Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The clinical and pharmacogenomic determinants of interferon beta induced liver injury in multiple sclerosis Kowalec, Kaarina 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_may_kowalec_kaarina.pdf [ 10.71MB ]
Metadata
JSON: 24-1.0228891.json
JSON-LD: 24-1.0228891-ld.json
RDF/XML (Pretty): 24-1.0228891-rdf.xml
RDF/JSON: 24-1.0228891-rdf.json
Turtle: 24-1.0228891-turtle.txt
N-Triples: 24-1.0228891-rdf-ntriples.txt
Original Record: 24-1.0228891-source.json
Full Text
24-1.0228891-fulltext.txt
Citation
24-1.0228891.ris

Full Text

THE CLINICAL AND PHARMACOGENOMIC DETERMINANTS OF INTERFERON BETA INDUCED LIVER INJURY IN MULTIPLE SCLEROSIS  by Kaarina Kowalec M.Sc., The University of Manitoba, 2011 B.Sc. (Hons.), The University of Manitoba, 2008      A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Neuroscience)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  April 2016  © Kaarina Kowalec, 2016 ii  Abstract   Drug-induced liver injury is a common cause of acute liver failure; it is also the leading reason for a drug’s withdrawal from the North American market. Interferon-beta (IFN-β) is a commonly used disease-modifying drug for multiple sclerosis (MS) and is generally considered safe. However, 30-60% of IFN-β exposed patients experience liver aminotransferase enzyme elevations, with an unknown proportion experiencing more severe, medically significant elevations. To date, there are no means of predicting who will experience this adverse drug reaction (ADR), although the application of pharmacogenomics could assist with identifying predictive genomic factors of the ADR. The purpose of this dissertation was to identify predictive factors associated with IFN-β induced liver injury to mitigate toxicity.  The research in this dissertation commenced with a review article summarizing the potential application of pharmacogenomics to severe ADRs in MS and a case report of a patient experiencing a hepatic autoimmune-like complication of IFN-β therapy. An original research article followed; utilizing 942 IFN-β exposed MS patients, primarily from Canada. This population-based study examined the rate of liver injury due to IFN-β in MS patients from British Columbia. Approximately 1 in 50 (or 1.9%) IFN-β exposed patients’ experienced liver injury.  A pharmacogenomic case-control study followed, involving 182 patients and a genome-wide scan of 785,230 genomic variants, to identify predictors of IFN-β induced liver injury. A genetic variant within synaptotagmin-14 was strongly associated with the ADR (odds ratio 9.83, 95% confidence interval 4.01-24.10, P-value 9.39 x 10-9) and was specific for liver injury from IFN-β iii  and has been previously correlated with hepatic expression of interferon regulatory factor 6. This represents the first genome-wide association study of an ADR from an MS drug and of drug induced liver injury due to a biological therapy.  The clinical, demographic and genomic characteristics identified here could modify the risk of experiencing a clinically significant ADR that often results in the cessation of a potentially useful treatment. Predictive characteristics of those at an increased risk will direct preventative strategies, such as enhanced monitoring for early signs of toxicity or alternative treatment regimens. This dissertation contributes towards the personalization of MS therapy and to the broader pharmacogenomic literature on biological therapies.               iv  Preface I wrote the entire dissertation with direction and input from Drs. Helen Tremlett, Bruce Carleton, Colin Ross and Anthony Traboulsee. These studies were conducted under ethical approval from the University of British Columbia Research Ethics Board (Certificate No: H10-00494 (Children’s & Women’s Hospital) & V10-0104 (Vancouver Coastal Health Research Institute). Anne Smith and Anna-Marie Bueno initiated the first ethics application at UBC. I performed the chart reviews to supplement the data in the BCMS database and all subsequent annual ethics renewal with the UBC Research Ethics Board and all other sites – in Manitoba, the University of Manitoba Health Research Ethics Board (Certificate No. H2011: 281) approved the study. Western University approved the study in London, Ontario (Certificate No. 18698E), Capital Health (now Nova Scotia Health Authority) approved the study in Halifax, Nova Scotia (Certificate No. CDHA-RS/2012-343), and the University of Montréal approved the study in Montréal (Certificate No. 12.158). Partners HealthCare approved the study at the Brigham & Women’s Hospital in Boston, MA (Certificate No. 2010P002504/BWH).  I completed all the data analysis (with guidance from thesis supervisors, Dr. Helen Tremlett and Dr. Bruce Carleton), interpretation and writing of all the published and to-be-published manuscripts in this dissertation. All co-authors involved in the different manuscripts assisted with data interpretation, manuscript revision and/or study conceptualization. Copyright permission has been obtained for all publications included in this dissertation.  CHAPTER 1: A version of this chapter is published as a peer-reviewed review article. v  Kowalec K, Carleton BC, Tremlett H. (2013) The potential role of pharmacogenomics in the prevention of serious adverse drug reactions in multiple sclerosis. Mult Scler Rel Dis, 2 (3): 183-192.  CHAPTER 2: A version of this chapter is published as a peer-reviewed case report. Kowalec K, Yoshida EM, Traboulsee A, Carleton BC, Tremlett, H. (2013) Suspected autoimmune hepatitis and primary biliary cirrhosis unmasked by interferon-beta in a multiple sclerosis patient. Mult Scler Rel Dis, 2 (1): 57-59.  CHAPTER 3: A version of this chapter is published as a peer-reviewed original research article. Kowalec K, Kingwell E, Yoshida EM, Marrie RA, Kremenchutzky M, Campbell T, Wadelius M, Carleton BC, Tremlett H. (2014) Characteristics associated with drug induced liver injury from interferon beta in multiple sclerosis patients. Expert Opin Drug Saf, 13(10): 1305-17.         vi  Table of Contents Abstract .......................................................................................................................................... ii	Preface ........................................................................................................................................... iv	Table of Contents ......................................................................................................................... vi	List of Tables ................................................................................................................................ ix	List of Figures .............................................................................................................................. xii	List of Symbols and Abbreviations .......................................................................................... xiv	Acknowledgements .................................................................................................................... xvi	Dedication ................................................................................................................................. xviii	Chapter 1: Introduction ............................................................................................................... 1	1.1	 Adverse drug reactions .................................................................................................. 1	1.2	 Pharmacogenomics ........................................................................................................ 2	1.3	 Multiple sclerosis ........................................................................................................... 4	1.4	 Drug induced liver injury ............................................................................................. 11	1.5	 Risk factors for IFN-β induced liver injury ................................................................. 18	1.6	 Pharmacogenomics of drug induced liver injury ......................................................... 20	1.7	 Genome-wide association studies ................................................................................ 23	1.8	 Hypothesis and thesis objectives ................................................................................. 30	Chapter 2: Suspected autoimmune hepatitis and primary biliary cirrhosis unmasked by interferon-beta in a multiple sclerosis patient .......................................................................... 32	2.1	 Introduction .................................................................................................................. 32	vii  2.2	 Case report ................................................................................................................... 32	2.3	 Discussion .................................................................................................................... 35	Chapter 3: Characteristics associated with drug induced liver injury from interferon beta in multiple sclerosis patients ...................................................................................................... 37	3.1	 Objectives .................................................................................................................... 37	3.2	 Introduction .................................................................................................................. 37	3.3	 Patients & methods ...................................................................................................... 38	3.4	 Results .......................................................................................................................... 45	3.5	 Discussion .................................................................................................................... 55	3.6	 Conclusions .................................................................................................................. 60	Chapter 4: Genetic determinants of interferon beta induced liver injury in multiple sclerosis ........................................................................................................................................ 61	4.1	 Objectives .................................................................................................................... 61	4.2	 Methods........................................................................................................................ 61	4.3	 Results .......................................................................................................................... 79	4.4	 Discussion .................................................................................................................... 96	4.5	 Conclusions ................................................................................................................ 110	Chapter 5: Overall Discussion ................................................................................................. 111	5.1	 Summary of findings ................................................................................................. 111	5.2	 Strengths and limitations ............................................................................................ 115	5.3	 Significance and implications of research ................................................................. 118	5.4	 Future work ................................................................................................................ 121	5.5	 Conclusions ................................................................................................................ 121	viii  References………………………...……………………………………………………………123 Appendix……………………………………………………………………………………….145 ix  List of Tables Table 1.1 Major adverse reactions associated with immunomodulatory treatments for multiple sclerosis ........................................................................................................................................... 5	Table 1.2 Clinical chemistry criteria for drug induced liver injury .............................................. 12	Figure 2.1 Longitudinal biochemical liver test results of an MS patient treated over 68 months with different formulations of interferon-beta. ............................................................................. 33	Table 3.1 Inclusion criteria to identify multiple sclerosis patients who developed or did not develop drug induced liver injury associated with interferon- treatment in British Columbia, Canada ........................................................................................................................................... 40	Table 3.2 Characteristics of multiple sclerosis patients exposed to interferon- and fulfilling study inclusion criteria in British Columbia, Canada ................................................................... 46	Table 3.3 Clinical features of the 18 cases of interferon- associated liver injury in British Columbia, Canada ......................................................................................................................... 48	Table 3.4 Univariable and multivariable Cox proportional hazards regression analyses of potential predictors of drug induced liver injury in multiple sclerosis patients exposed to interferon- ................................................................................................................................. 51	Table 3.5 Characteristics of multiple sclerosis patients reaching criteria for drug induced liver injury during interferon- exposure from the different sites – British Columbia, rest of Canada and the nationwide (USA/Sweden) adverse drug reaction surveillance networks ....................... 55	Table 4.1 Sample size estimation to detect a specified odds ratio with 80% power over predefined statistical significance thresholds ................................................................................ 63	Table 4.2 Inclusion and exclusion criteria for study cases and controls for genomic analyses .... 65	x  Table 4.3 Assessment of clinical and demographic characteristics (discovery, replication and combined patient characteristics) .................................................................................................. 81	Table 4.4 Principal components of discovery and replication stage patients. .............................. 83	Table 4.5 Case-control analysis results for the candidate gene analysis (discovery, replication and combined) ..................................................................................................................................... 87	Table 4.6 Case-control analysis results for the genome-wide analysis (discovery, replication and combined) ..................................................................................................................................... 87	Table 4.7 Fine mapping of genetic association signals in the 1q32.2 genomic region (reached P < 1 x 10-5 after adjusting for multiple sclerosis disease course in discovery cohort) ...................... 90	Table A.1The grades of severity of drug induced liver injury (DILI) ........................................ 145	Table A.2 Data collection form for patients experiencing drug induced liver injury due to interferon-beta ............................................................................................................................. 146	Table A.3 Structured Questionnaire ........................................................................................... 148	Table A.4 The Naranjo adverse drug reaction probability scale ................................................ 151	Table A.5 Roussel-Uclaf causality assessment method (RUCAM) hepatocellular injury scale 152	Table A.6 Genome-wide array quality control metrics .............................................................. 153	Table A.7 Candidate Genes, Chromosome Number, Tagging SNP information and Rationale for Inclusion ...................................................................................................................................... 154	Table A.8 Assessment of clinical and demographic characteristics associated with interferon- induced liver injury (discovery, replication and combined) ....................................................... 157	Table A.9 Clinical and biochemical variables of autoantibody positive interferon- induced liver injury cases ......................................................................................................................... 159	xi  Table A.10 Summary of genome-wide association studies of genetic susceptibility to drug induced liver injury ..................................................................................................................... 160	 xii  List of Figures Figure 3.1 Selection of relapsing-onset multiple sclerosis (MS) patients with first exposure to an interferon-beta (IFN-) between July 1995 and June 2013 from British Columbia, Canada. .... 45	Figure 3.2 Kaplan-Meier curve indicating the time from interferon- (IFN-) initiation to drug induced liver injury (DILI) in the British Columbian multiple sclerosis cohort. ......................... 47	Figure 3.3 Kaplan-Meier curve indicating the time from interferon- (IFN-) initiation to drug induced liver injury (DILI) in the British Columbian multiple sclerosis patients, by sex. ........... 52	Figure 3.4 Kaplan-Meier curve indicating the time from interferon- (IFN-) initiation to drug induced liver injury (DILI) in the British Columbian multiple sclerosis patients, by IFN- product. ......................................................................................................................................... 53	Figure 4.1 Study design for the determination of genomic and genetic contributions to interferon- induced liver injury in multiple sclerosis. ................................................................................ 62	Figure 4.2 Variation in minimum sample size estimation (prospective power calculation) to detect the specified odds ratio with 80% power and the predefined statistical significance thresholds for the candidate gene and genome wide association study (GWAS) of IFN- induced liver injury. ...................................................................................................................... 64	Figure 4.3 Study populations (Canadian and non-Canadian) included for the determination of genomic and genetic contributions to interferon- induced liver injury in multiple sclerosis in both discovery and replication stages. .......................................................................................... 71	Figure 4.4 Ancestry determined by principal component analysis (PCA) vs. self-reported ethnicity. ........................................................................................................................................ 83	Figure 4.5 Quantile-quantile plot of the Discovery Cohort. ......................................................... 84	xiii  Figure 4.6 Manhattan plot showing the observed distribution of –log10 (P-values) against SNP chromosome location (GRCH37.p13). ......................................................................................... 88	Figure 4.7 Regional association plot of chromosome 1q32.2. ...................................................... 91	Figure 4.8 Linkage disequilibrium plots of 41-kb and 257-kb regions containing rs2205986. .... 92	Figure 4.9 Receiver operating characteristics (ROC) curves of the clinical and clinical/genetic models for predicting interferon-beta induced liver injury in the combined cohort. .................... 94	Figure 4.10 Kaplan-Meier curves showing time (days) since treatment initiation to drug-induced liver injury (DILI) in interferon-beta treated multiple sclerosis (MS) patients from the combined cohort for rs2205986 carriers and non-carriers. ............................................................................ 96	Figure 4.11 Schematic depiction of the postulated mechanism of rs2205986 variant in a genetic predisposition to interferon (IFN)- induced liver injury. ........................................................ 103	         xiv  List of Symbols and Abbreviations 95%CI 95% confidence interval ADRs Adverse drug reactions AIH Autoimmune hepatitis ALP Alkaline phosphatase ALT Alanine aminotransferase AST Aspartate aminotransferase AUC Area under the curve BCMS British Columbia Multiple Sclerosis (database) BILI Bilirubin CPNDS Canadian Pharmacogenomic Network for Drug Safety CTCAE Common Terminology Criteria for Adverse Events DILI Drug induced liver injury eQTL Expression Quantitative Trait Loci EUR European (population) HWE Hardy Weinberg equilibrium IFN-β Interferon beta GWAS Genome wide association study HLA Human leukocyte antigen HR Hazard ratio IRF Interferon regulatory factor IM Intramuscular xv  LD Linkage disequilibrium MAF Minor allele frequency MS Multiple sclerosis NHGRI National Human Genome Research Institute NPV Negative predictive value OR Odds ratio PBC Primary biliary cirrhosis PCA Principal Component Analysis PCs Principal Components PD Pharmacodynamic PK Pharmacokinetic PPV  Positive Predictive Value QC Quality Control RRMS Relapsing remitting multiple sclerosis SC Subcutaneous SNP Single Nucleotide Polymorphism SPMS Secondary progressive multiple sclerosis ULN Upper limit of normal λGC Genomic Control Inflation Factor     xvi  Acknowledgements This dissertation was possible because of many contributors. I want to thank my thesis committee members, Dr. Colin Ross and Dr. Anthony Traboulsee for providing their invaluable perspectives and insights into the clinical and genomic aspects of this dissertation. My co-supervisor, Dr. Bruce Carleton has been an exceptional mentor and continually pushed me to achieve a higher scientific intellect and along the way provided some fantastic “dad” jokes. I thank my supervisor, Dr. Helen Tremlett for her limitless enthusiasm of multiple sclerosis research, her ability to look at anything related to science and not only be critical, but also see its potential. Drs. Tremlett and Carleton have provided a fantastic training environment for me to thrive and reach my scientific independency.   I want to thank all the collaborators and coauthors who made this work possible: Dr. Elaine Kingwell, Dr. Eric Yoshida, Dr. Marcelo Kremenchutzky, Dr. Ruth Ann Marrie, Trudy Campbell, Dr. Pierre Duquette and Dr. Mia Wadelius. I also wish to thank my local collaborators at PiMS and CPNDS who provided much insight and support for my studies: Dr. Elaine Kingwell, Dr. Yinshan Zhao, Feng Zhu, Bonnie Leung, Kyla McKay, Dr. Neda Razaz, Dr. Hilda de Jong, Anna-Marie Bueno, Dr. Charity Evans, Dr. Britt Drögemöller, Dr. Galen Wright, Dr. Amit Bhavsar, Dr. Folefac Aminkeng, Dr. Ursula Amstutz, Kaitlyn Shaw, Claudette Hildebrand, Gabriella Groeneweg, Shawna Abel, Jessica Stortz, Adrienne Borrie, Nasim Massah, Michelle Higginson, Fudan Miao, Rachel Bader, Tessa Bendyshe-Walton, Prasadani Gunaretnam, and Duncan Waltrip.  xvii  Beyond my research, I owe special thanks to my loved ones. To my parents, Linda and Gary, for inspiring me to always try my best and pursue my dreams. To my little Benny, thank you for always putting a smile on my face, even after long hours of thesis writing. To my husband, Max for being patient and supporting me throughout my graduate training, and most importantly, being my teammate.  This work was supported by PhD studentships from the Canadian Institutes of Health Research (Banting and Best Canada Graduate Scholarship), MS Society of Canada (Waugh Family PhD studentship), the CIHR Drug Safety and Effectiveness Cross-Disciplinary Training Program and the UBC. A pilot grant from the British Columbia Clinical Genomics Network was obtained by Drs. Helen Tremlett (Principal Investigator), Bruce Carleton, Colin Ross, Anthony Traboulsee, Michael Hayden and Mrs. Anne Smith (co-investigators).  xviii  Dedication To my family.  1 Chapter 1: Introduction 1.1 Adverse drug reactions Adverse drug reactions (ADRs) are defined as an unfavorable event following the therapeutic use of a drug [1]. ADRs differ from adverse events, as events are not necessarily attributed to the drug. ADRs are a public health issue and account for approximately 7% of hospital admissions [2, 3]. This estimate is likely conservative, owing to low reporting rates [3, 4]. These reactions can occur immediately following drug exposure or years later and are often identified after the drug is approved for the market [5]. As with many licensed drugs, it is not unusual to observe higher incidences of ADRs in the post marketing (clinical practice) setting than anticipated based on clinical trial results [6, 7]. This is likely due to the small, highly selected, homogeneous group of patients exposed during a clinical trial.  1.1.1 Types of ADRs There are two broad categories of ADRs, including type A (pharmacological) and type B (idiosyncratic) [1]. Type A reactions are more common than type B, and can be further classified based on the primary and secondary pharmacology of the drug. Type A ADRs due to the primary pharmacology of the drug are a result of an enhancement of the drugs therapeutic action. For example, with beta-blockers, bradycardia is a primary pharmacological ADR and bronchospasm is a secondary pharmacological effect.   Type B (idiosyncratic) ADRs are less common than type A but highly clinically relevant [1]. These ADRs are dose independent and not commonly predictable from the drugs pharmacology [8]. For example, carbamazepine-induced Stevens-Johnson syndrome (SJS)  2 and toxic epidermal necrolysis (TEN) are highly severe, rare skin reactions with an underlying immunological mechanism of action towards the drug. Many type B ADRs are thought to be mediated via the involvement of the immune response [1]. Type A and B ADRs can have severe and potentially fatal outcomes, with both types possibly having a genetic predisposition.  1.1.2 Genetic variation in drug response In 1957, Motulsky suggested drug reactions were a model for demonstrating the link between hereditary information and environment in disease pathogenesis [9]. Gonzalez et al (1988) characterized a genetic defect in cytochrome P450 (CYP)-2D6, predisposing individuals to an increased rate of ADRs upon exposure to debrisoquine, an antihypertensive drug [10]. However, the incomplete human genome sequence was a principal hurdle in investigating drug response and genetic variability. This was rectified in 2003, with the initial draft of the human genome sequence [11, 12], and followed shortly by a completed map of human genomic variation from the International HapMap Project [13]. These advances permitted many scientific developments in studying the genetic variability of drug response, including azathioprine induced leukopenia [14], and codeine induced central nervous system depression [15].  1.2 Pharmacogenomics Pharmacogenomics research involves the investigation of single nucleotide polymorphisms (SNPs) predictive of therapeutic response or occurrence of an ADR, thus improving drug effectiveness and safety [16]. Pharmacogenomics can be applied to assist clinicians in  3 selecting an appropriate drug or dose for a patient based on the individuals’ genetic makeup. For example, genetic variants within CYP2C9 and vitamin K epoxide reductase account for up to 27-35% of the total warfarin dose variation [17, 18]. These genetic variants, alongside clinical factors, can be incorporated into the clinical dose allocation for patients requiring warfarin.  1.2.1 Genetic variation Although the first draft of the human genome permitted many scientific advances, the extent to which each individuals sequence varies has been studied more recently. The human genome is approximately 3.2 gigabases (Gb) in length, with an estimated 19,042 protein coding genes [19]. Humans are approximately 99.6% similar in DNA sequence [20], with the remaining 1% of DNA sequence known as “variants” [19]. There are three events contributing to human genetic variation: point mutations, insertions or deletions and structural rearrangements [19]. Single base pair changes or SNPs, occur at approximately every 800 bp and are the most common form of genetic variation [19], and form the basis of genetic association studies into rare and complex disorders.  1.2.2 The use of pharmacogenomic biomarkers in preventing ADRs There is a growing recognition of severe ADRs caused by genetic differences that alter a drug’s biotransformation [16]. Unlike other factors influencing drug response, inherited determinants remain stable throughout a person’s lifetime and might be used to provide an unprecedented means of predicting and preventing serious ADRs. For example, the human leukocyte antigen (HLA)-B*57:01 allele causes hypersensitivity to abacavir in HIV-infected  4 patients, and screening for this allele is recommended on the drug monograph [21]. Similarly, biomarker information is included on the labels of >200 drugs [22]. However, uptake and application of pharmacogenomic testing has been suboptimal. There are a variety of reasons for this, such as lack of training in pharmacogenomics, poor clinical phenotyping and inadequate assessment of comorbidities and other clinical factors [16]. The availability of pharmacogenomic testing in Canada is becoming a reality, with many academic and private sector companies now offering testing prior to treatment with thiopurines (chemotherapy agent)[23] or psychiatric medications [24]. In addition, the lack of clinical practice guidelines has hindered the uptake of pharmacogenomic testing in clinics, but various groups are rectifying this [25-28].   1.3 Multiple sclerosis Multiple sclerosis (MS) is a neurodegenerative disease characterized by inflammation, demyelination and axonal destruction within the central nervous system [29]. The worldwide median incidence of MS is 2.5 per 100,000 and prevalence is 30 per 100,000 [30]. It is one of the most common causes of non-traumatic neurological disability in young adults, with an average onset age of around 30 years of age [31], with more women affected than men [32]. There are two main disease courses in MS: relapsing onset (incorporating relapsing remitting (RRMS) and the subsequent secondary progressive (SPMS) phase) and primary progressive [33]. Foreseeing the extent of disease activity in MS is difficult; the disease varies widely between individuals, ranging from so-called ‘benign’ to ‘aggressive,’ although these terms are not always well defined.   5 1.3.1 Immunomodulatory treatment for MS  First-line therapies for Canadian MS patients include interferon-beta (IFN-β) (Betaseron® Bayer Pharmaceuticals, Montville, NJ; Extavia® Novartis Pharmaceuticals Corporation, East Hanover, NJ; Rebif® EMD Serono Canada, Inc., Oakville, ON; Avonex® Biogen Idec, Inc., Cambridge, MA) and glatiramer acetate (Copaxone® Teva Canada Limited, Toronto, ON). In addition, there are second-line therapies, such as natalizumab (Tysabri® Biogen Idec, Inc., Cambridge, MA), fingolimod (Gilenya™ Novartis Pharmaceuticals Corporation, East Hanover, NJ), mitoxantrone (Novantrone® EMD Serono Canada, Inc., Oakville, ON), teriflunomide (Aubagio®, Genzyme Canada, Mississauga, ON), dimethyl fumarate (Tecfidera™ Biogen Idec, Inc., Cambridge, MA) and alemtuzumab (Lemtrada™ Genzyme Canada, Mississauga, ON). These treatments modulate the abnormal immune response exhibited in MS and are associated with a variety of ADRs (Table 1.1).  Table 1.1 Major adverse reactions associated with immunomodulatory treatments for multiple sclerosis Drug Adverse reactions  Interferon-β Flu-like symptoms, injection site reactions, asthenia, fever, leukopenia, headache, abnormal liver function [34-36] Glatiramer acetate Lipoatrophy, asthenia, injection site pain, back pain, anxiety, rash [37]  Natalizumab Progressive multifocal leukoencephalopathy, headache, fatigue, depression, abdominal discomfort [38] Mitoxantrone Decreased left ventricular ejection fraction, congestive heart failure, acute leukaemia, amenorrhea [39] Fingolimod Infections, headache, diarrhea, back pain, liver enzyme elevations, nasopharyngitis, fatigue, melanocytic nevus, bradycardia, sinus bradycardia, first-degree/ second-degree atrioventricular block [40]1                                                 1 Modified from Kowalec K.A. et al., The potential role of pharmacogenomics in the prevention of serious adverse drug reactions in multiple sclerosis, Multiple Sclerosis and Related Disorders, 2013. 2(3): p.183-92. Reprinted with permission from RightsLink.  6 IFN-β is a type I IFN believed to modulate MS disease by increasing anti-inflammatory and decreasing pro-inflammation components [41]. The use of IFN-β in MS was first investigated in the 1980s, when extracted from human fibroblasts and administered intrathecally to RRMS patients, it was noted that relapse rates were reduced [42]. Recombinant versions of IFN-β are now used, with the first version of IFN-β-1b from Escherichia coli cells [43], followed by IFN-β-1a from Chinese hamster ovary cells [44].   1.3.2 Pharmacokinetics of IFN-β There are limited pharmacokinetic data on IFN-β due to the scarce amount of drug remaining in the serum following intramuscular (IM) or subcutaneous (SC) injection [45]. From healthy controls, the mean residence time of IFN-β-1a is 1.4 h, with an estimated total clearance of 100 l/h, a distribution half time of 5 min and terminal half time of 5 h [46]. The initial volume of distribution is 12l, with a volume of distribution at steady state of 120l [46]. The two formulations of IFN-β (-1a and -1b) do not appear to differ significantly with respect to peak serum concentrations (Cmax) or area under the curve (AUC) [47]. However, there is greater systemic exposure depending on the route of administration, with greater exposure (AUC and Cmax values) after IM than SC administration for IFN-β-1a [35]. Additionally, IFN-β-1a drug concentrations and bioavailability are higher following IM injection, compared to SC injection [48]. The SC injection is therefore given three times weekly, compared to IM injection, which is given once weekly. IFN-β-1a has higher in vitro antiviral activity compared to -1b due to near-identical glycosylation and amino acid sequence to the endogenous form of IFN-β [49, 50] and thus higher dosages of IFN-β-1b are needed to achieve the same level of bioavailability to that of -1a [51].   7 1.3.3 Pharmacodynamics of IFN-β  IFN-β-1a appears to be transported to a lymphatic compartment and then absorbed into the central plasma compartment [48]. Larger doses of IFN-β-1a may saturate the primary pathway, making alternate pathways for elimination such as renal catabolism or proteolytic degradation more important [48].   IFN-β displays receptor-mediated disposition upon binding with high affinity to the type-1 IFN α/β receptor and undergoing rapid internalization into the cytoplasm [52]. Upon binding to the IFN-α/IFN-β receptor, tyrosine kinase 2 (Tyk2) and Janus kinase 1 (Jak1) are phosphorylated [45]. These kinases then phosphorylate signal transducer and activator of transcription (STAT)-1 and STAT-2, triggering the release of the STAT1-2 heterodimer from the receptor. This heterodimer then associates with interferon regulatory factor-9 (IRF9) to form the interferon-stimulated gene factor 3 (ISGF3), enters the nucleus and activates transcription of interferon-stimulated genes, driving the cell into an antiviral state. Even though IFN-α and IFN-β bind the same receptor, certain genes are more responsive to either IFN, with additional cross talk between the two IFNs. Other pathways can be targeted by IFN-β, including the cAMP-response element binding protein (CREB)/p300 transcription factor and the mitogen activated protein kinase pathways, although this has not been fully elucidated.   Upon IFN-β binding to its receptor, there are a number of molecules expressed from white blood cells, including 2,5-oligoadenylate synthetase (2,5-OAS), B2-microglobulin, IL-10, neopterin and myxovirus resistance protein (HuMx) [35], and peak 48 hours following dosing [35]. 2,5-OAS activation inhibits viral replication by destroying single stranded RNA. B2- 8 microglobulin is involved in antigen-specific activation of cytotoxic T cells and major histocompatibility class-1 related proteins [46]. Neopterin is a metabolite of pteridine biosynthesis and is only secreting upon immune activation [53]. HuMx is a homolog of the murine myxovirus protein, which protects murine cells from myxoviruses by inhibiting cytoplasmic viral replication [46].   