapsing encephalomyelitis, a model of multiple sclerosis. Proc Natl Acad Sci U S A. Jul 23 1996;93(15):7861-7864. 63. DeLuca HF, Plum LA. Vitamin D deficiency diminishes the severity and delays onset of experimental autoimmune encephalomyelitis. Arch Biochem Biophys. Sep 15 2011;513(2):140-143. 64. Yang S, Smith C, Prahl JM, Luo X, DeLuca HF. Vitamin D deficiency suppresses cell-mediated immunity in vivo. Arch Biochem Biophys. May 15 1993;303(1):98- 106. 65. Munger KL, Zhang SM, O'Reilly E, et al. Vitamin D intake and incidence of multiple sclerosis. Neurology. Jan 13 2004;62(1):60-65. 66. Munger KL, Levin LI, Hollis BW, Howard NS, Ascherio A. Serum 25- hydroxyvitamin D levels and risk of multiple sclerosis. JAMA. Dec 20 2006;296(23):2832-2838. 67. Fukazawa T, Yabe I, Kikuchi S, et al. Association of vitamin D receptor gene polymorphism with multiple sclerosis in Japanese. J Neurol Sci. Jun 15 1999;166(1):47-52. 68. Niino M, Fukazawa T, Yabe I, Kikuchi S, Sasaki H, Tashiro K. Vitamin D receptor gene polymorphism in multiple sclerosis and the association with HLA class II alleles. J Neurol Sci. Aug 1 2000;177(1):65-71. 69. Merlino LA, Curtis J, Mikuls TR, Cerhan JR, Criswell LA, Saag KG. Vitamin D intake is inversely associated with rheumatoid arthritis: results from the Iowa Women's Health Study. Arthritis Rheum. Jan 2004;50(1):72-77. 70. Cantorna MT, Hayes CE, DeLuca HF. 1,25-Dihydroxycholecalciferol inhibits the progression of arthritis in murine models of human arthritis. J Nutr. Jan 1998;128(1):68-72. 71. O'Regan S, Chesney RW, Hamstra A, Eisman JA, O'Gorman AM, Deluca HF. Reduced serum 1,25-(OH)2 vitamin D3 levels in prednisone-treated adolescents with systemic lupus erythematosus. Acta Paediatr Scand. Jan 1979;68(1):109- 111. 72. Monticielo OA, Teixeira TD, Chies JA, Brenol JC, Xavier RM. Vitamin D and polymorphisms of VDR gene in patients with systemic lupus erythematosus. Clin Rheumatol. Jun 13 2012. 73. Amital H, Szekanecz Z, Szucs G, et al. Serum concentrations of 25-OH vitamin D in patients with systemic lupus erythematosus (SLE) are inversely related to

122

disease activity: is it time to routinely supplement patients with SLE with vitamin D? Ann Rheum Dis. Jun 2010;69(6):1155-1157. 74. Mathieu C, Waer M, Laureys J, Rutgeerts O, Bouillon R. Prevention of autoimmune diabetes in NOD mice by 1,25 dihydroxyvitamin D3. Diabetologia. Jun 1994;37(6):552-558. 75. Zella JB, McCary LC, DeLuca HF. Oral administration of 1,25-dihydroxyvitamin D3 completely protects NOD mice from insulin-dependent diabetes mellitus. Arch Biochem Biophys. Sep 1 2003;417(1):77-80. 76. Poon AH, Laprise C, Lemire M, et al. Association of vitamin D receptor genetic variants with susceptibility to asthma and atopy. Am J Respir Crit Care Med. Nov 1 2004;170(9):967-973. 77. Raby BA, Lazarus R, Silverman EK, et al. Association of vitamin D receptor gene polymorphisms with childhood and adult asthma. Am J Respir Crit Care Med. Nov 15 2004;170(10):1057-1065. 78. Bosse Y, Lemire M, Poon AH, et al. Asthma and genes encoding components of the vitamin D pathway. Respir Res. 2009;10:98. 79. Cantorna MT, Munsick C, Bemiss C, Mahon BD. 1,25-Dihydroxycholecalciferol prevents and ameliorates symptoms of experimental murine inflammatory bowel disease. J Nutr. Nov 2000;130(11):2648-2652. 80. Zhu Y, Mahon BD, Froicu M, Cantorna MT. Calcium and 1 alpha,25- dihydroxyvitamin D3 target the TNF-alpha pathway to suppress experimental inflammatory bowel disease. Eur J Immunol. Jan 2005;35(1):217-224. 81. Daniel C, Radeke HH, Sartory NA, et al. The new low calcemic vitamin D analog 22-ene-25-oxa-vitamin D prominently ameliorates T helper cell type 1-mediated colitis in mice. J Pharmacol Exp Ther. Nov 2006;319(2):622-631. 82. Laverny G, Penna G, Vetrano S, et al. Efficacy of a potent and safe vitamin D receptor agonist for the treatment of inflammatory bowel disease. Immunol Lett. Jun 15 2010;131(1):49-58. 83. Froicu M, Weaver V, Wynn TA, McDowell MA, Welsh JE, Cantorna MT. A crucial role for the vitamin D receptor in experimental inflammatory bowel diseases. Mol Endocrinol. Dec 2003;17(12):2386-2392. 84. Froicu M, Zhu Y, Cantorna MT. Vitamin D receptor is required to control gastrointestinal immunity in IL-10 knockout mice. Immunology. Mar 2006;117(3):310-318. 85. Yu S, Bruce D, Froicu M, Weaver V, Cantorna MT. Failure of T cell homing, reduced CD4/CD8alphaalpha intraepithelial lymphocytes, and inflammation in the gut of vitamin D receptor KO mice. Proc Natl Acad Sci U S A. Dec 30 2008;105(52):20834-20839. 86. Simmons JD, Mullighan C, Welsh KI, Jewell DP. Vitamin D receptor gene polymorphism: association with Crohn's disease susceptibility. Gut. Aug 2000;47(2):211-214. 87. Naderi N, Farnood A, Habibi M, et al. Association of vitamin D receptor gene polymorphisms in Iranian patients with inflammatory bowel disease. J Gastroenterol Hepatol. Dec 2008;23(12):1816-1822.

