UNIVERSITY OF CALGARY Non-medical reasons for colectomy among ulcerative colitis patients by María Eugenia Negrón A THESIS SUBMITTED TO THE FACULTY OF GRADUATE STUDIES IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY VETERINARY MEDICAL SCIENCES GRADUATE PROGRAM CALGARY, ALBERTA July, 2014 © María E. Negrón 2014 Abstract Approximately 16% of ulcerative colitis (UC) patients will have a colectomy within 10 years of their diagnosis of UC. Due to the high morbidity and mortality associated with colectomy procedures, clinicians and patients strive to avoid surgery. The decline of colectomy rates over the last 6 decades has been mainly attributed to advances in medical management of UC. Clostridium difficile infection and colorectal neoplasia have also been associated with driving the risk of colectomy. This dissertation focuses on describing the effect of C. difficile and colorectal neoplasia on the risk of colectomy for UC patients. Chapters 2 and 3 establish the cumulative risk of acquiring C. difficile infection after the diagnosis of UC and demonstrated that C. difficile infections increase the short- and long-term risk of colectomy. Further, both Chapters 2 and 3 demonstrate that C. difficile infections increased the risk of postoperative complications following colectomy for UC. Chapter 4 demonstrate that the incidence of colectomy for colorectal dysplasia and cancer has remained stable over time, and showed that the observed decrease in colectomy rates is most likely due to a decrease in colectomies from medically refractory disease. Finally, Chapter 5 determined the cost-effectiveness of different surveillance colonoscopy intervals for colorectal dysplasia among a subgroup of inflammatory bowel disease (IBD) patients with a higher risk of developing colorectal dysplasia and cancer; i.e., patients with IBD and primary sclerosing cholangitis (PSC). ii Preface This dissertation consists of four manuscripts – two of which have been accepted for publication, and two are ready for submission. For all four manuscripts, the first author was involved with study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, and critical revision. This was done under the guidance of the supervisor and co-supervisor. All authors provided critical reviews of the manuscripts and contributed intellectual content. After receiving written permission from the publishers and co-authors, all four manuscripts were reproduced in their entirety as chapters in this dissertation. Manuscript I) Negrón ME, Barkema HW, Rioux K, De Buck J, Checkley S, Proulx MC, Frolkis A, Beck PL, Dieleman L, Panaccione R, Ghosh S, Kaplan GG. Clostridium difficile infection worsens prognosis of ulcerative colitis. Accepted: Canadian Journal of Gastroenterology and Hepatology. Manuscript II) Negrón ME, Barkema HW, Kaplan GG. Clostridium difficile infection diagnosed early in ulcerative colitis patients impacts disease related outcomes: A population-based inception cohort study. Will be submitted to: Gut. Manuscript III) Negrón ME, Barkema HW, Kaplan GG. Changes in the annual incidence of colectomy for colorectal neoplasia among UC patients: A population-base study. Will be submitted to Clinical Gastroenterology and Hepatology iii Manuscript IV) Negrón ME, Kaplan GG, Barkema HW, Eksteen B, Clement F, Manns BJ, Coward S, Panaccione R, Ghosh S, Heitman SJ. Colorectal cancer surveillance in patients with inflammatory bowel disease and primary sclerosing cholangitis: An economic evaluation. Accepted: Inflammatory Bowel Disease. iv Acknowledgements I would like to express my most sincere gratitude to my supervisor and cosupervisor, Drs. Herman W. Barkema and Gilaad G. Kaplan. Herman and Gil, I admire both of you as my mentors, colleagues and as a people. Thank you for your patience, for believing in me and guiding me during these last 4 years. You have no idea how much I learned from both of you! I also want to thank my committee members: Drs. Sylvia Checkley, Jeroen De Buck, Bertus Eksteen and Kevin Rioux for their guidance and support. Additionally, I want to thank Drs. Fiona Clement and Steven Heitman who introduced me to the world of health economics. I hope I make you both proud! I would like to show my deepest appreciation to Karin Orsel, who has been a friend and a mentor during these last 4 years. Thank you for being available to chat and for giving guidance when I needed it. I would like to acknowledge my lab members, Alex, Ellen and Stephanie for all those stimulating discussions and for all the fun we had during these last few years. In particular, Alex, I admire your wisdom and your attitude towards life. Thank you for taking time to brainstorm with me. To the amazing people I get to call my friends in Calgary. You guys are the epitome of what a family away from home means. Most importantly, Andrew, your care, support and confidence in me were all that kept me going at times during these last few months. Thank you, love. I would like to acknowledge the DIMR departments for providing the data for the patients with ulcerative colitis living in the Calgary Health Zone, and Calgary Laboratory Services for providing the Clostridium difficile test results. Finally, I would like to recognize the financial assistance of the University of Calgary (Teaching Assistantships, v Graduate Research Scholarship and Travel Grants) and the Alberta Inflammatory Bowel Disease Consortium (Scholarship). vi Dedication To my family… the one I was given and the one I chose along the way. vii Table of Contents Abstract ................................................................................................................................ii Preface ............................................................................................................................... iii Table of Contents............................................................................................................. viii List of Tables .....................................................................................................................xii List of Symbols, Abbreviations and Nomenclature.......................................................... xiv Epigraph............................................................................................................................ xvi CHAPTER ONE: INTRODUCTION ................................................................................. 1 1.1 General Overview ...................................................................................................... 2 1.2 Background of Ulcerative Colitis .............................................................................. 6 1.2.1 Burden of ulcerative colitis: incidence, prevalence and economic impact ....... 7 1.2.2 Risk factors........................................................................................................ 8 1.2.3 Natural history of the disease ............................................................................ 9 1.2.4 Medications ..................................................................................................... 10 1.2.5 Surgery ............................................................................................................ 11 1.3 Clostridium difficile ................................................................................................. 13 1.4 Colorectal Dysplasia and Cancer............................................................................. 16 1.5 Thesis Objectives and Hypothesis ........................................................................... 20 CHAPTER TWO: CLOSTRIDIUM DIFFICILE INFECTION WORSENS THE PROGNOSIS OF ULCERATIVE COLITIS ........................................................... 24 2.1 Abstract.................................................................................................................... 25 2.2 Introduction ............................................................................................................. 26 2.3 Materials and Methods ............................................................................................ 27 2.3.1 Data sources .................................................................................................... 27 2.3.2 Study population ............................................................................................. 27 2.3.3 Selection of cases and controls ....................................................................... 28 2.3.4 Exposure .......................................................................................................... 29 2.3.5 Covariates ........................................................................................................ 29 2.3.6 Statistical analysis ........................................................................................... 30 2.4 Results ..................................................................................................................... 32 2.5 Discussion................................................................................................................ 34 2.6 Acknowledgements ................................................................................................. 37 2.7 Financial Disclosures............................................................................................... 38 2.8 References ............................................................................................................... 39 2.9 Appendix 2.A Post-operative complications classification ..................................... 50 2.10 Appendix 2.B Comorbidity classifications............................................................ 51 2.11 Appendix 2.C Postoperative complications observed among patients tested for C. difficile. ............................................................................................................. 