1.3.4 Antibodies against IFN-β therapy IFN-β therapy is capable of inducing the formation of binding (BAbs) and neutralizing antibodies (NAbs), in order of decreasing risk: -1b > -1a SC > -1a IM [54]. BAbs alone are not associated with a decrease in IFN-β response markers [55] and are present in the majority of patients treated with IFN-β [56]. However, NAbs can interfere with the biological activity of the drug, and occur in approximately 5-30% of IFN-β-treated patients (varying on IFN-β type) [57]. NAb-positivity status can change over time and has been associated with both benefit (short-term) [58] and lack of benefit [55] from IFN-β therapy. NAb-positive patients have attenuated mRNA gene expression of IFN-β therapy response markers (STAT-1, MxA, TRAIL, MxB and B2-microglobulin) [59].   1.3.5 Endogenous levels of IFN-β in MS The endogenous levels of IFN-β in MS patients have been reported as being lower than healthy controls [60, 61]. In MS patients with low endogenous IFN-β activity, IFN-β therapy allows a stronger induction of the IFN-β pathways, as opposed to those with high endogenous IFN-β activity [62]. As IFN-β also displays anti-tumour and anti-viral activities, one might expect MS patients with lower endogenous IFN-β levels to be at increased risk for cancer and  9 viral infections. Yet a recent publication shows MS patients are at a decreased risk for cancer when compared to an age-, sex- and calendar year-matched general population [63].   1.3.6 Adverse drug reactions associated with the use of IFN-β  The use of IFN-β is associated with a variety of ADRs (Table 1.1) [64], including injection site reactions, flu-like symptoms, liver enzyme elevations and haematological abnormalities [34-36]. Injection site reactions can occur with the subcutaneous preparations of IFN-β (IFN-β-1b and IFN-β-1a SC), but there are a variety of techniques available to minimize their occurrence, including employing ‘auto-injectors’ and allowing the medications to warm to room temperature prior to use [64]. ‘Flu-like’ symptoms typically emerge within hours of injection, but appear to resolve within 3 months of treatment initiation (>80%)[65]. Mood disorders, such as depression, and suicide were initially reported in the randomized control trials of IFN-β-1b [66], although later observational studies have been inconclusive [67]. Blood work abnormalities are common during IFN-β use and require frequent monitoring, [64] with leukopenia occurring in approximately 5-14% of IFN-β treated patients [58, 68]. Liver aminotransferase elevations are discussed later in this chapter.  1.3.7 Pharmacogenomics of IFN-β for MS  There is considerable heterogeneity in response to IFN-β in MS, with more than 50% of patients failing to achieve an adequate response [69, 70]. This variability prompted the investigation of predictive genomic markers of IFN-β response to enhance therapeutic management [71]. A clinical response criterion typically involves prevention of disease progression and/or a reduction in relapse frequency over 1-2 years [72]. Magnetic resonance  10 imaging (MRI) metrics are often incorporated [73], but not always [74]. Initial investigations of IFN-β response involved candidate gene studies of IFN-β biological pathways, with more recent studies employing genome-wide association approaches [74, 75].   Initial studies of IFN-β response focused on candidate variants within the likely mechanistic pathway of IFN-β and have identified a number of significantly associated SNPs within LMP7 (OR: 6.37, 95%CI: 1.84-24.1, P = 0.002) [76], and MX1 (OR: 3.37, 95%CI: 1.11-11.4, P = 0.0015) [76]. Unfortunately, no genetic determinants have yet been applied for use in clinical practice [72]. Genome-wide association studies (GWAS) of IFN-β response have revealed multiple genes and variants as probable candidates [74, 75]. Of the two GWAS completed to date, two variants were identified: GPC5 (OR: 0.51, 95% CI: 0.29-0.89, P: 0.007)[74] and ADAR (OR 2.1, 95%CI: 1.1-4.0, P: 0.02) [75]. Both GWA studies pooled patient samples, which comes with the caveat of lost individual genotype data and decreased study power. By pooling samples, it was then not possible to assess population stratification, nor for imputation to take place. Different genotyping arrays were utilized in both studies, with Byun et al utilizing a 100,000 variant panel and Comabella et al using a 500,000 variant panel (covering 31% and 65% of the Caucasian genome, respectively). The responder groups were defined similarly between studies, however non-responders were characterized differently between the two studies. Comabella et al defined non-responders as ≥ 1 relapse and increase of 1 point in EDSS over two visits (6 months apart) [75], whereas Byun et al defined non-response as ≥ 2 relapses and increase of 1 point in EDSS over three visits [74]. There was heterogeneity between the two GWAS of IFN-β response and additional studies utilizing larger cohorts and greater genomic coverage are necessary.  11 Additional studies have used microarrays to investigate gene expression differences between responders and non-responders to IFN-β therapy. One study showed a rapid induction of MxA, MxB, guanylate binding protein-1 (GBP-1) and GBP-2 and slower induction of IFNAR1, B2-microglobulin, and STAT1, following IFN-β administration [77]. A 2009 cDNA macro-array study investigated the expression of 186 interferon-stimulated genes in IFN-β treated patients and demonstrated inter- and intra-individual differences [78]. The heterogeneity within-subjects was stable over time, which lends evidence to a personalized response within each patient.  Pharmacogenomics research of other MS therapies suggests immune-related genes are associated with response to glatiramer acetate or natalizumab [79, 80] and adenosine-triphosphate-binding cassette-transporters with mitoxantrone in MS patients [81].   1.4 Drug induced liver injury  Drug induced liver injury (DILI), or drug hepatotoxicity, is defined as liver injury associated with therapeutic drug use, occurring as either a hepatocellular, cholestatic or mixed pattern of injury [82]. DILI can manifest as a idiosyncratic drug reaction, independent of dosage, with the most common offending agents including antimicrobials, immunomodulatory agents, analgesics and nonsteroidal anti-inflammatory drugs (NSAIDs) [82]. However, DILI can also present as a dose-dependent ADR, the most well known example being that of liver injury due to acetaminophen [83]. DILI is the most common cause of acute liver failure in the USA [84], and is also the most common reason for drug withdrawal from the market [85]. Estimates of  12 DILI range from 2.4/100,000 person-years from a retrospective UK study [86], to 13.9/100,000 of French inhabitants from a population-based study [87].   1.4.1 Diagnosis of DILI To facilitate the comparison of DILI cases across studies, a standardized phenotype was published in 2011 [88]. The diagnosis of DILI begins with meeting one of three criteria listed in Table 1.2 [88], characterized by elevations in alanine aminotransferase (ALT), bilirubin or alkaline phosphatase (ALP) above the upper limit of normal (ULN). Minor elevations (i.e. ALT <5x ULN or ALP <2x ULN) in these biochemical enzymes are common and transient and do not represent clinically important liver injury. Raising the threshold of elevations for ALT to 5x the ULN increases the likelihood of including only clinically important and non-self limiting drug events [88].   Table 1.2 Clinical chemistry criteria for drug induced liver injury Modified from Aithal G.P. et al., Case Definition and Phenotype Standardization in Drug Induced Liver Injury, Clinical Pharmacology and Therapeutics, 2011. 89(6): p.806-815. Reprinted with permission from RightsLink.  Raising the DILI threshold to ALT 5x ULN in patients with atrial fibrillation limited the occurrence to 1.4% of patients, versus a threshold of 2x ULN, which resulted in an estimated Any one of the following: • More than or equal to fivefold elevation above the upper limit of normal (ULN) for alanine aminotransferase (ALT) (or AST, if ALT unavailable) • More than or equal to twofold elevation above the ULN for alkaline phosphatase (ALP) (particularly with accompanying elevations in concentrations of 5′-nucleotidase or γ-glutamyl transpeptidase in the absence of known bone pathology driving the rise in ALP level)  • More than or equal to threefold elevation in ALT concentration and simultaneous elevation of bilirubin concentration exceeding 2× ULN  13 6-8% affected [89]. In addition, patients treated with anti-tuberculosis therapies are recommended to remain on drug in the event ALT between 2x ULN and 5x ULN occurs, with no symptoms, implying that ALT elevations in this range are deemed clinically acceptable [90]. The upper threshold of 5x ULN for ALT to indicate DILI has been agreed upon by the international DILI Expert Working Group [88].   The pattern of liver injury, also known as the R-value is determined utilizing the relative ALT and ALP levels [91]. An R-value ≥ 5 indicates a hepatocellular pattern of liver injury and may or may not present with clinical symptoms [92]. Hepatocellular liver injury (R-value ≥ 5), in conjunction with bilirubin >2x ULN, is associated with a mortality rate of 10% and is used by the Food and Drug Administration (FDA) as a predictor of severe toxicity in clinical trials (“Hy’s Law”) [93]. Cholestatic liver injury (R-value ≤ 2) is defined as damage to the bile ducts and has reported mortality rates up to 7.8% [94]. In cases of mixed liver injury (R-value >2 and <5), patients typically present with a mixture of hepatitis and cholestasis, with a lower degree of severity than either hepatocellular or cholestatic. Overall, hepatocellular injury is more commonly associated with worse outcomes than either cholestasis or mixed injury, partly owing to an association with increased international normalized ratio (INR) [92]. Liver biopsy is not essential for a DILI diagnosis but can provide additional evidence to strengthen the association. The DILI Expert Working Group included four degrees of liver injury severity, ranging from mild to fatal or requiring transplantation [88] (Table A.1). The various levels differ on the degree of bilirubin elevation; presence of symptoms and other clinical criteria indicating increased severity, such as elevated INR and encephalopathy.   14 A diagnosis of DILI is pursued alongside an exclusion of other possible causes of liver enzyme elevations, including but not limited to, viral hepatitis, alcoholism, biliary disorders and metabolic disorders [92]. Various scales for assessing causality in DILI exist, including the Roussel Uclaf Causality Assessment Method (RUCAM) [91], the Naranjo scale [95] and the Maria & Victorino scale [96]. Studies have favoured the RUCAM over the Naranjo scale [97, 98] and the Maria & Victorino scale [99] due to increased specificity, discriminative power and similar assessment to a specialist. The Council of International Organizations of Medical Sciences endorses the RUCAM for assessing DILI causality [100]. Unlike the Naranjo scale, the RUCAM incorporates data elements to exclude other causes of liver enzyme elevations, such as viral hepatitis and excessive alcohol use. However, the RUCAM is problematic when minimal data is available or when being applied retrospectively [88, 101], nonetheless, it is the most commonly used causality tool for DILI [88].  1.4.2 Epidemiology of biochemical liver enzyme elevations associated with IFN-β for MS  The risk for elevations in the biochemical liver enzymes during IFN-β exposure in MS was poorly characterized during the pivotal clinical trials [102]. One clinical trial reported a “significant increase” of ALT in 5.4% of IFN-β-1a 44 mcg subcutaneous (SC) exposed patients [103], whereas another trial reported “mild to moderate” ALT elevations in 11% of IFN-β-1b exposed patients [66]. Finally, there was no evidence reported of liver enzyme elevations induced by IFN-β-1a IM [104]. These double blind, placebo controlled clinical trials for IFN-β utilized vague terminology to describe liver enzyme elevations, such as “mild to moderate” or “increased”, without indicating which quantitative results were used in these  15 categories. As such, comparisons between these trials was difficult and estimating the true risk was not possible.   In response to these criticisms, one of the pharmaceutical manufacturers of IFN-β-1a SC applied the National Cancer Institute Common Toxicity Criteria (CTC, version 2.0)[105] to grade liver transaminase elevations within a dataset of multiple IFN-β-1a IM and SC clinical trials [106]. Grade 3 ALT elevations (“medically significant event”: >5 to ≤10x ULN) were reported in 2.8% of IFN-β-1a SC (44 mcg) and 0.3% of IFN-β-1a IM [106]. Utilizing the same grading scale, a post-marketing investigation reported similar estimates of grade 3 ALT elevations for IFN-β-1a SC 44 mcg (2.5%) and IFN-β-1a IM (0%) in clinical practice [107]. In addition, 3.1% of those exposed to IFN-β-1b experienced grade 3 ALT elevations.   Overall, the risk estimate of IFN-β induced liver injury following these studies is unclear in the post-market setting. Two concepts exist which suggest the rate of IFN-β induced liver injury would be lower during clinical trials. The application of an undefined phenotype and vague terminology in clinical trials would prohibit the evaluation of a “severe” phenotype between trials and post-market studies [103], and estimates of DILI from IFN-β in the clinical trial and post-marketing phases could not consider the standardized case definition of DILI published in 2011 [88]. Secondly, in general, clinical trials are conducted in highly standardized conditions and exclude certain patients and/or situations (e.g. without comorbid conditions or pre-existing liver enzyme elevations) to improve the efficiency of statistical analysis to show clinical efficacy of the study drug [108]. This contrasts with the post-market  16 scenario, where wider groups (i.e. different ages, different comorbidities) of patients are treated and the drug reaction rate can differ [108].   On the contrary, there are two pieces of evidence that suggests an increased likelihood of liver injury events during clinical trials. During clinical trials, liver testing is performed more frequently than post-market [109], with an increased likelihood of detecting a mild elevation after four tests (28.3%), compared to two tests (13-15.3%) in placebo-treated MS patients [110]. However, these are mild elevations, not necessarily clinically relevant, severe medical events. Lastly, clinical trials require extensive safety reporting to gain market approval, whereas post-marketing surveillance of ADRs is on a voluntary basis. Greater than 95% of adverse drug reactions are unreported to Health Canada and this is likely due to the voluntary nature of reporting [4].   1.4.3 Clinical outcomes of IFN-β associated liver enzyme (aminotransferase) elevations The clinical implications of liver enzyme elevations associated with therapeutic drug use are not fully understood. Of the few liver biopsy results published in cases of MS patients experiencing IFN-β induced liver enzyme elevations, varying degrees of severity and autoimmune features typical of an immunological mechanism of action have been found [111, 112], although only an estimated 4% of MS patients developed de novo auto-antibodies during IFN-β-1b therapy [113]. However, testing for auto-antibodies such as anti-smooth muscle or anti-mitochrondrial antibodies is generally infrequent at the time of the ADR. Of the three reported liver transplants following DILI from IFN-β, one later reported use of nefazodone, which raised suspicion on the role of IFN-β in being the sole causative agent [114, 115].  17 When MS patients present with aminotransferase elevations during IFN-β exposure, a clinical decision is made dependant upon the level of ALT elevations reached [116], with Health Canada recommending 5x ULN warrants either dose reduction or discontinuation [117]. One study retrospectively examined outcomes within a clinical setting, where 77% of those reaching grade 1 toxicity (aminotransferases <2.5x ULN) had returned to normal levels with continued treatment, whereas only 64% normalized after reaching grade 2 with continued exposure. The actions of physicians differs when presented with a patient experiencing grades 1-3 aminotransferase elevations, with the majority of those reaching grade 1 permitted to continue treatment, whereas the opposite was true for those reaching grade 3 [116]. A higher proportion of grade 2 elevations did not return to normal levels with continued treatment, compared to grade 1. Dose reductions are uncommon and were reported in only 2% of IFN-β exposed patients with elevated liver enzymes.   1.4.4 Mechanism of action of IFN-β induced liver injury The mechanism(s) of action of IFN-β induced liver injury remains to be elucidated, although hepatotoxicity in traditional non-biological therapies have been associated with variation in drug metabolism and immune mediated mechanisms [118]. Biological therapies are often immunogenic [119], including IFN-β [51], and as MS patients exhibit an aberrant immune response, this could hypothetically increase susceptibility to experiencing liver injury from IFN-β. Either IFN-β or a metabolite might act as a hapten by binding to a liver protein, stimulating an immune response, leading to hepatocyte death [120]. This alone may be insufficient to induce an immune reaction, yet the presence of additional risk factor(s), such as  18 an inflammatory disease like MS coupled with other clinical risk factors may tip the balance, resulting in liver injury.   1.5 Risk factors for IFN-β induced liver injury Several risk factors have been reported for liver enzyme elevations during exposure to IFN-β in MS. These included IFN-β product (which differ by route, dose and frequency of administration), time on treatment, body mass index (BMI), sex, age, MS disease course and concomitant drug use.  1.5.1 IFN-β product type and length of treatment  More severe ALT elevations (>5x ULN) are noted with higher dose and increased injection frequency of IFN-β in MS [6]. The significance of injection frequency was also noted in hepatitis C patients treated with IFN-β, as more ALT elevations were noted upon increasing the injection frequency but maintaining the same dose [121]. It was postulated that decreasing the time between injections might prevent the regeneration of damaged hepatocytes between injections and lead to further damage [102]. MS patients have the greatest risk of aminotransferase elevations within the first six months of IFN-β treatment [6], although this often coincides with the most frequent monitoring of the liver enzymes. Approximately 30% of MS patients exposed to IFN-β-1b (SC) experienced grade 1 liver aminotransferase elevations between 0-6 months, with 15-20% at 6-12 months and similar percentages after 12-24 months [6]. Elevations can also occur years later [112]. Prolonged latency from drug exposure to liver enzyme elevations is also characteristic of idiosyncratic DILI, in general [92].  19 1.5.2 Sex  Men were more likely to have elevated ALT prior to starting IFN-β, as well as during treatment [6]. However, this could be due to men in general being at an increased risk of low grade elevations in ALT [109], whereas women were more susceptible to more severe liver enzyme elevations, especially when autoimmune related [122]. Similarly, others have reported 90% of IFN-β exposed MS patients with ‘severe hepatic dysfunction’ were women [106], compared to 64-76% of women in an IFN-β treated MS population [123, 124].  1.5.3 Age  Younger age is a risk factor for de novo ALT elevations in IFN-β exposed MS patients,[106, 107] with similar findings in hepatocellular DILI from therapies used to treat other conditions [125]. However, serum ALT levels decrease with age in healthy untreated individuals [126], which may lower the probability that older adults will reach the threshold for severe ALT elevations of >5x ULN. It is not currently known whether there is a biologic basis for increased risk of hepatocellular DILI in younger adults [127].  1.5.4 Other clinical risk factors One study showed individuals with an RRMS disease course to be at increased risk for liver aminotransferase abnormalities and could be due to increased immunological activity during this disease course [110]. This is also seen in those with active autoimmune rheumatoid diseases [128], and may be due to increased inflammation and cytokine activity. Higher BMI is associated with increased ALT and AST levels in the general public [109], as well as in placebo-treated MS patients [110]. Francis et al found more MS patients exposed to IFN-β  20 (62%) with elevated ALT were ≥ 66 kg than < 66 kg [106]. Fatty liver diseases are known to impair mitochondrial functioning, with DILI believed to confound this [129] and this may point to the mechanism of increased weight modifying the risk of DILI.  The use of concomitant medications, such as acetaminophen, alongside IFN-β could theoretically increase risk of DILI, in particular those medications that are also associated with liver injury [85]. Interestingly, acetaminophen use alongside IFN-β in MS was reported in one study as being associated with a decreased risk of ALT elevations [106]. In contrary, the concurrent use of acetaminophen in oncology patients treated with IFN-α, is associated with hepatotoxicity [130, 131]. In placebo-treated MS patients, over 90% of the cohort reported use of antibiotics or analgesics (such as acetaminophen, NSAIDs or acetylsalicylic acid), which were unexpectedly associated with a decreased risk of ALT and AST elevations [110]. The use of propionic acid derivatives (naproxen, ibuprofen) increased the risk of liver dysfunction in IFN-β exposed MS patients, but an increased risk was also seen in placebo-treated patients [106].   1.6 Pharmacogenomics of drug induced liver injury Recent pharmacogenomic studies of DILI have identified a number of immune-related genes to be of importance [132-136]. Utilizing various genotyping approaches, including GWAS, HLA typing or candidate gene arrays, strong genetic associations with liver injury induced by ximelagatran, nevirapine, amoxicillin-clavulanate, lapatinib, lumiracoxib or flucloxacillin have been identified and replicated. Flucloxacillin induced liver injury was one of the first GWA studies performed in investigating the genetic determinants of DILI [132]. Daly et al  21 (2009) identified the HLA-B*57:01 allele to be associated with the ADR in 51 cases and 282 population controls (OR = 45, 95%CI: 19.4-105, P = 8.7 x 10-33) and replicated in 23 cases and 64 drug treated controls [132]. An association was identified and replicated in 2011 between HLA-DQA1*02:01 and lapatinib induced liver injury (OR = 9.0, 95%CI: 3.2-27.4, P = 0.03) [136]. A prospective evaluation utilizing the associated HLA allele to predict DILI from lapatinib was published in 2014 and demonstrated a high negative predictive value (NPV), but a low positive predictive value (PPV) [137]. These results may impede the widespread application of this test in clinical practice since patients who test positive for the HLA allele may be excluded from lapatinib therapy but might not actually develop DILI. Nevertheless, a genetic test may assist in enhancing safety monitoring and early identification of DILI cases from lapatinib. In 2015, the most comprehensive evaluation of DILI associated genetic determinants was published with the first whole-genome sequencing study [138]. However, other than the HLA allele associated during the initial GWAS in 2011, no additional genetic markers were detected. Functional evidence has also linked lapatinib, amoxicillin-clavulanate and flucloxacillin with their respective associated HLA genes in inducing liver injury [139-141].    1.6.1 Functional validation of genetic variants associated with DILI A strong genetic association was discovered in two HLA alleles with the presence of elevated liver aminotransferases (≥ 3x ULN) in patients exposed to ximelagatran (a direct thrombin inhibitor) during a GWAS [133]. The immunological mechanism of this ADR was further observed upon the generation of a positive lymphocyte transformation test, which is an in vitro validation method indicating the presence of drug-specific immune cells. This was  22 further confirmed utilizing a competitive binding assay between ximelagatran and the associated HLA alleles. Together, these results suggested that ximelagatran acts as a contact sensitizer and is capable of stimulating an adaptive immune response. In another study, an immunological mechanism of flucloxacillin induced liver injury was initially identified during a GWAS between the ADR and HLA-B*57:01[142] and further confirmed using flucloxacillin-responsive T-cells in patients with liver injury [139]. Additionally, naïve T-cells were activated in the presence of flucloxacillin from volunteers expressing HLA-B*57:01 and upon flucloxacillin activation, a number of pro-inflammatory cytokines were secreted from activated T-cells. A 2012 study on rifampicin hepatotoxicity demonstrated SLCO1B1 genetic variants modified the bile acid transport in human embroynic kidney (HEK)-293 cells in the presence of rifampicin [143].  The investigation of in vivo characteristics of DILI has been poor owing to the lack of proper models, with few studies available. In vivo animal model systems complement in vitro systems to characterize metabolic pathways and the contribution of specific genetic variants within the context of the whole body with multiple transporters, metabolising enzymes, protein binding, and blood flow. One study found significantly elevated plasma ALT, AST and total bilirubin levels in flucloxacillin-administered female BALB/cCrSlc mice [144]. These authors also found a significant increase in the hepatic mRNA levels of inflammatory cytokines (e.g. IL-17), toll-like receptor 4 ligands upon flucloxacillin administration. Additionally, the co-administration of recombinant IL-17 and flucloxacillin resulted in an exacerbation of DILI, as measured using histological staining of the mouse liver [144]. Another in vivo study demonstrated mitochondrial and oxidative stress in male BALB/cCrSLc  23 mice experiencing amiodarone induced liver injury using plasma triglyceride levels and hepatic mRNA levels of inflammatory cytokines [145].  Recently, a zebrafish model was proposed to be a novel high-throughput in vivo method for studying DILI, with notable similarities to the mammalian hepatic system, including similar drug metabolism and histological patterns of DILI [146]. However, the zebrafish model requires wider validation in DILI prior to establishing its use for translation of findings to humans.  1.7 Genome-wide association studies  Pharmacogenomic studies initially utilized a hypothesis-driven, candidate gene approach to investigate the genetic determinants of ADRs. Since 2005, this approach was modified due to the increasing availability of data on human genomic variability. Genome-wide association studies (GWAS) are responsible for genotyping much of the common variability (>5%) in the human genome, using arrays typically covering 500,000 to 2,000,000+ SNPs. GWA studies are beneficial in their hypothesis-free nature since genotyping is performed in all genes and not only those of theoretical importance. The National Human Genome Research Institute  (NHGRI) and European Bioinformatics Institute have recorded 1,751 published GWA studies between 2007 and 2014, with associations between 11,912 SNPs and complex traits [147].  1.7.1 Statistical considerations in GWAS  GWA studies typically involve either a binary categorical outcome (case versus control) or a quantitative phenotype [148]. Quantitative measures can be more precise than designating  24 individuals as either “cases” or “controls”, however a quantitative outcome is not necessary for a successful GWAS. Sample sizes will vary greatly and depends on a number of metrics, including desired P-value significance threshold, the estimated effect size and the baseline rate of disease. Investigations of common complex genetic diseases such as type II diabetes or breast cancer, typically 1,000 cases and 1,000 controls are necessary for adequate study power [149].   Statistical analyses during a GWAS involve testing each SNP for an association to the desired phenotype [148]. Binary phenotype is analyzed using logistic regression to measure the hypothesis of no association between case/control status and genotypes. Genotyping data can be encoded in five ways: additive, dominant, recessive, genotypic and allelic. The additive model assumes a linear risk increase for each additive risk allele. This model has been used previously to detect significant genetic differences associated with cases of DILI [134, 142], and is the most common genetic association test in which the underlying genetic model is unknown [150]. The dominant model assumes increased risk when >1 copy of the dominant allele is present [148]. Recessive modeling assumes two copies of the recessive allele are required for risk modification. Allelic tests examine the SNP and phenotype association, whereas genotypic tests examine the association between genotype (or genotype classes) and phenotype.  1.7.2 Linkage disequilibrium (LD) Imputation of data is often necessary if different genotyping platforms have been used within the same study. Imputation is the process by which known linkage disequilibrium (LD)  25 patterns are exploited to determine SNP genotypes not directly typed during the study [148]. LD is the association between SNPs at single loci to variants at nearby loci. LD results from SNPs closer to one another on the same chromosome that display a high probability of being inherited together. LD is highly dependent upon ancestry, with African populations known to exhibit small degrees of LD relative to Asian or European populations [151]. GWA genotyping panels are specifically developed using tagging SNPs in high LD with other variants to capture majority of the genetic variation. For European populations, 500,000 to 1 million SNPs capture >80% of the commonly occurring SNPs [152]. These SNPs are often within intergenic regions and thus indicate likely regions where the causal variant may exist.  1.7.3 Multiple testing  For each statistical test conducted during a GWAS, in some cases up to 5 million, a P-value is generated. If the null hypothesis is true, the P-value is the probability of observing a test statistic equal or greater than the observed statistic. The threshold for significance is usually set at 0.05 (alpha value), meaning 5% of the time; the null hypothesis is rejected when it is true (indicating a false positive association) during one test of association. The cumulative possibility of detecting a false positive during a GWAS is extremely high, given the number of independent tests performed [148]. Thus, correction should be applied to adjust for multiple testing [153]. This Bonferroni correction involves adjusting the alpha value from 0.05 to 0.05/k, where k is the total number of tests completed. This form of correction would modify the P-value threshold for a GWAS involving 500,000 SNPs to P < 1 x 10-7. However, this conservative correction assumes all tests are independent of one another, which is generally false, given the amount of LD amongst the genetic variants being tested.   26  There are three other means of correcting for multiple testing: false discovery rate, permutation testing and applying a standard level of genome-wide significance [148]. The false discovery rate estimates and corrects for the number of significant false positive result, thus estimating the number of true significant results. Permutation testing can be computationally difficult for GWA studies. This form of correction reassigns phenotypes to random individuals within the dataset, assuming the null hypothesis is true. The reassignments are repeated a defined number of times to produce a P-value within 1/1000th of a decimal. Lastly, many genomic variants are not independent of one another (due to LD), therefore correction can take place using a predetermined “genome-wide significance threshold”. The threshold correction should incorporate only the number of independent tests, with the current threshold for GWAS at 5 x 10-8 [154], with also the threshold set by the NHGRI at 1 x 10-5 also possible [147].  1.7.4 Adjusting for ancestral, clinical and demographic patient characteristics There are a number of clinical and demographic factors, including sex, age and clinical factors influencing an individual’s risk of disease or specified outcome. Adjusting for covariates during GWA studies is integral to reduce unauthentic associations when investigating genetic risk factors. This is highly relevant to pharmacogenomic studies, in particular, since concomitant medications and other drug-specific factors may be different between cases and controls. On the contrary, clinical factor adjustment can reduce study power.    27 Importantly, adjustment of genetic data for population stratification is necessary during GWA studies. Certain SNPs are ancestry-specific, therefore in a multi-ethnic study; a SNP significantly different between cases and controls may be due to the underlying ancestral differences. There are two means for adjusting GWA data for population substructure: genomic inflation factor [155] and principal components analysis (PCA). The genomic inflation factor adjusts the test statistic of each marker to an overall, uniform inflation factor, where a value < 1.05 indicates minimal population stratification [155]. However, some SNPs are highly variable in frequency and genomic inflation factor is inadequate at its regulation.  PCA is the most popular and powerful method for population stratification correction in GWAS [156]. Principal components are linearly uncorrelated variables generated from the GWA genotype data to reduce the systemic ethnic variation. The first two principal components can be plotted to visualize population substructures and assist the removal of outliers. Adjusting for population stratification can occur either via logistic regression, where principal components are included as covariates or adjust the exact genotype during the genetic association tests [156].  1.7.5 Replication Once a genetic variant has been discovered during a GWAS, investigations should incorporate clinical replication in a separate, comparable cohort to validate and increase the generalizability of any genetic variants found in terms of clinical utility [157]. Replication assists to minimize false positive results and is often necessary for scientific publication. Non-replicating associations from GWAS are an issue [158] and emphasize the importance of  28 replicating the original GWAS study design and population ascertainment [159]. A successful replication study requires minimal study heterogeneity and adequate statistical power to detect the true effect size [160].  1.7.6 Imputation and fine mapping Fine mapping assists in identifying the true causal variant within a GWAS-associated genetic region. Using known LD patterns from online databases, such as the 1000 Genomes Project, SNPs are imputed that are not directly typed by the original GWAS panel [161]. This serves to unveil the causal variant for further investigation either by functional validation or subsequent sequencing. For genotypes with a minor allele frequency greater than 5%, inference accuracy is estimated at 90-95% [162]. Targeted sequencing of phenotype-associated genetic regions is necessary for variants with a minor allele frequency of less than 1% [163], since GWAS typically offer poor coverage of rare variants. However, genome-wide arrays are constantly being improved and expanded to include rare variants.   1.7.7 Functional validation Upon establishing a sufficient associated between a genetic variant and phenotype, functional validation is a multi-step and systematic process encompassing first an in vitro assessment. Using in vitro functional assays, the role of specific genes and variants are characterized using tissue-specific cell lines. These functional studies are conducted in cell culture and then extended to animal models, if available and applicable. Cell culture studies offer the advantage of reasonable throughput and high reproducibility. Specific cell lines, for which gene overexpression and silencing protocols have been optimized to assess cellular  29 phenotypes such as cytotoxicity, in the presence of drug can be employed. Phenotypes are compared between cells expressing wild type and polymorphic gene variants. Where appropriate, characterization will be extended to tissue-specific cell culture lines, e.g. HepG2 cells for liver injury due to drug therapies.   