123

88. Dresner-Pollak R, Ackerman Z, Eliakim R, Karban A, Chowers Y, Fidder HH. The BsmI vitamin D receptor gene polymorphism is associated with ulcerative colitis in Jewish Ashkenazi patients. Genet Test. Winter 2004;8(4):417-420. 89. Hughes DJ, McManus R, Neary P, O'Morain C, O'Sullivan M. Common variation in the vitamin D receptor gene and risk of inflammatory bowel disease in an Irish case-control study. Eur J Gastroenterol Hepatol. Sep 2011;23(9):807-812. 90. Jostins L, Ripke S, Weersma RK, et al. Host-microbe interactions have shaped the genetic architecture of inflammatory bowel disease. Nature. Nov 1 2012;491(7422):119-124. 91. Eloranta JJ, Wenger C, Mwinyi J, et al. Association of a common vitamin D- binding protein polymorphism with inflammatory bowel disease. Pharmacogenet Genomics. Sep 2011;21(9):559-564. 92. Wang TT, Dabbas B, Laperriere D, et al. Direct and indirect induction by 1,25- dihydroxyvitamin D3 of the NOD2/CARD15-defensin beta2 innate immune pathway defective in Crohn disease. J Biol Chem. Jan 22 2010;285(4):2227-2231. 93. Armitage EL, Aldhous MC, Anderson N, et al. Incidence of juvenile-onset Crohn's disease in Scotland: association with northern latitude and affluence. Gastroenterology. Oct 2004;127(4):1051-1057. 94. Wolters FL, Joling C, Russel MG, et al. Treatment inferred disease severity in Crohn's disease: evidence for a European gradient of disease course. Scand J Gastroenterol. Mar 2007;42(3):333-344. 95. Nerich V, Monnet E, Etienne A, et al. Geographical variations of inflammatory bowel disease in France: a study based on national health insurance data. Inflamm Bowel Dis. Mar 2006;12(3):218-226. 96. Nerich V, Jantchou P, Boutron-Ruault MC, et al. Low exposure to sunlight is a risk factor for Crohn's disease. Aliment Pharmacol Ther. Apr 2011;33(8):940- 945. 97. Joseph AJ, George B, Pulimood AB, Seshadri MS, Chacko A. 25 (OH) vitamin D level in Crohn's disease: association with sun exposure & disease activity. Indian J Med Res. Aug 2009;130(2):133-137. 98. Shivananda S, Lennard-Jones J, Logan R, et al. Incidence of inflammatory bowel disease across Europe: is there a difference between north and south? Results of the European Collaborative Study on Inflammatory Bowel Disease (EC-IBD). Gut. Nov 1996;39(5):690-697. 99. Nerich V, Monnet E, Weill A, et al. Fine-scale geographic variations of inflammatory bowel disease in France: correlation with socioeconomic and house equipment variables. Inflamm Bowel Dis. May 2010;16(5):813-821. 100. Khalili H, Huang ES, Ananthakrishnan AN, et al. Geographical variation and incidence of inflammatory bowel disease among US women. Gut. Dec 2012;61(12):1686-1692. 101. Sonnenberg A, Genta RM. Geographic distributions of microscopic colitis and inflammatory bowel disease in the United States. Inflamm Bowel Dis. Dec 2012;18(12):2288-2293. 102. Ananthakrishnan AN, Khalili H, Higuchi LM, et al. Higher predicted vitamin D status is associated with reduced risk of Crohn's disease. Gastroenterology. Mar 2012;142(3):482-489.

124

103. Ananthakrishnan AN, Cagan A, Gainer VS, et al. Normalization of Plasma 25- Hydroxy Vitamin D Is Associated with Reduced Risk of Surgery in Crohn's Disease. Inflamm Bowel Dis. Aug 2013;19(9):1921-1927. 104. Jorgensen SP, Agnholt J, Glerup H, et al. Clinical trial: vitamin D3 treatment in Crohn's disease - a randomized double-blind placebo-controlled study. Aliment Pharmacol Ther. Aug 2010;32(3):377-383. 105. Miheller P, Muzes G, Hritz I, et al. Comparison of the effects of 1,25 dihydroxyvitamin D and 25 hydroxyvitamin D on bone pathology and disease activity in Crohn's disease patients. Inflamm Bowel Dis. Nov 2009;15(11):1656- 1662. 106. Yang L, Weaver V, Smith JP, Bingaman S, Hartman TJ, Cantorna MT. Therapeutic effect of vitamin d supplementation in a pilot study of Crohn's patients. Clin Transl Gastroenterol. 2013;4:e33. 107. Pappa HM, Langereis EJ, Grand RJ, Gordon CM. Prevalence and risk factors for hypovitaminosis D in young patients with inflammatory bowel disease. J Pediatr Gastroenterol Nutr. Oct 2011;53(4):361-364. 108. Ulitsky A, Ananthakrishnan AN, Naik A, et al. Vitamin D deficiency in patients with inflammatory bowel disease: association with disease activity and quality of life. JPEN J Parenter Enteral Nutr. May 2011;35(3):308-316. 109. Jahnsen J, Falch JA, Mowinckel P, Aadland E. Vitamin D status, parathyroid hormone and bone mineral density in patients with inflammatory bowel disease. Scand J Gastroenterol. Feb 2002;37(2):192-199. 110. Siffledeen JS, Siminoski K, Steinhart H, Greenberg G, Fedorak RN. The frequency of vitamin D deficiency in adults with Crohn's disease. Can J Gastroenterol. Aug 2003;17(8):473-478. 111. Sinnott BP, Licata AA. Assessment of bone and mineral metabolism in inflammatory bowel disease: case series and review. Endocr Pract. Nov-Dec 2006;12(6):622-629. 112. Tajika M, Matsuura A, Nakamura T, et al. Risk factors for vitamin D deficiency in patients with Crohn's disease. J Gastroenterol. Jun 2004;39(6):527-533. 113. McCarthy D, Duggan P, O'Brien M, et al. Seasonality of vitamin D status and bone turnover in patients with Crohn's disease. Aliment Pharmacol Ther. May 1 2005;21(9):1073-1083. 114. Lo CW, Paris PW, Clemens TL, Nolan J, Holick MF. Vitamin D absorption in healthy subjects and in patients with intestinal malabsorption syndromes. Am J Clin Nutr. Oct 1985;42(4):644-649. 115. Farraye FA, Nimitphong H, Stucchi A, et al. Use of a novel vitamin D bioavailability test demonstrates that vitamin D absorption is decreased in patients with quiescent Crohn's disease. Inflamm Bowel Dis. Oct 2011;17(10):2116-2121. 116. de Lange KM, Moutsianas L, Lee JC, et al. Genome-wide association study implicates immune activation of multiple integrin genes in inflammatory bowel disease. Nat Genet. Feb 2017;49(2):256-261. 117. Halfvarson J. Genetics in twins with Crohn's disease: less pronounced than previously believed? Inflamm Bowel Dis. Jan 2011;17(1):6-12.