52 2.12 Appendix 2.D Sensitivity analysis: Logistic regression results of emergent colectomy with C. difficile diagnosis in hospital or 14 days prior to admission as primary exposure. .............................................................................................. 54 2.13 Appendix 2.E Sensitivity analyses: Logistic regression results of any postoperative complication and infectious post-operative complication with C. viii difficile diagnosis in hospital or 14 days prior to admission as primary exposure. ................................................................................................................ 55 2.14 Appendix 2.F Sensitivity analysis: Logistic regression results of emergent colectomy including those not tested for C. difficile. ............................................ 56 2.15 Appendix 2.G Sensitivity analyses: Logistic regression results of any postoperative complication and infectious post-operative complication including those not tested for C. difficile. .............................................................................. 57 CHAPTER THREE: CLOSTRIDIUM DIFFICILE INFECTION DIAGNOSED EARLY IN ULCERATIVE COLITIS PATIENTS IMPACTS DISEASE RELATED OUTCOMES: A POPULATION-BASED INCEPTION COHORT STUDY ..................................................................................................................... 58 3.1 Abstract.................................................................................................................... 59 3.2 Introduction ............................................................................................................. 61 3.3 Materials and Methods ............................................................................................ 62 3.3.1 Study population ............................................................................................. 62 3.3.1.1 Validation cohort ................................................................................... 62 3.3.1.2 Inception cohort of UC patients ............................................................ 63 3.3.2 Data sources .................................................................................................... 63 3.3.2.1 Validation of the ICD-10 code for C. difficile diagnosis among UC patients .................................................................................................... 63 3.3.2.2 Inception cohort of patients with UC .................................................... 64 3.3.3 Development of the UC inception cohort ....................................................... 65 3.3.4 Outcome and predictor variables .................................................................... 66 3.3.5 Analyses .......................................................................................................... 67 3.3.5.1 Validation of the ICD-10 code for C. difficile diagnosis among UC patients .................................................................................................... 67 3.3.5.2 Effect of C. difficile on UC incident cases ............................................ 68 3.4 Results ..................................................................................................................... 69 3.4.1 Validation of the ICD-10 code for C. difficile diagnosis among UC patients ............................................................................................................. 69 3.4.2 Effect of C. difficile on UC incident cases ...................................................... 70 3.5 Discussion................................................................................................................ 72 3.6 References ............................................................................................................... 76 3.7 Appendix 3.A List of postoperative complications and the respective ICD-10 code by category .................................................................................................... 88 3.8 Appendix 3.B Charlson comorbidities .................................................................... 93 3.9 Appendix 3.C Classification of colectomy approach and ICD-10 codes ................ 95 3.10 Appendix 3.D Types of postoperative complications ........................................... 97 3.11 Appendix 3.E Characteristics of colectomy patients stratified by developing infectious complications in hospital ...................................................................... 98 3.12 Appendix 3.F Results of multilevel logistic regression model for the development of infectious complications with C. difficile as primary predictor (hospital was included as a random effect in the model) ...................................... 99 ix CHAPTER FOUR: CHANGES IN THE ANNUAL INCIDENCE OF COLECTOMY FOR COLORECTAL NEOPLASIA AMONG PATIENTS WITH ULCERATIVE COLITIS: A POPULATION-BASED COHORT ........................ 100 4.1 Abstract.................................................................................................................. 101 4.2 Introduction ........................................................................................................... 102 4.3 Materials and Methods .......................................................................................... 103 4.3.1 Data sources and study population ................................................................ 103 4.3.1.1 Data Integration, Measurement, and Reporting Hospital Discharge Abstract ................................................................................................. 103 4.3.1.2 Alberta Health Care Insurance Plan (AHCIP)..................................... 104 4.3.2 Identification of UC colectomy patients in CHZ hospitals and data collection ....................................................................................................... 104 4.3.3 Identification of UC patients at risk for colorectal cancer in CHZ using AHCIP ........................................................................................................... 106 4.3.4 Analyses ........................................................................................................ 107 4.3.4.1 Risk factors for colectomy for colorectal dysplasia or cancer ............ 107 4.3.4.2 Incidence of colectomy for colorectal dysplasia or cancer and medically refractory disease ................................................................. 107 4.4 Results ................................................................................................................... 108 4.4.1 Risk factors for colectomy for colorectal dysplasia or cancer ...................... 108 4.4.2 Incidence of colectomy for colorectal neoplasia and medically refractory disease ............................................................................................................ 109 4.5 Discussion.............................................................................................................. 110 4.6 References ............................................................................................................. 114 4.7 Appendix 4.A Comorbidity classifications ........................................................... 123 CHAPTER FIVE: COLORECTAL CANCER SURVEILLANCE IN PATIENTS WITH INFLAMMATORY BOWEL DISEASE AND PRIMARY SCLEROSING CHOLANGITIS: AN ECONOMIC EVALUATION .................. 124 5.1 Abstract.................................................................................................................. 125 5.2 Introduction ........................................................................................................... 126 5.3 Materials and Methods .......................................................................................... 127 5.3.1 Model overview ............................................................................................ 127 5.3.2 Model assumptions........................................................................................ 128 5.3.3 Model validation ........................................................................................... 129 5.3.4 Parameters ..................................................................................................... 130 5.3.4.1 Dysplasia and cancer related risk ........................................................ 130 5.3.4.2 Median time to liver transplant and graft loss ..................................... 130 5.3.4.3 Mortality .............................................................................................. 130 5.3.4.4 Colonoscopy performance characteristics. .......................................... 131 5.3.4.5 Costs .................................................................................................... 131 5.3.4.6 Utilities ................................................................................................ 132 5.3.5 Analyses ........................................................................................................ 132 5.4 Results ................................................................................................................... 134 5.4.1 Base-case analysis ......................................................................................... 134 5.4.2 Sensitivity analyses ....................................................................................... 135 5.4.3 Probabilistic sensitivity analysis ................................................................... 136 x 5.5 Discussion.............................................................................................................. 137 5.6 References ............................................................................................................. 142 5.7 Appendix 5.A Summary of health states included in the model ........................... 