1.7.8 GWAS of ADRs  The relative rarity or challenges in identifying and collating serious ADRs pose a roadblock for any single research group to investigate their genetic associations; thus national and international collaborations are often needed to gain a sufficient number of well characterized cases. Successful networks and collaborations in Canada [164], United States [165], Europe [166] and Asia [167] all share common goals of either preventing ADRs and/or enhancing treatment effectiveness for a variety of therapies. For example, the Canadian Pharmacogenomics Network for Drug Safety (CPNDS) actively employs clinicians (pharmacists, nurses and physicians) at teaching hospitals across Canada to identify and characterize patients with suspected ADRs [164]. The CPNDS has been successful in identifying genetic variants associated with codeine-induced infant mortality[168], anthracycline-induced cardiotoxicity [169] and cisplatin-induced deafness [170, 171]. To the best of our knowledge, no network investigating MS-related serious ADRs exists, although networks currently focused on the pharmacogenomics of treatment response have been created, and could provide opportunity as well as the framework to expand into the field of ADRs [172]. Alternatively, linking existing ADR-focused networks (such as the CPNDS) with MS clinic networks provides other practical opportunity. For example, in North America, the Canadian Network of MS clinics (http://www.cnmsc.org) and the Consortium of MS  30 Centers (http://www.mscare.org) represent examples of well-established MS clinic networks, both of which are focused on the advancement of clinical care, patient education and research. MS clinic-based database networks exist in Europe [173], internationally (MSBase) [174], as well as in specific patient cohorts, e.g. paediatric MS [175, 176]. Specific drug-focused registries have also been developed, typically funded by the relevant pharmaceutical manufacturer [177].  Of upmost importance in predicting and preventing serious ADRs to therapies using genetics is access to extremely well characterized ADRs [164] and accurately phenotyped patients [178]. While very specific and detailed phenotyping guidelines exist for ADRs such as DILI [88] or drug-induced skin injury [179], many serious ADRs are not well defined. To accurately characterize a serious ADR, a thorough examination of the patients’ clinical history is necessary [180] and might be facilitated with the standardization of clinical data in electronic medical records. Databases such as EDMUS [173] or MSBase [174] are used by over 60 countries around the world and could be powerful resources to study and identify patients with serious ADRs to MS therapies.   1.8 Hypothesis and thesis objectives  This study hypothesizes demographic, clinical and genetic factors exist which modify the risk of experiencing IFN-β induced liver injury in MS. The overall aim of this study was to discover clinical and genetic markers of clinical significance that predict the occurrence of IFN-β induced liver injury in MS.   31 Specific objectives of this study were: 1. To identify demographic and clinical factors associated with IFN-β induced liver injury in a cohort of people with MS.  2. To perform a hypothesis-driven, candidate gene study to identify if genetic variants previously associated with drug-induced liver injury or IFN-β response are also relevant in modifying the risk of IFN-β induced liver injury in a cohort of people with MS.  3. To perform a hypothesis-free, genome-wide association study to identify novel genetic variants associated with IFN-β induced liver injury in a cohort of people with MS.          32 Chapter 2: Suspected autoimmune hepatitis and primary biliary cirrhosis unmasked by interferon-beta in a multiple sclerosis patient  2.1 Introduction The use of interferon-beta in multiple sclerosis is associated with various forms of hepatotoxicity, including autoimmune hepatitis and liver failure. We describe a case with features of autoimmune liver disease and primary biliary cirrhosis occurring during long-term treatment with interferon-beta in a patient with relapsing-remitting multiple sclerosis. This case highlights the importance of monitoring biochemical liver test results throughout interferon-beta treatment of multiple sclerosis.    2.2 Case report A 42-year-old woman with clinically definite RRMS (Poser criteria)[181] and symptom onset 14 years previously, was initially treated with her first MS disease-modifying therapy, IFN-β-1a (44 mcg subcutaneous injection, thrice weekly; Rebif® EMD Serono Canada, Inc., Oakville, ON) for 5 months, but discontinued due to injection site reactions. This was followed by 28 months of IFN-β-1b (250 mcg subcutaneous injection, alternate days; Betaseron® Bayer Pharmaceuticals, Montville, NJ, USA) but discontinued due to perceived lack of effectiveness. The biochemical liver test results were within normal range pre- IFN-β treatment (‘baseline’) and during the first 33 months of IFN-β-1a (44 mcg) and IFN-β-1b treatment (Figure 2.1).      33  Figure 2.1 Longitudinal biochemical liver test results of an MS patient treated over 68 months with different formulations of interferon-beta.  Key: ALT, alanine aminotransferase; AST, aspartate aminotransferase; ALP, alkaline phosphatase; TBILI, total bilirubin. TBILI levels were within normal range throughout.  The patient was restarted on IFN-β-1a (44 mcg) due to increasing MS disease activity. No biochemical liver testing was done immediately prior to switching to IFN-β-1a; the last liver test results prior to the switch had been normal (12 months prior), as were the subsequent results five months after re-starting. Approximately 34 months into this second course of IFN-β-1a (44 mcg), the patients’ biochemical liver test results were: alanine aminotransferase (ALT) = 260 (Normal (N) < 36 U/l; 7 times the upper normal limit); aspartate aminotransferase (AST) = 221 (N < 36 U/l; 6 times the upper normal limit); alkaline phosphatase (ALP) = 157 (N < 125 U/l; 1.3 times the upper normal limit); total bilirubin (Tbili) = 8 (N < 25 µmol/l). This would translate to an R-value of 5.4, meaning a hepatocellular pattern of liver injury. No clinical symptoms of acute liver injury were reported, but she was advised to discontinue IFN-β-1a. Three weeks later, her liver  34 biochemistry was repeated and remained similarly elevated. The liver biochemistry normalized within two months, following drug cessation.    Serologic autoimmune marker testing was completed two months following the highest liver test elevations and was strongly positive for anti-mitochondrial antibodies (titer = 1:1280). In addition, she was positive at a 1:1280 dilution with a centromere pattern for antinuclear antibodies (ANA), but was negative for parietal cell and anti-tissue transglutaminase antibodies. No baseline serologic autoimmune marker testing was done, as this was not considered standard of care in our MS clinic and was not clinically indicated at baseline.  Concomitant medications around the time of the elevated biochemical liver test results included occasional acetaminophen (unknown quantity) and modafinil (alertec® 100 mg daily; Shire Canada Inc., Saint-Laurent, QC), for MS-related fatigue (unlicensed indication). The patient had been on modafinil for four years with no other adverse reactions reported. She was negative for hepatitis viruses A, B and C and denied any substance abuse, with the exception of tobacco cigarettes (unknown quantity per day). She reported occasional alcohol intake when she would consume 1-2 beers (3-4 units of alcohol) on a given day.   A liver biopsy was not considered justifiable, in part due to normalization of liver biochemistry, although an abdominal ultrasound performed three months following drug discontinuation demonstrated mild fatty infiltration of the liver, with no calculi, obstruction or focal hepatic abnormalities. Her medical history was negative for acute liver injury and no family history of liver disease was found. To date, the patient has not been re-challenged with  35 IFN-β-1a. Two doses of ursodiol (Urso® 500 mg BID; Aptalis Pharma Canada Inc., Mont-Saint-Hilaire, QC) for primary biliary cirrhosis (PBC) were given, but ursodiol was not continued because of stabilizing liver biochemistry. The patients liver biochemistry remains normal, as of five months after stopping IFN-β-1a, except for a mild increase in serum gamma-glutamyltransferase (GGT) at 80 U/L (N < 55 U/L).  2.3 Discussion The use of IFN-β in MS is associated with autoimmune hepatitis [182-184], hepatotoxicity [6] and fulminant liver failure [114, 115] and can occur years into treatment [183]. To the best of our knowledge, the use of IFN-β for MS has not been associated with unmasking primary biliary cirrhosis. We describe a case with features of autoimmune liver disease occurring during treatment with IFN-β for MS, presumably exposing pre-existing, but clinically silent PBC. The strong presence of antimitochondrial antibodies are suggestive of PBC [185, 186], while the antinuclear antibody along with ALT and AST elevations were consistent with autoimmune hepatitis [187]. In the absence of a liver biopsy, which could not be justified on clinical grounds, or any remarkable elevations in ALP, the accepted criteria of PBC-autoimmune hepatitis overlap syndrome are not fulfilled [188], although the suggestion is of PBC with features of autoimmune hepatitis [189].   Similar to the development of MS, PBC and autoimmune hepatitis are thought to begin once a genetically susceptible individual encounters an unknown environmental trigger [187, 190]; for this patient, the trigger may have been long-term exposure to IFN-β. Although no HLA typing was performed on this patient, therefore conclusions on the patients’ genetic  36 predisposition to PBC cannot be made [191]. The exact mechanism behind the induction of autoimmunity by IFN-β in MS patients is unknown [182, 184]. PBC may have remained subclinical, although fatigue was present in this patient, which, in hindsight could have also been related to PBC [190]. The long-term use of IFN-β may have aggravated the presumed pre-existing PBC, triggering the elevations in ALT and AST, appearing as autoimmune hepatitis.   Other potential causes of the patients’ elevated liver enzymes include consumption of acetaminophen and alcohol. However, the AST/ALT ratio is typically much higher in cases of alcoholic liver damage [192] and neither the history of alcohol nor acetaminophen consumption appeared significant enough to account for the degree of liver biochemical derangement or the presence of autoantibodies. In conclusion, this case suggests IFN-β-related liver toxicity may occur years into treatment (apparently without any overt worrisome signs or symptoms) and further exhibits the importance of monitoring biochemical liver test results. The recommended frequency of testing varies, with some recommending baseline screening, followed by monthly tests for the first 6 months, then 6 monthly thereafter during IFN-β treatment [117]. 37 Chapter 3: Characteristics associated with drug induced liver injury from interferon beta in multiple sclerosis patients  3.1 Objectives The purpose of these analyses was to discover clinical factors predictive of IFN-β induced liver injury and determine its rate in the province of British Columbia, Canada.  3.2 Introduction A neurodegenerative disease of the central nervous system, multiple sclerosis (MS) is characterized by inflammation, demyelination and axonal injury. The interferon betas (IFN-β) represent the first disease-modifying drugs ever to be licensed for MS and they remain the most widely used world-wide [30]. While they are generally considered safer than the newer biological agents approved for MS, the IFN-βs are nonetheless associated with a variety of adverse drug reactions (ADR), requiring regular blood work, including biochemical liver enzyme tests.   Post-marketing studies or re-analysis of clinical trial data suggest that 30-60% of MS patients exposed to an IFN-β will experience elevations in their biochemical liver enzyme tests (typically alanine aminotransferase (ALT) and/or aspartate aminotransferase (AST)) [6, 106, 107]. While the long-term implication of liver injury as indicated by any liver enzyme elevations is unclear, around 1-2% of IFN-β exposed MS patients [6, 107] experience severe elevations. Case reports of fulminant liver failure requiring a transplant have been published in MS patients exposed to IFN-β [112, 114, 193], highlighting the potential severity of this reaction. In addition, several of  38 the new therapies for MS have also been implicated in liver injury, including natalizumab, fingolimod, teriflunomide and alemtuzumab [40, 194-197]. Drug induced liver injury (DILI) is the most common cause of acute liver failure in the United States [84], and is also the most common reason for drug withdrawal from the market [85]. A better understanding of serious ADRs such as DILI is extremely important, not just because of the associated morbidity and high socio-economic costs borne by the patients and society [3], but also once a serious ADR occurs, that drug is no longer a therapeutic option for the affected patient, limiting future choice.  DILI is a challenging condition, and attempts to predict and prevent this condition have fallen short in other clinical situations [198]; it also remains poorly understood in MS. Previous studies in MS have focused on describing the more common low grade or transient elevations in ALT and/or AST [6, 106, 107]. Few studies have attempted to clinically characterize DILI or explore factors that might identify those at greatest risk of developing DILI in patients exposed to IFN-β [112]. We applied the recommended DILI laboratory criteria [88] to people with MS treated with an IFN-β and examined possible clinical factors associated with this severe ADR.   3.3 Patients & methods 3.3.1 Study population and setting This was a retrospective observational study, using a mixed methodological approach. We primarily conducted a cohort study, which included IFN-β exposed MS patients who developed DILI as well as those who did not in British Columbia, Canada. As a secondary aspect of the study, we identified and characterized additional cases of IFN-β associated DILI from other MS  39 clinics in Canada, as well as from national ADR surveillance networks from the USA and Sweden and compared them to the BC cases.  3.3.2 Cohort study in British Columbia, Canada The source population for the cohort study consisted of all MS patients in the British Columbia (BC) MS database who were prescribed their first IFN-β between July 1995 (the first IFN-β licensing date in Canada) and June 2013. The BC MS database [199, 200] contains demographic and clinical data collected prospectively since 1980 on patients visiting the four core MS clinics in BC. Patient chart reviews were conducted in 2001 (all charts from all four clinics),[6, 116] 2009[107] and again in June 2013 (the latter two reviews included all charts at the main University of British Columbia clinic only) to generate study-specific data (ALT, AST, alkaline phosphatase [ALP], total bilirubin [TBILI], symptoms of liver injury, outcome/management of liver injury) that were not routinely collected in the BC MS Database [6, 107, 116]. The chart reviews conducted in 2001 reviewed the period of July 1995-June 2001, whereby 846 patients were eligible, 844 charts reviewed and 835 patients had at least 2 liver enzymes results recorded with these patients published previously [116]. For the chart review in 2009, the period reviewed included July 1995-December 2005, with 1064 charts reviewed, 802 patients had baseline and post-treatment liver enzyme results (262 were excluded because of absence of either baseline or post-treatment ALT) and were published previously [107]. For the final chart review in 2013, the period reviewed included January 2006-February 2011, with 487 charts reviewed.  All patients with definite MS and a relapsing-onset disease course with a first prescription of any IFN-β, who had a normal baseline biochemical liver test (including an ALT measurement) within  40 180 days prior to starting IFN-β and completed at least one biochemical liver test within 13 months after initiation of IFN-β therapy were selected (Table 3.1). Patients with primary-progressive MS were not included because no IFN-β has been licensed for use in this disease course. Table 3.1 Inclusion criteria to identify multiple sclerosis patients who developed or did not develop drug induced liver injury associated with interferon-β treatment in British Columbia, Canada Inclusion criteria • Registered in the BC MS database, and hence assessed by a neurologist • Definite MS (Poser or McDonald criteria) • Relapsing-remitting or secondary-progressive disease course • First prescription of an IFN-β (IFN-β-1b subcutaneous, SC; IFN-β-1a SC (high and low dose) or IFN-β-1a intramuscular, IM) for MS between July 1995-June 2013 • Normal baseline liver test result* (within the normal range for the reporting laboratory) <180 days prior to IFN-β start • Minimum of one liver test within 13 months after starting IFN-β*  *Consists of at least an alanine aminotransferase (ALT) measurement. Each individual liver test result had to include the corresponding ‘normal range,’ as reported by the relevant laboratory. IFN-β: interferon-beta  3.3.3 Data collection For all patients in the cohort study, the data was comprised of: biochemical liver test results (including ALT and when available, AST, ALP and TBILI with the corresponding normal ranges), demographics (age, sex), MS clinical data (MS disease course, MS symptom onset date, and start and stop dates of the IFN-β product used). In addition, for patients meeting the DILI criteria (as defined in section 2.4), additional data collected included: symptoms of liver injury (for determining the severity of liver injury[88]) and the outcome/management of liver injury. No restriction was placed on which laboratory the patients had chosen to complete their routine biochemical liver testing (i.e. local community lab, hospital lab etc.). The schedule for liver testing in Canadian patients, as recommended by Health Canada, was modified over the study  41 period (1995-2013), with an increase in the frequency of tests from 2003 [117], although the actual frequency of testing was at the discretion of the treating physician.   3.3.4 Study event Patients had to fulfill recognized criteria for DILI [88], defined as one of: (1) ALT or AST ≥ 5x upper limit of normal (ULN), or (2) ALP >2x ULN or (3) simultaneous ALT ≥ 3x ULN and bilirubin >2x ULN. Competing etiologies associated with liver enzyme elevations (i.e. viral hepatitis, etc.) were excluded, when possible, for patients fulfilling the study event.  3.3.5 Follow-up All patients were followed from the start of their first IFN-β to the earliest of: (1) a documented liver test result that fulfilled criteria for DILI, or (2) the last liver test result prior to end of drug exposure or the study end (June 2013), or (3) the last liver test result that fell within the ‘regular’ liver testing requirement. ‘Regular’ testing was broadly defined as evidence of at least one yearly liver test (with no gap greater than 13 months between tests). The end of drug exposure occurred if a patient switched from their original IFN-β, or between low and high-dose IFN-β-1a SC, or stopped IFN-β treatment (i.e. only the first IFN-β treatment period was considered).  3.3.6 Statistical analyses The following characteristics were described: sex, age at IFN-β initiation, age at MS onset, MS disease duration at IFN-β initiation, MS disease course at IFN-β initiation (relapsing-remitting vs. secondary-progressive) and IFN-β product. The time to DILI and patient characteristics that were possibly associated with DILI were explored using Kaplan Meier curves and Cox  42 proportional hazard regression models. MS disease course was not included in the models because the majority of patients had a relapsing-remitting disease course at IFN-β initiation. A multivariable Cox regression model was developed, including sex, age at IFN-β initiation and IFN-β product (chosen a priori as independent variables based on the previous DILI literature which has identified both sex and age as factors influencing DILI [125]). Other potential covariates (MS disease duration at IFN-β initiation and age at MS onset) were added to the multivariable models if statistical significance was reached in univariate model (P < 0.05).   3.3.7 Sensitivity analysis We assessed the sensitivity of our estimate by lowering the threshold of a “DILI” event with the reasoning that potential DILI cases may have reached a lower threshold and the drug was stopped due to an elevated liver test result before the DILI threshold was reached. For the alternative study event definition, the threshold was lowered to ALT or AST ≥ 3x ULN. A Cox regression analyses was performed and findings were compared to the results from the main analysis.  3.3.8 Identifying additional cases of IFN-β associated DILI Ethical approval was sought from the centre/university. Upon receiving ethical approval, a data transfer agreement must be approved between the receiving institution (UBC) and the sending institution. Two separate groups of DILI cases were identified from other Canadian MS clinics and from other nationwide ADR surveillance networks. These cases were assessed to identify similarities or differences to the BC cases of DILI.   43 Other Canadian sites: For the Winnipeg MS clinic location, the period reviewed for charts was January 1998 (opening of MS clinic) - May 2012, with 82 charts reviewed between January-August 2012, resulting in 12 cases. The clinic database was queried for all patients on IFN-β from 1998 (opening of MS clinic) to date of initial ethics approval. This search was performed again in May 2012 using electronic medical records for patients who came off IFN-β due to elevated liver test results and then confirmed for eligibility by chart review. At the London MS clinic, 82 charts were reviewed in June 2012 covering the period of July 1995-June 2012, with 1 case identified for contacting. Additional cases were identified between June 2012-December 2014 by MS neurologists. The clinic database was queried for all patients on IFN-β and then identified only patients who had been seen in the clinic since 2010. Cases were preliminarily identified using ‘adverse event: elevated liver test’, then confirmed using chart reviews. Finally, for the Halifax MS clinic, 1000 charts were reviewed in October 2012, spanning July 1995-October 2012, with 12 cases identified for contacting. An MS clinic data analyst queried the Dalhousie MS Clinic Database for patients experiencing elevated liver test results during IFN-β treatment. The charts of these patients were then reviewed to confirm eligibility.  Other nationwide ADR surveillance networks (USA and Sweden): Two ADR networks [112, 201] contributed cases: the United States-based Drug-Induced Liver Injury Network (DILIN) and the Sweden-based SWEDEGENE. These networks are described elsewhere [112, 201], and the clinical characteristics of the DILIN IFN-β cases have been reported [112]. The DILIN is an ongoing prospective study, which enrolls adults and children with suspected DILI. It enrolled 8 subjects between September 2004 and July 2012 with MS who developed well characterized DILI due to IFN-β [112]. Inclusion criteria matched the BC and other Canadian sites, except that  44 two consecutive elevations of ALT, AST or ALP were required [112]. For SWEDEGENE, which is a national biobank for ADRs, adult MS patients were identified retrospectively from the Swedish Drug Information System (SWEDIS) ADR reports (time period: 1990 to 2013) [201]. The case definitions were: (i) ALT or AST >5× ULN or (ii) ALP >2× ULN, or (iii) ALT >3× ULN and total bilirubin >2× ULN. In addition, both the DILIN and SWEDEGENE employed causality assessments as part of their enrollment process (the Roussel Uclaf Causality Assessment Method (RUCAM) [100] and Naranjo ADR probability scale [95], respectively).   Descriptive comparisons between the cases derived from BC and the other Canadian MS clinics, and between BC and the nationwide ADR networks were performed focusing on the following characteristics: sex, age at IFN-β initiation, time to DILI onset from IFN-β initiation, and IFN-β product (available for all cases); and age at MS onset, MS disease duration at IFN-β initiation, and disease course at IFN-β initiation (available for the BC and other Canadian cases only). Characteristics were compared using the Mann-Whitney U test for continuous variables and the Pearson’s chi-squared test or Fisher exact test for categorical variables.   Statistical significance was defined as P < 0.05, and all analyses were performed using SPSS 22.0 (SPSS Inc., Chicago IL, USA). The study was approved by each of the appropriate institutional ethics review boards.    45 3.4 Results 3.4.1 The British Columbia MS cohort In the BC cohort, a total of 1883 patients with relapsing-remitting or secondary-progressive MS were exposed to their first IFN-β between July 1995 and June 2013, and 942 patients met the full inclusion criteria, as shown in Figure 3.1. There were no significant differences between those included (n = 942) and those excluded (n = 941) based on sex (P = 0.59), age at IFN-β start (P = 0.91), age at MS onset (P = 0.62), IFN-β product (P = 0.35), MS disease course (P = 0.85) or MS disease duration (P = 0.73).   Figure 3.1 Selection of relapsing-onset multiple sclerosis (MS) patients with first exposure to an interferon-beta (IFN-β) between July 1995 and June 2013 from British Columbia, Canada.   Definite relapsing-remitting or secondary-progressive MS patients exposed to IFN-β between July 1995 and June 2013 (n=1883) Baseline liver test result available (n=1314) Normal baseline occurring <180 days prior to starting  IFN-β (n=1218) ≥ 1 liver test result available (including corresponding range of normal) within 13 months of starting IFN-β  (n=942)  46 The baseline characteristics of the 942 patients are shown in Table 3.2. Overall, 72.6% were women. The mean age at the start of IFN-β was 39.9 years and most patients were in the relapsing remitting phase of disease.   Table 3.2 Characteristics of multiple sclerosis patients exposed to interferon-β and fulfilling study inclusion criteria in British Columbia, Canada Characteristic N=942* Sex, n (%) Female Male Unknown  684 (72.6) 220 (23.4) 38 (4.0) Age at MS onset, mean y (SD) 31.2 (9.10) Age at IFN-β start, mean y (SD) 39.9 (9.68) Disease course at IFN-β start, n (%) Relapsing remitting  Secondary progressive Unknown  889 (94.4) 2 (0.2) 51 (5.4) MS disease duration at IFN-β start, mean y (SD) 8.6 (8.12) IFN-β product, n (%)  1a IM (30 mcg weekly)  1a SC (22 mcg 3x weekly)  1a SC (44 mcg 3x weekly)  1b (250 mcg every other day)  149 (15.8) 162 (17.2) 238 (25.3) 393 (41.7) *Date of birth, and therefore age at IFN-β start was unknown for 38 patients; age at MS onset, and therefore MS disease duration, was unknown for 86 patients. MS: multiple sclerosis; IFN-β: interferon-beta; y, years; SD, standard deviation.   DILI and IFN-β exposure: Among the 942 IFN-β exposed patients, 18 (1.9%) developed DILI. All cases developed DILI within two years (692 days) of starting IFN-β (see Figure 3.2), and all were identified by the DILI criterion of ALT or AST ≥ 5x ULN. The clinical features of these 18 DILI cases are summarized in Table 3.3. All cases for which an R-value could be calculated (11/18) displayed the hepatocellular liver injury pattern (R ≥ 5). Three patients displayed symptoms of DILI, two of which had severity scores rated as ‘moderate.’ Most of the cases stopped their current IFN-β drug treatment in response to the elevated liver enzyme results, although two cases remained on therapy and four switched to a lower dosed IFN-β. Normalization of aminotransferases was observed for the majority of the cases at the last follow- 47 up (Table 3.3). The remaining ‘non-DILI’ patients were censored at the last documented regular liver test result prior to the end of either: follow up (n=229) or drug exposure (n=695).   Time (Days) 0 50 100 150 200 350 400 700 N of cumulative events (cases)  0 4 8 11 12 14 16 18 N of remaining patients  942 862 802 751 678 496 445 300 Figure 3.2 Kaplan-Meier curve indicating the time from interferon-β (IFN-β) initiation to drug induced liver injury (DILI) in the British Columbian multiple sclerosis cohort. Of the 942 patients, 18 reached the outcome (DILI). The remaining ‘non-DILI’ patients were censored at the last documented regular liver test result prior to the end of either: follow up (n=229) or drug exposure (n=695).   48 Table 3.3 Clinical features of the 18 cases of interferon-β associated liver injury in British Columbia, Canada Pati-ent# Agea (y) Sex Product Latencyb  Initial ALT  (x  ULN) Initial ALK (x ULN) Initial BILI (x ULN) R valuec Length of follow upd Symp. Severity scoree Outcome/ Management 1 47 M 1a 22 mcg 17 5.5 ND ND ND 3665 none  1 Remained on drug; ALT =1.45x ULN at last follow up (3665 days) 2 43 M 1b 250 mcg 91 7.4 ND ND ND 445 Abdom-inal pain/ numbness 1 Stopped drug; ALT=1.6x ULN at last follow up (445 days)  3 51 F 1b 250 mcg 36 8.3 0.8 0.3 10.4 591 none  1 Stopped drug; ALT normalized within 6 months 4 36 F 1a 44 mcg 54 12 ND ND ND 57 none  1 Stopped drug; ALT normalized within 30 days 5 24 F 1a 44 mcg 19 6.8 ND ND ND 15 none  1 Stopped drug; ALT normalized within 2 weeks  6 19 F 1b 250 mcg 25 5.1 0.5 0.5 10.2 1206 none  1 Reduced dose (to unknown amount); ALT normalized within 11 months 7 42 F 1a 44 mcg 64 5.5 0.4 0.1 13.8 141 none  1 Reduced dose to -1a SC 22 mcg; ALT normalized within 3 months 8 35 F 1a 44 mcg 90 6.1 0.7 0.4 8.7 54 none  1 Reduced dose to -1a SC 22 mcg; ALT=1.06x ULN at last follow up (54 days) 9 28 F 1b 250 mcg 102 6.3 0.7 0.4 9.0 146 none  1 Stopped drug; ALT normalized within 3 months  49 Pati-ent# Agea (y) Sex Product Latencyb  Initial ALT  (x  ULN) Initial ALK (x ULN) Initial BILI (x ULN) R valuec Length of follow upd Symp. Severity scoree Outcome/ Management 10 29 F 1a 44 mcg 126 11 0.9 0.5 12.2 299 Vomiting/nausea 2 Stopped drug; AST normalized within 9 months 11 50 F 1b 250 mcg 169 5.8 0.9 0.1 6.4 349 none  1 Stopped drug; ALT normalized within one year; abdominal ultrasound showed slightly coarse texture and increase echogenicity of liver, but no focal lesion; other structures normal. 12 49 F 1a 44 mcg 343 5.7 0.5 0.4 11.4 341 ND ND ND 13 47 F 1a 44 mcg 679 6.5 0.8 0.5 8.1 477 none  1 Switched drug to 1a IM; ALT normalized within 15 months 14 35 F 1b 250 mcg 289 8.2 ND ND ND 872 Pruritus 2 Stopped drug; ALT normalized within 4 months 15 36 F 1b 250 mcg 362 12f ND 1.1 ND 657 ND ND ND 16 20 F 1a 44 mcg 109 6.2 0.7 0.4 8.9 654 none  1 Remained on drug; ALT normalized within 1 year  17 40 F 1b 250 mcg 692 5.2f ND ND ND 2408 none  1 Stopped drug; AST normalized within 2 weeks 18 45 F 1b 250 mcg 379 7 0.5 0.7 14.0 2037 none  1 ALT normalizedg Key: ALT: alanine aminotransferase; ULN: upper limit of normal; ALK: alkaline phosphatase; ND: data unavailable; Symp.=symptoms; ‘none’ indicates no symptomology recorded by the treating physician which was documented as being associated with the liver injury / elevated liver test results. All cases met criteria for DILI1; 2 remained on the original IFN-β at study end, and 4 switched to a lower dose. a: Within 3 months of DILI, b: Time from drug start to DILI (in days), c: R is the ratio of serum activity of alanine  50 aminotransferase to serum activity of alkaline phosphatase [(ALT/ULN)/(ALP/ULN)]. R≥5 indicates a hepatocellular pattern of liver injury, R>2 to <5 indicates a mixed liver injury pattern and R≤2 indicates a cholestatic pattern.17, d: Time from DILI to most recent ALT (or AST) measurement (in days), e: Severity Score17: 1 = mild (meets DILI criteria but TBILI<2x ULN), 2=moderate (meets DILI criteria and (a) TBILI≥2x ULN or (b) symptomatic hepatitis). No patient was scored 3 or higher [3=severe (meets DILI criteria and TBILI >2x ULN and at least one of the following: (a) international normalized ratio ≥1.5 or (b) ascites and/or encephalopathy, disease duration <26 weeks and absence of underlying cirrhosis or (c) other organ failure due to DILI), 4=fatal or transplantation], f: Aspartate aminotransferase level (ALT unavailable), g: Time to normalization unavailable 51 Patient characteristics associated with IFN-β DILI: From the survival analyses, while none of the characteristics examined were significantly associated with an increased hazard of reaching DILI (see Table 3.4, and Figures 3.3 and 3.4, depicting sex and IFN-β product, respectively), some trends were observed. There was a trend towards an increased risk for women compared to men (adjusted hazard ratio [adjHR]: 3.15; 95% CI: 0.72-13.72, P = 0.13) and for those exposed to IFN-β-1a SC (44 mcg) (adjHR: 6.26, 95% CI=0.78-50.39, P = 0.08) in comparison to IFN-β-1a SC (22 mcg). None of the 149 included patients who were exposed to IFN-β-1a IM developed DILI.  Table 3.4 Univariable and multivariable Cox proportional hazards regression analyses of potential predictors of drug induced liver injury in multiple sclerosis patients exposed to interferon-β  No. of patientsa Univariable Multivariable Variable  HR (95% CI) p-value HR (95% CI) p-value Female sex  904 2.26 (0.51-9.96) 0.28 3.15 (0.72-13.72) 0.13 Age at MS onsetb   856 1.00 (0.95-1.05) 0.99   Age at IFN-β startc  904 0.95 (0.91-1.01) 0.11 0.97 (0.92-1.02) 0.23 MS disease duration at IFN-β starta  856 0.92 (0.85-1.00) 0.06   IFN-β productc  1a SC (22 mcg 3x weekly) 1a SC (44 mcg 3x weekly) 1b (250 mcg every other day) 793  Ref. 4.55 (0.57-36.4) 2.94 (0.37-23.2)   0.15 0.31  Ref. 6.26 (0.78-50.39) 3.71 (0.47-29.59)   0.08 0.22 a: Number of patients included, per variable for univariable analysis. For the multivariable model, n=793. b: Hazard ratio expressed as per 1-year increase. c: None of the patients who were exposed to IFN-β-1a IM (30 mcg) reached the event (developed DILI); these patients were not included in the models that included IFN-β product. MS: multiple sclerosis; IFN-β: interferon-beta; HR: hazard ratio, 95% CI: 95% confidence interval.         52  Figure 3.3 Kaplan-Meier curve indicating the time from interferon-β (IFN-β) initiation to drug induced liver injury (DILI) in the British Columbian multiple sclerosis patients, by sex.    53  Figure 3.4 Kaplan-Meier curve indicating the time from interferon-β (IFN-β) initiation to drug induced liver injury (DILI) in the British Columbian multiple sclerosis patients, by IFN-β product.   Sensitivity analysis: There were 27/942 (2.9%) patients who reached the alternative outcome of ALT or AST ≥ 3x ULN during IFN-β exposure. Findings from the Cox regression analysis (uni- and multi-variable) were in the same direction to those seen in the main analyses, and as before, no patient characteristic was significantly associated with the outcome (data not shown).   54 3.4.2 Additional cases of IFN-β associated DILI: A total of 24 additional DILI cases were identified from the other sites; 14 from three other MS clinics in Canada, and 10 from the ADR surveillance networks (8 from the USA, which have previously been described as a case series [112], and 2 from Sweden reported in 2009 and 2010). While additional cases from the ADR surveillance networks were reported, only 10 cases were enrolled in the respective networks.  There was a high preponderance of females across all sites (see Table 3.5). The average age at the start of IFN-β was similar between BC and the other Canadian sites, but the BC cases were younger (by around 10 years) compared to those identified through the ADR surveillance networks in the USA and Sweden. The BC and other Canadian sites were also comparable for the age at MS onset. The median time to DILI onset was comparable between BC and the other Canadian sites (105 and 90 days respectively), but it was significantly longer for cases from the nationwide ADR networks compared to those from BC (P = 0.006). The longest time period from initiation of IFN-β to onset of DILI was 4900 days (around 13 years), as identified through one of the ADR surveillance networks. Of note, this patient experienced normalization of liver aminotransferases 4 months after stopping IFN-β. The longest time period to onset of DILI among the cases from Canada was 2639 days (approximately 7 years); this patient experienced a positive rechallenge and subsequently stopped IFN-β therapy. At the last follow up (3 months after meeting DILI criteria) ALT levels had not yet normalized.      