125

118. Shaw SY, Blanchard JF, Bernstein CN. Association between the use of antibiotics and new diagnoses of Crohn's disease and ulcerative colitis. Am J Gastroenterol. Dec 2011;106(12):2133-2142. 119. Blanck S, Aberra F. Vitamin d deficiency is associated with ulcerative colitis disease activity. Dig Dis Sci. Jun 2013;58(6):1698-1702. 120. Autier P, Boniol M, Pizot C, Mullie P. Vitamin D status and ill health: a systematic review. Lancet Diabetes Endocrinol. Jan 2014;2(1):76-89. 121. Froicu M, Cantorna MT. Vitamin D and the vitamin D receptor are critical for control of the innate immune response to colonic injury. BMC Immunol. 2007;8:5. 122. Kong J, Zhang Z, Musch MW, et al. Novel role of the vitamin D receptor in maintaining the integrity of the intestinal mucosal barrier. Am J Physiol Gastrointest Liver Physiol. Jan 2008;294(1):G208-216. 123. Bours PH, Wielders JP, Vermeijden JR, van de Wiel A. Seasonal variation of serum 25-hydroxyvitamin D levels in adult patients with inflammatory bowel disease. Osteoporos Int. Nov 2011;22(11):2857-2867. 124. Kimlin MG, Lucas RM, Harrison SL, et al. The contributions of solar ultraviolet radiation exposure and other determinants to serum 25-hydroxyvitamin D concentrations in Australian adults: the AusD Study. Am J Epidemiol. Apr 1 2014;179(7):864-874. 125. Frederick JE, Lubin D. Possible long-term changes in biologically active ultraviolet radiation reaching the ground. Photochem Photobiol. Apr 1988;47(4):571-578. 126. Long CS, Miller AJ, Lee HT, Wild JD, Przywarty RC, Hufford D. Ultraviolet index forecasts issued by the National Weather Service. Bulletin of the American Meteorological Society. Apr 1996;77(4):729-748. 127. Gies P, Roy C, Javorniczky J, Henderson S, Lemus-Deschamps L, Driscoll C. Global Solar UV Index: Australian measurements, forecasts and comparison with the UK. Photochem Photobiol. Jan 2004;79(1):32-39. 128. Wier LM PA, Maeda J, Stranges E, Ryan K, Jagadish P, Collins Sharp B, Elixhauser A. HCUP Facts and Figures: Statistics on Hospital-Based Care in the United States, 2009. Rockville (MD)2011. 129. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-383. 130. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. Jun 1992;45(6):613- 619. 131. Nguyen GC, Steinhart AH. Nationwide patterns of hospitalizations to centers with high volume of admissions for inflammatory bowel disease and their impact on mortality. Inflamm Bowel Dis. Dec 2008;14(12):1688-1694. 132. Houchens R, Elixhauser A. Final Report on Calculating Nationwide Inpatient Sample (NIS) Variances, 2001.2005. #2003-2. 133. Eide MJ, Weinstock MA. Association of UV index, latitude, and melanoma incidence in nonwhite populations--US Surveillance, Epidemiology, and End Results (SEER) Program, 1992 to 2001. Arch Dermatol. Apr 2005;141(4):477- 481.