158 CHAPTER SIX: DISCUSSION OF WORK PRESENTED ........................................... 159 6.1 Overview of main findings .................................................................................... 160 6.2 Limitations ............................................................................................................. 163 6.2.1 Limitations of observational studies ............................................................. 163 6.2.2 Limitations of decision analysis study .......................................................... 167 6.3 Implications on Clinical Management and Public Health ..................................... 168 6.4 Future Research ..................................................................................................... 169 6.5 Conclusions ........................................................................................................... 171 REFERENCES ................................................................................................................ 173 xi List of Tables Table 2.1. Characteristics of ulcerative colitis patients admitted to hospital stratified by flare (medically responsive) and emergent colectomy. Emergent colectomy patients were further stratified into development of any postoperative complication and development of infectious complication ....................................... 44 Table 2.2. Logistic regression results of emergent colectomy with C. difficile diagnosis as primary exposure. .................................................................................. 47 Table 2.3. Logistic regression results for emergent colectomy patients having any post-operative complication and infections post-operative complication with C. difficile diagnosis as primary exposure. .................................................................... 48 Table 3.1. Characteristics of patients diagnosed with ulcerative colitis in Alberta, Canada, between April 1st, 2003 and March 31st, 2010 stratified by colectomy. ...... 81 Table 3.2. Results of the competing risk regression and the Cox proportional hazard models for time to colectomy and time to death since UC diagnosis with prior C. difficile diagnosis as the primary predictor................................................................ 83 Table 3.3. Characteristics of colectomy patients stratified by developing any postoperative complication in hospital ...................................................................... 84 Table 3.4. Results of multilevel regression models for the development of any postoperative complication and a postoperative infectious complication with C. difficile as primary predictor (facility was included as a random effect in the models) ...................................................................................................................... 85 Table 4.1. Characteristics of ulcerative colitis patients admitted to Calgary Health Zone hospitals that had a colectomy for colorectal dysplasia or cancer versus colectomy for medically refractory UC. .................................................................. 118 Table 4.2. Logistic regression results comparing colectomy patients’ characteristics (colorectal dysplasia or cancer vs. medically refractory) with year of colectomy as primary exposure. ................................................................................................ 120 Table 5.1. Base-case parameter estimates and distributions and ranges used in the probabilistic sensitivity analysis .............................................................................. 150 Table 5.2. The number of patients who developed dysplasia and cancer, as well as the number of colonoscopies performed over a lifetime horizon for a hypothetical 1,000 average IBD-PSC patient cohort ................................................................... 152 Table 5.3. Base case-analysis results (in CAN$) for cost-utility analysis ....................... 153 Table 5.4. Scenario-analysis results (in CAN$) for cost-utility analysis ........................ 154 xii List of Figures and Illustrations Figure 2.1. Flow diagram illustrating the inclusion and exclusion criteria for identifying patients admitted with a flare for ulcerative colitis and tested Clostridium difficile infection in hospital and up to 90 days prior to admission....... 49 Figure 3.1. Flow diagram of patient selection criteria ....................................................... 86 Figure 3.2. Survival curves for the proportion of colectomy free UC patients (a) and survival of UC patients (b) stratified by having C. difficile diagnosis (red line) and no C. difficile diagnosis (blue line) ..................................................................... 87 Figure 4.1. Annual incidence of colectomy for colorectal dysplasia or cancer (green line), and medically refractory disease (red line) among UC patients living in the Calgary Health Zone between 1997 and 2009. The graph presents annual adjusted incidence of colectomy per 1,000 UC patients at risk* ............................. 122 Figure 5.1. Bubble diagram summarizing the structure of the Markov model, flow of patients through the health states, and the outcomes considered in this cost-utility analysis. ................................................................................................................... 155 Figure 5.2. Incremental cost-effectiveness scatterplots with 95% confidence ellipses for annual surveillance compared to biennial surveillance. The diagonal line represents a willingness to pay threshold of CAN$50,000 per quality-adjusted life-year. ................................................................................................................... 156 Figure 5.3. Cost-effectiveness acceptability curve. ......................................................... 157 xiii List of Symbols, Abbreviations and Nomenclature Symbol Definition 5-ASA ACCS AHCIP CAN CCI CD CI CLS CRC D DAD DALM DIMR 5-aminosalycilic acid Ambulatory care classification system Alberta Health Care Insurance Plan Canadian dollar Canadian Classification of Health Intervention Crohn’s disease Confidence interval Calgary Laboratory Services Colorectal cancer Days Discharge abstract database Dysplasia associated lesion or mass Data integration, measurement and reporting hospital discharge abstract Enzyme immunosorbent assay Expert opinion Generalized linear model Hazard ratio Incremental cost-effectiveness ratio International Classification of Diseases, 9th edition International Classification of Diseases, 10th edition International Classification of Diseases, 10th edition, Canadian Classification International Classification of Diseases, 10th edition, Canadian Classification International Classification of Diseases, 9th edition, Clinical Modification International Classification of Diseases, 10th edition, Clinical Modification Interquartile range Liver transplant National Institute for Health and Care Excellence Negative predictive value Not significant Odds ratio Alberta’s Physician Claims Positive predictive value Primary sclerosing cholangitis Probabilistic sensitivity analysis EIA EO GLM HR ICER ICD-9 ICD-10 ICD-9-CA ICD-10-CA ICD-9-CM ICD-10-CM IQR LT NICE NPV NS OR PC PPV PSC PSA xiv QALY RR sHR SIR SMR STROBE TNF UC UDCA USA WTP X2 Y Quality adjusted life year Risk ratio sub Hazard ratio Standardized incident rate Standardized mortality ratio Strengthening and Reporting of Observational Studies in Epidemiology Tumour necrosis factor Ulcerative colitis Ursodeoxycolic acid United States Willingness to pay Chi-squared Year xv Epigraph “Whatever you want to do, if you want to be great at it, you have to love it and be able to make sacrifices for it” - Maya Angelou xvi Chapter One: Introduction 1 1.1 General Overview Ulcerative colitis (UC) is an inflammatory bowel disease (IBD) that is characterized by chronic inflammation of the large bowel. This disease is common in industrialized nations with North America and Europe reporting the highest incidences. The increasing incidence throughout developed and developing countries has situated UC as an emerging global disease. Canada has the highest prevalence of IBD in the world.1 There are ~104,000 people living with UC and ~4,500 new diagnosis every year.2 UC is relapsing and remitting disease. The clinical course of UC is varied ranging from mild disease severity that is controlled with medications (e.g. aminosalicylates) to a severe course where patients undergo surgery to remove the colon. Overtime the prognosis for UC has improved,3 which has been attributed to better understanding of the disease, earlier diagnosis and advancements in patient management and medications. The treatment goal of UC is to optimize quality of life by inducing and maintaining remission, and preventing hospital admissions and surgery. Colectomy for UC is curative in most patients; however, physicians and patients aim to avoid colectomy due to the potential short- and long-term complications.4,5 A systematic review and meta-analysis of population-based studies demonstrated that 16% of UC patients will undergo a colectomy within 10 years of their diagnosis. The metaanalysis showed that the over the past several decades the 10-year risk of colectomy has decreased over time.6 However, this meta-analysis was not designed to explain the causes for the reduced risk of colectomy within 10 years of diagnosis. This manuscript, as well as others,7 suggests that advances in therapeutic treatments for UC may explain the decreasing incidence of colectomy in UC. However, non-medication factors, such as 2 infections (e.g. Clostridium difficile) and colorectal neoplasia, may influence the risk of colectomy. This thesis will evaluate the effect of C. difficile and colorectal neoplasia on the risk of colectomy for patients with UC. Due to the high morbidity and