55 Table 3.5 Characteristics of multiple sclerosis patients reaching criteria for drug induced liver injury during interferon-β exposure from the different sites – British Columbia, rest of Canada and the nationwide (USA/Sweden) adverse drug reaction surveillance networks Characteristic BC (n=18) [1] Rest of Canada (n=14) [2] p-value [1] vs. [2]  USA & Sweden  (n=10) [3] p-value [1] vs. [3]  Female sex, n (%)a 16 (88.9) 14 (100) 0.20 9 (90.0) 0.93 Age at MS onset, mean y (SD)b 31.4 (9.14) 31.5 (7.53) 0.78 NA  Age at IFN-β start, mean y (SD)b 36.8 (10.0) 39.0 (11.9) 0.55 46.6 (9.81) 0.03 Disease course at IFN-β start, n (%)a Relapsing remitting  Secondary progressive  16 (88.9) 2 (11.1)  13 (92.9) 1 (7.1)  0.70   NA   Disease duration at IFN-β start, mean y (SD)b  5.17 (7.80)  8.64 (7.85)  0.14  NA   Days to DILI onset, median (range)b 105.5 (17-692) 90.5 (31-2639) 0.84 590.5 (101-4900) 0.006    IFN-β product, n (%)c 1a IM (30 mcg weekly) 1a SC (22 mcg 3x weekly) 1a SC (44 mcg 3x weekly) 1b (250 mcg every other day)  0  1 (5.6) 8 (44.4) 9 (50.0)  1 (7.1) 2 (14.3) 6 (42.9) 5 (35.7)  0.58  3 (30) 0 3 (30) 4 (40)  0.11 a: Comparison using Pearson’s chi-squared test. b: Comparison using the Mann-Whitney U test. c: Comparison using the Fisher’s exact test. MS: multiple sclerosis; DILI: drug induced liver injury; IFN-β: interferon-beta; BC, British Columbia; SD: standard deviation; NA: data not available.    3.5 Discussion We applied standardized DILI laboratory parameters and inclusion criteria to people with MS who were treated with an IFN-β. In British Columbia, Canada, we found approximately one in 50 MS patients treated with an IFN-β reached criteria for DILI. In comparison, a previous Icelandic population-based study of the incidence of DILI (defined as AST > 3x ULN or ALT > 2x ULN) reported rates that were considerably lower; the highest observed incidence rate was 1 in 133 treated patients (for azathioprine) [202], although this estimate was based on only 4 cases and the 95% confidence interval included 1 in 50. The difference may also be due to differences in the case ascertainment methods and a potential underestimation of cases in the Icelandic study, as acknowledged by the authors [202]. We were unable to find another study using the same explicit criteria in a population of MS patients. One study using data from clinical trials  56 found a higher proportion of cases when investigating only those treated with IFN-β-1a SC 44 mcg (1 in 35, or 2.8%) [106], however, this study was not able to employ the more recently published DILI criterion [88].    Our current study applies internationally recommended DILI criteria [88] to a population of MS patients; it uses a more stringent cut-off to define ‘liver injury’ than has been used in previous studies of DILI in MS and includes a minimum frequency of liver testing. These methods could provide a standardized approach to identifying and characterizing DILI in future studies of MS.   Similar case definitions to the one used here have been employed by other research groups to investigate predictive factors in clinical and pharmacogenomic studies in non-MS, clinical populations [134, 142]. Utilization of a recommended case definition and accurate phenotyping of patients is considered of primary importance in reliably determining whether clinical risk factors are associated with the ADR [178]. All cases identified here met the case definition [88], including the six patients that either continued treatment or switched to a lower dose; likely representing a balance between risk versus possible benefit from IFN-β. This case definition is known to represent patients experiencing a severe or medically significant and potentially disabling event, as per the Common Terminology Criteria for Adverse Events (v4.0) [203].  The employment of stricter criteria to define DILI comes at a cost, as it results in a relatively small group of cases within which to explore potential clinical characteristics that might be useful as predictors of DILI in IFN-β exposed individuals. Although no single characteristic reached statistical significance in our analyses, we observed some interesting trends. Women had  57 an approximately three-fold greater risk than men (albeit not significant). Using a lower threshold to define liver injury (i.e. any elevation in ALT or AST), others have reported the reverse; that men with MS are at a greater risk than women with MS [109, 110]. These seemingly contradictory findings actually concur with the broader literature indicating that men in general are at an increased risk of low grade elevations in their liver test results [109], whilst women are more susceptible to the more severe reaction of DILI, especially when it is autoimmune-related [122, 204]. This higher than expected preponderance of women observed among the BC cases was also evident in the DILI cases from the other MS clinics in Canada, and the national ADR surveillance networks (USA and Sweden). Similarly, others have reported that 90% of IFN-β exposed MS patients with symptomatic and ‘severe hepatic dysfunction’ are women [106], as compared to the expected 64-76% of women in an IFN-β-treated MS population [123, 124].   There was also a trend (non-significant) towards a greater risk of DILI for the higher-dose IFN-β product. This observation concurs with other studies [106] (including that of a smaller sample of the BC MS cohort) [107], even when using a lower threshold to identify liver injury (i.e. any elevations in ALT or AST) [6, 106, 107].   Younger age has previously been reported as a risk factor for de novo ALT elevations in IFN-β exposed MS patients [106, 107]. Although the association was not statistically significant, the direction of our hazard ratio for age concurred with these previous observations [106, 107]. Younger age has similarly been reported to be associated with hepatocellular DILI from other therapies used to treat other conditions [125]. However, serum ALT levels decrease with age in  58 healthy untreated individuals [126], which may lower the probability that older adults will reach the threshold for DILI as currently defined.  DILI cases from BC, other Canadian and international sites were not only ascertained from different populations, but they were collected during different time periods, utilizing different methods. However, this approach enabled us to identify a broad range of DILI patients, and the age at MS onset, age at IFN-β start and sex distribution were all similar between the cases from BC and from the rest of Canada. Even so, an understanding of the main differences between the sites is important to contextualize findings. For instance, findings might be affected by local prescribing practices [205] or even the licensing (hence availability) or the local cost of specific drugs. These factors may alter the number of DILI cases attributed to one IFN-β product versus another. DILI in the BC and other Canadian MS patients developed fairly early during the course of IFN-β treatment (medians of 90.5 and 105.5 days, respectively) and differed from the ADR surveillance network cases (median of 590.5 days). This might relate to the types of patients clinicians refer to a surveillance network. For instance, the rates of DILI in clinical practice may be up to 10-fold higher than that reported to any centralized data registry or to regulatory bodies [86, 87]; indicating only select, and perhaps more unusual cases are reported to surveillance networks. A relatively short time to event was also observed in clinical trial data (trial participants are typically monitored more frequently than patients in clinical practice), where 75% of ALT or AST elevations occurred within the first 6 months among those exposed to IFN-β-1a SC (44 mcg) [106]. A wide range in time to onset of idiosyncratic hepatotoxicity is not uncommon among other drugs [120].    59 In this patient cohort, cases typically displayed a hepatocellular liver injury pattern, as evidenced by the R-value ≥ 5, which is associated with worse prognosis in the presence of jaundice, than the other two patterns [92]. The three patterns of liver injury associated from drugs (hepatocellular, cholestatic and mixed) are associated with different prognoses [94]. However, only 60% of BC cases had alkaline phosphatase tests, so the other patterns of liver injury were less likely to be detected.  This study employed an internationally recommended and rigorous definition of DILI [88], facilitating a standardized approach and potential replication of findings [206]. We also purposely selected individuals from the BC cohort with a previous ‘normal’ liver test result (in the six months prior to IFN-β initiation) who maintained a regular frequency of testing. While this resulted in the exclusion of a number of the treated BC patients, this minimum data requirement was a strength of the study, as it reduced the likelihood of misclassification of cases, and minimized the inclusion of individuals with a prior history of liver comorbidity. Reassuringly, we found that, excluded patients did not differ demographically from those that were included, however we do not know if they were more or less at risk of DILI than the included patients due to differences in unmeasured factors. We were unable to consider certain potential confounders, such as body mass index. DILI is more frequent in obesity-related non-alcoholic fatty liver disease (NAFLD) [129] and an increase in ALT/AST could be caused or amplified by NAFLD. However, ensuring that all individuals had an ALT/AST within the normal range pre-treatment would have helped to mitigate this by potentially screening out people with NAFLD at baseline.    60 It is known that the application of retrospective causality assessments such as the RUCAM scale [88] is problematic, if not impossible, unless exhaustive clinical data is collected at the time of the acute injury [101]. This is largely because the confidence in the diagnosis increases as other potential causes are excluded (such as viral hepatitis, biliary obstruction, hypotension etc.) [88]. The Canadian MS patients who met the DILI case definition with biochemical liver enzyme parameters were not subject to expert panel causality assessments [207] (in contrast to the international cases) which might have contributed to qualitative differences between the DILI cases among the IFN-β exposed Canadian patients and those identified by the surveillance networks.   3.6 Conclusions In summary, we found the risk of DILI in the IFN-β exposed BC MS cohort to be approximately 1 in 50. We did not identify any significant predictors of DILI associated with IFN-β exposure, although there was a trend towards a greater risk for women and those exposed to IFN-β-1a SC (44 mcg). The ‘at risk’ window appears wide; while the greatest risk appeared in the first few years after treatment initiation, the reaction can also occur over a decade after initiation of IFN-β therapy. It is recommended that MS patients undergo regular monitoring of ALT and/or AST during exposure to IFN-β [117]. However, regular liver testing by itself is not expected to prevent every case of acute liver failure [114, 206]; the identification of additional risk factors, such as genetic determinants, that are predictive of this severe reaction would be of high clinical utility.    61 Chapter 4: Genetic determinants of interferon beta induced liver injury in multiple sclerosis  4.1 Objectives The purpose of these analyses was to discover genetic markers of clinical significance that predict the occurrence of IFN-β DILI using both candidate gene (hypothesis-driven) and GWAS (hypothesis-free) genotyping analytical approaches. Additionally, there are currently no reliable means of predicting this ADR using only demographic or clinical information, thus the discovery of genomic biomarkers would be fulfill an unmet need in the clinic.  4.2 Methods 4.2.1 Overview and rationale of study design This analysis examined the allele/genotype frequency distribution of genetic markers in a case-control association study design, using both candidate and genome-wide approaches (Figure 4.1). Initially, a set of predetermined candidate variants in high priority genes identified by previous genome-wide and candidate gene studies of either DILI or IFN-β response were analyzed [74-76, 132-136, 143, 208-218]. This was followed by an analysis encompassing all genome-wide variants included within the genotyping array.   Initially, we undertook a hypothesis-driven; candidate gene approach given there is evidence to suggest DILI due to different small molecule drugs may share genetic risk factors [219-221]. Previously identified genetic variants associated with liver injury due to other drugs were analyzed to assess whether they were also associated with liver injury from IFN-β. Whether  62 known IFN-β response variants are linked with IFN-β induced liver injury was also investigated. The mechanism of IFN-β induced liver injury is postulated to be immunological [111, 112, 114]; therefore IFN-β immunological response markers may also increase susceptibility to experiencing liver injury. This is shown by the overlap between the pathways of IFN-β response and DILI. For example, IFN-β increases STAT4 activity [222] and variants within STAT4 have been associated with an increased risk of DILI (odds ratio for cases vs. controls = 1.45, P = 4.5 x 10-4)[223] and chronic liver diseases (odds ratio for cases vs. controls = 1.65, P = 4.67 × 10 − 5)[224].  Figure 4.1 Study design for the determination of genomic and genetic contributions to interferon-β induced liver injury in multiple sclerosis.  SNPs: single nucleotide polymorphisms, GWAS: genome wide association study, EUR: Europeans, LD: linkage disequilibrium.  Finally, a genome-wide analysis of genomic biomarkers associated with this ADR was also conducted. This hypothesis-free approach is justified in part because there is no prior literature Genetic Fine Mapping Analyses Cohort: Discovery Population (Canadian) Analyses: Genotype Imputation (EUR 1000 Genomes Population), LD Analysis Stage 2 - Replication Population   Cohort: Patients recruited in Sweden / USA  Analyses: Clinical Characterization, Candidate Gene and GWAS  Stage 1 - Discovery Population   Cohort: Patients recruited in Canada Analyses: Clinical Characterization, Candidate Gene and GWAS   63 on the genetic basis of IFN-β induced liver injury. Second, the exact mechanism of action of IFN-β is unknown [41], which makes it difficult to predict which genes are potentially involved in this ADR. Third, we were unable to find any pharmacogenomic or pharmacogenetic studies of liver injury induced by a biological therapy. Lastly, GWAS offers the opportunity to uncover novel variants that will improve our understanding of the mechanism of action of DILI [225].  4.2.2 Sample size estimation (prospective power calculation) A priori power calculations were performed to determine the minimum number of ADR cases and control patients required to detect a genetic marker with a per-allele odds ratio (OR) and minor allele frequency in a population of people with MS and exposed to IFN-β (Table 4.1, Figure 4.2). All power calculations were performed using Quanto [226]. Table 4.1 and Figure 4.2 demonstrate the relatedness between sample size, P-value significance thresholds, the estimated effect size and the rate of IFN-β induced liver injury.   Table 4.1 Sample size estimation to detect a specified odds ratio with 80% power over predefined statistical significance thresholds Odds Ratio Study arm Patient type IFN-β induced liver injury rate 1% 2% 3% MAF 5% MAF 10% MAF 15% MAF 5% MAF 10% MAF 15% MAF 5% MAF 10% MAF 15% > 3.0 Candidate Gene Case  50 29 23 51 30 23 52 31 24 Control 200 116 92 204 120 92 208 124 96 GWAS  Case  254 149 115 259 152 117 264 154 119 Control 1016 596 460 1036 608 468 1056 616 476 > 5.0 Candidate Gene  Case  20 12 10 21 13 11 21 14 11 Control 80 48 40 84 52 44 84 56 44 GWAS  Case  99 63 52 103 66 54 107 68 56 Control 396 252 208 412 264 216 428 272 224 MAF: minor allele frequency, GWAS: genome wide association study; 4 controls: 1 case. For the candidate gene study, the specified P-value is 0.05, for GWAS, the specified P-value is 5 x 10-8.   64 Three minor allele frequencies (MAF) were examined (5%, 10% and 15%), based on previous significant associations in DILI GWA studies [133, 135, 136]. The sample size estimate differs depending on the MAF employed, but given there were no known pharmacogenomic studies of IFN-β induced liver injury, the MAF could only be estimated from previous genomic studies in DILI.   Figure 4.2 Variation in minimum sample size estimation (prospective power calculation) to detect the specified odds ratio with 80% power and the predefined statistical significance thresholds for the candidate gene and genome wide association study (GWAS) of IFN-β induced liver injury. Controls = 4n cases. MAF: minor allele frequency, OR: odds ratio  The previous pharmacogenomic-DILI studies demonstrated similar patterns of liver injury (indicated by the relative ALT/ALP levels) and immunological mechanisms to that of IFN-β induced liver injury [133, 135, 136]. However, it should be noted these previous studies were not biologic response modifying therapies. The likely rate of IFN-β liver injury is ~2-3%, however, employing a lower rate of 1-2%, the sample size estimate does not drastically change within this  65 range of liver injury incidence (Table 4.1). Our calculations indicate in order to detect a per-allele OR of > 5 with a MAF of 0.10, we would require 63 cases of DILI from IFN-β in a population of people with MS (80% power).  4.2.3 Study populations 4.2.3.1 Inclusion and exclusion criteria All patients with a definite MS diagnosis and relapsing-onset disease course documented as exposed to any of the MS licensed IFN-β products and a normal baseline liver test result (consisting of at least ALT) were included (Table 4.2). Patients with primary-progressive MS were excluded because no IFN-β has been licensed for use in this disease course.  Table 4.2 Inclusion and exclusion criteria for study cases and controls for genomic analyses Inclusion criteria Exclusion criteria • Definite MS (Poser or McDonald criteria) • Registered at the site-specific MS clinic (and hence assessed by a neurologist) • Relapsing-onset disease course (relapsing-remitting or secondary-progressive) • Prescribed an IFN-β (IFN-β-1b SC, IFN-β-1a SC (high or low dose) or IFN-β-1a IM) • Normal baseline liver test (within the normal range for the reporting laboratory) • Presence of primary-progressive MS  • Presence of an elevated baseline liver test result • Presence of a known risk factor for liver injury (viral hepatitis, hereditary liver disease, autoimmune hepatitis, alcoholic liver disease and fatty liver disease)  4.2.3.2 Case and control definitions  Cases were MS patients treated with IFN-β meeting the standardized DILI phenotype definition of at least one of the following: 1) alanine aminotransferase (ALT) or aspartate aminotransferase (AST) ≥ 5x upper limit of normal (ULN) or 2) alkaline phosphatase (ALP) > 2x ULN or 3) simultaneous ALT ≥ 3x ULN and bilirubin > 2x ULN [88].   66 Controls were defined as MS patients exposed to the same IFN-β for ≥ 2 years [107] and have documented results of liver testing, which were all within the normal range (based on the normal ranges for the relevant testing laboratory). Two years was deemed a sufficient length of time to determine if a patient was truly a control as previous studies have reported the first 15 months of IFN-β therapy to be the greatest risk period for developing de novo ALT elevations during IFN-β therapy [107]. Additionally, 100% of cases developed DILI from IFN-β therapy within 700 days of beginning therapy (Figure 3.2, Chapter 3). To account for the large variation in the frequency of liver biochemistry testing between institutes, there was no minimum number of tests required for study entry.  4.2.3.3 Discovery stage The study population for the discovery cohort was recruited from one of five MS clinics in Canada (Vancouver, Winnipeg, Montréal, Halifax and London) (Figure 4.3). The patients recruited for the discovery stage of genetic analyses were similar to that presented in Chapter 3 (Vancouver, Winnipeg, Halifax and London), with one additional Canadian MS clinic (Montréal, Québec).  A total of 170 MS patients (42 cases, 128 controls) from Canada were recruited for the discovery stage between 2010 and 2014. The following five sites contributed: UBC Hospital MS Clinic (Vancouver, BC): 112 patients (20 cases, 92 controls); Winnipeg Health Sciences Centre MS clinic (Winnipeg, MB): 28 patients (9 cases, 19 controls); L’Hôpital Notre-Dame MS Clinic (Montréal, QC): 6 patients (2 cases, 4 controls); University Hospital MS Clinic (London, ON): 5  67 patients (3 cases, 2 controls) and the Dalhousie MS Clinic (Halifax, NS): 19 patients (8 cases, 11 controls).    The details for identification of participants from Vancouver, Winnipeg, Halifax and London are reported in Chapter 3. The Montréal, Québec site was located at the Hôpital Notre-Dame. Ethical approval was sought from the University of Montréal on 31 October 2012 and was approved by the local REB on 8 November 2012. Patients were identified using iMED, Logibec or vOACIS electronic medical records (EMR) encompassing the time period of January 1990-October 2012 and were reviewed in November 2012. All patients that were prescribed an IFN-β between January 1990 and 22 October 2012 (n = 489 EMR records) were identified. Patients experiencing an “adverse event" or “lack of tolerance” and those who decreased from IFN-β-1a 44 mcg to IFN-β-1a 22 mcg were selected for further chart review for possible inclusion as a case. Upon review of the EMR records, 12 patients had mentions of “liver transaminases 5x ULN” or “hepatitis”. Further review of these 12 patients encompassed the paper chart records (in the context of this study, all available or accessible information for an individual patient would have been located either in the paper chart or EMR), with two cases of confirmed DILI for contacting and study enrolment. Identification of controls was similar utilizing patients located in the iMED database, prescribed an IFN-β between January 1990 and 22 October 2012 (the same 489 records), except that no liver enzyme elevations must be present and the patient must have ≥ 2 years exposure to the same IFN-β.  The MS clinic research coordinator at L’Hôpital Notre-Dame contacted patients to assess interest in study participation. Once confirmed, the patients contact information was forwarded to  68 CPNDS in Vancouver for further contact and sending of saliva collection kits. Saliva was returned to CPNDS/Vancouver and the signed consent form was sent to the Montréal site. All consent forms for this site were translated into French. The MS clinical research coordinator also administered the questionnaires to patients due to language barriers.   4.2.3.4 Replication stage The replication cohort consisted of 34 non-Canadian MS patients (21 cases and 13 controls) (Figure 4.3). Participants were identified from the Partners HealthCare MS Clinic (Boston, MA, USA) (11 cases, 13 controls), the United States-based Drug-Induced Liver Injury Network (DILIN) (8 cases) [112], the Sweden-based SWEDEGENE (2 cases) [201]. The clinical characteristics of the DILIN IFN-β cases have been reported [112].  DNA samples were provided from DILIN and SWEDEGENE and were genotyped alongside the discovery stage patients (n = 10 cases). Imputed genomic data was available for participants (n = 24 patients, 11 cases, 13 controls) from the Partners HealthCare MS Clinic. These participants were genotyped on either the Illumina Human 660-Quad chip (Illumina, San Diego, CA, USA) or the Affymetrix Genome-Wide Human SNP array 6.0 (Affymetrix, Santa Clara, CA, USA). Genotyped data was fully imputed for each chromosome using the 1000 Genome Project EUR as the reference population.  All cases met the same definition and inclusion/exclusion criteria as the patients included in the discovery cohort, except for the DILIN (n = 8 cases), where two consecutive elevations of liver enzymes were required, as opposed to only one elevation for all remaining patients. One  69 elevation is deemed sufficient to meet the DILI case definition by the international DILI Expert Working Group [88].   Patients from the Partners HealthCare site were enrolled for another study at the Harvard Medical School (Boston, MA) and only their imputed genetic and accompanying clinical data was shared with the Vancouver site. Patient EMRs were extracted from the “Informatics for Integrating Biology and the Bedside” (i2b2) MS database (i2b2 is an National Institutes of Health-funded National Center for Biomedical Computing program based at the Partners HealthCare site) [227], using the following search methodologies: MS patients with at least one electronic prescription of interferon beta-1a or -1b or at least three narrative mentions of ‘avonex’, ‘betaseron’, ‘extavia’, ‘rebif’ or ‘interferon’ (in natural language processing). The period reviewed was all available records up to February 2011, which isolated 5495 EMRs. This list of patients was cross-referenced with blood work results for liver enzyme elevations occurring around the time (+/- 3 months) of IFN-β mentions (any of: ALT >5x ULN; AST >5x ULN; ALP >2x ULN; or TBILI >2x ULN & ALT >3x ULN). This search identified 45 potential cases requiring an in-depth review of each individual’s EMR. Controls were selected from the same 5495 EMRs, but had no elevations in the liver aminotransferases, with 564 potential controls and 70 reviewed, resulting in 13 controls included for the replication stage.  4.2.3.5 Additional sites for recruitment The Toronto, Ontario site was located at the St. Michael’s Hospital. Ethical approval was sought on 15 March 2013 and was gained on 8 August 2014 (time for completion by the St. Michael’s Hospital: 512 days). Given the lengthy turnaround for ethical approval, this site never recruited  70 patients for this study and the study was closed on 17 June 2015 due to the permanent departure of the study site-specific principal investigator. The Ottawa, Ontario site was located at the Ottawa Hospital. Ethical approval submission was prepared, however the site was unable to commit to this study and thus approval was never sought. Additional European sites were contacted for inclusion in this study but were unable to participate, typically due to logistical reasons e.g., lack of available resources or inability to access the required information in a timely manner; thus no patients were recruited from these sites.  4.2.3.6 Combined cohort This cohort combined all patients from the discovery and replication stages, consisting of 182 European (determined using principal components analysis in Section 4.1.6.3) MS patients (56 cases and 126 controls) recruited from Canada, USA or Sweden. More specifically, the cohort was comprised of combining all patients recruited from five Canadian MS clinics, Partners HealthCare MS Clinic (Boston, MA), DILIN [112] and SWEDEGENE [201]. Any genetic associations identified during the discovery stage (both candidate variant & GWAS) were assessed within this combined cohort.  4.2.4 Clinical characterization The following clinical information was collected for all patients: demographics (sex, date of birth, self-reported ancestry), MS disease characteristics (date of MS symptom onset, MS disease course), medications (IFN-β product type, concurrent medication usage (name, route of administration, dose, frequency, where possible), start and end of all drug treatments) and biochemical liver enzyme test results (date of test, liver enzyme parameters, reporting laboratory  71 upper limit of normal). Additional variables were collected pertaining to the DILI, including the date DILI criteria was met, reasons for discontinuation and autoantibody (auto-AB) testing (Table A.1 and A.2).     Figure 4.3 Study populations (Canadian and non-Canadian) included for the determination of genomic and genetic contributions to interferon-β induced liver injury in multiple sclerosis in both discovery and replication stages.  *Three samples removed due to genotyping call rates < 0.95 (i.e. 170-3=167 for the Canadians). Blue box represents the discovery cohort and the red box are those patients included during the replication stage.    Questionnaires were administered to the discovery cohort cases (Table A.3). Smoking status was assessed using validated questions from the Behavioural Risk Factor Surveillance Survey [228]. This is a reliable method for assessing self-reported smoking status in MS [229]. A smoker was MS patients exposed to IFN-β  (N = 204) Canadian             (n = 170)* European            (n = 151, 38 cases, 113 controls)   Non European      (n = 16, 4 cases, 12 controls) Non-Canadian (Sweden / USA)    (n = 34) European            (n = 31, 18 cases, 13 controls) Non European     (n = 3 cases)  72 defined as someone who had smoked at least 100 cigarettes in their lifetime [228]. Cases were also queried on their use of concomitant drugs, alcohol consumption, vaccinations and infections to exclude competing aetiologies associated with liver enzyme elevations and further characterize risk behaviours [92]. Cases were also asked for their weight, height (to calculate body mass index, BMI) and self-reported ancestry.  All cases was subjected to a causality assessment using validated tools: the Naranjo scale [95] (Table A.4) and the Roussel-Uclaf Casuality Assessment Method (RUCAM) scale [91] (Table A.5). These scales assist with excluding competing aetiologies associated with liver enzyme elevations, such as viral hepatitis and alcohol induced hepatitis.   4.2.4.1 Matching of patients on clinical features Previous studies of IFN-β induced liver aminotransferase elevations have revealed a number of clinical risk factors for experiencing this ADR, including sex, high-frequency dosing, MS disease course and age [6, 106, 107]. Thus, we initially sought to recruit controls that would be matched to cases on the following five criteria: age at start of treatment (+/- 5 years) [106, 107], sex [6, 106, 107], IFN-β product [6, 106, 107] disease course (relapsing remitting vs. secondary progressive at start of treatment)[110] and disease duration (+/-5 years) [107]. Following the identification and contact of patients from chart reviews, many patients could not be enrolled and this resulted in some cases having no matching controls or controls lacking a matching case. Thus, matching was inefficient and n = 31 patients (8 cases, 23 controls) would be excluded if performing a matched analysis. While it is generally considered advantageous to include matched controls in a case-control study to increase efficiency (and not validity [230]), others  73 have shown the reduction in statistical power due to sample exclusion (because no matching sample was available) is more detrimental to the validity of results than performing an unmatched analysis, with adjustment for confounders [231]. However, it should be noted that clinical differences in these five variables would be diminished because of the recruitment methods employed.  4.2.5 Statistical analyses of the clinical variables  Categorical variables (sex, MS disease course at IFN-β initiation (relapsing-remitting vs. secondary-progressive), IFN-β product, self-reported ancestry and concomitant hepatotoxic medications (concomitant hepatotoxic medications were defined as such by the Spanish Drug Induced Liver Injury Registry [232]) use were summarized by frequency (percent) and analyzed using the Fisher exact test. Continuous variables (age at IFN-β initiation, age at MS onset, MS disease duration at IFN-β initiation, BMI) were summarized using the median (IQR and range) or mean (standard deviation) and analyzed using the Mann-Whitney U test or student’s t-test, depending on the distribution of the data. Additional variables were analyzed for data only available for cases (pertaining to the drug reaction), including age at DILI onset, outcome of DILI (discontinued therapy, continued therapy, etc.), auto-AB testing, pattern of liver injury (hepatocellular, cholestatic or mixed), and severity (Table A.1). Exploratory analyses were performed; one to compare the clinical and demographic characteristics of cases between the discovery and replication cohort; the second was to compare cases of DILI who had received auto-AB testing and finally the time to DILI and any DILI associated genetic markers identified were explored using Kaplan Meier curves. These secondary analyses will assist with describing the cases of severe liver injury in greater detail; in particular the auto-AB. Statistical analyses  74 involving clinical variables were performed using SPSS (version 22.0) with statistical significance defined as P < 0.05.   4.2.6 Genetic methods and quality control (QC) procedures 4.2.6.1 DNA sampling, extraction and genotyping Saliva samples were collected using Oragene collection tubes (DNA Genotek, Kanata, ON). Patients recruited from London and Winnipeg sent their saliva samples to the CPNDS surveillance team member located in the respective city and were then forwarded on to the Vancouver site for genotyping. Patients recruited from Montréal and Halifax sent their saliva samples directly to the Vancouver site for genotyping. DNA was extracted from saliva samples using the QIAmp DNA purification system (Qiagen, Toronto, ON). The Quanti-iT PicoGreen assay (Invitrogen, Eugene, OR, USA) was used to quantify the DNA samples, which were subsequently normalized to 50 ng/ul for genotyping.  DNA samples from patients were genotyped using the Illumina Multi-Ethnic Genotyping Array (MEGA) array containing 1,705,969 SNPs (Illumina, San Diego, CA, USA). The MEGA array was selected because of enhanced content compared to the Illumina OmniExpress Array, including 2.5x more SNPs, 3.5x more variants located within the major histocompatibility complex (this region has been previously associated with DILI [132-136]) and analogous coverage of variants with MAF > 5% (European population), at a lower cost than the OmniExpress Array.   75 Samples were processed using the Illumina Tecan Freedom EVO 150 (Illumina, San Diego, CA, USA) and scanned using the Illumina HiScan System (Illumina, San Diego, CA, USA). Every set of 96 samples included a negative (1x Tris-EDTA buffer) and positive control.  To validate the genotyping calls from the Illumina MEGA array, significantly associated SNPs in the GWAS, or those prioritized by fine mapping, were genotyped in all discovery cohort patients using the TaqMan SNP genotyping assay (Life Technologies, Streetsville, ON, Canada). The concordance rate between the TaqMan assay and the Illumina MEGA array was 100%.  4.2.6.2 QC procedures for genetic variants and samples All SNPs genotyped using the Illumina MEGA array were clustered and marker statistics were calculated. The quality control (QC) for genotype data was performed using Illumina GenomeStudio software (Illumina, San Diego, CA, USA). We used the following combination of thresholds for QC metrics with visual inspection of cluster plots for markers at the boundaries of these thresholds. SNPs that clustered poorly (call rate < 95%) were filtered out, re-clustered and evaluated using the metrics detailed in Table A.6, with visual inspection of markers at the boundaries of the thresholds. Any SNPs with a MAF < 1% in both cases and controls and those deviating from Hardy-Weinberg equilibrium (HWE) genotype distributions (P-value < 1.0 x 10-4 in controls) were excluded using Golden Helix SVS software (v.8.4, Bozeman, MT, USA). A total of 785,230 SNPs were retained for the GWA discovery analysis.  The QC for samples was performed using Golden Helix SVS (v.8). Patient samples were excluded if: call rate < 95% (n = 3 samples), the reported and genotypically inferred biological  76 sex did not match (no samples), and if they were related (using an identity by descent estimation metric < 0.25) (no samples). X- and Y-chromosomes and mitochondria SNPs were excluded from the association analyses in keeping with recent GWAS quality control practices [233].  