126

134. Hu S, Ma F, Collado-Mesa F, Kirsner RS. UV radiation, latitude, and melanoma in US Hispanics and blacks. Arch Dermatol. Jul 2004;140(7):819-824. 135. Hiller R, Giacometti L, Yuen K. Sunlight and cataract: an epidemiologic investigation. Am J Epidemiol. May 1977;105(5):450-459. 136. Beretich BD, Beretich TM. Explaining multiple sclerosis prevalence by ultraviolet exposure: a geospatial analysis. Mult Scler. Aug 2009;15(8):891-898. 137. Kimlin MG, Olds WJ, Moore MR. Location and vitamin D synthesis: is the hypothesis validated by geophysical data? J Photochem Photobiol B. Mar 1 2007;86(3):234-239. 138. Thirumurthi S, Chowdhury R, Richardson P, Abraham NS. Validation of ICD-9- CM diagnostic codes for inflammatory bowel disease among veterans. Dig Dis Sci. Sep 2010;55(9):2592-2598. 139. Mozoaowski W. Jedrzej Sniadecki (1768-1838) on the cure of rickets. Nature. 1939;143(3612):121. 140. Molodecky NA, Soon IS, Rabi DM, et al. Increasing incidence and prevalence of the inflammatory bowel diseases with time, based on systematic review. Gastroenterology. Jan 2012;142(1):46-54 e42; quiz e30. 141. Ross AC. The 2011 report on dietary reference intakes for calcium and vitamin D. Public Health Nutr. May 2011;14(5):938-939. 142. Bailey RL, Dodd KW, Goldman JA, et al. Estimation of total usual calcium and vitamin D intakes in the United States. J Nutr. Apr 2010;140(4):817-822. 143. Pappa HM, Bern E, Kamin D, Grand RJ. Vitamin D status in gastrointestinal and liver disease. Curr Opin Gastroenterol. Mar 2008;24(2):176-183. 144. Limketkai BN, Bayless TM, Brant SR, Hutfless SM. Lower regional and temporal ultraviolet exposure is associated with increased rates and severity of inflammatory bowel disease hospitalisation. Aliment Pharmacol Ther. Jun 18 2014. 145. Higgins JP, Green S. Cochrane Handbook for Systematic Reviews of Interventions. 2011; http://handbook.cochrane.org. Accessed October 20, 2014. 146. Baars JE, Vogelaar L, Wolfhagen FH, Biermann K, Kuipers EJ, van der Woude CJ. A short course of corticosteroids prior to surveillance colonoscopy to decrease mucosal inflammation in inflammatory bowel disease patients: results from a randomized controlled trial. J Crohns Colitis. Dec 2010;4(6):661-668. 147. Bernstein CN, Seeger LL, Anton PA, et al. A randomized, placebo-controlled trial of calcium supplementation for decreased bone density in corticosteroid-using patients with inflammatory bowel disease: a pilot study. Aliment Pharmacol Ther. Oct 1996;10(5):777-786. 148. Boothe D, Lakehomer H, Jacob V, Scherl E. High dose vitamin D3 improves clinical activity in Crohn's disease. Am J Gastroenterol. 2011;106(S2):S458. 149. Dendrinos K, Stucchi A, Becker JM, et al. Efficacy of two different weekly doses of vitamin D (400 IU and 50,000 IU) in patients with Crohn's disease in improving vitamin D nutritional status. Gastroenterology. 2008;134(4):A-495. 150. Kitazaki S, Mitsuyama K, Masuda J, et al. Clinical trial: comparison of alendronate and alfacalcidol in glucocorticoid-associated osteoporosis in patients with ulcerative colitis. Aliment Pharmacol Ther. Feb 15 2009;29(4):424-430.

127

151. Mathur J, Mills P, Naing S, Limsui D. Supplementation of vitamin D3 (cholecalciferol) in patients with ulcerative colitis and hypovitaminosis D: a prospective randomized trial. Gastroenterology. 2014;146(5):S-454. 152. Pappa HM, Mitchell PD, Jiang H, et al. Treatment of vitamin D insufficiency in children and adolescents with inflammatory bowel disease: a randomized clinical trial comparing three regimens. J Clin Endocrinol Metab. Jun 2012;97(6):2134- 2142. 153. Pappa HM, Mitchell PD, Jiang H, et al. Maintenance of optimal vitamin D status in children and adolescents with inflammatory bowel disease: a randomized clinical trial comparing two regimens. J Clin Endocrinol Metab. Sep 2014;99(9):3408-3417. 154. Raftery T, Martineau AR, Greiller CL, et al. Effects of vitamin D supplementation on intestinal permeability, cathelicidin and disease markers in Crohn's disease: Results from a randomised double-blind placebo-controlled study. United European Gastroenterol J. Jun 2015;3(3):294-302. 155. Vogelsang H, Ferenci P, Resch H, Kiss A, Gangl A. Prevention of bone mineral loss in patients with Crohn's disease by long-term oral vitamin D supplementation. Eur J Gastroenterol Hepatol. Jul 1995;7(7):609-614. 156. Wingate KE, Jacobson K, Issenman R, et al. 25-Hydroxyvitamin D concentrations in children with Crohn's disease supplemented with either 2000 or 400 IU daily for 6 months: a randomized controlled study. J Pediatr. Apr 2014;164(4):860- 865. 157. Abitbol V, Mary JY, Roux C, et al. Osteoporosis in inflammatory bowel disease: effect of calcium and vitamin D with or without fluoride. Aliment Pharmacol Ther. May 2002;16(5):919-927. 158. Abitbol V, Briot K, Roux C, et al. A double-blind placebo-controlled study of intravenous clodronate for prevention of steroid-induced bone loss in inflammatory bowel disease. Clin Gastroenterol Hepatol. Oct 2007;5(10):1184- 1189. 159. Bendix-Struve M, Bartels LE, Agnholt J, Dige A, Jorgensen SP, Dahlerup JF. Vitamin D3 treatment of Crohn's disease patients increases stimulated T cell IL-6 production and proliferation. Aliment Pharmacol Ther. Dec 2010;32(11-12):1364- 1372. 160. Bentley RW, Keown D, Merriman TR, et al. Vitamin D receptor gene polymorphism associated with inflammatory bowel disease in New Zealand males. Aliment Pharmacol Ther. Apr 2011;33(7):855-856. 161. Klaus J, Reinshagen M, Herdt K, Adler G, von Boyen GB, von Tirpitz C. Intravenous ibandronate or sodium-fluoride--a 3.5 years study on bone density and fractures in Crohn's disease patients with osteoporosis. J Gastrointestin Liver Dis. Jun 2011;20(2):141-148. 162. Pappa HM, Saslowsky TM, Filip-Dhima R, et al. Efficacy and harms of nasal calcitonin in improving bone density in young patients with inflammatory bowel disease: a randomized, placebo-controlled, double-blind trial. Am J Gastroenterol. Aug 2011;106(8):1527-1543.