mortality associated with colectomy, a major research objective for clinicians and investigators is the prevention of colectomy for UC patients.8-10 Treatment for UC involves suppressing the immune system in order to control colonic inflammation. However, when these medications fail (i.e. medically refractory disease) then a colectomy is indicated. A subset of UC patients undergoes colectomy for gastrointestinal infections (e.g. C. difficile), and colorectal dysplasia and/or cancer. Clostridium difficile and colorectal neoplasia are potential modifiable risk factors for colectomy for UC. By understanding the clinical importance of these drivers of colectomy, clinicians may use this information to mitigate the burden of colectomy for UC. For this reason, this dissertation addresses important gaps in the literature on the epidemiology of C. difficile infections and colorectal neoplasia among UC patients. UC patients have a higher risk of acquiring a C. difficile infection when compared to the non-IBD population.11-15 Additionally, C. difficile infection increases the risk of surgery among UC patients, but the magnitude of the risk and its effect on short- and long-term risk of colectomy is controversial in the literature.13,16-18 In addition, most of these studies used administrative healthcare databases to identify patients with UC with C. difficile infection. While these data sources are fairly accessible and allow researchers to capture a large sample of patients, the accuracy of the C. difficile diagnostic code has not been validated for UC.11,15,18-21 Chapters 2 and 3 uses a multi-method approach to address these gaps in the literature. In Chapter 2 the effect of C. difficile infection on 3 colectomy is studied in a cohort of patients whereby every chart is reviewed to confirm the diagnosis of C. difficile. In Chapter 3 the effect of C. difficile infection on colectomy is studied in a larger sample using an administrative healthcare database after the accuracy of the diagnostic code for C. difficile was validated. Overall, these two chapters assess the risk of C. difficile infections following UC diagnosis, evaluate the effect of C. difficile infection on the short- and long-term risk of colectomy among UC patients, and determine the impact of C. difficile infections on postoperative complications. Chapter 2, titled “Clostridium difficile infection worsens the prognosis of ulcerative colitis patients” was accepted in the Canadian Journal of Gastroenterology. Chapter 3, titled “Clostridium difficile infection diagnosed early in ulcerative colitis patients impacts disease related outcomes: A population-based inception cohort study” will be submitted to Gut. UC patients also have a higher risk of developing colorectal cancer when compared to the general population. Meta-analyses have demonstrated the cumulative risk of developing colon cancer among patients with UC is 1% at 10-years and 8% at 20years.22 However, the risk overtime of colorectal cancer for UC patients is variable with some studies reporting no changes and others demonstrating that the risk of colorectal cancer has decreased over time.22-26 Inflammation is one of the most important risk factors of developing colorectal neoplasia as several studies have demonstrated a direct relationship between histological inflammation on previous biopsy specimens and the development of colorectal dysplasia or cancer.22-24 As a result, advances in the medical management of UC that have reduced chronic inflammation, also may have affected the risk of dysplasia and colon cancer among patients. As a result, the decreasing risk of 4 surgery over time among UC patients may in part be explained by the reduced need for colectomy for dysplasia and cancer. However, studies have not adequately evaluated whether the reduced risk of colectomy overtime has been influenced by a reduction in the need for colectomy for dysplasia and/or cancer. Chapter 4 titled “Changes in the annual incidence of colectomy for colorectal neoplasia among UC patients: A population-base study” addresses this gap in the literature by evaluating if the incidence of colectomy for colorectal dysplasia or cancer has changed over time when compared to colectomy due to medically refractory disease. This study will be submitted to Clinical Gastroenterology and Hepatology. The risk of colorectal dysplasia or cancer is considered to be minimal within the first 8 to 10 years after UC diagnosis.25 However, patients with UC diagnosed with concomitant primary sclerosing cholangitis (PSC) have 10 times higher the risk of developing colorectal cancer and it occurs early in the disease.26 Due to this difference in risk, contemporary guidelines for the management of IBD patients recommend surveillance colonoscopy every 1-2 year, starting 8-10 years after diagnosis.27,28 On the other hand, annual colonoscopy is recommended for IBD-PSC patients immediately after both diseases are concurrently recognized.27,28 However, the cost-effectiveness of this recommendation has never been determined and the implication of different time intervals of surveillance colonoscopies has never been assessed. Chapter 5, titled “Colorectal cancer surveillance in patients with inflammatory bowel disease and primary sclerosing cholangitis: An economic evaluation” is the first cost-utility analysis that evaluates different colorectal cancer surveillance intervals among IBD-PSC. This paper was accepted for publication in the Inflammatory Bowel Disease. 5 UC is a chronic and potentially debilitating disease that is primarily diagnosed in individuals at the prime of their lives.1 Consequently, UC imparts a significant burden to patients, their families, and to the healthcare system. Colectomy remains among the greatest burden to patients with UC. Thus, patients and physicians strive to avoid colectomy. Important advances in the medical management of UC have led to a decrease in the risk of colectomy overtime. However, other non-medication factors influence the risk of colectomy including C. difficile infection and colorectal neoplasia. Collectively the 4 studies in this thesis dissertation evaluate the impact of C. difficile infection and colorectal neoplasia on colectomy for UC with the goal of improving understanding these modifiable risk factors in order to reduce the burden of UC. 1.2 Background of Ulcerative Colitis UC is thought to be a complex polygenic disorder, meaning that it is likely associated with an interaction of multiple genes, in combination with the person’s lifestyle and environmental factors.29 Characteristic lesions of UC include continuous inflammation of colonic mucosa (beginning in the rectum and continuing proximally).30 Typically, patients with UC present with a history of rectal bleeding, tenesmus, diarrhoea and abdominal pain. The diagnosis of UC is based on the patient’s clinical history in conjunction with endoscopy and confirmation by histopathology of the affected tissue. Most patients require daily medications to induce and maintain clinical remission and when these drugs fail a colectomy (i.e. resection of the colon) is often required. The following sections will overview the epidemiology, natural history, and management of UC. 6 1.2.1 Burden of ulcerative colitis: incidence, prevalence and economic impact UC is common in industrialized nations, with Northern Europe and North America reporting the highest incidences.1 In Europe, the incidence of UC ranges from 0.7 to 9.8 cases per 100,000 person-years.1 North America has the highest incidence of UC, ranging from 3.1 to 14.6 cases per 100,000 persons-years.1 Approximately 104,000 people live with UC in Canada, and ~4,500 people are diagnosed with UC every year.2 The fact that the incidence is increasing overtime in both, developed and developing countries, suggests that UC is a global disease.1 Diagnosis occurs commonly between the second and third decade of life; however, some studies suggest a bimodal age distribution with a second smaller peak later in life.1 The prognosis of UC has improved dramatically with better patient management. The overall mortality is similar to the general population (standardized mortality ratio (SMR): 1.1; 95% confidence interval (CI): 0.9-1.2).3 The global prevalence is expected to continue to rise due to the fact that UC is often diagnosed in the young population and mortality is low. Because of the chronic nature of the disease where most patients require continuous or intermittent treatment throughout their lifetime, UC is a costly disease to patients and society. A recent meta-analysis reported that the annual direct medical costs (in-patient, out-patient, medications, physician consults, diagnostics and emergency visits) ranged from $3,374 to 11,477 and €2,210 to 10,395 per patient in the US and Europe, respectively.31 Further, between $16,442 and 55,541 is spent per surgery (includes all inpatient costs) in the USA for UC patients.31 In 2012, annual costs of IBD in Canada were estimated to be $2.8 billion, with $1.2 billion being related to direct medical costs (medications, hospitalizations, physician costs and procedures) and the remainder $1.6 7 billion is due to indirect costs (to society and the patient, i.e. productivity losses due to disability and sickness, early retirement or death).2 1.2.2 Risk factors UC arises from an interaction of multiple genes in combination with the person’s lifestyle and environmental exposures. The first clues suggesting genetic susceptibility came from studies showing that IBD occurs in “clusters” within families.32 Since then numerous genes associated with the development of UC have been identified. A genomewide analysis by Jostins et al. identified 163 susceptibility genes for IBD.33 One hundred and ten of these genes are associated with both UC and Crohn’s disease, whereas only 30 are specific for Crohn’s disease and 23 are UC-specific.33 