4.2.6.3 Principal component analysis  Investigation of genetic ancestry in the discovery and the combined (partial analysis only) cohorts was performed using principal component analysis (PCA) (the EIGENSTRAT method) in Golden Helix SVS. As only part of the combined cohort was genotyped by our research centre, a statistical comparison of the PCs between cases and controls was completed for those with full GWAS data available. The first 10 principal components were calculated and genetic analysis was restricted to European only. Correction of genetic data was conducted for significant differences (P < 0.05) between cases and controls on the 10 principal components.  4.2.6.4 Selection of candidate variants  We searched for all candidate gene or GWA studies of either DILI or IFN-β response, published up to and including 25 June 2015 in MEDLINE. The following search terms were used: “interferon beta AND pharmacogenomic or pharmacogenetic or genetic or genomic or gene”, “interferon beta AND response or nonresponse”, “DILI AND pharmacogenomic or pharmacogenetic or genetic or genomic or gene”, “drug induced liver injury AND pharmacogenomic or pharmacogenetic or genetic or genomic or gene”, “drug induced liver toxicity AND pharmacogenomic or pharmacogenetic or genetic or genomic or gene”. We found 35 articles through MEDLINE, 25 of which reported significant variants (significance threshold determined by each study) associated with either DILI (n = 34 SNPs) or IFN-β response (n=54  77 SNPs). If a SNP was not directly genotyped by the MEGA array, a SNP in strong LD (r2 > 0.8 - using the 1000 Genomes, Phase 3 v5 European reference population) was utilized, as determined using SNiPA [234]. A list of candidate variants is available in Table A.7.  4.2.6.5 Statistical analysis for genetic association  The association for each genetic marker with case/control status was tested using adjusted logistic regression with an additive model in candidate gene and GWA stages (Odds ratio and 95% confidence intervals). For each test in the discovery stage, the specific significance level to indicate genetic variants from the GWAS to be investigated in the replication cohort was P-value = 1 x 10-5 (The National Human Genome Research Institute GWAS Catalog genome-wide threshold), after adjustment for any significant clinical covariates [147]. For the candidate gene study, the Bonferroni-corrected significance threshold was P < 0.0009 (0.05 / 55 variants tested). Finally, in the replication cohort, the pre-specified significance level was P < 0.05. Statistical analyses were performed using SVS/HelixTree 8.4.0 (Golden Helix) and SPSS Version 22 (IBM). Visual plots were generated using Golden Helix SVS (Bozeman, MT, USA), Haploview [235] and LocusZoom [236].  4.2.6.6 Fine-mapping, imputation and annotation Imputation was performed using BEAGLE 4.0 and 503 European individuals from the 1000 Genomes Project as the reference haplotype population [237]. Imputed SNPs with a BEAGLE allelic R2 ≥ 0.5 and MAF ≥ 0.01 was utilized for QC of the imputation. LD analysis (r2 and D’) was conducted using the European reference population from the 1000 Genomes Project. Any significant variants (P < 1 x 10-5 in discovery, or P < 0.05 in the replication after adjusted  78 logistic regression with the additive model) were annotated using the Ensembl Variant Effect Predictor (VEP) tool [238], the Combined Annotation Dependent Depletion (CADD) score [239] and expression Quantitative Trait Loci (eQTL) information from the Genotype-Tissue Expression project [240].   4.2.6.7 Clinical prediction of IFN-β induced liver injury Markers confirmed at the discovery and replication (P < 1 x 10-5 in discovery and P < 0.05 in the replication after adjusted logistic regression with the additive model) were evaluated for potential clinical utility using estimates of positive predictive value (PPV), negative predictive value (NPV), specificity and sensitivity. Receiver operating characteristic (ROC) curves and area under the curve (AUC) estimates (and 95% confidence intervals) were generated using the predicted probabilities of each patient for a model encompassing only clinical variables and a separate model combining clinical and genetic variables. The following variables were included for the clinical only model: age at IFN-β start (continuous), IFN-β product (IFN-β-1b, IFN-β-1a SC high dose, IFN-β-1a SC low dose or IFN-β-1a IM) and sex. The combined clinical and genetic model included the same variables in the clinical-only model, but with the addition of the top genetic variant (rs2205986). The ROC curves of these two prediction models (clinical model and clinical/genetic model) were compared using the DeLong’s test [241] and the R-package pROC [242].   79 4.3 Results 4.3.1 Patient clinical and demographic characteristics  Baseline patient characteristics are provided in Table 4.3 for the discovery, replication and combined cohorts. In the discovery cohort, the majority of patient characteristics were similar between cases and controls. There was a preponderance of females in the control group (91.2% controls vs. 78.9% cases, P > 0.05), although over 85% of all discovery stage patients were female. Compared with cases, controls were marginally more likely to be relapsing remitting MS disease course at the start of IFN-β therapy (98.2% vs. 89.5%, P = 0.035). Majority of the participants self-identified as European, with ten patients of unknown self-report and were later confirmed by PCA (genetically ascertained) to cluster with the Europeans. The use of concomitant hepatotoxic medications and medications associated with acute liver failure was higher in cases than controls, although this was not significantly different (P = 0.654, P = 0.184, respectively) and may represent a reporting bias. The number of post IFN-β liver aminotransferases tests did not differ between cases and controls, although if weighted across the length of IFN-β treatment, cases were tested every 1.75 months, with controls tested approximately once annually. As expected, cases were exposed to IFN-β for a significantly shorter duration (median: 4.0 months; IQR: 2.5-27.5 months) than controls (median: 82 months, IQR: 51-110.5 months) (P = 6.0 x 10-15), as per the inclusion criteria for controls (IFN-β exposure ≥ 2 years or 24 months).   Similarly, the clinical and demographic characteristics of the cases closely resembled that of the controls in the replication stage. Cases were significantly older at the start of IFN-β therapy (mean age: 38.5 years old), compared to controls (mean age: 29.9 years old, P = 0.026). No  80 patients were exposed to pegylated-IFN-β and a higher proportion of patients were exposed to IFN-β-1a, compared to the discovery cohort. Median IFN-β exposure was again shorter in cases than controls (IFN-β exposure: 17.5 months (IQR: 4.5-53 months) vs. 59 months (IQR: 32-87.5 months), P = 0.014).   81 Table 4.3 Assessment of clinical and demographic characteristics (discovery, replication and combined patient characteristics) Variable Discovery                                    (n = 151) Replication                             (n = 31) Combined                                         (n = 182) Cases          (n = 38) Controls (n = 113) Pa Cases (n = 18) Controls (n = 13) Pa Cases (n = 56) Controls    (n = 126) Pa Females, no. (%) 30 (78.9) 103 (91.2) 0.078 15 (83.3) 12 (92.3) 0.462 45 (80.4) 115 (91.3) 0.051 Age at IFN-β start, median y (IQR) 40.5 (30.75, 50.25) 40 (34, 47.5) 0.857 38.5 (10.24) 29.9 (8.97) (n=12)d 0.026 40 (31.25, 47.75) 40 (33, 46.5) (n=125) 0.978 Age at MS onset, median y (IQR) 30 (26, 37) (n=37) 32 (25, 39) (n=111) 0.349 31.61 (9.45) (n=13) 25.81 (9.26) (n=11)d 0.145 30 (26, 37.25),     (n=50) 32 (27, 39) (n=122) 0.542 MS disease duration at IFN-β start, median y (IQR) 6 (1, 15) (n=37) 4 (2, 10) (n=111) 0.762 4.08 (5.66) (n=13) 3.63 (5.27) (n=11)d 0.847 5 (1, 12), (n=50) 4 (2, 10)    (n=122) 0.626 Relapsing remitting disease at IFN-β start, no. (%) 34 (89.5) 111 (98.2) 0.035 11 (100) (n=11) 12 (92.3) 0.347 45 (91.8) (n=49) 123 (97.6) 0.08 IFN-β product, no. (%)   0.195   0.105   0.464 1a IM (30 mcg once weekly) 1 (2.6) 12 (10.6)  7 (38.9) 10 (76.9)  8 (14.3) 22 (17.5)  1a SC (22 mcg 3x weekly) 5 (13.2) 13 (11.5)  0 0  6 (8.9) 13 (10.3)  1a SC (44 mcg 3x weekly) 14 (36.8) 32 (28.3)  5 (27.8) 1 (7.7)  19 (33.9) 33 (26.2)  1b (250 mcg every other day) 17 (44.7) 56 (49.6)  6 (33.3) 2 (15.4)  23 (41.1) 58 (46)  Pegylated 1a SC 1 (2.6) 0  0 0  1 (1.8) 0  Self-reported ancestry, no. (%)   0.453   N/A   0.500 European 34 (89.5) 107 (94.7)  18 (100) 13 (100)  52 (92.9)  120 (95.2)  Unknown 4 (10.5) 6 (5.3)  0 0  4 (7.1) 6 (4.8)  BMI (kg/m2), median (IQR)  26.6 (22.5, 30.2) 25.6 (22.5, 30.2) 0.973 26.2 (21.3, 29) (n=7) 25.5 (25.5) (n=1) 0.408 26.6 (22.0, 29.5), (n=45) 25.6 (22.5, 30.1), (n=114) 0.919 Underweight - normal, no. (%)b 10 (38.5) 50 (46.3) 0.516 3 (42.9) 0  13 (39.4) 50 (45.9) 0.512 Overweight - obese, no. (%)b 16 (61.5) 58 (53.7)  4 (57.1) 1 (100)  20 (60.6) 59 (54.1)  Concomitant hepatotoxic medication, no. (%)c 31 (81.6) 87 (77) 0.654 16 (88.9) 10 (76.9) 0.37 47 (83.9) 97 (77) 0.287 Medication associated with acute liver failure, no. (%) 20 (50.6) 44 (38.9) 0.184 9 (50) 6 (46.2) 0.833 29 (51.8) 50 (39.7) 0.128  82 Variable Discovery                                    (n = 151) Replication                             (n = 31) Combined                                         (n = 182) Cases          (n = 38) Controls (n = 113) Pa Cases (n = 18) Controls (n = 13) Pa Cases (n = 56) Controls    (n = 126) Pa Number of liver tests completed, post-treatment start, median (IQR) 7 (5, 10) 8 (6, 13) 0.079 2.5 (1.75, 6) 4 (2, 6) 0.387 6 (4, 9) 8 (6, 13) 0.006 Categorical data: Fisher Exact Test. Continuous data: Mann-Whitney U Test or Student’s t Test (t test was used only for variables with a normal distribution, i.e. those designated by d, all other continuous variables were compared using the Mann-Whitney U test). a: Statistically significant differences (P < 0.05) between cases and controls are highlighted in bold. b: Underweight and normal: BMI = < 24.9 kg/m2; overweight and obese: BMI >25 kg/m2. c: Any hepatotoxic medication for cases are those taken at the same time as DILI; for controls, concomitant medications are those taken during their entire IFN-β exposure period, Hepatotoxicity is determined by the Spanish Drug Induced Liver Injury Registry. d: Mean (+/-SD) are reported for these variables. The number of patients is specified if less than the total for cases/controls for the respective stage of analysis.  83  Figure 4.4 Ancestry determined by principal component analysis (PCA) vs. self-reported ethnicity.  The first two principal components are plotted to visualize the distribution of population ancestry in the patients genotyped by the Vancouver site. Each point represents one patient, colored by their self-reported ancestry. The circled patients are the Canadian patients used in the discovery stage of analysis (n =151).  Table 4.4 Principal components of discovery and replication stage patients.   Discovery Europeans                                                             (n=151, 38 cases, 113 controls) Replication Europeansa     (n=7 cases) PC No. Eigenvalue P-value Eigenvalue 1 0.77 0.62 0.45 2 0.66 0.87 0.44 3 0.57 0.55 0.44 4 0.56 0.16 0.44 5 0.54 0.83 0.44 6 0.51 0.76 0.44 7 0.51 0.69 b 8 0.50 0.61  9 0.49 0.07  10 0.49 0.18  PC: principal component. aPC analysis and statistical comparison was not completed for the entire replication cohort since full GWAS data was only available for a subset of cases and no controls, bNo further PCs computed because only n-1 (i.e. 7 samples-1 = 6 samples) PCs can be calculated   84                                    Figure 4.5 Quantile-quantile plot of the Discovery Cohort. Quantile-quantile plot showing the distributions of adjusted observed –log10 (P-values) plotted against expected –log10 (P-values) (adjusted for MS disease course). Smaller P-values than would be expected by chance were observed at the tail of the plot. λGC = 1.03384 indicates no obvious population stratification.   4.3.2 Candidate association variant results Assessment of the literature identified 88 candidate gene variants, of which 28 SNPs were directly genotyped by the Illumina array and a further 27 variants were sufficiently tagged (r2 > 0.8) by genotyped SNPs on the array for a total of 55 candidate SNPs in the initial analysis (Table A.7). Thus, 33 candidate SNPs were not included because either they were not genotyped or had no SNPs in high LD (r2 > 0.8) that were genotyped on the Illumina array.   The genetic association test between 55 SNPs and case/controls status revealed no significant associations (P > 0.001) (Table 4.5). Although no significant associations were identified, the most highly associated SNP, rs6788472, conferred a protective effect against IFN-β induced liver  85 injury (P = 0.015, OR: 0.69, 95%CI: 0.32-1.51). This SNP was in high LD (r2 = 0.82) with rs1368576, which has been previously associated as a protective factor for ximelagatran (thrombin inhibitor) induced liver injury. The next most significant association was an intronic variant (rs10492199) within the gene coding for interferon gamma (IFNG) (P = 0.036, OR: 2.33, 95%CI: 0.97-5.6).   4.3.3 Genome-wide association results  A genome-wide analysis of 151 patients revealed five significant associations with IFN-β induced liver injury (P < 1 x 10-5) (Table 4.6, Figure 4.6). The most highly associated SNP (rs2205986) was an intronic variant located on chromosome 1 (P = 1.9x10-7, OR: 11.1, 95%CI: 4.19-29.43). Additionally, one missense variant also located on chromosome 1 (rs200839898) was significantly associated with IFN-β induced liver injury (P = 7.3 x 10-6). There were three additional intronic or intergenic SNPs that reached the significance threshold for follow up in the replication cohort.   4.3.4 Replication of genetic variants  The replication of the genetic findings was examined in an independent cohort to minimize the likelihood of false positive associations, as an important step towards the validation of the associations. Five SNPs were selected for follow up in the replication cohort based on the pre-determined threshold of association (P < 1 x 10-5) in the discovery stage. As there was a significant difference in age at IFN-β start (P = 0.038) in the replication cohort, adjustment of the genetic data for differences in age was necessary. Following adjusted logistic regression, the association of rs2205986 was confirmed in an independent cohort of 31 European patients with  86 MS (P = 0.02) (Table 4.6). Combining the discovery and replication together, the association with rs2205986 was highly significant (P = 9.39 x 10-9, OR = 9.83, 95%CI = 4.01-24.10), beyond the widely adopted threshold of P < 5 x 10-8 for GWAS findings.  Two other genetic variants (rs17115661 and rs62403705) were not replicated in the replication stage (P > 0.05). Finally, the remaining two variants identified in the discovery stage (rs72693229 and rs200839898) were not located in the replication control dataset (and also had no SNPs in high LD in the 1000 Genomes Project [237]) and could not be investigated further at this time. There exists an insertion-deletion just upstream of rs200839898 and may have caused alignment issues, thus preventing its inclusion in the 1000 Genomes Project data.  To validate the GWAS genotyping data, the rs2205986 SNP was re-genotyped in all discovery cohort patients using a TaqMan assay, resulting in a concordance rate of 100%.  4.3.5 Imputation and fine mapping  Imputation of the 1q32.2 locus, centered on rs2205986 (chromosome 1: 208,116,112 bp-212,116,112 bp), used 1,118 genotyped SNPs to impute an additional 43,083 SNPs (total of 44,201 SNPs). Following QC, 11,979 variants were available for analysis in the discovery cohort (Table 4.7, Figure 4.7). Following logistic regression (additive model) with adjustment for MS disease course, three additional variants within this region were associated with DILI (P < 1 x 10-5) (rs72741075, rs12086774, rs12569320).   87 Table 4.5 Case-control analysis results for the candidate gene analysis (discovery) Biomarker Pharmacogenomic Analyses Adjusted logistic regression (additive model) a Variant Gene Prev. association Function Study population MAF cases MAF controls P OR (95% CI) rs6788472 WNT7A Ximelagatran induced liver injury Intron Stage 1-discovery  0.289 0.451 0.021 0.52 (0.3-0.92) rs10492199  IFNG IFN-β Response Intron Stage 1-discovery  0.487 0.350 0.048 1.7 (0.99-2.9) aThe covariate for logistic regression in the discovery stage was MS disease course (n = 151 European MS patients). WNT7A: Wingless-type MMTV integration site family member 7A, IFNG: Interferon gamma, MAF: minor allele frequency.   Table 4.6 Case-control analysis results for the genome-wide analysis (discovery, replication and combined) Biomarker Pharmacogenomic Analyses Logistic regression (additive model)a Variant Gene Function Study population MAF cases MAF controls P OR (95% CI) rs2205986 SYT14 Intron Stage 1-discovery GWAS  0.237 0.035 1.9 x 10-7 11.10 (4.19-29.43) Stage 2-replication 0.166 0 0.02 N/A All European MS patients  0.214 0.032 9.39 x10-9 9.83 (4.01-24.10)    Stage 1-discovery GWAS 0.513 0.226 4.4 x 10-6 3.68 (2.03-6.65) rs17115661 NAP1L1 Intergenic Stage 2-replication 0.33 0.35 0.86 0.92 (0.34-2.46)    All European MS patients  0.46 0.24 1.17 x 10-4 3.22 (1.57-6.62) rs72693229 FLRT2 Intron Stage 1-discovery GWAS 0.171 0.022 5.4 x10-6 11.7 (3.79-36.57) rs200839898 SYT14 Missense Stage 1-discovery GWAS 0.289 0.088 7.3 x 10-6 5.95 (2.66-13.34) rs62403705 ETV7 Intron Stage 1-discovery GWAS 0.592 0.305 9.6 x 10-6 3.84 (2.01-7.33) Stage 2-replication 0.31 0.42 0.41 0.59 (0.16-2.10)   All European MS patients  0.50 0.32 6.47 x 10-4 2.15 (1.02-4.54) aAdjusted logistic regression was performed in the discovery (adjustment for MS disease course) and replication cohorts (adjusted for age at the start of IFN-β), with unadjusted logistic regression performed in the combined cohort due to insignificant clinical/demographic differences between cases and controls. MAF: minor allele frequency, N/A: not applicable as rs2205986 was absent in controls.   88    Figure 4.6 Manhattan plot showing the observed distribution of –log10 (P-values) against SNP chromosome location (GRCH37.p13).  P-values are for logistic regression (additive model) with adjustment for MS disease course. Red line, P = 1.0 x 10-5 [147]. 89 These variants were all either intronic or intergenic variants and were in moderate LD (D’ = 1.00, r2 > 0.74) with rs2205986. All four variants, including rs2205986, were located within the same 257-kb LD block (European population) on chromosome 1 (Figure 4.8). Upon adjusting the logistic regression for rs2205986, the additional 3 variants were non-significant (P > 0.05).   4.3.6 Annotation of significant variants  The initial annotation of the genome-wide significant variant (rs2205986) and imputed variants was done using the Ensembl Variant Effect Predictor, eQTL information from the GTex portal (v.6) [240] and the Combined Annotation Dependent Depletion (CADD) score [239]. The variant with genome-wide significance in the combined cohort, rs2205986, is an intronic variant in SYT14 (synaptotagmin 14). Interestingly, this variant is an eQTL for interferon-regulatory factor 6 (IRF6) in the cerebellum (decreased expression, P = 6 x 10-8) and in the liver (increased expression, P = 0.07) (Table 4.7) and had a CADD score of 5.744 (CADD scores ≥ 15 indicate that this is predicted to be one of the 10% most deleterious mutations that you can do to the human genome [239]). The variant in high LD (D’ = 1.00, r2 = 0.89) with rs2205986, namely rs72741075, was also a SYT14 regulatory variant, and an eQTL for IRF6 (cerebellum - decreased expression, P = 1.6 x 10-7 and in the liver with increased expression, P = 0.24). This variant also had a relatively high CADD score of 17.19.  90 Table 4.7 Fine mapping of genetic association signals in the 1q32.2 genomic region (reached P < 1 x 10-5 after adjusting for multiple sclerosis disease course in discovery cohort) Biomarker Pharmacogenomic Analyses Adjusted Logistic Regression (additive)a Annotation SNP rsIDb Position (bp) Type LD with rs2205986 D' (r2)c MAF P OR       (95% CI) Conditional Analysis on rs2205986 P CADD eQTL Cases Controls rs2205986 210,116,112 Intron 1.00 (1.00) 0.24 0.04 1.9 x 10-7 11.1      (4.19-29.43) - 5.744 IRF6 rs72741075* 210,074,833 Regulatory modifier 1.00 (0.89) 0.25 0.07 8.9 x 10-6 6.38     (2.74-14.86) 0.85 17.19 IRF6 rs12086774* 210,079,167 Intergenic 1.00 (0.74) 0.25 0.07 8.9 x 10-6 6.38     (2.74-14.86) 0.85 3.428 - rs12569320* 210,084,705 Intergenic 1.00 (0.74) 0.25 0.07 8.9 x 10-6 6.38     (2.74-14.86) 0.85 11.45 - We imputed 44,201 additional variants on chromosome 1 into the discovery cohort using the EUR component of the 1000 Genomes Project population as a reference. MAF: minor allele frequency, OR: odds ratio, CADD: combined annotation dependent depletion, eQTL: expression quantitative trait loci (from the Genotype-Tissue Expression project, GTex v.6. aThe covariates for logistic regression were MS disease course, and rs2205986 where indicated. bAssociation analyses for imputed SNPs were restricted to those with Beagle allelic value R2 > 0.5 and MAF > 0.01. cLD calculated using EUR component of the 1000 Genomes Project population. *Imputed SNP (vs. genotyped).        91   Figure 4.7 Regional association plot of chromosome 1q32.2.  Association results (primary y-axis) are shown for genotyped (circles) and imputed (squares) SNPs along with recombination rates (secondary y-axis) for a 1500-kb region on chromosome 1. Each circle or square represents the P-value from the logistic regression analysis using an additive model, adjusted for MS disease course. SNPs are coloured according to their pairwise correlation (r2) with rs2205986 (purple circle) using the 1000 Genomes Project EUR reference population.      02468−log10(p−value)020406080100Recombination rate (cM/Mb)rs2205986●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●●●●●● 0.20.40.60.8r2MIR205HGMIR205CAMK1GLAMB3MIR4260G0S2HSD11B1TRAF3IP3C1orf74IRF6DIEXFSYT14SERTAD4−AS1SERTAD4HHATKCNH1209.4 209.6 209.8 210 210.2 210.4 210.6 210.8Position on chr1 (Mb)Plotted SNPs 92   Figure 4.8 Linkage disequilibrium plots of 41-kb and 257-kb regions containing rs2205986.  (A): An LD plot of a 41-kb region encompassing rs2205986 is shown. Three SNPs (rs72741075, rs12086774, rs12569320) (boxed in black at the top) are in LD with rs2205986 (r2 > 0.74). (B): A LD plot of the entire 257-kb block encompassing all four variants. Colours in plots are coded as follows: light blue (D’ = 1, LOD < 2), red (D’ = 1, LOD ≥ 2), white (D’ < 1, LOD < 2), purple lines indicate the LD block (257-kb) calculated by Haploview. LD analysis (r2 and D’) was conducted using the European reference from the 1000 Genomes Project. 93 4.3.7 Clinical prediction of IFN-β induced liver injury To examine the predictability of risk using any genetic variants identified, the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were estimated among our cases and controls (combined cohort). The clinical prediction of DILI due to IFN-β therapy using only clinical variables had an AUC of 0.61 (95%CI: 0.52-0.70) (Figure 4.9). The addition of the genetic marker (rs2205986) to the clinical prediction significantly increased the AUC from 0.61 (95%CI: 0.52-0.70) to 0.72 (95%CI: 0.64-0.81, P = 0.0039). As a predictive marker of IFN-β induced liver injury capable of differentiating cases from controls, rs2205986 had a high specificity (93.7%, 95%CI: 87.9%-97.2%) and moderate sensitivity (41.1%, 95%CI: 28.1%-55.0%). The marker also had a moderate PPV of 74.2% (95%CI: 55.4%-88.1%) and NPV of 78.1% (95%CI: 70.7%-84.5%).  4.3.8 Exploratory analyses 4.3.8.1 Characteristics of IFN-β induced liver injury cases  Characteristics specifically pertaining to the ADR in people with MS experiencing IFN-β induced liver injury are noted in Table A.8. The median time to DILI, after starting IFN-β therapy was 3 months (IQR: 1-17.5 months) for the discovery cohort and significantly longer for cases in the replication cohort at 12.5 months (IQR: 3-52.25 months) (P = 0.012). More than 75% of discovery cases (n = 33/38 cases) discontinued the implicated IFN-β agent following the DILI episode, with over half (57.9%, 22/38 cases) experiencing a normalization of liver enzymes following the cessation of therapy. In the replication cohort, most cases had post-IFN-β cessation follow up information available (n = 16/18), with all 16 cases halting therapy due to DILI. Of these 16 cases, nine had available liver enzyme testing post-IFN-β cessation, with one-third (3/9 cases) experiencing a normalization of liver enzyme elevations. Peak absolute levels of ALT and  94 ALP were significantly higher in the replication cohort than the discovery (P < 0.05), although the opposite was true for total bilirubin levels (i.e. significantly higher total bilirubin was documented in the discovery cohort cases).     Model AUC (95% CI) P-value vs. Clinical Model Clinical 0.61 (0.52-0.70) - 0.0039 Clinical and Genetic 0.72 (0.64-0.81)  Figure 4.9 Receiver operating characteristics (ROC) curves of the clinical and clinical/genetic models for predicting interferon-beta induced liver injury in the combined cohort.  The area under the curves (AUC) and corresponding 95% confidence intervals (95% CI) are also presented. Clinical variables included age at interferon beta start, interferon beta product and sex and the only genetic variable included was rs2205986 (SYT14 marker).   95  In the combined cohort, the DILI cases who were rs2205986 carriers (n =23) had a longer, albeit non-significant, median time to DILI onset than rs2205986 non-carrier DILI cases (n = 33) (Figure 4.10) (6 months, 95%CI: 0-20.1 months vs. 3 months, 95%CI: 1.8-4.1 months, P = 0.086).   4.3.8.2 IFN-β induced liver injury cases receiving autoantibody testing The role of auto-AB in predisposing or modifying the risk of DILI from IFN-β was investigated (Table A.9) because liver injury due to other therapies has been considered immune-mediated [132-135]. All cases with auto-AB testing results available (n = 24) for any of anti-nuclear (n = 22 cases tested), anti-mitochondrial (n = 8 cases tested) or anti-smooth muscle antibodies (n = 16 cases tested) were selected and a comparison between DILI cases positive for any auto-AB were compared to DILI cases that were auto-AB negative. All testing for autoantibody markers was completed after the DILI onset. Over half of auto-AB positive DILI cases were positive for antinuclear antibodies. A greater proportion of auto-AB positive were more severe cases of DILI and had significantly (P = 0.045) higher maximum ALT x ULN (Table A.9). Finally, a higher proportion of auto-AB positive patients were rs2205986 carriers, compared to those who were auto-AB negative (50% vs. 18.8%, P = 0.121).  96    Figure 4.10 Kaplan-Meier curves showing time (days) since treatment initiation to drug-induced liver injury (DILI) in interferon-beta treated multiple sclerosis (MS) patients from the combined cohort for rs2205986 carriers and non-carriers.  Key: rs2205986 carriers (green line) and non-carriers (blue line). Log-rank test: P = 0.086.   4.4 Discussion We have identified a new genetic biomarker with the first genome-wide association study of IFN-β induced liver injury in MS patients. Exposure to IFN-β is marked by a considerable proportion of patients experiencing an ADR during IFN-β exposure, with 1-2% experiencing severe, medically significant elevations in the liver aminotransferases [6, 107, 203]. A genetic predisposition to ADRs has been well characterized in other therapeutic areas, such as oncology,  97 but to the best of our knowledge, has not been investigated in the context of individuals with MS. To the best of our knowledge, no other group has performed a GWAS of IFN-β induced liver injury, nor of liver injury induced by a biological agent.   4.4.1 Genomic determinants associated with DILI We describe here the results of a genome-wide analysis of 785,230 genetic variants genotyped in an initial discovery of 151 people with MS, which identified a significant association with a variant in SYT14 (rs2205986, adjusted OR = 11.1, P = 1.88 x 10-7). This initial discovery was followed by replication in an independent cohort of 31 people with MS, where the risk variant was only detected in the cases (P = 0.02). Over the combined cohort, carriers of rs2205986 had a ~ten-fold increased risk of experiencing IFN-β induced liver injury and reached genome-wide significance (P = 9.39 x10-9).  The identified marker (rs2205986) is an intronic variant located within a membrane trafficking protein, synaptotagmin-14 (SYT-14). In general, synaptotagmins are involved in the exocytosis of synaptic and non-synaptic vesicles [244], with SYT-14 specifically involved in trafficking both inside and outside the brain [245]. To the best of our knowledge, SYT-14 has not been associated with DILI, nor have any known implications to MS. Importantly, the identified genetic variant was also an eQTL for interferon regulatory factor 6 (IRF6) and has been previously associated with increased expression of IRF6 in the liver, compared to its expression in rs2205986 non-carriers in the Genotype-Tissue Expression project (GTex) [240]. Although findings in that study did not reach significance, that may have been a result of the small number of liver tissue samples included in the GTex project (n = 97 for liver tissue samples compared to,  98 for example, 278 lung tissue samples) [240]. In addition, RNA stability can be reduced in human liver samples even a few hours following tissue harvesting [246]. Furthermore, in the context of our study with a specific interest in IFN-β therapy, it is unlikely the GTEx liver samples were exposed to IFN-β and hence the effect on IRF6 expression may have been even greater in our MS patients cohorts treated with IFN-β.  The IRF family is comprised of nine members (IRF1-9) and are responsible for the regulation of the type-I interferon system, encompassing both IFN-β and IFN-α [247]. Although its function has not been firmly established, IRF6 was recently proposed to be associated with modifying IFN-β response [248]. Other IRFs have been implicated in IFN-β therapy, with a significantly increased expression of IRF9 (Adjusted P-value: 2.93 x 10-13) and increased IRF7 (Adjusted P-value: 1.29 x 10-13) upon IFN-β stimulation in peripheral blood mononuclear cells of healthy controls [249]. The implication of IRFs in IFN-β response and now, in IFN-β related liver injury, may suggest an overlap between those experiencing an ADR and also differential IFN-β responsiveness. The normal induction of the type-I interferon pathway by IFN-β therapy in MS alone may be insufficient to induce liver injury, yet the hepatic overexpression of IRF6 in rs2205986 carriers may “tip the balance”, resulting in liver injury (Figure 4.11). The overexpression of IRF6 could lead to increased levels of hepatocyte apoptosis, as IRF6 is known to promote apoptosis, following brain injury [250]. DILI worsens in the presence of apoptotic cells [251], as the release of ALT from liver cells is due to apoptotic processes, with DILI drugs capable of inducing caspase (apoptotic) activity [252]. Of the few liver biopsies reported to date in MS patients experiencing IFN-β induced liver injury, the presence of apoptosis was noted [253]. The administration of IFN-β to HepG2 cells (a human hepatoma cell  99 line) has been shown to increase cell death [254]; thus normal hepatic IRF6 expression during IFN-β exposure could lead to subclinical (or no) liver injury. However, rs2205986 carriers exhibiting increased expression of IRF6 in the liver could lead to increased liver cell death and clinically significant elevations in ALT and liver injury.   The IRFs have not been directly implicated in DILI, however the members of the IRF family are well known contributors to hepatic ischemia/reperfusion injury. Liver injury from ischemia/reperfusion and liver injury from drug therapies are both models of liver cell death, albeit from different causes [255]. Upregulation of IRF1 results in worsening of ischemia/reperfusion liver injury, and its deficiency result in less liver damage as the presence of IRF1 begins the cascade of inflammatory processes in ischemia/reperfusion injury [256]. IRF3-knock out mice have improved liver function maintenance, following ischemia/reperfusion injury than wild type and IRF3-overexpressing mice [257]. In this same study, IRF3 was associated with activating toll-like receptor 4 (TLR4), mediating hepatocyte damage [257]. Following ischemia/reperfusion injury, the overexpression of IRF9 induced hepatocyte apoptosis, increased inflammatory cytokine levels and serum ALT levels and enhanced hepatic necrosis, with IRF9-deficient mice displaying markedly reduced levels of these factors [258]. Together, this collection of data highlight the complexity of IRF functions, but also develops biological rationale as to why their upregulation may be associated with DILI.   To the best of our knowledge, the risk variant (rs2205986) represents the first time either of SYT14 or IRF6 have been associated with DILI, as well as this being the ninth GWAS of DILI to be conducted (Table A.10) (as of December 1, 2015). Schaid et al (2014) published a  100 prospective validation of the top genomic risk marker in predicting lapatinib DILI in 1,194 breast cancer patients [137]. The study reported that carriers of HLA-DQA1*02:01 were at a 14-fold greater risk for liver injury, defined as ALT > 3x ULN, than non-carriers (95%CI: 6.35-21.24, P = 2.4 x 10-13), or over 17-fold when using the stricter criteria of ALT > 5x ULN (OR: 17.77, 95%CI: 5.99-52.75). This 2014 study highlighted the potential utility of genomic markers in prospectively predicting DILI, but also that using an increasingly robust phenotype, as indicated by higher thresholds for defining DILI (i.e. ALT > 5x ULN not > 3x ULN) may strengthen the associations. From the prospective lapatinib DILI study, information on the genomic marker and its increased risk for DILI was subsequently included in the drug monograph, although there were no explicit suggestions or recommendations for baseline genetic testing [259]. The high NPV reported from this prospective lapatinib-DILI study (99.5%) may impede the widespread application of this test in clinical practice since patients who test positive for the HLA allele may be excluded from lapatinib therapy but might not actually develop DILI. Nevertheless, implementing genetic testing for DILI due to lapatinib could be beneficial not only to prevent cases of DILI, but also to support causality assessments, earlier identification of high-risk individuals and subsequent stratification of liver chemistry monitoring [137].   Imputation of the SYT14 locus was undertaken to statistically infer unobserved genotypes using known haplotypes of the European component of the 1000 Genomes Project as a reference population [237, 260]. The 1000 Genomes Project provides a higher resolution of the human genome sequence variation, compared to the HapMap [260]. Sung et al (2012) investigated the performance of imputation using 1000 Genomes Project reference panels (specifically European) and reported a higher imputation yield (i.e. more successful imputation) compared to the  101 HapMap Phase 2 data, with high accuracy [260]. Employing imputation can be used to uncover causal or additional variants located within a region previously identified from a GWAS [260]. We imputed the SYT14 region to uncover additional (or causal) variants and found three additional significant associations, however these became non-significant upon adjusting for rs2205986, similarly to other pharmacogenomic studies of ADRs [261, 262]. Interestingly, all four variants (including rs2205986) were located within the same 257-kb haplotype block. Expanding the number of SNPs tested past a traditional GWAS (to the whole genome sequencing level) revealed no additional variants associated with lapatinib (a chemotherapeutic agent)-induced liver injury [138] and may highlight the utility of performing a GWAS alone, as opposed to the much higher cost (and additional risk of false positives) of whole genome sequencing.  4.4.2 Clinical and demographic differences between patient groups The clinical and demographic attributes of the cases and controls were relatively similar in both the analytical stages. Approximately 80% of all cases were female, as expected based on other IFN-β-treated MS populations [124, 263], with others reporting 90% of severe hepatic dysfunction in IFN-β-treated MS patients being female [106]. Over 90% of the discovery and replication patient cohorts were within the relapsing-remitting disease course at the start of therapy, as expected [124]. Cases in the replication stage were significantly older at IFN-β start than controls (P = 0.026); a finding previously reported for IFN-β induced liver injury [106]. Increasing age is considered a risk factor for DILI in other patient populations for unknown reasons [92, 264]. A moderately higher proportion of cases in the combined cohort were exposed  102 to IFN-β-1a 44 mcg than controls, concurring with previously studies of IFN-β associated liver aminotransferase elevations [6, 106, 107].   