128

163. Siffledeen JS, Fedorak RN, Siminoski K, et al. Randomized trial of etidronate plus calcium and vitamin D for treatment of low bone mineral density in Crohn's disease. Clin Gastroenterol Hepatol. Feb 2005;3(2):122-132. 164. Soo I, Siffledeen J, Siminoski K, McQueen B, Fedorak RN. Risedronate improves bone mineral density in Crohn's disease: a two year randomized controlled clinical trial. J Crohns Colitis. Aug 2012;6(7):777-786. 165. van Bodegraven AA, Bravenboer N, Witte BI, et al. Treatment of bone loss in osteopenic patients with Crohn's disease: a double-blind, randomised trial of oral risedronate 35 mg once weekly or placebo, concomitant with calcium and vitamin D supplementation. Gut. Sep 2014;63(9):1424-1430. 166. von Tirpitz C, Klaus J, Bruckel J, et al. Increase of bone mineral density with sodium fluoride in patients with Crohn's disease. Eur J Gastroenterol Hepatol. Jan 2000;12(1):19-24. 167. von Tirpitz C, Klaus J, Steinkamp M, et al. Therapy of osteoporosis in patients with Crohn's disease: a randomized study comparing sodium fluoride and ibandronate. Aliment Pharmacol Ther. Mar 15 2003;17(6):807-816. 168. Hastie T, Tibshirani R, Friedman JH. The elements of statistical learning : data mining, inference, and prediction. 2nd ed. New York, NY: Springer; 2009. 169. Giovannucci E, Liu Y, Rimm EB, et al. Prospective study of predictors of vitamin D status and cancer incidence and mortality in men. J Natl Cancer Inst. Apr 5 2006;98(7):451-459. 170. Escudero-Hernandez C, Pena AS, Bernardo D. Immunogenetic Pathogenesis of Celiac Disease and Non-celiac Gluten Sensitivity. Curr Gastroenterol Rep. Jul 2016;18(7):36.

129

APPENDICES

Appendix 1. Electronic search strategy for systematic review on randomized trials evaluating vitamin D intervention in inflammatory bowel disease.

EMBASE [years covered: from 1947 to date] 1. random$.tw. 2. factorial$.tw. 3. (crossover$ or cross over$ or cross-over$).tw. 4. placebo$.tw. 5. single blind.mp. 6. double blind.mp. 7. triple blind.mp. 8. (singl$ adj blind$).tw. 9. (double$ adj blind$).tw. 10. (tripl$ adj blind$).tw. 11. assign$.tw. 12. allocat$.tw. 13. crossover procedure/ 14. double blind procedure/ 15. single blind procedure/ 16. triple blind procedure/ 17. randomized controlled trial/ 18. or/1-17 19. (exp animal/ or animal.hw. or nonhuman/) not (exp human/ or human cell/ or (human or humans).ti.) 20. 18 not 19 21. exp Crohn disease/ or crohn*.mp. 22. (colitis and ulcerat*).mp. 23. ulcerative colitis.mp. or exp ulcerative colitis/ 24. inflammatory bowel disease*.mp. 25. IBD.mp. 26. extraintestinal.mp. 27. 21 or 22 or 23 or 24 or 25 or 26 28. 20 and 27 29. exp Vitamin D/ 30. (vitamin D or vitamin D2 or vitamin D3).tw. 31. (al*acalcidol or c?olecalciferol or calcitriol or calcidiol or calcifediol or calciferol or calciol or calderol or dihydrotachysterol or dedrogyl or dihydrotachysterol or dihydroxycolecalciferol or dihydroxycholecalciferol or dihydroxyvitamin D or dihydroxyvitamin D2 or dihydroxyvitamin D3 or doxercalciferol or eldecalcitol or ercalcidiol or ergocalciferol* or hidroferol or hydroxycalciferol or hydroxylcalciferol or hydroxycolecalciferol or hydroxycholecalciferol or hydroxyergocalciferol* or hydroxyvitaminDor hydroxyvitamin D2 or hydroxyvitamin D3 or paricalcitol or tachystin).tw. 32. or/29-31

130

33. 28 and 32

MEDLINE [years covered: from 1945 to date] 1. random$.tw. 2. factorial$.tw. 3. (crossover$ or cross over$ or cross-over$).tw. 4. placebo$.tw. 5. single blind.mp. 6. double blind.mp. 7. triple blind.mp. 8. (singl$ adj blind$).tw. 9. (double$ adj blind$).tw. 10. (tripl$ adj blind$).tw. 11. assign$.tw. 12. allocat$.tw. 13. crossover procedure/ 14. double blind procedure/ 15. single blind procedure/ 16. triple blind procedure/ 17. randomized controlled trial/ 18. or/1-17 19. (exp animal/ or animal.hw. or nonhuman/) not (exp human/ or human cell/ or (human or humans).ti.) 20. 18 not 19 21. exp Crohn disease/ or crohn*.mp. 22. (colitis and ulcerat*).mp. 23. ulcerative colitis.mp. or exp ulcerative colitis/ 24. inflammatory bowel disease*.mp. 25. IBD.mp. 26. extraintestinal.mp. 27. 21 or 22 or 23 or 24 or 25 or 26 28. 20 and 27 29. exp Vitamin D/ 30. (vitamin D or vitamin D2 or vitamin D3).tw. 31. (al*acalcidol or c?olecalciferol or calcitriol or calcidiol or calcifediol or calciferol or calciol or calderol or dihydrotachysterol or dedrogyl or dihydrotachysterol or dihydroxycolecalciferol or dihydroxycholecalciferol or dihydroxyvitamin D or dihydroxyvitamin D2 or dihydroxyvitamin D3 or doxercalciferol or eldecalcitol or ercalcidiol or ergocalciferol* or hidroferol or hydroxycalciferol or hydroxylcalciferol or hydroxycolecalciferol or hydroxycholecalciferol or hydroxyergocalciferol* or hydroxyvitaminD or hydroxyvitamin D2 or hydroxyvitamin D3 or paricalcitol or tachystin).tw. 32. or/29-31 33. 28 and 32

131

Cochrane Central Register of Controlled Trials (CENTRAL) #1 crohn* or IBD or (inflammatory bowel disease*) or (ulcerative colitis) or colitis #2 Vitamin D or calcitriol or calcidiol or calcifediol or calciferol or calciol or calderol or dihydrotachysterol or dedrogyl or dihydrotachysterol or dihydroxycolecalciferol or dihydroxycholecalciferol or dihydroxyvitamin Dor dihydroxyvitaminD2 or dihydroxyvitamin D3 or doxercalciferol or eldecalcitol or ercalcidiol or ergocalciferol* or hidroferol or hydroxycalciferol or hydroxylcalciferol or hydroxycolecalciferol or hydroxycholecalciferol or hydroxyergocalciferol* or hydroxyvitamin D or hydroxyvitamin D2 or hydroxyvitamin D3 or paricalcitol or tachystin. #3 #1 and #2

132

CURRICULUM VITAE

BERKELEY N. LIMKETKAI, M.D.