In UC, most of these genes are associated with the immune system and intestinal barrier function.34 However, the fact that these genes are not present among all UC patients and that the incidence of IBD is rising in developing nations as they become industrialized suggest that individual lifestyle and environmental exposures influence the development of UC. One theory to explain the rise of UC in industrialized nations is the Hygiene Hypothesis, which suggests that reduced exposure to enteric microorganisms during early childhood may lead to an inappropriate immunological response upon exposure to gastrointestinal infection later in life.35 Similarly, the “hit and run” hypothesis suggests that pathogens may initiate an initial inflammatory response that, in the presence of genetic alterations is not fully controlled.36 Several environmental risk factors have been shown to influence the development of UC including smoking, appendectomy and breastfeeding.37 Environmental risk factors (e.g. diet or smoking) might influence the development of UC by shifting microbiota composition and diversity.38-41 8 1.2.3 Natural history of the disease The disease course of UC is characterized by periods of disease exacerbations (i.e., flares) followed by periods of remission. Disease activity varies from mild to severe. The majority of the patients have moderate UC, which has the classical intermittent disease course. A subset of the UC population has mild UC, experiencing one relapsing episode followed by long-term remission.42 Lastly, some patients with UC experience a severe form of UC, characterized by fulminant medically refractory disease with a need of colectomy within a short period form their diagnosis. In addition to these flares, one third of UC patients will develop extraintestinal manifestations including arthritis (e.g. ankylosing spondylitis and sacroiliitis), erythema nodosum, pyoderma gangrenosum, uveitis and PSC.43 Primary sclerosing cholangitis is a chronic cholestatic liver disease, characterized by chronic inflammation, and destruction and fibrosis of the intrahepatic and/or extrahepatic biliary tree.44 Liver transplant is the only curative option for these patients. Approximately 68% of PSC patients will also have a diagnosis of IBD with the most common type being UC.45 Most often the IBD diagnosis precedes PSC diagnosis. Their prognosis is worse than those with only UC, with a mortality risk 3 times higher than the general population.46 The median time to liver transplant or death among these patients is 12-13 years.26,47 Both disease severity and the number of relapses are influenced by the extent of disease activity at presentation and +/- progression during follow up. Based on the Montreal Classification, disease extent is stratified into 3 subgroups: proctitis, left-sided colitis, and pancolitis or extensive UC. Proctitis is limited to the rectum. Left-sided UC is 9 the most common disease extent at diagnosis and extends up to the splenic flexure. Pancolitis extends proximal to the splenic flexure and is the most severe of the three forms.48 Patients presenting at diagnosis with left-sided UC often progress to pancolitis. Additionally, age at diagnosis is an important determinant on disease extent and progression. For example, in one study, diagnosis at >40 years of age was protective for developing pancolitis.49 1.2.4 Medications Disease severity varies with some patients having a mild course that is successfully controlled by medications, whereas others suffer from refractory disease unresponsive to medication leading to surgical removal of the colon. Even among those with successfully controlled UC, the response to medications is quite varied.50 The first goal of medical treatment of UC is to suppress inflammation and induce remission while allowing the tissues to heal. The second goal is to maintain remission and prevent a relapse. Inducing and maintaining remission improves quality of life, and reduces the likelihood of colectomy. The choice of treatment varies according to disease severity, disease extent, and disease course. Additionally, physicians also take into consideration patient’s extraintestinal manifestations, age, comorbidities and patient’s preferences when determining which therapeutic approach is optimal of each patient. Five-ASA is the treatment of choice for patients with mild to moderate UC for induction as well as for remission therapy.51 A recent Cochrane Systematic Review of 11 trials involving 1,598 patients confirmed the effectiveness of 5-ASA for mild to moderate UC demonstrating it was superior to placebo for maintaining remission.52 However, several studies have reported compliance issues because patients are required to take 10 multiple doses every day. Poor adherence to medications have been associated with relapse of disease activity53 Steroids are used for patients unresponsive to 5-ASA.51 Corticosteroids are effective therapy in inducing remission of moderate-to-severe UC.54 However, corticosteroids are not effective for maintenance of remission and prolonged use of steroids is associated with multiple serious adverse events.27 Immunomodulatory drugs such as azathioprine and 6-mercaptopurine are used for patients unresponsive to 5-ASA or for patients with chronic active disease that is refractory or dependent to corticosteroids.51 However, the major limitation of immunomodulators is that they are not effective at induction of remission and are associated with toxicity (e.g. leukopenia and pancreatitis).55 Infliximab and adalimumab are monoclonal antibodies against tumour necrosis factor-alpha (anti-TNF) that is indicated for the induction and remission of UC. A network meta-analysis demonstrated similar efficacy across anti-TNF medications.56 Overall, patients were 2.45 times more likely to achieve remission (RR=2.45; 95% CI: 1.72-3.47) and 1.65 times more likely to improve their disease activity (RR=1.65; 95% CI: 1.37-1.99) when compared to patients treated with placebo.56 Additionally, another meta-analysis demonstrated that infliximab was associated with lower risk of surgery (OR=0.57; 95% CI: 0.037-0.88).57 1.2.5 Surgery Colectomy is indicated for patients who do not respond to medical treatment or for those who develop colorectal dysplasia or cancer. A restorative proctocolectomy is considered to be curative. Additionally, following colectomy, the chance of developing colorectal dysplasia or cancer is reduced dramatically. However, the procedure is associated with short-term postoperative complications including a 3.2% risk of 11 postoperative mortality.58 Additionally, long-term complications may arise including pouchitis, reduced fertility, sexual dysfunction, fecal incontinence and pouch complications.59,60 It is, therefore, not surprising that patients prefer medical management as opposed to undergoing surgery.4,5 Colectomies are commonly divided into elective or emergent procedures. Elective colectomies are performed among those with refractory disease or due to colorectal dysplasia or cancer. On the other hand, emergent colectomies are done in life-threatening situations where hospitalized patients fail to respond to medical management or experience complications and are associated with higher risk of developing postoperative complications and mortality. By 10 years following a UC diagnosis approximately 16% of the patients have undergone a colectomy.6 The highest colectomy risk is observed during the first years after diagnosis, most likely due to patients presenting with fulminant colitis.61 Still, the overall risk of surgery among UC patients is decreasing over time suggesting that earlier diagnosis, better patient management, and the development of new medications play an important role.6 However, the incidence of emergent colectomies (which account for ~50% of colectomies performed) has remained stable in the last decade, whereas elective procedures decreased significantly.7 The reason why the risk of colectomy is decreasing over time is not completely understood. Advances in medical management from the introduction of immunosuppressants in the 1990’s to the advent of anti-TNF therapies in the 2000’s likely explains the reduction in the risk of colectomy for patients with UC. However, other non-medication factors may also, in part, explain the reduced risk of colectomy. 12 There previous advances in medical management that led to reduced chronic inflammation may have also decreased the risk of developing dysplasia or cancer among patients with UC. In addition, C difficile infections influence the risk of colectomy for UC patients. Studying factors such as colorectal neoplasia and C difficile infection is important because these are potentially modifiable risk factors of the colectomy. 