Non-alcoholic fatty liver disease and obesity are risk factors for DILI in other patient populations [129]. A higher number of our cases were in the overweight-obese category (BMI > 25 mg/m2) relative to controls (61.5% vs. 53.7%, P > 0.05)). However, BMI suffers from the inability to distinguish between lean body mass and body fat, and may overestimate the obesity risk as those with higher lean body mass would also have a high BMI [265]. Nonetheless BMI is a commonly used tool for estimating obesity, in part owing to the ease of measurement. The median BMI was similar between cases and controls (combined cohort - cases: 26.6 kg/m2 vs. controls: 25.6 kg/m2, P > 0.05), which were comparable to a 2015 study where the mean BMI for MS patients starting IFN-β therapy was 25.7 kg/m2 (+/- standard deviation: 0.47) [266]. In addition, we required a normal baseline measurement of ALT/AST prior to IFN-β initiation, which would have minimized the confounding effects of pre-existing fatty liver disease on increasing risk of DILI.  More than 90% of the patient cohort (cases and controls) reported the use of a concomitant medication. The use of concomitant hepatotoxic medications was higher in cases than controls (cases: 83.9% vs. controls: 77%), albeit non-significant (P > 0.05). A previous study from 2003 of IFN-β associated liver injury did not indicate an increased risk of hepatic injury with the use of acetaminophen, nor of any other concomitant medication (P > 0.05) [106]. From the broader DILI literature, the use of concomitant medications may increase ones’ susceptibility to DILI, but most often, a single concurrent medication is unlikely to be identified [85].  103     Figure 4.11 Schematic depiction of the postulated mechanism of rs2205986 variant in a genetic predisposition to interferon (IFN)-β induced liver injury.  Left panel: rs2205968 non-carrier and right panel: rs2205986 variant carrier. IFNAR1/2: Interferon receptor 1 and 2, TYK2: Tyrosine kinase 2, JAK1: Janus kinase 1, STAT1/2: Signal transducers and Activators of Transcription 1 and 2, IRF9/6: Interferon regulatory factor 9 or 6, ISGF3: Interferon-stimulated gene factor 3.  IFNAR1 IFNAR2 IFNβ JAK1 TYK2 STAT1 or STAT2 PSTAT1 PSTAT2 PSTAT1 PSTAT2 PIRF9 or IRF6? ISGF3  STAT1 STAT2 PPISGF3  IFNβ IFNβ Nucleus rs2205986 non-carrier Liver uninjured IRF9 or IRF6? IRF9 or IRF6? Normal induction of IFN-β pathway IFNAR1 IFNAR2 IFNβ JAK1 TYK2 STAT1 or STAT2 PSTAT1 STAT2 PSTAT1 STAT2 PIRF9 or IRF6? IRF9 or IRF6? ISGF3  STAT1 STAT2 PIRF9 or IRF6? PISGF3  IFNβ IFNβ Nucleus Cell lysis / apoptosis Nuclear fragmentation and over induction of IFN-β pathway? rs2205986 carrier Injured liver IRF9 or IRF6? IRF9 or IRF6? IRF9 or IRF6? IRF6 overexpression PP 104 Participants of this study were recruited from a range of sources and time periods and used different collection methods, to maximize efficiency. Understanding these differences is important when extrapolating these findings to a wider context for generalizability. While the majority of patients from both stages were identified as a result of a retrospective chart review, subsets of cases in the replication stage were identified from ADR surveillance networks. Being so, some differences in clinical characteristics of cases between the discovery and replication stage are to be expected and may play a role in the lack of replication of some variants. Cases in the replication cohort had increased DILI severity scores, significantly longer time to DILI onset, higher peak absolute ALT and ALP levels and higher maximum ALT x ULN than cases in the discovery (P < 0.05).   4.4.3 Ancestral differences between cohorts The discovery cohort was fully genotyped by our research centre and thus population stratification was assessed using PCA and the genomic inflation factor. The lack of significant population stratification in the discovery was indicated by a genomic inflation factor of 1.0334 and no significant differences in the top 10 principal components. MS is known to be most prevalent in people of Northern European ancestry [29], with our patients also of self-reported and genetically determined European ancestry, thereby limiting the degree of population stratification.   4.4.4 Non-replicating associations We had ≥ 85% power (post-hoc calculation) to detect genome-wide associations at P < 1 x 10-5 with per-allele OR ≥ 8 and a MAF of ≥ 0.10 in the discovery stage (n = 151 patients). There are  105 potentially additional genetic variants that contribute to modifying the risk of IFN-β induced liver injury discovered here, such as variants within FLRT2 (intronic, rs72693229) and an additional SYT14 variants (e.g., missense, rs200839898). Owing to data restrictions in the replication cohort, these SNPs were not investigated and should be examined in future studies. An intronic variant in ETV7 (rs62403705, OR = 4.12, P = 9.6 x 10-6) and an intergenic variant (rs17115661, OR = 5.25, P = 4.4 x 10-6) were both associated with IFN-β induced liver injury in the discovery stage, but did not replicate (P > 0.05). Genetic variants identified during GWA studies require replication to minimize false positive associations, with only robust findings replicating. In our replication cohort (post-hoc power calculation), we had 80% power to detect genetic associations at P < 0.05 with per-allele OR ≥ 6 and a MAF of 0.15. Our study suffers from a small sample size in the replication cohort and possible clinical differences, as we were only powered to detect a per-allele OR ≥ 6 with a MAF of 0.15. The two non-replicating variants are relatively common in the 1000 Genomes European population (rs62403705: 38%, rs17115661: 32%) and in our replication cohort, [267] making it unlikely to replicate these associations with the sample size employed.   We did not identify any significant candidate gene variants associated with case/control status in the discovery stage. The candidate gene variants were selected either because they were previously associated with DILI or with IFN-β response in MS. There were two SNPs with a significance threshold close to an uncorrected p-value of 0.05. The most significant variant from the discovery stage was located in the WNT7A gene, which is a member of the Wnt signaling pathway, and is expressed in the liver and contributes to liver regeneration [268]. This SNP was originally associated with ximelagatran (an anticoagulant) induced ALT elevations >3x ULN  106 (OR = 0.37, P = 2.5 x 10-4) in a cohort of 74 cases and 130 ximelagatran treated controls, but did not replicate in 10 cases and 16 ximelagatran treated controls [133]. Patients with this protective variant could be at a lower risk for IFN-β induced liver injury as they may be able to adequately regenerate damaged liver cells, preventing injury [268]. The second significant association (discovery stage) from the candidate variant analysis is an intronic variant located in the gene coding for IFN-γ (IFNG), a pro-inflammatory cytokine. Compared to untreated MS patients, IFN-β-1b exposure in MS is associated with lower IFN-γ serum levels in an attempt to shift from pro-inflammatory to anti-inflammatory conditions [269, 270]. Mice expressing a genetic mutant associated with an increased risk of flucloxacillin (antibiotic)-induced liver injury were found to have higher IFN-γ levels compared to wild-type mice [271]. In the non-MS literature, IFNG genetic variation is capable of modulating IFN-γ expression levels [272]. In a 2008 GWAS, a marginal difference was detected between this SNP and IFN-β responders / non-responders (P = 0.046) [74]. This variant could be associated with an increased level of IFN-γ leading to a higher risk of IFN-β induced liver injury. Neither of these genetic variants reached the Bonferroni p-value threshold in the discovery stage, likely due to the study being underpowered to detect common variants with small effect sizes [148].  4.4.5 Ability of rs2205986 to predict IFN-β induced liver injury In evaluating genetic variants associated with IFN-β induced liver injury, the determination of rs2205986 allele status of patients will help inform the clinical decision making for those exposed to IFN-β therapy (rs2205986 variant OR: 9.83, 95%CI: 4.01-24.1). This marker had a high specificity (93.7%), with a patient positive for rs2205986 being approximately ten times more likely to develop DILI from IFN-β. Notably, the only patient who required a liver  107 transplant in this study was an rs2205986 carrier. The specificity is also known as the rate of true negatives, meaning the high specificity (93.7%) generated here indicates a high proportion of negatives (i.e. patients who test negative and do not develop DILI) are correctly identified as such. We have shown that the incorporation of a genetic marker (rs2205986) to the clinical model significantly improves the prediction of this ADR (P = 0.0039). It should be noted the predictive performance of the model is likely overestimated since its performance was determined using the subjects initially used to construct the model, thus internal and external validation is necessary. The small sample size of the replication cohort prohibits an internal validation. Nonetheless, MS patients who are rs2205986 carriers are highly likely to develop DILI from IFN-β and thus may warrant a modified dose of IFN-β therapy or an alternative non-IFN-β based first-line MS therapy with similar efficacy, such as glatiramer acetate [273]. Although adverse liver effects are not thought common with glatiramer acetate, such that routine liver testing is not required, case reports of glatiramer acetate associated liver injury have been reported [274, 275].  4.4.6 Autoimmune related DILI Drugs causing DILI may also trigger an immune response to hepatic proteins, resulting in features of autoimmune hepatitis (AIH) [276]. DILI occurring with features of AIH is typically demonstrated by the presence of serum auto-AB, hepatocellular injury and immune mediated histological features on liver biopsy [277]. Although only exploratory, ANA and hepatocellular injury pattern was the most frequent in auto-AB positive patients, similar to the wider DILI literature [277]. Notably, a higher proportion of auto-AB positive cases were more severe, which also concurs with current drug induced AIH literature [278]. A genetic predisposition in those  108 with non-drug induced AIH has been previously investigated [279], but has not been investigated in those with drug induced AIH. Our preliminary analysis showed half of the auto-AB positive cases were rs2205986 carriers (P = 0.121), however the low number of patients (8 cases, 16 controls) clearly limits our ability to define this association. Of note, a liver biopsy is needed to definitively diagnose drug induced AIH [277], however, biopsies were not clinically indicated in any of the auto-AB patients included, particularly as liver aminotransferase levels normalized in most cases, upon stopping drug therapy.   4.4.7 Future directions Here we describe a novel variant associated with increasing the risk of IFN-β induced liver injury. Future studies may encompass a larger cohort of people with MS, who have experienced this ADR, as well as drug tolerant controls to further replicate these findings and provide an external validation of its diagnostic prediction. In addition, the investigation of rs2205986 in non-European ancestral populations experiencing the ADR could be of value since MS disease is also prevalent in other ethnicities [280], and would assist with the generalizability of this genetic marker in predicting liver injury due to IFN-β. As not all DILI cases were rs2205986 carriers, there remains unexplained genetic variability and future investigations may encompass only extreme (i.e. liver transplantation or fatal) cases to identify different variants (with an appropriate sample size) or variants that are known to have pathogenic effects. Moreover, whole exome sequencing may also prove useful to uncover rare, functionally relevant variants; with the additional use of whole genome sequencing to identify supplementary novel variants.    109 There are likely other clinical, demographic or environmental factors contributing to experiencing this ADR in MS, that our small sample size prohibited us from investigating. Therefore, employing a larger number of patients in the future could assist with detecting significant differences in the clinical and demographic characteristics of patients. Employing validated scales for investigating confounders to delineating their role, if any, in IFN-β induced liver injury might be advantageous. As an example, the modified Skinner Alcohol Dependence Scale would increase the accuracy of alcohol consumption reporting [281] and is used by the DILIN [282]. Collecting information on the use of concomitant medications is difficult via retrospective chart review, and over 90% of patients included here reported use of a concomitant medication. As an example, in British Columbia, future work could query patients for their consent to link their Personal Health Numbers (PHN) to PharmaNet (all pharmacy dispensed prescriptions in the province of British Columbia) to confirm the use of any concomitant medications alongside IFN-β. For example, antibiotics are a leading cause of DILI [283], but owing to their typically short course of therapy and initiation in the community, it is possible they are poorly recorded in the specialist medical records (e.g. MS clinic or hospital derived) accessed for this study.    As a statistically significant pharmacogenomic association between the ADR and a genetic variant was identified here, even after adjustment for clinical factors; functional validation will help to further delineate the mechanistic basis of rs2205986. Using in vitro functional assays, one could characterize the role of SYT14 and IRF6 in IFN-β induced liver injury, using tissue-specific cell lines, e.g. HepG2 cells for DILI. This could then be extended to animal models to complement the in vitro systems and assess the contribution of specific variants within the  110 context of the whole body. Unfortunately, in vivo modeling of DILI has been poor owing to the lack of proper models, although some studies exist for flucloxacillin and amiodarone DILI in BALB/cCrSlc mice [144, 145]. Zebrafish as a novel high-throughput in vivo model for studying DILI has been proposed because of its histological similarities with DILI in humans [146].   4.5 Conclusions The identification of rs2205986, which was genetically linked to experiencing IFN-β induced liver injury and was correlated with liver-specific IRF6 expression levels, provides the first ever investigation into the genetic determinants of this ADR. IFN-β induced liver injury is a clinically significant ADR, often resulting in cessation of a potentially useful treatment. Carriers of rs2205986 have ten-fold increased odds of experiencing liver injury due to IFN-β. This highly significant finding will provide a substantial reduction in the toxicity of IFN-β treatment, and allow for its safer use in MS.    111 Chapter 5: Overall Discussion 5.1 Summary of findings Gaining a better understanding of the post-market safety profile of a therapy remains an unmet need across many clinical areas, including MS. Identifying those at a modified risk of experiencing an adverse reaction to an MS drug therapy would be beneficial to patients and clinicians. However, the characterization of ADRs and subsequent identification of predictive markers face many obstacles and challenges, such as low patient recruitment, selection of an appropriate genomic panel (in particular when the drugs mechanism of action is ambiguous) and strict phenotyping (also leading to lower patient numbers). The overall objective of this dissertation was to address the gaps in understanding IFN-β induced liver injury in MS, with an overall aim of discovering clinical and genetic markers that predict its occurrence.  Multiple sclerosis patients occupying the risk variant, rs2205986, had a nearly ten-fold (OR: 9.83, 95%CI: 4.01-24.10) increased risk of experiencing IFN-β induced liver injury compared to non-carriers, even when adjusting for significant clinical (or demographic) variables. A strong genetic predisposition for DILI to a range of drugs and drug classes has been shown (Table A.10). However, to date no predictive tests for DILI have been developed, primarily due to the rare occurrence of DILI [284]. For example, one would need to screen 13,513 people exposed to flucloxacillin (an antibiotic) in order to prevent one case of flucloxacillin DILI [284]. Although in another example, if the DILI is more common, as was the case with ximelagatran (a thrombin inhibitor) induced liver injury (7.9% of those exposed to ximelagatran experience DILI vs. only 0.0085% from flucloxacillin), only 24 people would require screening to prevent one case of ximelagatran DILI [284]. However, ximelagatran was withdrawn from the market due to  112 hepatotoxicity concerns, hence precluding any future prospective studies. From our population-based study in Chapter 3, we have shown ~2% of IFN-β exposed experienced DILI, meaning a predictive test might be more feasible for this ADR, as it is approximately 200 times more common than DILI due to flucloxacillin. Additionally, the genetic marker identified here had a high specificity (93.7%); meaning rs2205986 carriers are very likely to develop DILI from IFN-β and may warrant a different course of therapy or modified dose.  The occurrence of IFN-β associated liver injury, as described in Chapter 3, revealed approximately 1 in 50 IFN-β exposed patients developed DILI in British Columbia, Canada. Investigating the rate of a rare ADR is difficult owing to the small number of cases, especially when a stringent cut-off such as ALT ≥ 5x ULN is employed as the case definition. Other estimates of DILI due to IFN-β have been reported to be between 0.3-3.1% and varies depending on the IFN-β product [106, 107]. The advantage our estimate has over these previous studies is the requirement of a stringent cut-off for indicating liver injury (ALT ≥ 5x ULN), the application of an internationally accepted case definition of DILI [88] and also the use of a population-based cohort. The possibility remains that our rate evaluation of DILI from IFN-β may have been overestimated as increasing ALT testing frequency is known to play a role in increasing the chance of detecting an abnormal ALT result in MS [110]. However, applying the cut-off of ≥ 5x ULN should mitigate the inclusion of false positive cases of liver injury [206].   Chapters 3 and 4 showed some interesting trends with respect to the clinical factors of IFN-β induced liver injury and may assist in informing clinicians and patients as to who might be at an altered risk of experiencing this ADR (ALT ≥ 5x ULN). Chapter 3 showed an increased risk for  113 liver injury due to IFN-β therapy for women (adjHR: 3.15; 95% CI: 0.72-13.72, P = 0.13). While in Chapter 4, using a different study design and patients from different countries, but the same case definition, a moderately increased proportion of cases were male (19.6% of cases vs. 8.7% of controls from the combined cohort: P = 0.051). This may be due to the comparably smaller sample size in Chapter 4 (n = 182 patients vs. n = 942 patients for Chapter 3) or to geographical differences, as Chapter 3 was primarily comprised of British Columbian MS patients, and Chapter 4 had worldwide representation of participants. Finally, there were differences in study design, as Chapter 3 was a mixed methods study including a population-based component, whereas Chapter 4 involved a case-control design. Also, there may have been sex-differences in those recruited for the Chapter 4 study (compared to Chapter 3 individuals). For example, men are less likely to participate in personalized genomics research [285] and this may have led to a smaller number of male controls than Chapter 3, which was an epidemiological study. An additional trend detected in Chapter 3 was an increased risk for those exposed to IFN-β-1a (44 mcg) (adjHR: 6.26, 95% CI: 0.78-50.39, P = 0.08) or IFN-β-1b (adjHR; 3.71, 95%CI: 0.47-29.59, P = 0.22), compared to IFN-β-1a (22 mcg). Findings were comparable in Chapter 4, with a higher proportion of cases exposed to the higher dosed interferons [IFN-β-1b and IFN-β-1a (44 mcg) relative to IFN-β-1a (22 mcg)]. Similar findings have been previously demonstrated [106], including a study using a smaller sample of the BC MS cohort [107]. The rationale behind an increased DILI risk for those exposed to the higher dosed interferons may be due to insufficient recovery of hepatocytes prior to the next injection, as increasing the frequency of doses, but not the total dose increases the risk of elevated aminotransferases in people with hepatitis C exposed to IFN-β [121].    114 Chapters 3 and 4 highlight the obstacles and advantages of identifying DILI cases from various sources. One hindrance was the presence of clinical and demographic differences between cases identified from Canada vs. USA / Sweden. In Chapter 3 (clinical/demographic study), we demonstrated the median time to DILI was comparable between the Canadian cases (BC: 105 days and rest of Canada: 90 days), but significantly longer for the ADR network cases sourced from the USA and Sweden compared to those from BC (590 days, P = 0.006). Additional cases from the USA were compiled with the USA/Sweden cases from Chapter 3 to serve as the replication cohort for Chapter 4. Collectively, Chapter 4 replication cases (i.e. USA/Sweden) were more severe (indicated by higher ALT elevations and severity scores), compared to the cases in the discovery stage (Canadian patients). Exploratory analyses demonstrated a higher proportion of cases who were rs2205986 carriers in the discovery compared to the replication (44.7% vs. 33.3%, P = 0.415), indicating additional or different genetic variants may be associated with more severe DILI, although studies with a larger sample size are needed. Clinical and demographic differences in cases from various locations were apparent, nonetheless, national and international collaboration were necessary to achieve a sufficient number of participants. The research presented in this dissertation highlights the necessity and advantages of establishing a pan-Canadian (or Global, if replication is required) to investigating rare ADRs. There have been many successful examples of networks around the world attempting to improve drug safety and effectiveness due to non-MS therapies utilizing genetics [164-167]. Along with these previous established networks and the research presented in this dissertation, further work into establishing global networks to unveil genetic and clinical predictive markers of severe ADRs is proving to be a meaningful endeavor in enhancing drug safety.   115 The studies presented in Chapters 2 and 4 highlight the possibility of a biological agent such as IFN-β, in potentially inducing liver autoimmunity in MS. It should be noted auto-AB testing was completed after the liver enzyme elevations of ≥ 5x ULN; and thus it is unknown whether the auto-AB were pre-existing or if they were induced by IFN-β therapy. Additionally, this analysis was exploratory, owing to the small sample size (n = 24). Nonetheless, the use of IFN-β in MS has been previously associated with AIH [113, 182, 184], with a publication describing two MS patients exposed to IFN-β with subsequent AIH, who were auto-AB negative prior to IFN-β therapy [182]. A prospective, multi-center study reported approximately 4% of MS patients developed de novo auto-AB positivity during IFN-β-1b therapy (i.e. these individuals were auto-AB negative prior to therapy) [113]. The definitive diagnosis of AIH alongside liver enzyme elevations is highly valuable, since a different course of therapy is warranted if AIH is also present (primarily oral corticosteroids, as opposed to ursodiol for persistently elevated aminotransferases) [189]. IFN-β exposure is hypothesized to promote the development of auto-AB via the induction of the BAFF (B-cell activating factor of the tumor necrosis factor family)/APRIL (a proliferation-inducing ligand) system [286]. Only half of the auto-AB positive cases were rs2205986 carriers but this may be due to the limited sample size of auto-AB positive cases (n = 8) or that these cases have different genetic markers associated with their occurrence.  5.2 Strengths and limitations This study has a number of strengths including, to the best of our knowledge, the first GWAS to investigate an ADR associated with an MS therapy and the first GWAS to investigate DILI due to a biological agent. Second, our patient cohorts were selected to minimize heterogeneity. For example, we applied a standardized case definition [88], since ensuring an accurate phenotype  116 plays a large role in determining the genetic determinants of an ADR [178]. As per the Common Terminology Criteria for Adverse Events (v4.03), the case definition utilized in Chapters 3 and 4 represents a severe or medically significant and potentially disabling event [203]. The requirement of a normal baseline liver aminotransferase also minimized the possibility of including individuals with a pre-existing liver comorbidity. In addition, we employed drug-tolerant controls (as opposed to population controls), since they provide the greatest power and lowest confounding in GWAS studies of ADRs [287]. The patients used to replicate the genomic marker in Chapter 4 were selected to minimize heterogeneity. The replication cohort was drawn from individuals with genetically ascertained European ancestry with MS; all patients met the same inclusion criteria for study entry and also met the same case and drug-tolerant control definitions (apart from cases recruited from DILIN where two consecutive ALT elevations were required [282]). Lastly, we capitalized on two successful and established pan-Canadian networks, the CPNDS [164] and the Canadian Network of MS Clinics [288]. This enabled us to achieve the study results and recruit a sufficient number of patients in a timely and cost-effective manner, while also establishing a novel study within both of these networks.   Limitations of our study include potential issues arising from the use of a case-control study design, such as selection bias (non-random sampling of the population) [289]. Since we have recruited cases and controls representative of the same population (all MS patients exposed to IFN-β), we have attempted to minimize this bias. Another potential limitation relates to the assessment of causality, as this is known to be difficult when retrospective collection of data is performed [101]. However, we employed several tactics to combat this, such as a requirement of a normal baseline liver aminotransferase test, a severe elevation of ALT (≥ 5x ULN) and the  117 application of two different causality scales. It is possible that controls could develop IFN-β induced liver injury in the future, although by selecting controls exposed to IFN-β for ≥ 2 years with no biochemical liver test elevations, we have limited this. This was shown in Chapter 3, whereby all cases of DILI occurred within less than 2 years, with the greatest odds of developing liver enzyme elevations (≥ 2x ULN) known to be within the first 15 months of IFN-β treatment [107]. The case definition of DILI used here depends predominantly on ALT elevations and although its use as a biomarker for DILI has limitations. For instance, elevations in ALT appear after injury has already occurred and represent markers of potential liver injury and not actual liver function. Nonetheless testing for ALT levels is considered the gold standard, alongside levels of total bilirubin (also collected, when possible) [206].   A limitation in Chapters 3 and 4 were the sample sizes of the studies. No significant differences in the clinical and demographic factors were detected in Chapter 3, only trends for those at greater risk for DILI from IFN-β. This is not surprising with only 18 cases of DILI due to IFN-β included in Chapter 3 (Clinical Study). Post-hoc power calculations demonstrated the minimum significantly different hazard ratio we could have detected was > 12 (80% power, using a two-tailed test with 5% probability of a type-I error, i.e. α = 0.05). To detect a hazard ratio > 5, we would have required 158 DILI cases, which would have been prohibitive from one site and would have required world-wide collaboration, as recruiting from five MS clinics in Canada resulted in only 38 cases of DILI from IFN-β (Chapter 4 – Genetics Study). In Chapter 4, perhaps as a consequence of the small size of the replication cohort, we were only able to replicate one of three genetic variants identified in the discovery stage (rs2205986). Although it is possible the other two variants were false positive associations or that we simply did not have  118 the power to detect these more common alleles in such a small replication cohort (n = 31 patients). Additionally, there were two other significant genetic associations in the discovery cohort, including a SYT14 missense variant. However, we were limited by the genetic data provided for the controls and some of the cases in the replication cohort as these two variants were neither genotyped nor imputed in this cohort and hence we could not assess their effects. Another limitation noted in Chapter 4 is the lack of genetically ascertained non-European patients to assess for replication of the genetic variants to enhance generalizability of the findings.   5.3 Significance and implications of research This research is relevant to the health of people with MS as the identified risk factors may help reduce the incidence of serious liver toxicity and improve outcomes for MS patients who receive IFN-β. The rs2205986 genomic variant contributing to this serious ADR should be used as predictive test, prior to administration of IFN-β, to identify rs2205986 carriers whom from the research undertaken during this dissertation, were shown to be at a high risk of liver injury from IFN-β. This test will direct preventative strategies towards rs2205986 carriers, including enhanced monitoring for early signs of liver toxicity, dose reductions, or alternative treatment regimens to IFN-β. The availability of a diagnostic test to assist in the selection of medications for individual patients will provide a vital, cost-effective tool that in rare cases could be life saving. The identification of SYT14 / IRF6 provides new insights into the mechanism underlying IFN-β induced liver injury, and in the future could lead to the development of preventative measures or better treatment options. Additionally, the implication of IRF6 / SYT14 in liver  119 injury due to IFN-β provides new acumen into the underlying mechanism of DILI, in general to other therapies, as neither have been previously associated with DILI.  The consequences for an MS patient experiencing liver toxicity due to IFN-β could include considerable anxiety, as well as symptoms of the liver injury, such as fever and jaundice. Other negative outcomes that are more rare can be catastrophic, such as liver transplant or liver failure. For many, it leads to the cessation of a potentially beneficial drug, restricted choice of future MS treatment, dose reduction, treatment delays, additional monitoring, tests and physician visits. The costs of managing IFN-β induced liver toxicity can be considerable, in part due to the continued monitoring required. The cost of liver transplant in Canada, including follow up, are estimated at $121,732, with a mean length of hospital stay at 43 days [290], although the occurrence (risk) of liver transplant from IFN-β induced liver toxicity is unknown and likely extremely rare. An anecdote from an MS patient not included in this study, who experienced IFN-β associated liver injury and was subsequently placed on natalizumab and then fingolimod, stated “Because it (the liver injury from IFN-β) was asymptomatic, the main reason liver toxicity was troublesome and stressful for me was because it meant I had finished with the front-line RRMS treatments; as you progress down the treatment line, the side effects tend to get more severe, and, in the case of Gilenya (fingolimod), less well-known. And I was also mad because I had abstained from alcohol and tylenol for my entire course of Betaseron, and I still had liver toxicity! Gilenya has only been on the market for 10 years so the side effects may not all be well known yet - and I don't like being a guinea pig.  I also had to come off the birth control pill to reduce strain on my liver” (Personal conversation, June 2015).    120 DILI in MS patients is not specific to IFN-β treatment, as liver injury occurring from the newer MS therapies (e.g. natalizumab, fingolimod, dimethyl fumarate, teriflunomide and alemtuzumab) has also been reported [40, 194-197]. As such, findings from this dissertation could help guide the search for genetic variation attributable to liver injury from these newer MS immunomodulatory therapies. However, while there is evidence to suggest DILI due to different small molecule drugs may share genetic risk factors [219-221], we were unable to find any literature on whether DILI due to biological therapies share common genetic risk factors. Additionally, patients who experienced DILI due to IFN-β presented in this dissertation were likely to stop therapy. Presumable, they may be placed on a second-line or alternative therapy, as clinically warranted. Whether these cases are more prone to experiencing a secondary DILI from a different therapy is unknown. In the non-MS literature, multiple DILI occurrences in the same patient from different drugs are possible, but fortunately infrequent [221].  During the course of work presented in this dissertation, two noticeable hurdles in research methodology were apparent. First, the lack of blood work (biochemical liver enzyme) results located in electronic medical records in the majority of Canadian MS clinics presented an issue as it increased the time necessary to ascertain participants for Chapter 4, as a thorough review of paper health records was subsequently needed. Increasing the use of electronic medical records and also linking these electronic records to other research sites and/or databases may improve pharmacogenomic studies by enhancing standardized reporting of ADRs and amassing larger cohort sizes [291]. Another issue during this research was related to the lack of a centralized research ethics approval board in Canada. Six separate ethics applications were submitted over the course of this research, taking approximately 24 months to acquire all approvals. A 2015  121 study proposed the use of centralized ethics boards would increase participation in large multi-site studies and would facilitate clinical innovation [292]. The British Columbia Ethics Harmonization Initiative [293] is developing a centralized ethics approval process for the province of B.C., with also the United States Food and Drug Administration recommending the creation of a single ethical review board to facilitate clinical trials [294].  5.4 Future work It would be useful to expand these investigations to include functional validation of the associated genetic variant discovered during this study and ultimately a therapeutically useful tool, as previously mentioned in Chapter 4. An evaluation of the cost-effectiveness of this genomic biomarker in preventing cases of DILI due to IFN-β would also be a useful and critical step in its development and to generate discussion on test reimbursement by public and private payers. Finally, the framework developed during this course of research could be expanded in the future to include ADRs associated with other and newer MS therapies, such as natalizumab or dimethyl fumarate and progressive multifocal leukoencephalopathy or cardiotoxicity and fingolimod.   