PERSONAL DATA

300 Pasteur Drive, Alway M211 Stanford, CA 94305 (650) 736-0431 E-mail: [email protected]

ACADEMIC HISTORY

Undergraduate 1993-1998 A.B., Molecular & Cell Biology, University of California, Berkeley, CA 1993-1998 A.B., French, University of California, Berkeley, CA

Graduate/Doctoral 1999-2003 M.S., Biomedical Laboratory Science, San Francisco State University, San Francisco, CA 2003-2007 M.D., University of Cincinnati College of Medicine, Cincinnati, OH 2011-present Ph.D. Candidate, Clinical Investigation, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD

Postdoctoral 2007-2010 Residency, Internal Medicine, Johns Hopkins Hospital, Baltimore, MD 2010-2014 Clinical & Research Fellowship, Gastroenterology, Johns Hopkins University School of Medicine, Baltimore, MD 2013-2014 Theodore M. Bayless Fellowship, Inflammatory Bowel Diseases, Johns Hopkins University School of Medicine, Baltimore, MD 2015 Fellowship, Clinical Nutrition, Nestlé Nutrition Institute

PROFESSIONAL APPOINTMENTS

Current Appointments 2014-present Director, Gastrointestinal Nutrition, Stanford Health Care, Palo Alto, CA 2014-present Clinical Assistant Professor, Division of Gastroenterology & Hepatology, Department of Medicine, Stanford University School of Medicine, Stanford, CA 2014-present Adjunct Assistant Professor, Division of Gastroenterology & Hepatology, Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD 2014-present Staff Physician, Veterans Affairs Palo Alto Health Care System, Palo Alto, CA

133

HONORS & AWARDS

2001 First Place, San Francisco State University Research Competition 2001 First Place, California State University (statewide) Research Competition 2004 Second Place, Excellence in Medical Physiology Award 2004 McGraw-Hill Award. Top student in medical school class 2005 First Place, Adriano Essay Writing Competition in Pathology 2005 H. B. Weiss Award. Academic excellence in medical school 2006 Alpha Omega Alpha. National medical honor society 2006 First Place, University of Cincinnati Grand Rounds Poster Competition 2006 First Place, Ohio ACP Clinical Vignette Poster Competition 2007 Winner, National ACP Clinical Vignette Abstract Competition 2007 Honorable Mention, University of Cincinnati Trainees’ Research Poster Competition 2007 Outstanding Medical Student, American College of Physicians 2012 Clinical Research Award. Top score on JHSPH department exam 2012 Travel Award, IBD Mentoring Program for GI Fellows 2012 Best Young Investigator Poster, Crohn’s and Colitis Foundation of America 2013 Investing in the Future Award, American Gastroenterological Association 2014 Visiting IBD Fellowship, Crohn’s & Colitis Foundation of America 2014 Academic Skills Workshop Travel Award, American Gastroenterological Association 2014 Fellowship2Leadership Award, Salix Pharmaceuticals 2015 Poster of Distinction, American Gastroenterological Association

SCHOLARLY PUBLICATIONS

Peer-Reviewed Original Publications 1. Limketkai BN, Zucker SD. Hyperammonemic encephalopathy caused by carnitine deficiency. J Gen Intern Med 2008;23(2):210-3. 2. Limketkai BN, Chandrasekhara V, Milligan F. Primary amelanotic anorectal melanoma presenting as internal hemorrhoids. Gastroenterol Hepatol (N Y) 2009;5:516-8. 3. Dryer PD *, Limketkai BN *, Martin CM, Ma G, Sherman KE, Taylor LE, Mayer KH, Jamieson DJ, Blackard JT. Screening for hepatitis C virus non-nucleoside resistance mutations in treatment naïve women. J Antimicrob Chemother 2009;64(5):945-8. *First authors. 4. Blackard JT, Ma G, Limketkai BN, Welge JA, Dryer PD, Martin CM, Hiasa Y, Taylor LE, Mayer KH, Jamieson DJ, Sherman KE. Variability of the polymerase gene (NS5B) in HCV-infected women. J Clin Microbiol 2010;48(11):4256-9. 5. Limketkai BN, Mehta SH, Sutcliffe CG, Higgins YM, Torbenson MS, Brinkley SC, Moore RD, Thomas DL, Sulkowski MS. Relationship of liver disease stage and antiviral therapy with liver-related events and death in adults coinfected with HIV/HCV. JAMA 2012;308(4):370-8.