1.3 Clostridium difficile Currently, the most common gastrointestinal infectious microorganism associated with relapse of UC is C. difficile. Clostridium difficile is a Gram-positive spore forming anaerobe commonly associated with nosocomial acquisition and/or exposure to antibiotics. Clinical signs vary and can range from asymptomatic, to diarrhea, sepsis and even death. An increase in incidence of C. difficile infections during the last decade20,62 and outbreaks of a hyper-virulent strain in Quebec, Canada,63 have increased public awareness of the importance of this bacterium. IBD is not considered a risk factor for C.difficile infection,64 which is a completely different opinion from 30 years ago when the role of C. difficile in IBD was not considered to be important.65-69 Several epidemiological studies have consistently reported higher ocurrence of C. difficile infections among IBD patients when compared to the non-IBD population with the highest number of C. difficile infections reported among UC patients.11-15 Nguyen et al. reported an eight and three times higher prevalence of C. difficile infection among UC patients when compared to the non-IBD and CD patients, respectively.11 In the US, the incidence of C. difficile infections among UC patients has been increasing with the incidence nearly doubling between 1998 and 2004 (26.6 per 1,000 admissions in 1998 vs. 13 51.2 per 1,000 admissions in 2004). Similar findings have been reported among single center studies and other studies using administrative hospital databases from North America and the UK.12,13,16 In contrast, studies from The Netherlands and Germany reported a low prevalence of C. difficile infection among UC patients.70,71 This might indicate that geographic variation and the background C. difficile infection pressure may play a role. However, heterogeneity between studies may also be explained by the use of different diagnostic tests, case definitions and comparator groups. In addition, studies using administrative health databases have not validated the accuracy of the diagnostic code for C. difficile infection among UC patients.15 Among non-IBD patients, risk factors for C. difficile infection include antibiotic use, older age, presence of comorbidities, immunosuppression, admission to hospital and prolonged hospitalization.72 Alteration of the gut microbiota, as a result of antibiotic usage, allows C. difficile to proliferate and produce toxins causing the observed clinical signs.73 However, the association between antibiotic use and an increased risk of C. difficile infection is not commonly observed among UC patients,13 perhaps due to the already altered gut microbiota characteristic of UC.74 As in the non-IBD patients, the risk of C. difficile infection increases with age and comorbidities. However, IBD patients’ average age at C. difficile diagnosis is much lower than in the general population.15 Immunosuppression has been recognized as a risk factor for C. difficile infection among UC patients. The use of corticosteroids, thiopurines, and methotrexate has been consistently associated with an increased risk of C. difficile infection while results for infliximab have been inconsistent.13,75,76 Greater disease extent has been associated with C. difficile infections among UC patients.77 However, one could argue that this is as a 14 result of the underlying immunosuppression due to extensive medication use to control UC. Recently, Ananthakishnan et al. identified several IBD-related risk loci associated with C. difficile infection with several of them being associated with the host immune response.76 The gold standard for the detection of C. difficile is stool culture and toxin detection or cell culture cytotoxicity assay.78 However, this test is not clinically practical because it is resource intensive and has a slow turnaround time. As a result, enzyme immunoassays (EIA) that detect C. difficile toxins have replaced stool culture in a clinical laboratory setting. However, EIAs are less sensitive (range: 31-99%) while specificity ranges between 65 and 100%.79 A two-step approach is recommended to address the issue of low positive predictive value due to low prevalence of C. difficile infection and suboptimal specificity, which would result in a decrease of false-positive results.79 For example, during the study period in Chapter 2 (from 2000 to 2009), Calgary Laboratory Services used an enzyme immunosorbent assay (EIA) for toxins A and B (TechLab C. difficile TOX A/B II, TechLab Enteric Diagnostics, USA) as a screening method; and tested the stool sample with a second EIA (Triage Clostridium difficile panel, Biosite Diagnostics, USA) if the results of the first EIA were positive. Supportive care and antibiotics (metronidazole or vancomycin) are the mainstay treatments for C. difficile infections. Metronidazole is cheaper, can be given intravenously and is recommended for patients with mild and non-recurrent C. difficile infection. On the other hand, vancomycin is the drug of choice for severe and recurrent C. difficile infection. Fecal microbiota transplant has been proven to be effective for recurrent C. difficile infections.80,81 The American College of Gastroenterology 15 Clostridium difficile infection task force suggested that IBD patients may begin treatment for both C. difficile infection and the IBD flare while waiting for C. difficile test results; but escalation of immunosuppression should be avoided in the setting of C. difficile infection.82 However, these recommendations are based on low-quality evidence.82 It is not clear whether C. difficile infection causes an infectious colitis superimposed on UC or precipitates an UC flare. A C. difficile diagnosis among UC patients is associated with a higher rate of endoscopies, hospitalizations and complications such as toxic megacolon, colectomy, and mortality. Whether C. difficile infection increases the risk of colectomy has been a matter of debate. While some studies demonstrate that C. difficile infection is associated with a short-term colectomy risk,13,16 other studies demonstrate that a C. difficile infection increases the long-term colectomy risk17,18 and one study found a negative association.11 In addition, diagnosis of C. difficile infection among UC patients is associated with higher mortality rates.11,16,83 Thus, C. difficile infection is an important modifiable complication of UC that may increase the risk of colectomy. This dissertation will enhance the literature by validating the accuracy of the code for C. difficile used in an administrative database, determine the risk of acquiring C. difficile after the diagnosis of UC, assess the effect of C. difficile infection on the risk of colectomy and mortality, and evaluate the impact of C. difficile infection on postoperative outcomes following colectomy for UC. 1.4 Colorectal Dysplasia and Cancer Inflammatory bowel disease is one of the most important high-risk conditions for colorectal cancer. Among IBD patients, colorectal cancer arises from an inflammation16 dysplasia carcinoma sequence.84,85 However, in clinical practice, cancers can arise without observing this sequence. The incidence of colorectal cancer has been described extensively in the literature and summarized in multiple meta-analyses.25,86-89 Eaden et al. published the first metaanalysis describing the colorectal cancer risk among UC patients and reported the 10-, 20- and 30-year cumulative risk of colorectal cancer to be 1, 8 and 18%, respectively.86 Since then, many meta-analyses have been performed with contradicting reports most likely due to a variety of methodologies and inclusion criteria. Only one of these studies reported that the incidence of colorectal cancer is increasing with an overall incidence rate of 14 cases per 1,000 patient-years.87 Two studies reported that the incidence of colorectal cancer has decreased over time with only one performing a meta-regression, but the results were not statistically significant.88,89 In the setting of IBD, some of the factors that increase the risk of colorectal cancer are: age, gender, disease extent, disease duration, PSC, and familial history of colorectal cancer.25,90,91 Disease duration is the most important risk factor for the development of dysplasia or colorectal cancer – i.e. the longer duration the disease the higher the incidence of neoplasia. Individuals who are young at diagnosis are at increased risk of developing cancer or dysplasia because of the higher burden of chronic inflammation. Men have a higher risk of developing colorectal cancer than females.25 Inflammation plays an important role in the development colorectal cancer.92 Several studies have demonstrated a direct relationship between the degree of histological inflammation on previous biopsy specimens and the development of colorectal cancer.2224 Patients with pancolitis have a higher risk compared to patients with left-sided UC or 17 proctitis.88,93 The increased risk of development of colorectal cancer observed in longstanding colitis and in greater disease extent supports the role of inflammation as a predisposing factor. The cumulative risk of developing cancer during the first 10 years of UC diagnosis is minimal (<1%).25 According to a meta-analysis by Jess et al., the 20-year cumulative risk of colorectal cancer is between 1.1-5.3%.25 Patients with UC and PSC have an increased risk of developing colorectal neoplasia when compared to patients with UC alone (OR=4.79; 95% CI: 3.58-6.41), but the reason for this is not completely understood.94 A recent population-based study concluded that the risk of developing CRC among IBD-PSC patients is 10 times higher and their diagnosis of cancer/dysplasia occurs at a younger age when compared to the general IBD population.26 In addition, cancers diagnosed among IBD-PSC patients are more often located in the right colon with an overall worse prognosis.95 The risk of colorectal cancer continues after liver transplant with an incidence rate of 13.5 per 1000 person-years (95% CI: 8.7-18.2).96 Typically dysplasia is categorized as flat or dysplasia associated lesion or mass (DALM).30 The IBD dysplasia morphology group defined dysplasia as “an unequivocal neoplastic alteration of the colonic epithelium”.97 This same group developed a dysplasia grading system classifying them into: indefinite, low grade and high grade based on the extent of affected epithelium.97 Diagnosis of colorectal dysplasia and cancer is based on the traditional random biopsy-driven colonoscopy surveillance practice that most gastroenterologists currently apply. Novel techniques with better sensitivity of identifying colonic dysplasia are available in certain academic centers including mucosal dye spray during colonoscopy (sensitivity= 83.5% and specificity= 91.3%) and confocal 18 endomicroscopy (sensitivity= 81% and specificity= 88%).98-100 While these techniques may improve the detection rates of dysplasia and early CRC, they are more time consuming, require additional