5.5 Conclusions The work presented in this dissertation represents a major step in the pharmacogenomic investigation of ADRs in MS. This work was partly motivated by a case report of a patient experiencing the unmasking of an autoimmune disorder targeting the liver, following IFN-β exposure, revealing a complication of this first-line therapy for MS [204]. Next, we published a review article summarizing the potential for pharmacogenomics to predict and prevent severe  122 adverse drug reactions in MS and identified the requirements necessary for a genomic investigative study [295]. The subsequent retrospective, mixed methods study was the first to apply an internationally recommended case definition for DILI to identify and characterize DILI associated with IFN-β in MS on the basis of clinical factors [296]. Lastly, the first ever GWAS of an MS ADR resulted in the discovery of a highly specific genetic variant conferring a clinically significant ten-fold increased risk of liver injury from IFN-β. Overall, we demonstrated specific clinical, demographic and genomic characteristics increase the risk of experiencing this severe, sometimes fatal ADR. The results of this study contribute to the overall goal of improving drug safety for MS patients and specifically, alleviating the risk of liver injury due to IFN-β in MS.   123  References  1. Pirmohamed, M., et al., Adverse drug reactions. BMJ, 1998. 316(7140): p. 1295-8. 2. Lazarou, J., B.H. Pomeranz, and P.N. Corey, Incidence of adverse drug reactions in hospitalized patients: a meta-analysis of prospective studies. JAMA, 1998. 279(15): p. 1200-5. 3. Pirmohamed, M., et al., Adverse drug reactions as cause of admission to hospital: prospective analysis of 18 820 patients. BMJ, 2004. 329(7456): p. 15-9. 4. Mittmann, N., et al., Evaluation of the extent of under-reporting of serious adverse drug reactions: the case of toxic epidermal necrolysis. Drug Saf, 2004. 27(7): p. 477-87. 5. US General Accounting Office. FDA Drug Review: Postapproval Risks, 1976-85. 1990; Available from: http://161.203.16.4/d24t8/141456.pdf. 6. Tremlett, H.L., E.M. Yoshida, and J. Oger, Liver injury associated with the beta-interferons for MS: a comparison between the three products. Neurology, 2004. 62(4): p. 628-31. 7. Ajayi, F.O., H. Sun, and J. Perry, Adverse drug reactions: a review of relevant factors. J Clin Pharmacol, 2000. 40(10): p. 1093-101. 8. Aronson, J.K. and R.E. Ferner, Clarification of terminology in drug safety. Drug Saf, 2005. 28(10): p. 851-70. 9. Motulsky, A.G., Drug reactions enzymes, and biochemical genetics. J Am Med Assoc, 1957. 165(7): p. 835-7. 10. Gonzalez, F.J., et al., Characterization of the common genetic defect in humans deficient in debrisoquine metabolism. Nature, 1988. 331(6155): p. 442-6. 11. Lander, E.S., et al., Initial sequencing and analysis of the human genome. Nature, 2001. 409(6822): p. 860-921. 12. Venter, J.C., et al., The sequence of the human genome. Science, 2001. 291(5507): p. 1304-51. 13. International HapMap, C., The International HapMap Project. Nature, 2003. 426(6968): p. 789-96. 14. Lennard, L., et al., Thiopurine pharmacogenetics in leukemia: correlation of erythrocyte thiopurine methyltransferase activity and 6-thioguanine nucleotide concentrations. Clin Pharmacol Ther, 1987. 41(1): p. 18-25.  124 15. Koren, G., et al., Pharmacogenetics of morphine poisoning in a breastfed neonate of a codeine-prescribed mother. Lancet, 2006. 368(9536): p. 704. 16. Pirmohamed, M., Pharmacogenetics: past, present and future. Drug Discov Today, 2011. 16(19-20): p. 852-61. 17. Yang, J., et al., Contribution of 1173C > T polymorphism in the VKORC1 gene to warfarin dose requirements in Han Chinese patients receiving anticoagulation. Int J Clin Pharmacol Ther, 2011. 49(1): p. 23-9. 18. Herman, D., et al., Influence of CYP2C9 polymorphisms, demographic factors and concomitant drug therapy on warfarin metabolism and maintenance dose. Pharmacogenomics J, 2005. 5(3): p. 193-202. 19. Feero, W.G., A.E. Guttmacher, and F.S. Collins, Genomic medicine--an updated primer. N Engl J Med, 2010. 362(21): p. 2001-11. 20. Kidd, J.M., et al., Mapping and sequencing of structural variation from eight human genomes. Nature, 2008. 453(7191): p. 56-64. 21. Mallal, S., et al., HLA-B*5701 screening for hypersensitivity to abacavir. N Engl J Med, 2008. 358(6): p. 568-79. 22. PharmGKB. Drug label annotations. 2016 [Accessed 31 March 2016]; Available from: https://www.pharmgkb.org/view/drug-labels.do. 23. Bashir, N.S. and W.J. Ungar, The 3-I framework: a framework for developing public policies regarding pharmacogenomics (PGx) testing in Canada. Genome, 2015. 58(12): p. 527-40. 24. Centre for Addiction and Mental Health. IMPACT: Individualized Medicine: Pharmacogenetic Assessment & Clinical Treatment. 2016 [Accessed 4 February 2016]; Available from: http://impact.camhx.ca/en/home.php. 25. Shaw, K., et al., Clinical Practice Recommendations on Genetic Testing of CYP2C9 and VKORC1 Variants in Warfarin Therapy. Ther Drug Monit, 2015. 37(4): p. 428-36. 26. Amstutz, U., et al., Recommendations for HLA-B*15:02 and HLA-A*31:01 genetic testing to reduce the risk of carbamazepine-induced hypersensitivity reactions. Epilepsia, 2014. 55(4): p. 496-506. 27. Madadi, P., et al., Clinical practice guideline: CYP2D6 genotyping for safe and efficacious codeine therapy. J Popul Ther Clin Pharmacol, 2013. 20(3): p. e369-96. 28. Relling, M.V. and T.E. Klein, CPIC: Clinical Pharmacogenetics Implementation Consortium of the Pharmacogenomics Research Network. Clin Pharmacol Ther, 2011. 89(3): p. 464-7.  125 29. Compston, A., et al., McAlpine's Multiple Sclerosis. 2005: Churchill Livingstone. 30. World Health Organization; Multiple Sclerosis International Federation. Atlas: multiple sclerosis resources in the world. 2013. [Accessed 3 February 2014]; Available from: http://www.atlasofms.org/query.aspx. 31. Confavreux, C., G. Aimard, and M. Devic, Course and prognosis of multiple sclerosis assessed by the computerized data processing of 349 patients. Brain, 1980. 103(2): p. 281-300. 32. World Health Organization; Multiple Sclerosis International Federation. Atlas: multiple sclerosis resources in the world. 2008. [Accessed 28 October 2011]; Available from: http://www.who.int/mental_health/neurology/Atlas_MS_WEB.pdf. 33. Lublin, F.D., New multiple sclerosis phenotypic classification. Eur Neurol, 2014. 72 Suppl 1: p. 1-5. 34. Bayer HealthCare Pharmaceuticals Inc. Betaseron (Interferon Beta-1b) for SC Injection Drug Monograph. 2011. [Accessed 10 January 2012]; Available from: http://www.bayer.ca/files/BETASERON-PM-ENG-13JUL2011-133528.pdf. 35. Biogen Idec. Avonex (Interferon-beta-1a) IM injection drug monograph. 2012 [Accessed 15 March 2012]; Available from: http://www.avonex.com/pdfs/guides/Avonex_Prescribing_Information.pdf. 36. EMD Serono Inc. Rebif (Interferon-beta-1a) SC injection drug monograph. 2011 [Accessed 20 March 2012]; Available from: http://www.emdserono.com/cmg.emdserono_us/en/images/rebif_tcm115_19765.pdf. 37. Teva Pharmaceutical Industries Ltd. Copaxone (glatiramer acetate injection) drug monograph. 2011 [Accessed 20 March 2012]; Available from: http://tevacanadainnovation.com/downloads/Copaxone_Product_Monograph_EN.pdf. 38. Biogen Idec. Tysabri (Natalizumab) Drug Monograph. 2012. [Accessed 20 May 2012]; Available from: http://www.tysabrihcp.com/pdfs/I61061-13_PI.pdf. 39. EMD Serono Inc. Novantrone (mitoxantrone for injection concentrate) drug monograph. 2011 [Accessed 20 March 2012]; Available from: http://dailymed.nlm.nih.gov/dailymed/archives/fdaDrugInfo.cfm?archiveid=56726. 40. Novartis Pharmaceuticals. Gilenya (Fingolimod capsules) drug monograph. 2012 [Accessed 20 May 2012]; Available from: http://www.pharma.us.novartis.com/product/pi/pdf/gilenya.pdf. 41. Kieseier, B.C., The Mechanism of Action of Interferon-beta in Relapsing Multiple Sclerosis. Cns Drugs, 2011. 25(6): p. 491-502.  126 42. Jacobs, L., et al., Intrathecal interferon reduces exacerbations of multiple sclerosis. Science, 1981. 214(4524): p. 1026-8. 43. Mark, D.F., et al., Site-specific mutagenesis of the human fibroblast interferon gene. Proc Natl Acad Sci U S A, 1984. 81(18): p. 5662-6. 44. Chernajovsky, Y., et al., Efficient constitutive production of human fibroblast interferon by hamster cells transformed with the IFN-beta 1 gene fused to an SV40 early promoter. DNA, 1984. 3(4): p. 297-308. 45. Weinstock-Guttman, B., M. Ramanathan, and R. Zivadinov, Interferon-beta treatment for relapsing multiple sclerosis. Expert Opin Biol Ther, 2008. 8(9): p. 1435-47. 46. Salmon, P., et al., Pharmacokinetics and pharmacodynamics of recombinant human interferon-beta in healthy male volunteers. J Interferon Cytokine Res, 1996. 16(10): p. 759-64. 47. Munafo, A., et al., Comparative pharmacokinetics and pharmacodynamics of recombinant human interferon beta-1a after intramuscular and subcutaneous administration. Eur J Neurol, 1998. 5(2): p. 187-193. 48. Mager, D.E. and W.J. Jusko, Receptor-mediated pharmacokinetic/pharmacodynamic model of interferon-beta 1a in humans. Pharm Res, 2002. 19(10): p. 1537-43. 49. Runkel, L., et al., Structural and functional differences between glycosylated and non-glycosylated forms of human interferon-beta (IFN-beta). Pharm Res, 1998. 15(4): p. 641-9. 50. Antonetti, F., et al., A comparison of the biologic activity of two recombinant IFN-beta preparations used in the treatment of relapsing-remitting multiple sclerosis. J Interferon Cytokine Res, 2002. 22(12): p. 1181-4. 51. Hemmer, B., et al., Immune response to immunotherapy: the role of neutralising antibodies to interferon beta in the treatment of multiple sclerosis. Lancet Neurol, 2005. 4(7): p. 403-12. 52. Abraham, A.K., et al., Type I interferon receptor is a primary regulator of target-mediated drug disposition of interferon-beta in mice. J Pharmacol Exp Ther, 2010. 334(1): p. 327-32. 53. Lahoz, M., et al., Pharmacokinetics and pharmacodynamics of interferon beta 1a in Cebus apella. J Med Primatol, 2009. 38(3): p. 187-91. 54. Hartung, H.P., High-dose, high-frequency recombinant interferon beta-1a in the treatment of multiple sclerosis. Expert Opin Pharmacother, 2009. 10(2): p. 291-309.  127 55. Francis, G.S., G.P. Rice, and J.C. Alsop, Interferon beta-1a in MS: results following development of neutralizing antibodies in PRISMS. Neurology, 2005. 65(1): p. 48-55. 56. Bertolotto, A., Evaluation of the impact of neutralizing antibodies on IFNbeta response. Clin Chim Acta, 2015. 57. Ross, C., et al., Immunogenicity of interferon-beta in multiple sclerosis patients: influence of preparation, dosage, dose frequency, and route of administration. Danish Multiple Sclerosis Study Group. Ann Neurol, 2000. 48(5): p. 706-12. 58. Panitch, H., et al., Randomized, comparative study of interferon beta-1a treatment regimens in MS: The EVIDENCE Trial. Neurology, 2002. 59(10): p. 1496-506. 59. Santos, R., et al., Dynamics of interferon-beta modulated mRNA biomarkers in multiple sclerosis patients with anti-interferon-beta neutralizing antibodies. J Neuroimmunol, 2006. 176(1-2): p. 125-33. 60. Wandinger, K.P., et al., Diminished production of type-I interferons and interleukin-2 in patients with multiple sclerosis. J Neurol Sci, 1997. 149(1): p. 87-93. 61. Yamaguchi, K.D., et al., IFN-beta-regulated genes show abnormal expression in therapy-naive relapsing-remitting MS mononuclear cells: gene expression analysis employing all reported protein-protein interactions. J Neuroimmunol, 2008. 195(1-2): p. 116-20. 62. Hundeshagen, A., et al., Elevated type I interferon-like activity in a subset of multiple sclerosis patients: molecular basis and clinical relevance. J Neuroinflammation, 2012. 9: p. 140. 63. Kingwell, E., et al., Cancer risk in multiple sclerosis: findings from British Columbia, Canada. Brain, 2012. 64. Wingerchuk, D.M., Multiple sclerosis disease-modifying therapies: adverse effect surveillance and management. Expert Rev Neurother, 2006. 6(3): p. 333-46. 65. Khan, O.A. and J.R. Hebel, Incidence of exacerbations in the first 90 days of treatment with recombinant human interferon beta-1b in patients with relapsing-remitting multiple sclerosis. Ann Neurol, 1998. 44(1): p. 138-9. 66. Interferon beta-1b is effective in relapsing-remitting multiple sclerosis. I. Clinical results of a multicenter, randomized, double-blind, placebo-controlled trial. The IFNB Multiple Sclerosis Study Group. Neurology, 1993. 43(4): p. 655-61. 67. Feinstein, A., Multiple sclerosis, disease modifying treatments and depression: a critical methodological review. Mult Scler, 2000. 6(5): p. 343-8.  128 68. Durelli, L., et al., Every-other-day interferon beta-1b versus once-weekly interferon beta-1a for multiple sclerosis: results of a 2-year prospective randomised multicentre study (INCOMIN). Lancet, 2002. 359(9316): p. 1453-60. 69. La Mantia, L., et al., Interferon beta for secondary progressive multiple sclerosis. Cochrane Database Syst Rev, 2012. 1: p. CD005181. 70. Rice, G.P., et al., Interferon in relapsing-remitting multiple sclerosis. Cochrane Database Syst Rev, 2001(4): p. CD002002. 71. Sriram, U., et al., Pharmacogenomic analysis of interferon receptor polymorphisms in multiple sclerosis. Genes Immun, 2003. 4(2): p. 147-52. 72. Mahurkar, S., V. Suppiah, and C. O'Doherty, Pharmacogenomics of interferon beta and glatiramer acetate response: a review of the literature. Autoimmun Rev, 2014. 13(2): p. 178-86. 73. Weinstock-Guttman, B., et al., Pharmacogenetics of MXA SNPs in interferon-beta treated multiple sclerosis patients. J Neuroimmunol, 2007. 182(1-2): p. 236-9. 74. Byun, E., et al., Genome-wide pharmacogenomic analysis of the response to interferon beta therapy in multiple sclerosis. Arch Neurol, 2008. 65(3): p. 337-44. 75. Comabella, M., et al., Genome-wide scan of 500,000 single-nucleotide polymorphisms among responders and nonresponders to interferon beta therapy in multiple sclerosis. Arch Neurol, 2009. 66(8): p. 972-8. 76. Cunningham, S., et al., Pharmacogenomics of responsiveness to interferon IFN-beta treatment in multiple sclerosis: a genetic screen of 100 type I interferon-inducible genes. Clin Pharmacol Ther, 2005. 78(6): p. 635-46. 77. Weinstock-Guttman, B., et al., Genomic effects of IFN-beta in multiple sclerosis patients. J Immunol, 2003. 171(5): p. 2694-702. 78. Rani, M.R., et al., Heterogeneous, longitudinally stable molecular signatures in response to interferon-beta. Ann N Y Acad Sci, 2009. 1182: p. 58-68. 79. Grossman, I., et al., Pharmacogenetics of glatiramer acetate therapy for multiple sclerosis reveals drug-response markers. Pharmacogenet Genomics, 2007. 17(8): p. 657-66. 80. Millonig, A., et al., Natalizumab treatment reduces endothelial activity in MS patients. J Neuroimmunol, 2010. 227(1-2): p. 190-4. 81. Cotte, S., et al., ABC-transporter gene-polymorphisms are potential pharmacogenetic markers for mitoxantrone response in multiple sclerosis. Brain, 2009. 132(Pt 9): p. 2517-30.  129 82. Grove, J.I. and G.P. Aithal, Human leukocyte antigen genetic risk factors of drug-induced liver toxicology. Expert Opin Drug Metab Toxicol, 2014: p. 1-15. 83. Murray, K.F., et al., Drug-related hepatotoxicity and acute liver failure. J Pediatr Gastroenterol Nutr, 2008. 47(4): p. 395-405. 84. Ostapowicz, G., et al., Results of a prospective study of acute liver failure at 17 tertiary care centers in the United States. Ann Intern Med, 2002. 137(12): p. 947-54. 85. Navarro, V.J. and J.R. Senior, Drug-related hepatotoxicity. N Engl J Med, 2006. 354(7): p. 731-9. 86. de Abajo, F.J., et al., Acute and clinically relevant drug-induced liver injury: a population based case-control study. Br J Clin Pharmacol, 2004. 58(1): p. 71-80. 87. Sgro, C., et al., Incidence of drug-induced hepatic injuries: a French population-based study. Hepatology, 2002. 36(2): p. 451-5. 88. Aithal, G.P., et al., Case definition and phenotype standardization in drug-induced liver injury. Clin Pharmacol Ther, 2011. 89(6): p. 806-15. 89. Makar, G.A., et al., Incidence and prevalence of abnormal liver associated enzymes in patients with atrial fibrillation in a routine clinical care population. Pharmacoepidemiol Drug Saf, 2008. 17(1): p. 43-51. 90. Saukkonen, J.J., et al., An official ATS statement: hepatotoxicity of antituberculosis therapy. Am J Respir Crit Care Med, 2006. 174(8): p. 935-52. 91. Benichou, C., Criteria of drug-induced liver disorders. Report of an international consensus meeting. J Hepatol, 1990. 11(2): p. 272-6. 92. Abboud, G. and N. Kaplowitz, Drug-induced liver injury. Drug Saf, 2007. 30(4): p. 277-94. 93. Davidson C.S., L.C.M., Chamberlayne E.C., Guidelines for detection of hepatotoxicity due to drugs and chemicals, N.I.o. Health, Editor. 1978: Washington DC. 94. Bjornsson, E. and R. Olsson, Outcome and prognostic markers in severe drug-induced liver disease. Hepatology, 2005. 42(2): p. 481-9. 95. Naranjo, C.A., et al., A method for estimating the probability of adverse drug reactions. Clin Pharmacol Ther, 1981. 30(2): p. 239-45. 96. Maria, V.A. and R.M. Victorino, Development and validation of a clinical scale for the diagnosis of drug-induced hepatitis. Hepatology, 1997. 26(3): p. 664-9.  130 97. Teschke, R., W. Schmidt-Taenzer, and A. Wolff, Spontaneous reports of assumed herbal hepatotoxicity by black cohosh: is the liver-unspecific Naranjo scale precise enough to ascertain causality? Pharmacoepidemiol Drug Saf, 2011. 20(6): p. 567-82. 98. Miljkovic, M.M., S. Dobric, and V. Dragojevic-Simic, Consistency between causality assessments obtained with two scales and their agreement with clinical judgments in hepatotoxicity. Pharmacoepidemiol Drug Saf, 2011. 20(3): p. 272-85. 99. Lucena, M.I., et al., Comparison of two clinical scales for causality assessment in hepatotoxicity. Hepatology, 2001. 33(1): p. 123-30. 100. Danan, G. and C. Benichou, Causality assessment of adverse reactions to drugs--I. A novel method based on the conclusions of international consensus meetings: application to drug-induced liver injuries. J Clin Epidemiol, 1993. 46(11): p. 1323-30. 101. Lee, W.M. and J.R. Senior, Recognizing drug-induced liver injury: current problems, possible solutions. Toxicol Pathol, 2005. 33(1): p. 155-64. 102. Tremlett, H. and J. Oger, Hepatic injury, liver monitoring and the beta-interferons for multiple sclerosis. J Neurol, 2004. 251(11): p. 1297-303. 103. Randomised double-blind placebo-controlled study of interferon beta-1a in relapsing/remitting multiple sclerosis. PRISMS (Prevention of Relapses and Disability by Interferon beta-1a Subcutaneously in Multiple Sclerosis) Study Group. Lancet, 1998. 352(9139): p. 1498-504. 104. Jacobs, L.D., et al., Intramuscular interferon beta-1a for disease progression in relapsing multiple sclerosis. The Multiple Sclerosis Collaborative Research Group (MSCRG). Ann Neurol, 1996. 39(3): p. 285-94. 105. U.S. Department of Health and Human Services. Common Toxicity Criteria (CTC). 1999  [Accessed 13 July 2015]. Available from: http://ctep.cancer.gov/protocolDevelopment/electronic_applications/docs/ctcv20_4-30-992.pdf. 106. Francis, G.S., et al., Hepatic reactions during treatment of multiple sclerosis with interferon-beta-1a: incidence and clinical significance. Drug Saf, 2003. 26(11): p. 815-27. 107. Chan, S., et al., High-dose frequency beta-interferons increase the risk of liver test abnormalities in multiple sclerosis: a longitudinal study. Mult Scler, 2011. 17(3): p. 361-7. 108. Martin, K., et al., Differences between clinical trials and postmarketing use. Br J Clin Pharmacol, 2004. 57(1): p. 86-92.  131 109. Dufour DR, Laboratory guidelines for screening, diagnosis and monitoring of hepatic injury. 2000, Washington, D.C., USA: National Academy of Clinical Biochemistry. 110. Tremlett, H., et al., Liver test abnormalities in multiple sclerosis: findings from placebo-treated patients. Neurology, 2006. 67(7): p. 1291-3. 111. Duchini, A., Autoimmune hepatitis and interferon beta-1a for multiple sclerosis. Am J Gastroenterol, 2002. 97(3): p. 767-8. 112. Fontana, R.J., et al., Presentation and outcomes with clinically apparent interferon beta hepatotoxicity. Dig Dis Sci, 2013. 58(6): p. 1766-75. 113. Durelli, L., et al., Liver and thyroid function and autoimmunity during interferon-beta 1b treatment for MS. Neurology, 2001. 57(8): p. 1363-70. 114. Yoshida, E.M., et al., Fulminant liver failure during interferon beta treatment of multiple sclerosis. Neurology, 2001. 56(10): p. 1416. 115. Yoshida EM, R.S., Steinbrecher UP, Erb SR, Scudamore CH, Chung SW, Oger JJ, Hashimoto SA., Erratum: Fulminant liver failure during interferon beta treatment of multiple sclerosis. Neurology, 2001. 57(11). 116. Tremlett, H.L. and J. Oger, Elevated aminotransferases during treatment with interferon-beta for multiple sclerosis: actions and outcomes. Mult Scler, 2004. 10(3): p. 298-301. 117. Health Canada. Hepatic injury associated with beta-interferon treatment for multiple sclerosis. 2003. [Accessed 1 January 2012]; Available from: http://www.hc-sc.gc.ca/dhp-mps/medeff/advisories-avis/prof/_2003/beta_interferon_hpc-cps-eng.php. 118. Daly, A.K. and C.P. Day, Genetic association studies in drug-induced liver injury. Semin Liver Dis, 2009. 29(4): p. 400-11. 119. Lacana, E., et al., The emerging role of pharmacogenomics in biologics. Clin Pharmacol Ther, 2007. 82(4): p. 466-71. 120. Tujios, S. and R.J. Fontana, Mechanisms of drug-induced liver injury: from bedside to bench. Nat Rev Gastroenterol Hepatol, 2011. 8(4): p. 202-11. 121. Fujimori, K., et al., Possible mechanisms of elevation of serum transaminase levels during interferon-beta therapy in chronic hepatitis C patients. J Gastroenterol, 2002. 37(1): p. 40-6. 122. Bjornsson, E. and L. Davidsdottir, The long-term follow-up after idiosyncratic drug-induced liver injury with jaundice. J Hepatol, 2009. 50(3): p. 511-7. 123. Coppola, G., et al., Long-term clinical experience with weekly interferon beta-1a in relapsing multiple sclerosis. Eur J Neurol, 2006. 13(9): p. 1014-21.  132 124. Shirani, A., et al., Association between use of interferon beta and progression of disability in patients with relapsing-remitting multiple sclerosis. JAMA, 2012. 308(3): p. 247-56. 125. Lucena, M.I., et al., Phenotypic characterization of idiosyncratic drug-induced liver injury: the influence of age and sex. Hepatology, 2009. 49(6): p. 2001-9. 126. Dong, M.H., et al., Serum levels of alanine aminotransferase decrease with age in longitudinal analysis. Clin Gastroenterol Hepatol, 2012. 10(3): p. 285-90 e1. 127. Chalasani, N., et al., Features and Outcomes of 899 Patients With Drug-Induced Liver Injury: The DILIN Prospective Study. Gastroenterology, 2015. 148(7): p. 1340-52 e7. 128. Kojima, H., et al., Clinical features of liver disturbance in rheumatoid diseases: clinicopathological study with special reference to the cause of liver disturbance. J Gastroenterol, 2002. 37(8): p. 617-25. 129. Tarantino, G., et al., A prospective study of acute drug-induced liver injury in patients suffering from non-alcoholic fatty liver disease. Hepatol Res, 2007. 37(6): p. 410-5. 130. Kellokumpu-Lehtinen, P., E. Iisalo, and E. Nordman, Hepatotoxicity of paracetamol in combination with interferon and vinblastine. Lancet, 1989. 1(8647): p. 1143. 131. Fabris, P., M. Dalla Palma, and F. de Lalla, Idiosyncratic acute hepatitis caused by paracetamol in two patients with melanoma treated with high-dose interferon-alpha. Ann Intern Med, 2001. 134(4): p. 345. 132. Daly, A.K., et al., HLA-B*5701 genotype is a major determinant of drug-induced liver injury due to flucloxacillin. Nat Genet, 2009. 41(7): p. 816-9. 133. Kindmark, A., et al., Genome-wide pharmacogenetic investigation of a hepatic adverse event without clinical signs of immunopathology suggests an underlying immune pathogenesis. Pharmacogenomics J, 2008. 8(3): p. 186-95. 134. Lucena, M.I., et al., Susceptibility to amoxicillin-clavulanate-induced liver injury is influenced by multiple HLA class I and II alleles. Gastroenterology, 2011. 141(1): p. 338-47. 135. Singer, J.B., et al., A genome-wide study identifies HLA alleles associated with lumiracoxib-related liver injury. Nat Genet, 2010. 42(8): p. 711-4. 136. Spraggs, C.F., et al., HLA-DQA1*02:01 is a major risk factor for lapatinib-induced hepatotoxicity in women with advanced breast cancer. J Clin Oncol, 2011. 29(6): p. 667-73.  133 137. Schaid, D.J., et al., Prospective validation of HLA-DRB1*07:01 allele carriage as a predictive risk factor for lapatinib-induced liver injury. J Clin Oncol, 2014. 32(22): p. 2296-303. 138. Parham, L.R., et al., Comprehensive genome-wide evaluation of lapatinib-induced liver injury yields a single genetic signal centered on known risk allele HLA-DRB1*07:01. Pharmacogenomics J, 2015. 139. Monshi, M., et al., HLA-B*57:01-restricted activation of drug-specific T-cells provides the immunological basis for flucloxacillin-induced liver injury. Hepatology, 2012. 140. Kim, S.H., et al., Characterization of amoxicillin- and clavulanic acid-specific T cells in patients with amoxicillin-clavulanate-induced liver injury. Hepatology, 2015. 62(3): p. 887-99. 141. Hirasawa, M., et al., The Possible Mechanism of Idiosyncratic Lapatinib-Induced Liver Injury in Patients Carrying Human Leukocyte Antigen-DRB1*07:01. PLoS One, 2015. 10(6): p. e0130928. 142. Daly, A.K., et al., HLA-B*5701 genotype is a major determinant of drug-induced liver injury due to flucloxacillin. Nat Genet, 2009. 41(7): p. 816-9. 143. Li, L.M., et al., SLCO1B1 *15 haplotype is associated with rifampin-induced liver injury. Mol Med Rep, 2012. 6(1): p. 75-82. 144. Takai, S., et al., Involvement of immune- and inflammatory-related factors in flucloxacillin-induced liver injury in mice. J Appl Toxicol, 2015. 35(2): p. 142-51. 145. Takai, S., et al., Establishment of a mouse model for amiodarone-induced liver injury and analyses of its hepatotoxic mechanism. J Appl Toxicol, 2015. 146. Vliegenthart, A.D., et al., Zebrafish as model organisms for studying drug-induced liver injury. Br J Clin Pharmacol, 2014. 78(6): p. 1217-27. 147. Welter, D., et al., The NHGRI GWAS Catalog, a curated resource of SNP-trait associations. Nucleic Acids Res, 2014. 42(Database issue): p. D1001-6. 148. Bush, W.S. and J.H. Moore, Chapter 11: Genome-wide association studies. PLoS Comput Biol, 2012. 8(12): p. e1002822. 149. Daly, A.K., Using genome-wide association studies to identify genes important in serious adverse drug reactions. Annu Rev Pharmacol Toxicol, 2012. 52: p. 21-35. 150. Clarke, G.M., et al., Basic statistical analysis in genetic case-control studies. Nat Protoc, 2011. 6(2): p. 121-33.  134 151. De, R., W.S. Bush, and J.H. Moore, Bioinformatics Challenges in Genome-Wide Association Studies (GWAS). Methods in molecular biology (Clifton, N.J.), 2013. 1168: p. 63-81. 152. Li, M., C. Li, and W. Guan, Evaluation of coverage variation of SNP chips for genome-wide association studies. Eur J Hum Genet, 2008. 16(5): p. 635-43. 153. Abdi, H., Bonferonni and Sidak corrections for multiple comparisons. Encyclopedia of Measurement and Statistics, ed. N.J. Salkind. 2007, Thousand Oaks CA: Sage. 154. Jannot, A.S., G. Ehret, and T. Perneger, P < 5 x 10(-8) has emerged as a standard of statistical significance for genome-wide association studies. J Clin Epidemiol, 2015. 68(4): p. 460-5. 155. Zheng, G., B. Freidlin, and J.L. Gastwirth, Robust genomic control for association studies. Am J Hum Genet, 2006. 78(2): p. 350-6. 156. Price, A.L., et al., Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet, 2006. 38(8): p. 904-9. 157. Sadee, W., Pharmacogenomic biomarkers: validation needed for both the molecular genetic mechanism and clinical effect. Pharmacogenomics, 2011. 12(5): p. 675-80. 158. Ioannidis, J.P., Non-replication and inconsistency in the genome-wide association setting. Hum Hered, 2007. 64(4): p. 203-13. 159. Pirinen, M., P. Donnelly, and C.C. Spencer, Including known covariates can reduce power to detect genetic effects in case-control studies. Nat Genet, 2012. 44(8): p. 848-51. 160. Hirschhorn, J.N. and D. Altshuler, Once and again-issues surrounding replication in genetic association studies. J Clin Endocrinol Metab, 2002. 87(10): p. 4438-41. 161. Bush, W.S. and J.H. Moore, Chapter 11: Genome-wide association studies. PLoS computational biology, 2011. 8(12). 162. Genomes Project, C., et al., An integrated map of genetic variation from 1,092 human genomes. Nature, 2012. 491(7422): p. 56-65. 163. Via, M., C. Gignoux, and E.G. Burchard, The 1000 Genomes Project: new opportunities for research and social challenges. Genome Med, 2010. 2(1): p. 3. 164. Carleton, B., et al., Adverse drug reaction active surveillance: developing a national network in Canada's children's hospitals. Pharmacoepidemiol Drug Saf, 2009. 18(8): p. 713-21. 165. Giacomini, K.M., et al., The pharmacogenetics research network: from SNP discovery to clinical drug response. Clin Pharmacol Ther, 2007. 81(3): p. 328-45.  135 166. Henney, A., Ageing populations, complex diseases and their challenge to 21st century healthcare: the focus of the European Society for Pharmacogenomics and Theranostics. Drug Metabol Drug Interact, 2011. 26(2): p. 43-4. 167. Seguin, B., et al., Universal health care, genomic medicine and Thailand: investing in today and tomorrow. Nat Rev Genet, 2008. 9 Suppl 1: p. S14-9. 168. Sistonen, J., et al., Prediction of codeine toxicity in infants and their mothers using a novel combination of maternal genetic markers. Clin Pharmacol Ther, 2012. 91(4): p. 692-9. 169. Visscher, H., et al., Pharmacogenomic Prediction of Anthracycline-Induced Cardiotoxicity in Children. J Clin Oncol, 2011: p. Epub Sept 09. 170. Ross, C.J., et al., Genetic variants in TPMT and COMT are associated with hearing loss in children receiving cisplatin chemotherapy. Nat Genet, 2009. 41(12): p. 1345-9. 171. Pussegoda, K., et al., Replication of TPMT and ABCC3 genetic variants highly associated with cisplatin-induced hearing loss in children. Clin Pharmacol Ther, 2013. 172. Vandenbroeck, K., et al., United Europeans for development of pharmacogenomics in multiple sclerosis network. Pharmacogenomics, 2009. 10(5): p. 885-94. 173. European Database for Multiple Sclerosis. EDMUS. 2002. [Accessed 30 May 2012]; Available from: http://www.edmus.org/index.html. 174. Butzkueven, H., et al., MSBase: an international, online registry and platform for collaborative outcomes research in multiple sclerosis. Mult Scler, 2006. 12(6): p. 769-74. 175. Yeh, E.A., et al., Pediatric multiple sclerosis. Nat Rev Neurol, 2009. 5(11): p. 621-31. 176. O'Mahony, J., et al., Masquerades of Acquired Demyelination in Children: Experiences of a National Demyelinating Disease Program. J Child Neurol, 2012. 177. Vass K, S.C.o.t.A.N.R. Real-life use of natalizumab: safety data from the Austrian Nation-wide Natalizumab Registry. in ECTRIMS. 2011. Amsterdam, NL: Congrex Switzerland Ltd. 178. Gurwitz, D. and M. Pirmohamed, Pharmacogenomics: the importance of accurate phenotypes. Pharmacogenomics, 2010. 11(4): p. 469-70. 179. Pirmohamed, M., et al., Phenotype standardization for immune-mediated drug-induced skin injury. Clin Pharmacol Ther, 2011. 89(6): p. 896-901. 180. Pirmohamed, M., et al., The phenotype standardization project: improving pharmacogenetic studies of serious adverse drug reactions. Clin Pharmacol Ther, 2011. 89(6): p. 784-5.  136 181. Poser, C.M., et al., New diagnostic criteria for multiple sclerosis: guidelines for research protocols. Ann Neurol, 1983. 13(3): p. 227-31. 182. Durelli, L., et al., Interferon treatment for multiple sclerosis: autoimmune complications may be lethal. Neurology, 1998. 50(2): p. 570-1. 183. Grieco, A., et al., Severe acute hepatitis after resumption of interferon-Beta therapy for multiple sclerosis: a word of caution. Am J Gastroenterol, 2007. 102(11): p. 2606-7. 184. Pulicken, M., et al., Unmasking of autoimmune hepatitis in a patient with MS following interferon beta therapy. Neurology, 2006. 66(12): p. 1954-5. 185. Muratori, L., et al., Antimitochondrial antibodies and other antibodies in primary biliary cirrhosis: diagnostic and prognostic value. Clin Liver Dis, 2008. 12(2): p. 261-76; vii. 186. Muratori, P., et al., Characterization and clinical impact of antinuclear antibodies in primary biliary cirrhosis. Am J Gastroenterol, 2003. 98(2): p. 431-7. 187. Mieli-Vergani, G. and D. Vergani, Autoimmune hepatitis. Nat Rev Gastroenterol Hepatol, 2011. 8(6): p. 320-9. 188. Chazouilleres, O., et al., Primary biliary cirrhosis-autoimmune hepatitis overlap syndrome: clinical features and response to therapy. Hepatology, 1998. 28(2): p. 296-301. 189. Boberg, K.M., et al., Overlap syndromes: the International Autoimmune Hepatitis Group (IAIHG) position statement on a controversial issue. J Hepatol, 2011. 54(2): p. 374-85. 190. Bhandari, B.M., H. Bayat, and K.D. Rothstein, Primary biliary cirrhosis. Gastroenterol Clin North Am, 2011. 40(2): p. 373-86, viii. 191. Invernizzi, P., et al., Classical HLA-DRB1 and DPB1 alleles account for HLA associations with primary biliary cirrhosis. Genes Immun, 2012. 192. Hannuksela, M.L., et al., Biochemical markers of alcoholism. Clin Chem Lab Med, 2007. 45(8): p. 953-61. 193. Pietrosi, G., et al., Fulminant hepatic failure and autoimmune disorders in patient with multiple sclerosis on interferon beta 1a: a fatal combination? Transpl Int, 2008. 21(5): p. 502-4. 194. Bezabeh, S., et al., Clinically significant liver injury in patients treated with natalizumab. Aliment Pharmacol Ther, 2010. 31(9): p. 1028-35. 195. Brunetti, L., et al., Teriflunomide for the treatment of relapsing multiple sclerosis: a review of clinical data. Ann Pharmacother, 2013. 47(9): p. 1153-60.  137 196. Biogen Idec. Tecfidera (dimethyl fumarate) drug monograph. 2013. [Accessed 17  December 2013]; Available from: www.tecfidera.com/pdfs/full-prescribing-information.pdf. 197. Cohen, J.A., et al., Alemtuzumab versus interferon beta 1a as first-line treatment for patients with relapsing-remitting multiple sclerosis: a randomised controlled phase 3 trial. Lancet, 2012. 380(9856): p. 1819-28. 198. Tarantino, G., M.N. Di Minno, and D. Capone, Drug-induced liver injury: is it somehow foreseeable? World J Gastroenterol, 2009. 15(23): p. 2817-33. 199. Tremlett, H., et al., New perspectives in the natural history of multiple sclerosis. Neurology, 2010. 74(24): p. 2004-15. 200. Kingwell, E., et al., Cancer risk in multiple sclerosis: findings from British Columbia, Canada. Brain, 2012. 135(Pt 10): p. 2973-9. 201. SWEDEGENE. SWEDEGENE: Genetic causes of adverse drug reactions. 2013 9 Oct 2013 [cited 2014 1 April]; Available from: http://www.swedegene.se/methodology.html. 202. Bjornsson, E.S., et al., Incidence, presentation, and outcomes in patients with drug-induced liver injury in the general population of Iceland. Gastroenterology, 2013. 144(7): p. 1419-25, 1425 e1-3; quiz e19-20. 203. U.S. Department of Health and Human Services. Common Terminology Criteria for Adverse Events (CTCAE), v4.03 2010. [Accessed 12 October 2012]; Available from: http://evs.nci.nih.gov/ftp1/CTCAE/CTCAE_4.03_2010-06-14_QuickReference_5x7.pdf. 204. Kowalec, K., Yoshida, E., Traboulsee, A., Carleton, B.C., Tremlett, H., Suspected autoimmune hepatitis and primary biliary cirrhosis unmasked by interferon-beta in a multiple sclerosis patient. Mult Scler Rel Disord, 2013. 2: p. 57-59. 205. Marriott, J.J., et al., Multiple sclerosis disease-modifying therapy prescribing patterns in Ontario. Can J Neurol Sci, 2013. 40(1): p. 67-72. 206. Senior, J.R., Alanine aminotransferase: a clinical and regulatory tool for detecting liver injury-past, present, and future. Clin Pharmacol Ther, 2012. 92(3): p. 332-9. 207. Rockey, D.C., et al., Causality assessment in drug-induced liver injury using a structured expert opinion process: comparison to the Roussel-Uclaf causality assessment method. Hepatology, 2010. 51(6): p. 2117-26. 208. Torbati, S., et al., Association of CD58 Polymorphism with Multiple Sclerosis and Response to Interferon ss Therapy in A Subset of Iranian Population. Cell J, 2015. 16(4): p. 506-13.  138 209. O'Doherty, C., et al., Genetic polymorphisms, their allele combinations and IFN-beta treatment response in Irish multiple sclerosis patients. Pharmacogenomics, 2009. 10(7): p. 1177-86. 210. Court, M.H., et al., Candidate gene polymorphisms in patients with acetaminophen-induced acute liver failure. Drug Metab Dispos, 2014. 42(1): p. 28-32. 211. Vosslamber, S., et al., Interferon regulatory factor 5 gene variants and pharmacological and clinical outcome of Interferonbeta therapy in multiple sclerosis. Genes Immun, 2011. 12(6): p. 466-72. 212. Liang, X., et al., Specific genetic polymorphisms of IL10-592 AA and IL10-819 TT genotypes lead to the key role for inducing docetaxel-induced liver injury in breast cancer patients. Clin Transl Oncol, 2013. 15(4): p. 331-4. 213. Wergeland, S., et al., IL-10 promoter haplotype influence on interferon treatment response in multiple sclerosis. Eur J Neurol, 2005. 12(3): p. 171-5. 214. Guo, Y., et al., Polymorphisms in CTLA4 influence incidence of drug-induced liver injury after renal transplantation in Chinese recipients. PLoS One, 2012. 7(12): p. e51723. 215. Esposito, F., et al., A pharmacogenetic study implicates SLC9a9 in multiple sclerosis disease activity. Ann Neurol, 2015. 