134

6. Arnold CA, Limketkai BN, Illei PB, Montgomery E, Voltaggio L. Syphilitic and lymphogranuloma venereum (LGV) proctocolitis: clues to a frequently missed diagnosis. Am J Surg Pathol 2013;37(1):38-46. *Cover image. 7. Gultepe E, Yamanaka S, Laflin KE, Kadam S, Shim Y, Olaru AV, Limketkai BN, Khashab MA, Kalloo AN, Gracias DH, Selaru FM. Biologic tissue sampling with untethered microgrippers. Gastroenterology 2013;144(4):691-3. 8. Kim B, Woreta T, Chen PH, Limketkai B, Singer A, Dagher N, Cameron A, Lin MT, Kamel I, Gurakar A. Donor-transmitted malignancy in a liver transplant recipient: a case report and review of literature. Dig Dis Sci 2013;58(5):1185-90. 9. Limketkai BN, Chandrasekhara V, Kalloo AN, Okolo PI 3rd. Comparison of performance and safety of endoscopic retrograde cholangiopancreatography across pediatric age groups. Dig Dis Sci 2013;58(9):2653-60. 10. Swanson BJ, Limketkai BN, Liu T, Montgomery E, Nazari K, Park JY, Santangelo WC, Torbenson MS, Voltaggio L, Yearsley M, Arnold CA. Sevelamer crystals in the gastrointestinal tract (GIT): a new entity associated with mucosal injury. Am J Surg Pathol 2013;37(11):1686-93. 11. Arnold MA, Schoenfield L, Limketkai BN, Arnold CA. Diagnostic pitfalls of differentiating desmoplastic small round cell tumor (DSRCT) from Wilms tumor (WT): overlapping morphologic and immunohistochemical features. Am J Surg Pathol 2014;38(9):1220-6. 12. Limketkai BN, Bayless TM, Brant SR, Hutfless SM. Lower regional and temporal ultraviolet exposure is associated with increased rates and severity of inflammatory bowel disease hospitalisation. Aliment Pharmacol Ther 2014;40(5):508-17. 13. Nowacki NB, Arnold MA, Frankel WL, Harzman A, Limketkai BN, Yearsley MM, Arnold CA. Gastrointestinal tract pulse granulomata: clues to an under-recognized pseudotumor. Am J Surg Pathol 2015;39:84-92. 14. Ma C, Park JY, Montgomery EA, Arnold CA, McDonald OG, Liu TC, Salaria SN, Limketkai BN, McGrath KM, Musahl T, Singhi AD. A comparative clinicopathological study of collagenous gastritis in children and adults: the same disorder with associated immune-mediated diseases. Am J Surg Pathol 2015;39(6):802-12. 15. Kim B, Tan A, Limketkai BN, Pinney SP, Schiano TD. Comparison of outcome in patients with versus without ascites referred for either cardiac transplantation or ventricular assisted device placement. Am J Cardiol 2015;116(10):1596-600. 16. Arnold CA, Roth R, Arsenescu R, Harzman A, Lam-Himlin D, Limketkai BN, Montgomery EA, Voltaggio L. Sexually transmitted infectious colitis versus inflammatory bowel disease: distinguishing features from a case controlled study. Am J Clin Pathol 2015;144(5):771-81. 17. Chen PH, Limketkai BN, Trilianos P, Pirtini-Cetingul M, Woreta TA, Kim B, Gulsen MT, Segev DL, Cameron AM, Gurakar A. Effect of prior hepatitis B virus exposure on long-term risk of liver-related events after liver transplantation. Clin Transplant 2016;30(5):579-88. 18. Yokomizo L, Limketkai B, Park KT. Cost-effectiveness of adalimumab, infliximab, or vedolizumab as first-line biologic therapy in moderate-to-severe ulcerative colitis. BMJ Open Gastroenterol 2016;3(1):e93

135

19. Parian A, Koh J, Limketkai BN, Eluri S, Rubin DT, Brant SR, Ha CY, Bayless TM, Giardiello F, Hart J, Montgomery E, Lazarev MG. Association between serrated epithelial changes and colorectal dysplasia in inflammatory bowel disease. Gastrointest Endosc 2016;84(1):87-95. 20. Limketkai BN, Orandi BJ, Luo X, Segev DL, Colombel J. Mortality and rates of graft rejection or failure following intestinal transplantation in patients with vs without Crohn’s disease. Clin Gastroenterol Hepatol 2016;14(11):1574-1581. 21. Parian A *, Limketkai B *, Koh J, Bitton A, Brant SR, Cho JH, Duerr RH, McGovern DP, Proctor DD, Reguiero M, Rioux JD, Schumm PP, Taylor KD, Silverberg MS, Steinhart AH, Lazarev M. Appendectomy is an independent predictor of colectomy in ulcerative colitis: results from a large cross-sectional cohort. Gut 2016 (in press) *First authors

Invited Reviews 1. Limketkai BN, Lam-Himlin D, Arnold MA, Arnold CA. The cutting edge of serrated polyps: a practical guide to approaching and managing serrated colon polyps. Gastrointest Endosc 2013;77(3):360-75. 2. Voltaggio L, Lam-Himlin D, Limketkai BN, Singhi AD, Arnold CA. Message in a bottle: decoding medication injury patterns in the gastrointestinal tract. J Clin Pathol 2014;67(10):903-12. 3. Ma C, Limketkai BN, Montgomery EA. Recently highlighted non-neoplastic pathologic entities of the upper gastrointestinal tract and their clinical significance. Gastrointest Endosc 2014;80(6):960-9. 4. Shah ND, Parian A, Mullin GE, Limketkai BN *. Oral diets and nutrition support for inflammatory bowel disease: what is the evidence? Nutr Clin Pract 2015;30(4):462- 73. *Senior author 5. Parian AM, Limketkai BN, Shah ND, Mullin GE. Nutraceutical supplements for inflammatory bowel disease. Nutr Clin Pract 2015;30(4):551-8. 6. Shah ND, Limketkai BN *. Low residue vs. low fiber diets in inflammatory bowel disease: evidence to support vs. habit? Practical Gastroenterology 2015;39(7):48-57. *Senior author 7. Parian A, Limketkai BN *. Dietary supplement therapies for inflammatory bowel disease: Crohn’s disease and ulcerative colitis. Curr Pharm Des 2015;22(2):180-8. *Senior author 8. Limketkai BN, Parian AM, Shah ND, Colombel J. Short bowel syndrome and intestinal failure in Crohn’s disease. Inflamm Bowel Dis 2016;22(5):1209-18. 9. Limketkai BN, Bechtold ML, Nguyen DL. Vitamin D and the pathogenesis of inflammatory bowel disease. Curr Gastroenterol Rep 2016;18(10):52 10. Nguyen DL, Limketkai B, Medici V, Saire-Mendoza M, Palmer L, Bechtold M. Nutritional strategies in the management of adult patients with inflammatory bowel disease: dietary considerations from active disease to disease remission. Curr Gastroenterol Rep 2016;18(10):55. 11. Shah ND, Limketkai BN *. The use of medium-chain triglycerides in gastrointestinal disorders. Practical Gastroenterology 2016 (in press) *Senior author 12. Limketkai BN, Mullin GE, Limsui D, Parian AM. The role of vitamin D in Inflammatory Bowel Disease. Nutr Clin Pract 2016 (in press)

136

Invited Editorials 1. Limketkai BN, Bayless TM. Can stenosis in ileal Crohn’s disease be prevented by current therapy? Am J Gastroenterol 2013;108(11):1755-6.