training, and the costs to integrate these modalities into an endoscopy unit are higher. Further, the vast majority of practices do not have access to these advanced techniques for dysplasia surveillance. A complete proctocolectomy is indicated following the diagnosis of multi-focal low-grade dysplasia, high-grade dysplasia, dysplasia associated lesion or mass, and adenocarcinoma. However, the management of unifocal low-grade dysplasia is controversial. Though, the American College of Gastroenterology recommends colectomy also for low-grade dysplasia27 due to the increase likelihood of finding synchronous cancer in the resected specimen.101,102 Ullman et al. demonstrated that 23% of the patients who had a colectomy for low-grade dysplasia had synchronous advanced neoplasia.101 Further, 53% of the flat low-grade dysplasia progressed to advanced neoplasia by 5 years.101 However, despite the increased risk of developing cancer, 60% UC patients say that, if diagnosed with dysplasia, they would choose not to undergo a recommended elective colectomy and instead choose to be followed routinely with surveillance colonoscopy.103 Colorectal cancer prevention efforts have been focused around early detection (i.e. dysplasia detection). The identification of these risk factors and knowledge of the behaviour and progression of dysplasia in UC was crucial in the development of surveillance strategies. Several strategies have been recommended to improve survival among patients diagnosed with colorectal cancer. Contemporary guidelines for the management of IBD patients recommend surveillance colonoscopy every 1 to 2 years, 8 19 to 10 years after diagnosis based on several economic analyses assessing the costeffectiveness on different surveillance strategies.27,28 In contrast, annual colonoscopy is recommended for IBD-PSC patients immediately after both diseases are concurrently recognized but the cost-effectiveness of this intervention is unknown.27,28 Chemoprevention is defined as “the use of natural or synthetic chemical agents to reverse, suppress or delay the process of carcinogenesis”.90 However, chemoprevention has been a matter of controversy. Ursodeoxycholic acid (UDCA), 5-ASA and biologics are the most common chemoprotectants used among IBD patients. While a recent metaanalysis concluded that both, UDCA and 5-ASA, do not confer a protective effect against the development of colorectal cancer, recent experimental colitis models and case-control studies suggest that biologics could decrease the risk of colorectal cancer.104-107 Thus, this dissertation will enhance the literature by determining if the observed overall decreased risk of colectomy is in part due to a decrease in risk of colectomy due to colorectal neoplasia and evaluates the cost-effectiveness of different colorectal cancer surveillance intervals among IBD-PSC patients. 1.5 Thesis Objectives and Hypothesis The overall aim of this dissertation was to address important gaps on the epidemiology of C. difficile infections and colorectal neoplasia among UC patients in order to identify areas or ways of further decreasing the risk of colectomy among UC patients. Below are the specific objectives of each chapter along with their respective hypotheses. 20 1.5.1 Chapter 2: Clostridium difficile infection worsens the prognosis of ulcerative colitis 1.5.1.1 Objective 1: Determine if C. difficile diagnosis in hospital or 90-days prior to hospital admission among UC patients was associated with having an emergent colectomy. Hypothesis: UC patients hospitalized due to disease exacerbation with a diagnosis of C. difficile have higher odds of having an emergency colectomy when compared to patients who were C. difficile-negative. 1.5.1.2 Objective 2: Among UC patients who had an emergent colectomy, determine if C. difficile diagnosis in hospital or 90-days prior to hospital admission among UC patients was associated with an increased risk of postoperative complication. Hypothesis: UC patients who had an emergent colectomy and were diagnosed with C. difficile infection in-hospital or 90 days prior to hospital admission have higher odds of developing postoperative complications when compared to those UC patients who were C. difficile-negative. 1.5.2 Chapter 3: Clostridium difficile infection diagnosed early in ulcerative colitis patients impacts disease related outcomes: A population-based inception cohort 1.5.2.1 Objective 1: To validate the ICD-10 code for C. difficile diagnosis among UC patients. Hypothesis: The ICD-10 code C. difficile diagnosis is valid in detecting UC patients diagnosed with C. difficile. 1.5.2.2 Objective 2: Determine the 1-, 3-, and 5-year cumulative risk of a C. difficile diagnosis after being diagnosed with UC. Hypothesis: Among UC patients, the risk of acquiring a C. difficile infection increases with time. 21 1.5.2.3 Objective 3: Determine the effect of a C. difficile diagnosis on the risk of colectomy and mortality. Hypothesis 1: UC patients with a C. difficile diagnosis have a higher risk of colectomy when compared to UC patients without a C. difficile diagnosis. Hypothesis 2: The overall mortality risk among UC patients diagnosed with C. difficile is higher when compared to UC patients without a C. difficile diagnosis. 1.5.2.4 Objective 4: Determine the association between a C. difficile diagnosis and developing post-operative complications. Hypothesis: UC patients who had a colectomy and were diagnosed with C. difficile infection have higher odds of developing postoperative complications in hospital when compared to those UC patients who were C. difficile-negative. 1.5.3 Chapter 4: Changes in the annual incidence of colectomy for colorectal neoplasia among patients with ulcerative colitis: A population-based cohort 1.5.3.1 Objective 1: To evaluate the risk profile of patients who had a colectomy for dysplasia and cancer. Hypothesis: UC patients who have a colectomy for colorectal dysplasia and cancer have distinct characteristics compared to patients who had a colectomy for medically refractory disease. 1.5.3.2 Objective 2: Determine if the incidence of colectomy for colorectal neoplasia has decreased over time. Hypothesis: The incidence of colectomy for colorectal dysplasia has decreased over time. 22 1.5.4 Chapter 5: Colorectal cancer surveillance in patients with inflammatory bowel disease and primary sclerosing cholangitis: An economic evaluation 1.5.4.1 Objective 1: To determine whether annual colonoscopy among patients with IBDPSC is cost-effective compared to less frequent intervals from a publicly funded health care system perspective. Hypothesis: Annual colonoscopy is the most cost-effective strategy compared to less frequent intervals. 23 Chapter Two: Clostridium difficile infection worsens the prognosis of ulcerative colitis 24 2.1 Abstract The impact of Clostridium difficile infections among ulcerative colitis (UC) patients is well characterized. However, there is little knowledge about the association between C. difficile infections and postoperative complications among UC patients. The objective of this study was to determine if C. difficile infection was associated with having an emergent colectomy and postoperative complications. The present population based study identified UC patients admitted to Calgary Health Zone hospitals for a flare from 2000-2009. C. difficile toxin tests ordered in hospital or 90 days prior to hospital admission were provided by Calgary Laboratory Services (Calgary, Alberta). Hospital records were reviewed to confirm diagnoses and extract clinical data. Multivariate logistic regression analyses were perfomed among individuals tested for C. difficile to examine the association between C. difficile infection and emergent colectomy and diagnosis of any postoperative complications and, secondarily, an infectious postoperative complication. Estimates were presented as adjusted ORs with 95% CIs. C. difficile was tested in 278 (58%) UC patients and 6.1% were positive. C. difficile infection was associated with an increased risk for an emergent colectomy (adjusted OR=3.39; 95% CI: 1.02-11.23). Additionally, a preoperative diagnosis of C. difficile was significantly associated with the development of postoperative infectious complications (OR=4.76; 95% CI: 1.10-20.63). C. difficile diagnosis worsened the prognosis of UC by increasing the risk of colectomy and postoperative infectious complications following colectomy. Future studies are needed to explore whether early detection and aggressive management of C. difficile infection will improve UC outcomes. 25 2.2 Introduction Ulcerative colitis (UC) is characterized by periods of remission followed by periods of disease activity that decrease quality of life. In western nations, the prevalence of UC was as high as 500 per 100,000 persons,1 and approximately 15% of UC patients undergo surgical resection of the colon within the first 10 years of diagnosis.2,3 Furthermore, UC patients who undergo colectomy are at increased risk of postoperative morbidity and mortality.4,5 The increased incidence and severity of Clostridium difficile infections during the last decade,6,7 along with outbreaks of more virulent strains,8 have increased public and practitioner awareness of the importance of this pathogen. While antibiotic exposure is the primary risk factor for C. difficile infection, UC has become recognized as an independent risk factor.9,10 C. difficile infection risk among UC patients have increased over time, with C. difficile prevalence doubling from 26.6 to 51.2 per 1,000 discharges from 1998-2004.11 C. difficile infection may worsen the prognosis of UC because the infection has been associated with higher morbidity and increased risk of surgery up to one year after diagnosis of infection.11-13 Consequently, C. difficile diagnosis is associated with higher hospital costs among inflammatory bowel disease patients.11 These studies assessed only inpatient C. difficile test results and, thus, missed the impact of C. difficile infections diagnosed before hospital admission.11,14 We studied whether C. difficile diagnosis in hospital or 90 days before hospital admission among UC patients was associated with an emergent colectomy and, furthermore, the development of postoperative complications. 