78(1): p. 115-27. 216. Kim, S.H., et al., Superoxide Dismutase Gene (SOD1, SOD2, and SOD3) Polymorphisms and Antituberculosis Drug-induced Hepatitis. Allergy Asthma Immunol Res, 2015. 7(1): p. 88-91. 217. Ho, H.T., et al., The NAT2 tag SNP rs1495741 correlates with the susceptibility of antituberculosis drug-induced hepatotoxicity. Pharmacogenet Genomics, 2013. 23(4): p. 200-7. 218. Kulakova, O.G., et al., Allelic combinations of immune-response genes as possible composite markers of IFN-beta efficacy in multiple sclerosis patients. Pharmacogenomics, 2012. 13(15): p. 1689-700. 219. Andrade, R.J., et al., Drug-induced liver injury: insights from genetic studies. Pharmacogenomics, 2009. 10(9): p. 1467-87. 220. Daly, A.K., Drug-induced liver injury: past, present and future. Pharmacogenomics, 2010. 11(5): p. 607-11. 221. Lucena, M.I., et al., Recurrent drug-induced liver injury (DILI) with different drugs in the Spanish Registry: the dilemma of the relationship to autoimmune hepatitis. J Hepatol, 2011. 55(4): p. 820-7.  139 222. Kress-Bennett, J.M., et al., Preliminary study: treatment with intramuscular interferon beta-1a results in increased levels of IL-12Rbeta2+ and decreased levels of IL23R+ CD4+ T - Lymphocytes in multiple sclerosis. BMC Neurol, 2011. 11: p. 155. 223. Urban, T.J., et al., Limited contribution of common genetic variants to risk for liver injury due to a variety of drugs. Pharmacogenet Genomics, 2012. 22(11): p. 784-95. 224. Wang, Y., A. Qu, and H. Wang, Signal transducer and activator of transcription 4 in liver diseases. Int J Biol Sci, 2015. 11(4): p. 448-55. 225. Chan, S.L., et al., Progress in understanding the genomic basis for adverse drug reactions: a comprehensive review and focus on the role of ethnicity. Pharmacogenomics, 2015. 16(10): p. 1161-78. 226. Gauderman WJ, M.J. QUANTO 1.1: A computer program for power and sample size calculations for genetic-epidemiology studies. 2006. [Accessed 10 December 2012]; Available from: http://hydra.usc.edu/gxe. 227. Xia, Z., et al., Modeling disease severity in multiple sclerosis using electronic health records. PLoS One, 2013. 8(11): p. e78927. 228. Centers for Disease Control (CDC), Behavioral Risk Factor Surveillance System Survey Questionnaire, D.o.H.a.H. Services;, Editor. 1995: Atlanta. 229. Marrie, R.A., et al., Smoking status over two years in patients with multiple sclerosis. Neuroepidemiology, 2009. 32(1): p. 72-9. 230. Garay KW, The role of matching in epidemiological studies. Am J. Pharm. Edu., 2004. 3: p. 1-7. 231. Faresjo, T. and A. Faresjo, To match or not to match in epidemiological studies--same outcome but less power. Int J Environ Res Public Health, 2010. 7(1): p. 325-32. 232. Suzuki, A., et al., Drugs associated with hepatotoxicity and their reporting frequency of liver adverse events in VigiBase: unified list based on international collaborative work. Drug Saf, 2010. 33(6): p. 503-22. 233. Anderson, C.A., et al., Data quality control in genetic case-control association studies. Nat Protoc, 2010. 5(9): p. 1564-73. 234. Arnold, M., et al., SNiPA: an interactive, genetic variant-centered annotation browser. Bioinformatics, 2015. 31(8): p. 1334-6. 235. Barrett, J.C., et al., Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics, 2005. 21(2): p. 263-5.  140 236. Pruim, R.J., et al., LocusZoom: regional visualization of genome-wide association scan results. Bioinformatics, 2010. 26(18): p. 2336-7. 237. Consortium, G.P., et al., A global reference for human genetic variation. Nature, 2015. 526(7571): p. 68-74. 238. Cunningham, F., et al., Ensembl 2015. Nucleic Acids Res, 2015. 43(Database issue): p. D662-9. 239. Kircher, M., et al., A general framework for estimating the relative pathogenicity of human genetic variants. Nat Genet, 2014. 46(3): p. 310-5. 240. GTex Consortium. Human genomics. The Genotype-Tissue Expression (GTEx) pilot analysis: multitissue gene regulation in humans. Science, 2015. 348(6235): p. 648-60. 241. DeLong, E.R., D.M. DeLong, and D.L. Clarke-Pearson, Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics, 1988. 44(3): p. 837-45. 242. Robin, X., et al., pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics, 2011. 12: p. 77. 243. Visscher, H., et al., Application of principal component analysis to pharmacogenomic studies in Canada. Pharmacogenomics J, 2009. 9(6): p. 362-72. 244. Doi, H., et al., Exome sequencing reveals a homozygous SYT14 mutation in adult-onset, autosomal-recessive spinocerebellar ataxia with psychomotor retardation. Am J Hum Genet, 2011. 89(2): p. 320-7. 245. Fukuda, M., Molecular cloning, expression, and characterization of a novel class of synaptotagmin (Syt XIV) conserved from Drosophila to humans. J Biochem, 2003. 133(5): p. 641-9. 246. Lee, S.M., et al., RNA stability in human liver: comparison of different processing times, temperatures and methods. Mol Biotechnol, 2013. 53(1): p. 1-8. 247. Honda, K., A. Takaoka, and T. Taniguchi, Type I interferon [corrected] gene induction by the interferon regulatory factor family of transcription factors. Immunity, 2006. 25(3): p. 349-60. 248. Baranzini, S.E., et al., Prognostic biomarkers of IFNb therapy in multiple sclerosis patients. Mult Scler, 2015. 21(7): p. 894-904. 249. Clarelli, F., et al., Pharmacogenetic study of long-term response to interferon-beta treatment in multiple sclerosis. Pharmacogenomics J, 2015.  141 250. Lin, Y., et al., Upregulation of Interferon Regulatory Factor 6 Promotes Neuronal Apoptosis After Traumatic Brain Injury in Adult Rats. Cell Mol Neurobiol, 2015. 251. Luedde, T., N. Kaplowitz, and R.F. Schwabe, Cell death and cell death responses in liver disease: mechanisms and clinical relevance. Gastroenterology, 2014. 147(4): p. 765-783 e4. 252. Tomida, T., et al., Multiparametric assay using HepaRG cells for predicting drug-induced liver injury. Toxicol Lett, 2015. 236(1): p. 16-24. 253. Byrnes, V., et al., Drug induced liver injury secondary to interferon-beta (IFN-beta) in multiple sclerosis. Ann Hepatol, 2006. 5(1): p. 56-9. 254. Ethiraj, P., et al., Inhibitory effects of interferon-beta on hepatocellular carcinoma HepG2 via Akt/STAT phosphorylation. Fundam Clin Pharmacol, 2015. 29(3): p. 278-85. 255. Malhi, H., G.J. Gores, and J.J. Lemasters, Apoptosis and necrosis in the liver: a tale of two deaths? Hepatology, 2006. 43(2 Suppl 1): p. S31-44. 256. Dhupar, R., et al., Interferon regulatory factor 1 mediates acetylation and release of high mobility group box 1 from hepatocytes during murine liver ischemia-reperfusion injury. Shock, 2011. 35(3): p. 293-301. 257. Zhai, Y., et al., Cutting edge: TLR4 activation mediates liver ischemia/reperfusion inflammatory response via IFN regulatory factor 3-dependent MyD88-independent pathway. J Immunol, 2004. 173(12): p. 7115-9. 258. Wang, P.X., et al., Interferon regulatory factor 9 is a key mediator of hepatic ischemia/reperfusion injury. J Hepatol, 2015. 62(1): p. 111-20. 259. GlaxoSmithKline Inc. TYKERB®: Lapatinib Product Monograph. 2015  [cited 2015 December 3]; Available from: http://www.gsk.ca/english/docs-pdf/product-monographs/Tykerb.pdf. 260. Sung, Y.J., et al., Performance of genotype imputations using data from the 1000 Genomes Project. Hum Hered, 2012. 73(1): p. 18-25. 261. Aminkeng, F., et al., A coding variant in RARG confers susceptibility to anthracycline-induced cardiotoxicity in childhood cancer. Nat Genet, 2015. 47(9): p. 1079-84. 262. Behr, E.R., et al., Genome wide analysis of drug-induced torsades de pointes: lack of common variants with large effect sizes. PLoS One, 2013. 8(11): p. e78511. 263. Brown, M.G., et al., How effective are disease-modifying drugs in delaying progression in relapsing-onset MS? Neurology, 2007. 69(15): p. 1498-507.  142 264. Chalasani, N. and E. Bjornsson, Risk factors for idiosyncratic drug-induced liver injury. Gastroenterology, 2010. 138(7): p. 2246-59. 265. Okorodudu, D.O., et al., Diagnostic performance of body mass index to identify obesity as defined by body adiposity: a systematic review and meta-analysis. Int J Obes (Lond), 2010. 34(5): p. 791-9. 266. Kvistad, S.S., et al., Body mass index influence interferon-beta treatment response in multiple sclerosis. J Neuroimmunol, 2015. 288: p. 92-7. 267. Consortium, G.P., et al., A map of human genome variation from population-scale sequencing. Nature, 2010. 467(7319): p. 1061-73. 268. Xu, C.Q., et al., Chronic Ethanol-Induced Impairment of Wnt/beta-Catenin Signaling is Attenuated by PPAR-delta Agonist. Alcohol Clin Exp Res, 2015. 39(6): p. 969-79. 269. Boivin, N., et al., Interferon-beta suppresses murine Th1 cell function in the absence of antigen-presenting cells. PLoS One, 2015. 10(4): p. e0124802. 270. Trenova, A.G., et al., Cytokines and disability in interferon-beta-1b treated and untreated women with multiple sclerosis. Arch Med Res, 2014. 45(6): p. 495-500. 271. Nattrass, R., et al., Activation of Flucloxacillin-Specific CD8+ T-Cells With the Potential to Promote Hepatocyte Cytotoxicity in a Mouse Model. Toxicol Sci, 2015. 146(1): p. 146-56. 272. Kantarci, O.H., et al., Interferon gamma allelic variants: sex-biased multiple sclerosis susceptibility and gene expression. Arch Neurol, 2008. 65(3): p. 349-57. 273. La Mantia, L., et al., Interferons-beta versus glatiramer acetate for relapsing-remitting multiple sclerosis. Cochrane Database Syst Rev, 2014. 7: p. CD009333. 274. Subramaniam, K., et al., Glatiramer acetate induced hepatotoxicity. Curr Drug Saf, 2012. 7(2): p. 186-8. 275. Deltenre, P., et al., Acute hepatitis induced by glatiramer acetate. BMJ Case Rep, 2009. 2009. 276. Liu, Z.X. and N. Kaplowitz, Immune-mediated drug-induced liver disease. Clin Liver Dis, 2002. 6(3): p. 755-74. 277. US National Library of Medicine. LiverTox: Autoimmune Hepatitis. 2015 [Accessed 30 November 2015]; Available from: http://livertox.nih.gov/Phenotypes_auto.html. 278. Kuzu, U.B., et al., Clinical and histological features of idiosyncratic liver injury: Dilemma in diagnosis of autoimmune hepatitis. Hepatol Res, 2015.  143 279. Webb, G.J. and G.M. Hirschfield, Using GWAS to identify genetic predisposition in hepatic autoimmunity. J Autoimmun, 2015. 280. Langer-Gould, A., et al., Incidence of multiple sclerosis in multiple racial and ethnic groups. Neurology, 2013. 80(19): p. 1734-9. 281. Doyle, S.R. and D.M. Donovan, A validation study of the alcohol dependence scale. J Stud Alcohol Drugs, 2009. 70(5): p. 689-99. 282. Fontana, R.J., et al., Drug-Induced Liver Injury Network (DILIN) prospective study: rationale, design and conduct. Drug Saf, 2009. 32(1): p. 55-68. 283. Stine, J.G. and J.H. Lewis, Hepatotoxicity of antibiotics: a review and update for the clinician. Clin Liver Dis, 2013. 17(4): p. 609-42, ix. 284. Alfirevic, A. and M. Pirmohamed, Predictive genetic testing for drug-induced liver injury: considerations of clinical utility. Clin Pharmacol Ther, 2012. 92(3): p. 376-80. 285. Hensley Alford, S., et al., Participation in genetic testing research varies by social group. Public Health Genomics, 2011. 14(2): p. 85-93. 286. Krumbholz, M., et al., Interferon-beta increases BAFF levels in multiple sclerosis: implications for B cell autoimmunity. Brain, 2008. 131(Pt 6): p. 1455-63. 287. Nelson, M.R., et al., Genome-wide approaches to identify pharmacogenetic contributions to adverse drug reactions. Pharmacogenomics J, 2009. 9(1): p. 23-33. 288. Canadian Network of MS Clinics. Canadian Network of Multiple Sclerosis Clinics. 2016  [Accessed 5 January 2016]; Available from: http://www.cnmsc.ca/. 289. Grady, B.J. and M.D. Ritchie, Statistical Optimization of Pharmacogenomics Association Studies: Key Considerations from Study Design to Analysis. Curr Pharmacogenomics Person Med, 2011. 9(1): p. 41-66. 290. Sullivan J, New price tag put on liver transplants. Canadian Medical Association Journal,, 2003. 168(2): p. 206. 291. Wilke, R.A., et al., The emerging role of electronic medical records in pharmacogenomics. Clin Pharmacol Ther, 2011. 89(3): p. 379-86. 292. Stryjewski, T.P., et al., The Impact of Institutional Review Boards (IRBs) on Clinical Innovation: A Survey of Investigators and IRB Members. J Empir Res Hum Res Ethics, 2015. 10(5): p. 481-7. 293. Michael Smith Foundation for Health Research. British Columbia Ethics Harmonization Initiative. 2015 [Accessed 6 January 2016]; Available from: http://bcethics.ca/.  144 294. FDA. Guidance for industry: Using a centralized IRB review process in multicenter clinical trials. 2006. [Accessed 6 January 2016]; Available from: http://www.fda.gov/RegulatoryInformation/Guidances/ucm127004.htm. 295. Kowalec K, C.B., Tremlett H, The potential role of pharmacogenomics in the prevention of serious adverse drug reactions in multiple sclerosis. Mult Scler Rel Disord, 2013. 2 (3): p. 183-192. 296. Kowalec, K., et al., Characteristics associated with drug-induced liver injury from interferon beta in multiple sclerosis patients. Expert Opin Drug Saf, 2014. 13(10): p. 1305-17. 297. Cao, S., et al., Genome-wide Association Study on Platinum-induced Hepatotoxicity in Non-Small Cell Lung Cancer Patients. Sci Rep, 2015. 5: p. 11556. 298. Xu, C.F., et al., HLA-B*57:01 Confers Susceptibility to Pazopanib-Associated Liver Injury in Patients With Cancer. Clin Cancer Res, 2015.                145 Appendix    Table A.1The grades of severity of drug induced liver injury (DILI) Category Severity Description 1 Mild Elevated alanine aminotransferase/ alkaline phosphatase (ALT/ALP) concentration reaching criteria for DILI but bilirubin concentration < 2× upper limit of normal (ULN) 2 Moderate Elevated ALT/ALP concentration reaching criteria for DILI and bilirubin concentration ≥2× ULN, or symptomatic hepatitis 3 Severe   Elevated ALT/ALP concentration reaching criteria for DILI, bilirubin concentration ≥2× ULN, and one of the following: • International normalized ratio ≥ 1.5 • Ascites and/or encephalopathy, disease duration < 26 weeks, and absence of underlying cirrhosis • Other organ failure considered to be due to DILI 4 Fatal or transplantation Death or transplantation due to DILI From Aithal, G.P et al., Case Definition and Phenotype Standardization in Drug-Induced Liver Injury, Clinical Pharmacology and Therapeutics, 2011. 89(6): p.806-815. Reprinted with permission from RightsLink.      146 Table A.2 Data collection form for patients experiencing drug induced liver injury due to interferon-beta   147           148 Table A.3 Structured Questionnaire Canadian Pharmacogenomics Network for Drug Safety: Genetic Factors Associated with Multiple Sclerosis Treatment. Structured Questionnaire To be administered by designated trained research assistant either over the phone   Date of call:_________ ID:___________ 1. Which of the following MS medications have you ever taken: Avonex (IFNB -1a, 30mcg im weekly) 1 Betaseron  (IFNB-1b, 250mcg sc alt days) 2 Rebif  low (IFNB -1a, 22mcg sc x3 weekly) 3 Rebif  high (IFNB -1a, 44mcg sc x3 weekly) 4 Copaxone (glatiramer acetate, 20mg daily)5 Patient cannot recall    6  [*THANK & DISCONTINUE] *[rationale: if the patient is unable to recall which IMD they were taking, they are unlikely to recall what other medications they were taking]  2. Start and stop dates When did you start and stop taking [insert name of MS medication] When did you stop taking [insert name of MS medication] [IF not yet stopped their MS medication; draw a line through the stop date box]        Start date 1 Stop date 1 Start date 2 Stop date 2 Avonex     Betaseron     Rebif low     Rebif high     Copaxone     [ESTABLISH THE STOP DATE OF THE FIRST IMD____________________________]   3. We are interested in any other medication that you were taking within the three months prior to when you stopped [insert name of first MS medication stopped].  Please think back for a moment. We are interested in ANY medicines, vitamins, supplements, herbal or alternative remedies have you taken in the three months prior to [insert stop date of first IMD].   These could be from your doctor, pharmacy, supermarket, health food shop, mail order or internet.  Please include all tablets, liquids, drops, patches, sprays, inhalers, creams, ointments, injections or suppositories.  [IF THE PATIENT CANNOT RECALL, ASK IF THERE IS ANYONE AT HOME THEY WOULD LIKE TO ASK TO HELP REMIND THEM] Not taking any medicines at that time   1 Taking one or more medicines at that time   2 Patient cannot recall      3 Name of medicine Source (1-6) Dose and frequency (if known) Reason for taking (7-11) Start date  Stop date                                             149             Source: 1=prescription/ Dr; 2=pharmacy/over the counter; 3=supermarket; 4=health food shop; 5=internet/ mail order; 6=other (please specify) Reason for taking: 7=to alleviate a symptom of MS / to help their MS; 8=to alleviate a side effect of IMD therapy; 9=for another health issue; 10= for general health reasons; 11=other (specify)  4. Did you experience any infections in the year prior to when you stopped [insert name of first MS medication stopped]?  *Specifically antibiotics or antivirals Infection type Date of infection (MM/YYYY) Please rate seriousness of infection (10=most serious infection could imagine 1=mild infection) Infection required hospitalization? Received treatment from the Dr* for infection?  Please specify type of treatment Upper respiratory tract infection   YES  / NO YES / NO  Urinary tract infection (bladder)   YES  / NO YES / NO  Influenza (the ‘flu)   YES  / NO YES / NO  Bacterial pneumonia   YES  / NO YES / NO  Chronic sinus infection   YES  / NO YES / NO  Gastroenteritis   YES  / NO YES / NO  Cold sores (or eruption of herpes anywhere on the body)   YES  / NO YES / NO  Other:_________ ______________   YES  / NO YES / NO  Other:_________ ______________   YES  / NO YES / NO   5. We are interested in any vaccinations that you might have had in the three months prior to when you stopped [insert name of first MS medication stopped].  Please think back for a moment. These vaccinations could include your annual ‘flu shot or perhaps a booster vaccine such as polio or tetanus or vaccines associated with going on a foreign holiday.   [IF THE PATIENT CANNOT RECALL, ASK IF THERE IS ANYONE AT HOME THEY WOULD LIKE TO ASK TO HELP REMIND THEM]  Did not have any vaccines at that time   1 Had one or more vaccine at that time   2 Patient cannot recall      3  Name of vaccine (brand name if possible) Date of vaccination      150      6. Did you drink any alcoholic beverages at all in the three months prior to when you stopped [insert name of first MS medication stopped and date].  YES     1 [GO TO Q7]  NO     2 [GO TO Q9] DO NOT RECALL   3 [GO TO Q9]   7. How often did you drink alcoholic beverages in the three months prior to when you stopped [insert date of first MS medication stopped]? less than once a month   1 one to three times a month   2 one to three times a week   3 four times or more a week   4   8. On those days when you drank alcohol, how many drinks did you usually have?  One or two drinks    1 3 or 4 alcoholic drinks    2 5 drinks or more    3 (Note: 1 drink =1 small glass of wine, or 1 bottle of beer, or one 12oz liquor)  9. Have you smoked at least 100 cigarettes in your life?  YES     1 [GO TO Q10]  NO     2 [GO TO Q12]  10. In the three months before you stopped [insert IMD name and date] were you smoking cigarettes every day, occasionally or not at all?  Every day    1 [GO TO Q11] Occasionally    2[GO TO Q11] Not at all    3[GO TO Q12]  11. How many cigarettes were you smoking every day? Between 1-10 cigarettes per day  1 Between 11-19 cigarettes per day  2 20 cigarettes or more per day   3   12. What is your current weight: _________ and height___________? Has your weight changed at all from around [insert date of first MS medication stopped]? YES     1  Previous weight: ______________ NO     2  DO NOT RECALL   3   13. We would like to ask about your ethnic background as this can alter how some people react to medications. How would you describe your racial or cultural group? You may belong to more than one group from the following list. Are you…..  [Please circle each answer]  White/European? -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 1 Chinese?- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - - -  - - -2  151 Southest Asian? (for example, Vietnamese, Cambodian etc) - - - - - - - - - - - - - -3 South Asian? (for example, East Indian, Sri Lankan etc) - - - - - - - - - - - - - -- - 4 West Asian? (for example Iranian, Afghan etc?)- - - - - - - - - - -- - - - - - - - - - - 5 Filipino? - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - 6 Latin American? - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - - - - -7 Black?- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 8 Arab? - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 9 Japanese? - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - -  - 10 Korean? - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - -11 Aboriginal (that is, North American Indian, Metis or Inuit?) - - - - - - - - - - - -12 Or another group (please specify)____________________________ Refused - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 14 Don’t know - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 15  --  Table A.4 The Naranjo adverse drug reaction probability scale  The ADR is assigned to a probability category from the total score as follows: definite if the overall score is 9 or greater, probable for a score of 5-8, possible for 1-4 and doubtful if the score is 0.  From Naranjo C.A. et al., A method for estimating the probability of adverse drug reactions, Clinical Pharmacology and Therapeutics, 1981. 30(2): p.239-45. Reprinted with permission from RightsLink.     152 Table A.5 Roussel-Uclaf causality assessment method hepatocellular injury scale Subject Information   1. Temporal relationship of start of drug to ALT > 2x ULN   Score Initial treatment 5–90 days; subsequent treatment course: 1–15 days  2 Initial treatment <5 or >90 days; subsequent treatment course: >15 days  1 From cessation of drug: <15 days, or <15 days after subsequent treatment 1 Otherwise  0 2. After drug cessation- difference between peak ALT and upper limits normal  Decreases >50% within 8 days  3 Decreases >50% within 30 days  2 No information or decrease >50% after >30 days, or inconclusive 0 Decrease <50% after 30 days or recurrent increase -2 3. Risk factors  No alcohol use 0 Alcohol use 1 Age <55 years   0 Age >55 years  1 4. Concomitant drug  No concomitant drug administered 0 Concomitant drug with suggestive or compatible time of onset  -1 Concomitant known hepatotoxin with suggestive or compatible time of onset   -2 Concomitant drug with positive rechallenge or validated diagnostic test    -3 5. Nondrug causes: Six are primary: recent hepatitis A, B, or C, biliary obstruction, acute alcoholic hepatitis (AST > 2x ALT), recent hypotension. Secondary group: Underlying other disease; possible CMV, EBV or HSV infection   All primary and secondary causes reasonably ruled out:  2 All 6 primary causes ruled out  1 4 or 5 primary causes ruled out  0 < 4 primary causes ruled out (max. negative score for items 4 and 5: –4) -2 Nondrug cause highly probable  -3 6. Previous information on hepatotoxicity of the drug in question  Package insert or labelling mention  2 Published case reports but not in label  1 Reaction unknown  0 7. Rechallenge  Positive (ALT doubles with drug in question alone)  3 Compatible (ALT doubles with same drugs as given before initial reaction)   +1 1 Negative (Increase in ALT but <2x ULN, same conditions as when  reaction occurred)  -2 Not done, or indeterminate result  0 Total (range of algebraic sum: –8 to +14)  Score Interpretation: Highly probable >8; Probable 6–8;  Possible  3–5;  Unlikely 1–2;  Excluded  <0 From Toxicologic Pathology, 33 (1), Lee, W.M. & Senior, J.R., Recognizing drug-induced liver injury: current problems, possible solutions, p.155-64. Reprinted with permission from SAGE Publications.       153 Table A.6 Genome-wide array quality control metrics  * X and Y chromosome and mitochondrial SNPs ignored for this metric. SNP: single nucleotide polymorphism, AA/BB: homozygote, AB: heterozygote.     Metric Cut-off value Heterozygous Excess* (More heterozygotes expected under Hardy Weinberg Equilibrium) < -0.3 and > 0.2 Cluster Separation  < 0.3 Multiple Minor-Allele Homozygotes (AA or BB R Dev) ≥ 0.05 Mean Normalized Intensity  ≤ 0.25 Wide Homozygote Clusters (AA or BB T Dev)  ≥ 0.05 Wide Heterozygote Clusters (AB T Mean)  < 0.2 or ≥ 0.8 Replication errors > 3 Call Rate < 0.95  154 Table A.7 Candidate Genes, Chromosome Number, Tagging SNP information and Rationale for Inclusion  SNP Gene Chr. Tagging SNP (r2 valuea) Rationale for Inclusion Reference rs1136774 CTSS (Cathepsin S) 1 rs12125650 (1) IFN-β response [76] rs12044852 GPC5 (Glypican 5) 1 rs10924106 (0.97) IFN-β response [74, 208] rs9301789 GPC5 (Glypican 5) 1 rs9523546 (1) IFN-β response [74] rs12089335 GBP1 (Guanylate Binding Protein 1, Interferon-Inducible) 1 rs10922563 (0.98) IFN-β response [209] rs4131514 ADAR (Adenosine Deaminase, RNA-Specific) 1 rs4474240 (0.94) IFN-β response [74] rs2229857 ADAR (Adenosine Deaminase, RNA-Specific) 1  IFN-β response [75] rs1800871 IL10 (Interleukin-10) 1  Docetaxel DILI / IFN-β response DILI: [212]. IFN-β: [209, 213] rs1800896 IL10 (Interleukin-10) 1  IFN-β response [213] rs231775 CTLA4 (cytotoxic T-lymphocyte-associated protein 4) 2 rs231779 (0.97) Immunosuppressive DILI [214] rs728005 PRKR (Protein Kinase R) 2 rs890579 (0.96) IFN-β response [74] rs676210 APOB (Apolipoprotein B) 2  Ximelagatran DILI [133] rs1368576 WNT7A (Wingless-type MMTV integration site family, member 7A)  3 rs6788472 (0.81) Ximelagatran DILI [133] rs10510779 ERC2 (ELKS/RAB6-Interacting/CAST Family Member 2) 3 rs17825189 (0.96) IFN-β response [74] rs4855469 FAM19A1/TAFA1 (Family With Sequence Similarity 19 (Chemokine (C-C Motif)-Like), Member A1) 3 rs9833308 (0.93) IFN-β response [74] rs60834 SYN2 (Synapsin II) 3 rs92090 (1) IFN-β response [74] rs795000 SYN2 (Synapsin II) 3 rs310766 (0.98) IFN-β response [74] rs9828519 SLC9A9 (Solute Carrier Family 9, Subfamily A, Member 9) 3 rs2800 (1) IFN-β response (non response) [215] rs2019978 CASP3 (Caspase 3, Apoptosis-Related Cysteine Peptidase) 4  IFN-β response [209] rs4466137 HAPLN1 (Hyaluronan And Proteoglycan Link Protein 1) 5  IFN-β response [74] rs3129900 C6orf10 (Chromosome 6 Open Reading Frame 10) 6  Lumiracoxib DILI [135] rs2523822 HLA-A (Major Histocompatibility Complex, Class I, A) 6  Amoxicillin-clavulanate DILI [134] rs2395029  HLA-B (Major Histocompatibility 6  Flucloxacillin DILI [132]  155 SNP Gene Chr. Tagging SNP (r2 valuea) Rationale for Inclusion Reference Complex, Class I, B) rs9270986 HLA-DRB1 (Major Histocompatibility Complex, Class II, DR Beta 1) 6  Lumiracoxib DILI [135] rs9274407 HLA-DRB1 (Major Histocompatibility Complex, Class II, DR Beta 1) 6  Amoxicillin-clavulanate DILI [134] rs3135388  HLA-DRB1 (Major Histocompatibility Complex, Class II, DR Beta 1) 6  Amoxicillin-clavulanate DILI [134] rs9275141 HLA-DQB1 (Major Histocompatibility Complex, Class II, DQ Beta 1) 6  Ximelagatran DILI [133] rs9267992 NOTCH4 (Notch 4) 6  Amoxicillin-clavulanate DILI [134] rs1799964  TNF alpha (Tumor Necrosis Factor Alpha) 6  Flucloxacillin DILI [132] rs1800629  TNF alpha (Tumor Necrosis Factor Alpha) 6  Flucloxacillin DILI [132] rs1800630  TNF alpha (Tumor Necrosis Factor Alpha) 6  Flucloxacillin DILI [132] rs361525  TNF alpha (Tumor Necrosis Factor Alpha) 6  Flucloxacillin DILI [132] rs2227956  HSPA1L  (Heat Shock Protein 1-Like) 6  Flucloxacillin DILI [132] rs12153855 TNXB (Tenascin XB) 6  Lapatinib DILI [136] rs2071543 PSMB8 (Proteasome Subunit, Beta Type, 8) 6  IFN-β response [76] rs4722897 CHN2 (Chimerin 2) 7 rs2214569 (0.81) Ximelagatran DILI [133] rs776746 CYP3A5 (Cytochrome P450 3A5)  7 rs10242455 (1) Acetaminophen induced ALF [210] rs4728142 IRF5 (Interferon regulatory Factor 5) 7  IFN-β response [211] rs11787532 ZFHX4 (Zinc Finger Homeobox 4) 8 rs7822914 (0.96) IFN-β response [75] rs1495741 NAT2 (N-Acetly Transferase 2) 8  Anti-Tuberculosis DILI [217] rs733254 ZFAT (Zinc Finger And AT Hook Domain Containing) 8  IFN-β response [75] rs1051922 IFNB1 (Interferon, Beta 1, Fibroblast) 9 rs1857651 (0.97) IFN-β response [218] rs1121979 OGN (Osteoglycin) 9 rs10123342 (1) IFN-β response [74] rs10501154 TRAF6 (TNF Receptor-Associated Factor 6, E3 Ubiquitin Protein Ligase) 11 rs10501155 (1) IFN-β response [74] rs2306283 SLCO1B1 (Solute Carrier Organic Anion Transporter 12 rs6487213 (0.87) Rifampin DILI [143]  156 SNP Gene Chr. Tagging SNP (r2 valuea) Rationale for Inclusion Reference Family, Member 1B1) rs7308076 CIT (Citron Rho-Interacting Serine/Threonine Kinase) 12 rs4767848 (0.99) IFN-β response [75] rs2430561 IFNG (Interferon, Gamma) 12 rs971545 (0.92) IFN-β response [218] rs10492199 IFNG (Interferon, Gamma) 12  IFN-β response [74] rs9527281 STARD13 (Star-Related Lipid Transfer (START), Domain Containing 13) 13 rs9527285 (0.97) IFN-β response [75] rs4128599 NPAS3 (Neuronal PAS Domain Protein 3) 14  IFN-β response [74] rs10162905 PIAS1 (Protein Inhibitor Of Activated STAT, 1) 15 rs75395345 (0.83) IFN-β response [209] rs1800469 TGFB1 (Transforming Growth Factor, Beta 1) 19  IFN-β response [218] rs3736775 POLR3F (Polymerase (RNA) III (DNA Directed) Polypeptide F) 20 rs3818217 (0.92) IFN-β response [74] rs2070424 SOD1 (Superoxide Dismutase 1, Soluble) 21 rs2833479 (1) Anti-Tuberculosis DILI [216] rs1012335 IFNAR1 (Interferon (Alpha, Beta And Omega) Receptor 1) 21 rs2253413 (0.96) IFN-β response [218] rs2248202 IFNAR2 (Interferon (Alpha, Beta And Omega) Receptor 2) 21  IFN-β response [75] A: r2 value calculated using the 1000 Genomes Project EUR population             157 Table A.8 Assessment of clinical and demographic characteristics associated with interferon-β induced liver injury (discovery, replication and combined) Variable Discovery Cases              (n = 38) [1] Replication Cases               (n = 18) [2] P-value [1] vs. [2] Combined Cases                   (n = 56) [1+2] Female sex, n (%) 30 (78.9) 15 (83.3) 0.501 45 (80.4) Relapsing remitting disease at IFN-β start, no. (%) 34 (89.5) 11 (100) (n=11) 0.562 45 (91.8) (n=49) Age at DILI onset, median y (IQR) 42.5 (31, 51) 42 (32.5, 49) 0.902 42 (31.5, 50.5) Time to DILI onset, median months (IQR) 3 (1, 17.5) 12.5 (3, 52.25) 0.012 3.5 (2, 23.5) Outcome of liver injury, n (%)  (n = 16) 0.001 (n = 54) D/C IFN-β therapy, liver enzymes normalized 22 (57.9) 3 (18.8)  25 (46.3) D/C IFN-β therapy, liver enzymes did not normalize 10 (26.3) 6 (37.5)  16 (29.6) D/C IFN-β therapy, unknown liver enzymes outcome 1 (2.6) 7 (43.8)  8 (14.8) Cont. IFN-β therapy, enzymes normalized 3 (7.9) 0  3 (5.6) Cont. IFN-β therapy, enzymes did not normalize 2 (5.3) 0  2 (3.7) Pattern of liver injury, n (%)  (n = 28) (n = 7) 0.019 (n = 35) Hepatocellular 27 (96.4) 4 (57.1)  31 (88.6) Cholestatic 1 (3.6) 2 (28.6)  3 (8.6) Mixed 0 1 (14.3)  1 (2.9) RUCAM Scale Scoring, n (%) (n = 37) (n = 18) 0.302 (n = 55) Highly probable (>8) 3 (8.1) 1 (5.6)  4 (7.3) Probable (6-8) 15 (40.5) 6 (33.3)  21 (38.2) Possible (3-5) 19 (51.4) 9 (50)  28 (50.9) Unlikely (1-2) 0 2 (11.1)  2 (3.6) Naranjo Scale Scoring, n (%) (n = 37) (n = 18) 0.068 (n = 55) Definite (>9) 2 (5.3) 2 (11.1)  4 (7.3) Probable (5-8) 26 (68.4) 7 (38.9)  33 (60) Possible (1-4) 9 (23.7) 9 (50)  18 (32.7) DILI Severity, n (%)  (n = 16) 0.01 (n = 54) Mild  37 (97.4) 11 (68.75)  48 (88.8) Moderate  1 (2.6) 3 (18.75)  4 (7.4) Severe 0 1 (6.25)  1 (1.9) Transplantation 0 1 (6.25)  1 (1.9) Peak level, median (IQR)     ALT (U/l) 274 (197.5, 390.8) 366 (289, 895.5) 0.004 320 (212, 447) AST (U/l) 151.5 (109.5, 253.5) 242.5 (156.75, 536.75) 0.052 181 (112.25, 287.75) ALP (U/l) 93 (70.5, 107) 151 (95, 220) 0.004 97 (72.5, 119.75) TBILI (µmol/l) 12 (9.4, 15) 2.3 (0.6, 6.0) 0.002 11 (6, 15) Max ALT or AST x ULN, median (IQR) (range) 6.8 (5.9, 9.1)          (5, 28) 9.7 (6.9, 16.7)   (5.4, 57.1) 0.041 7.4 (6.07, 12.84) (5, 57.11)  158 Categorical data: Fisher Exact Test. Continuous data: Mann-Whitney U Test. DILI: Drug induced liver injury, D/C: discontinued, RUCAM: Roussel Uclaf Causality Assessment Method. A: Alcohol consumption (yes/no), B: Smoked 100 cigarettes in lifetime (yes/no). C: Questionnaires not administered to replication cohort. The number of patients is specified if less than the total for cases/controls for the respective stage of analysis.                   Variable Discovery Cases              (n = 38) [1] Replication Cases               (n = 18) [2] P-value [1] vs. [2] Combined Cases                   (n = 56) [1+2] Autoantibody testing available, n (%) 14 (36.8) 10 (55.6) 0.25 24 (42.9) Autoantibody positive, n (%) 5 (35.7) 3 (30) 0.703 8 (33.3) Alcohol consumption, no. (%)A 16 (57.1) (n=28) NAC NAC 16 (57.1) (n=28) Ever-smoker, no. (%)B 5 (55.6) (n=9) NAC NAC 5 (55.6) (n=9)  159 Table A.9 Clinical and biochemical variables of autoantibody positive interferon-β induced liver injury cases  Variable Auto-AB Positive      (n = 8) Auto-AB Negative (n = 16) P-value Female sex, n (%) 7 (87.5) 15 (93.8) 0.602 Age at DILI onset, median y (IQR) 38 (29.5, 45) 41 (35.25, 52.5) 0.320 Time to DILI onset, median months (IQR) 2.5 (1, 27.25) 3 (1.25, 19.75) 0.569 IFN-β product, no. (%)   0.323 1a IM (30 mcg once weekly) 0 4 (25)  1a SC (22 mcg 3x weekly) 0 1 (6.3)  1a SC (44 mcg 3x weekly) 3 (37.5) 6 (37.5)  1b (250 mcg every other day) 5 (62.5) 5 (31.3)  Outcome of liver injury, n (%)   0.076 D/C IFN-β therapy, liver enzymes normalized 5 (62.5) 2 (12.5)  D/C IFN-β therapy, liver enzymes did not normalize 2 (25) 6 (37.5)  D/C IFN-β therapy, unknown outcome 1 (12.5) 7 (43.8)  Cont. IFN-β therapy, liver enzymes normalized 0 1 (6.3)  Cont. IFN-β therapy, liver enzymes did not normalize 0 0  Pattern of liver injury, n (%)  n = 5 n = 10 0.562 Hepatocellular 5 (100) 8 (80)  Cholestatic 0 1 (10)  Mixed 0 1 (10)  RUCAM Scale Scoring, n (%)   0.306 Highly probable (>8) 0 1 (6.3)  Probable (6-8) 5 (62.5) 5 (31.3)  Possible (3-5) 3 (37.5) 10 (62.5)  Naranjo Scale Scoring, n (%)   0.231 Definite (>9) 0 3 (18.8)  Probable (5-8) 5 (62.5) 5 (31.3)  Possible (1-4) 3 (37.5) 8 (50)  DILI Severity, n (%)   0.127 Mild  5 (62.5) 15 (93.8)  Moderate  2 (25) 1 (6.3)  Severe 1 (12.5) 0  Peak level, median (IQR)    ALT (U/l) 399.5 (288, 666.25) 288 (212, 484) n = 15 0.357 AST (U/l) 278 (166, 386) 203 (127.75, 423) 0.769 ALP (U/l) 107 (88, 136) 93.5 (69.5, 135.5) 0.360 TBILI (µmol/l) 12 (2.3, 18) 6 (3.2, 4) 0.535 Max ALT or AST x ULN, median (IQR) (range) 12.87 (8.89, 25.4) 7.33 (5.75, 13.3) 0.045 Time for ALT to normalize, median days (IQR) 299 (129.5, 501.5) (n = 5) 57 (21, 112) (n =3) 0.036 Autoimmune antibody positive, n (%)    Antismooth muscle positive 1 (12.5) 0   160 Variable Auto-AB Positive      (n = 8) Auto-AB Negative (n = 16) P-value Antinuclear positive 5 (62.5) 0  Antimitochondrial positive 1 (12.5) 0  Antimitochondrial & antinuclear positive 1 (12.5) 0  Categorical data: Fisher Exact Test. Continuous data: Mann-Whitney U Test. DILI: Drug induced liver injury, D/C: discontinued, RUCAM: Roussel Uclaf Causality Assessment Method. The number of patients is specified if less than the total for cases/controls for the respective stage of analysis.   Table A.10 Summary of genome-wide association studies of genetic susceptibility to drug induced liver injury  No. of participants Most Significantly Associated Marker Ref Drug Cases ControlsA Gene MAF (%) OR (95%CI) P Ximelagatran                  (thrombin inhibitor) 74 130 treated HLA DRB1*07 8.5 4.4 (2.2-9.9) 7.3x10-8 [133] Flucloxacillin                            (antibiotic) 51 282 untreated HLA-B*5701 6 80.6 (22.8-284.9) 9.0x10-19 [132] Lumiracoxib                        (COX-2 inhibitor) 41 175 treated HLA DRB1*1501 15 5.0 (3.6-7.0) 6.8x10-25 [135] Lapatinib (kinase inhibitor) 37 286 treated HLA DQA1*02:01 40 9.0 (3.2-27.4) 8.0x10-5 [136] Amoxicillin-Clavulanate (antibiotic) 201 532 untreated HLA DRB1*1501 16 2.8 (2.1-3.8) 3.5x10-11 [134] Multi-drug  783 3001 untreated STAT4 B 1.5 (1.2-1.8) 4.5x10-4 [223] Platinum-based compounds 502 202 treated LRRC3 6.2 2.6 (1.7-3.9) 2.6x10-5 [297] Pazopanib (kinase inhibitor) 429 950 treated RAGE B 1.8 (1.5-2.1) 7.7x10-9 [298] Interferon-β (disease modifying agent)C 56 126 treated SYT14 / IRF6 21 9.8 (4.0-24.1) 9.4x10-9 NA OR: Odds ratio, Ref: reference, HLA: Human leukocyte antigen, COX: cyclooxygenase, WGS = whole genome sequencing, NA: not applicable. AControls are either treated with the drug listed in the first column “treated” or are population controls meaning they were not exposed to the drug “untreated”. BNo MAF reported. CRepresents the current study included in this dissertation (Chapter 4).    	

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0228891/manifest

Comment

Related Items