Non-Peer Reviewed Publications 1. Limketkai BN. Genetic Polymorphisms in Tat from PBMC of HIV-infected Patients on Antiretroviral Therapy [master’s thesis]. San Francisco: San Francisco State University; 2003.

Books and Chapters 1. Editor. Osler Residency Survival Guide 2008-2009. Baltimore, MD: Johns Hopkins; 2008. 2. Illustrator. In: Arnold CA, Lam-Himlin D, Montgomery EA. Atlas of Gastrointestinal Pathology: A Pattern Based Approach to Non-neoplastic Biopsies. 1st ed. Philadelphia, PA: Lippincott Williams & Wilkins; 2014. 3. Chan L, Limketkai BN. Electrolyte and acid-base disturbances. In: DiBaise JK, Parrish CR, Thompson JS, eds. Short Bowel Syndrome: Practical Approach to Management. Boca Raton, FL: CRC Press; 2016. 4. Limketkai BN, Hurt R, Palmer L. Short bowel syndrome. In: Mueller CM, Kovacevich DS, McClave SA, Miller SJ, Schwartz DB, eds. The A.S.P.E.N. Adult Nutrition Support Core Curriculum. 3rd ed. Silver Spring, MD: A.S.P.E.N. (accepted).

EDITORIAL SERVICE

Editorial Board Appointments 2017-present World Journal of Gastrointestinal Physiology

Journal Reviewer 2014-present Digestive Diseases & Sciences 2014-present Inflammatory Bowel Diseases 2015-present Alimentary Pharmacology & Therapeutics 2016-present PLOS ONE 2016-present Expert Review of Gastroenterology & Hepatology 2017-present World Journal of Gastroenterology

GRANTS

Extramural Sponsorship

Current Grants: 1/2017 – 12/2017 Probiotics in Inflammatory Bowel Disease Private donor Co-PIs: Limketkai BN, Habtezion A

137

7/3/15 – 6/30/18 Evaluating Safety and Efficacy of Herbal Treatment in Ulcerative Colitis Weston Havens Foundation PI: Limketkai BN

Previous Grants: 8/1/13-12/31/14 Vitamin D Levels and Genotype Interactions in Incident Crohn’s Disease GI Fellows Research Award IBD Working Group PI: Limketkai BN

7/1/13-6/30/16 Role of Vitamin D in Crohn’s Disease Pathogenesis Loan Repayment Program for Clinical Researchers National Institutes of Health PI: Limketkai BN

7/1/13-6/30/15 The Role of Vitamin D in Crohn’s Disease Pathogenesis Clinical Research Award American College of Gastroenterology PI: Limketkai BN

7/1/13-6/30/14 Advanced Fellowship in Inflammatory Bowel Diseases Janssen Biotech PI: Bayless TM

7/1/11-6/30/13 Basic Science Research in Digestive Diseases 5T32DK007632-24 NIH/NIDDK PI: Donowitz M

7/1/10-6/30/11 Morbidity and mortality among HBV-infected/exposed patients who undergo liver transplantation Viral Hepatology Fellowship Bristol-Myers Squibb PI: Limketkai BN

Intramural Sponsorship

Previous Grants: 7/1/08-6/30/10 Uroguanylin in cirrhotogenic ascites Osler Fund for Research Scholarship Johns Hopkins University School of Medicine PI: Limketkai BN

138

UNIVERSITY ADMINISTRATIVE SERVICE

Committee Service 2009-2010 Clinical Data and Documentation Committee, Johns Hopkins Hospital, Baltimore, MD 2015-present Nutrition Committee, Stanford Health Care, Stanford, CA 2015-present Fellowship Selection Committee, Division of Gastroenterology & Hepatology, Stanford University School of Medicine, Stanford, CA 2016-present Education Advisory Committee, Division of Gastroenterology & Hepatology, Stanford University School of Medicine, Stanford, CA 2017-present Leadership Council for Clinical Excellence, Stanford Health Care, Stanford, CA

SERVICE TO PROFESSIONAL ORGANIZATIONS

Membership 2003-2010 American College of Physicians 2006-present Alpha Omega Alpha 2009-present American College of Gastroenterology 2010-2015 American Society for Gastrointestinal Endoscopy 2010-present American Gastroenterological Association 2011-present Crohn’s & Colitis Foundation 2014-present American Society for Parenteral and Enteral Nutrition

Committee Service 2015-present Chair, Medical Advisory Committee, Crohn’s & Colitis Foundation (Northern California Chapter) 2015-present Abstract Review Committee, American Society for Parenteral and Enteral Nutrition 2016-present Clinical Nutrition Week Monitoring Committee, American Society for Parenteral and Enteral Nutrition 2016-present Publications Review Committee, American Society for Parenteral and Enteral Nutrition 2016-present Practice Parameters Committee, American College of Gastroenterology 2017-present Leadership Council for the Medical Practice Section, American College for Parenteral and Enteral Nutrition

CLINICAL ACTIVITIES

Certification 2010-present Advanced Cardiovascular Life Support, American Heart Association 2010-present Internal Medicine, American Board of Internal Medicine 2015-present Gastroenterology, American Board of Internal Medicine

Clinical (Service) Responsibilities 2014-present Attending physician, Stanford University School of Medicine

139

2014-present Attending physician, Veterans Affairs Palo Alto Health Care System

Clinical Program Building/Leadership 2014-present Gastrointestinal Nutrition, Stanford University School of Medicine 2015-present Fecal Microbiota Transplantation, Stanford University School of Medicine 2015-present IBD Clinic, Veterans Affairs Palo Alto Health Care System

140