26 2.3 Materials and Methods 2.3.1 Data sources The Data Integration, Measurement and Reporting Hospital Discharge Abstract Database was used to capture hospitalizations of UC patients in the Calgary Health Zone (CHZ) in Alberta. The CHZ is a population-based health authority that provides health care to Calgary residents and >20 nearby cities under a public, single-payer system.15 This database includes patients’ demographic data, admission and discharge date, and 42 diagnostic and 25 procedural fields using the International Classification of Disease, Ninth and Tenth Revisions, Clinical Modification (ICD-9-CM & ICD-10-CM) and the Canadian Classification of Health Intervention (CCI).4,16 The population-based Calgary Laboratory Services database was used to identify UC patients who underwent stool testing for C. difficile. Calgary Laboratory Services confirmed the C. difficile infection using a two-step approach: an enzyme immunosorbent assay (EIA) for toxins A and B (TechLab C. difficile TOX A/B II, TechLab Enteric Diagnostics, USA) was used as a screening method; and the stool sample was tested with a second EIA (Triage Clostridium difficile panel, Biosite Diagnostics, USA) if the results of the first EIA were positive. A sample was considered to be C. difficile positive if both tests were positive. 2.3.2 Study population The study population consisted of adults (≥ 18 years) admitted emergently to a CHZ hospital between January 1, 2000 and December 31, 2009 for a UC flare. The approach for identifying the study population was previously validated.16 First, the Data Integration, Measurement and Reporting Hospital Discharge Abstract database was used 27 to identify patients admitted to hospital with a diagnosis of UC (ICD-9-CM 556.X or ICD-10-CA K51.X) in any diagnostic position and a procedural code for colectomy (ICD-9-CM 45.7, 45.8 or CCI 1.NM.87, 1.NM.89, 1.NM.91, 1.NQ.89, 1.NQ.90). The colectomy admission was recorded as the index date. Second, from the remaining discharge abstracts, UC patients admitted to hospital for a flare were identified from admissions with a UC code in the primary diagnostic position. Among patients with UC who were admitted to hospital for a flare, but did not undergo a colectomy, all hospital admissions that occurred within the study period (2000-2009) were recorded and one admission was randomly selected as the index date.16 Patients who were not tested for C. difficile in hospital or within 90 days of hospitalization were not included in the study population. Figure 2.1 illustrates is the inclusion and exclusion criteria that defined the study population. Medical charts of all UC patients identified were reviewed to confirm that the UC patients were admitted to hospital emergently for a UC flare and to collect patients’ surgical and medical history relevant to the index admission. 2.3.3 Selection of cases and controls The primary case definition was having an emergent colectomy during the index admission. Colectomy was defined as emergent if the decision to perform colectomy was made during the admission and after failing to respond to medical management, or because the patient experienced a complication. Controls were the remaining UC patients discharged from hospital without a colectomy after responding to medical management. The secondary case definition was the development of any postoperative complication, and, specifically, an infectious postoperative complication among UC patients who underwent an emergent colectomy. Postoperative complications were 28 defined as an unexpected medical event that occurred between the start of the operation and discharge from the hospital. These complications were classified into seven complication categories including postoperative infection (Appendix 2.A). Complications were stratified by severity using the Clavien classification of surgical complications.17 The development of postoperative complications was recorded if the patient experienced at least one complication described under any of the categories graded as Clavien II or higher (ie, requiring medical or surgical intervention or leading to death). C. difficile infection was not considered to be a postoperative complication.4 The definition of postoperative complication has been previously validated for UC.16 2.3.4 Exposure The primary exposure was a diagnosis of C. difficile infection in hospital or 90 days before the index admission. Ninety days before hospital admission was defined a priori to ensure that patients defined as a negative test for C. difficile were not treated for a C. difficile infection prior to hospital admission. A patient was classified as diagnosed with C. difficile if at least one test result was positive during the predetermined period. The date of the first positive test was recorded as the index date of infection. C. difficile diagnosis was confirmed by medical chart review. Patients were classified as C. difficile negative if all test results during the predetermined period were negative. 2.3.5 Covariates Additional demographic and clinical data extracted from chart review included: age (stratified as: 18-32, 33-47, ≥ 48 year based on the tercile of the cohort); sex; disease extent (left-sided versus pancolitis); disease duration (defined as the interval between UC diagnosis and admission date), length of flare (< 2, 2 to 8, > 8 weeks); and smoking status 29 (current, ex-smoker, never). Inflammatory bowel disease medications (5-aminosalicylic acid or sulfasalazine, azathioprine, corticosteroids, and infliximab) taken at time of admission and/or administered in-hospital were recorded. Patients were classified as having comorbidities (i.e., health conditions occurring before hospitalization) if they had at least one of the comorbidities listed in Appendix 2.B. The definition of comorbidity has been previously validated for UC.16 2.3.6 Statistical analysis The associations between the outcome and categorical variables were tested using the Fisher’s exact test or the X2 test. Continuous variables were expressed as medians with interquartile ranges (IQR) and compared using the Wilcoxon rank-sum test. Multivariate logistic regression analysis was performed to determine the association between the need for an emergent colectomy and diagnosis of C. difficile (defined a priori) after adjusting for other covariates. Age was a priori forced into the regression model. For the other covariates, a backwards elimination approach was used to examine independent effects of additional variables on the need of emergent surgery with an entry P-value of < 0.20. Variables were kept in the model if: the two-sided P-value was < 0.05; or there was evidence of confounding because their removal resulted in a 30% change in the estimate of the primary exposure. The variance inflation factor was used to measure multicollinearity among the independent variables. Multicollinearity was considered negligible if the variance inflation factor was < 10. Interactions between C. difficile positivity and variables that were indepedently associated with colectomy were tested and included in the model if the likelihood ratio test was statistically significant (P < 0.05). Point estimates were presented as adjusted ORs with 95% CIs. 30 The association between C. difficile infection and the development of postoperative complications was also evaluated. Multivariate logistic regression was performed to examine the association between postoperative complications (and then postoperative infection separately) and C. difficile diagnosis (defined a priori) after adjusting for other covariates. A backwards elimination approach was used to examine the independent effects of variables on the development of postoperative complications (and postoperative infection) using the same procedure as described above. A sensitivity analysis was performed to determine whether the timing of C. difficile diagnosis affected the association with the outcomes. For this, logistic regression models were recalculated with C. difficile diagnosis defined as having at least one positive test result in hospital or 14 days before the index admission. Patients who were not tested for C. difficile in hospital or 90 days before admission were included in a second sensitivity analysis. Logistic regression models were recalculated for all three outcomes: emergent colectomy, any postopertative complication and infectious complication. All analyses were performed using STATA version 11 (STATA Corp, USA). The study was approved by the Conjoint Health Research Ethics Board at the University of Calgary (Calgary, Alberta). The present study was conducted in accordance with the strengthening of the reporting of observational studies in epidemiology (STROBE) statement.18 31 2.4 Results A total of 278 UC patients met the inclusion criteria. Table 2.1 summarizes the baseline characteristics of the study population. The indications for an emergent colectomy were bowel complication (n = 9), cancer/dysplasia (n = 1), and failed medical management in hospital (n = 92). C. difficile diagnosis was recorded in 11 (11%) of UC patients who underwent an emergent colectomy compared with six (3%) patients who responded to medical management (P = 0.01). Patients diagnosed with an infection in hospital or up to 90 days before hospitalization had higher odds of undergoing emergent surgery when admitted to hospital (adjusted OR=3.39; 95% CI: 1.02-11.23) (Table 2.2). Table 1 summarizes patients’ characteristics of the 102 patients who underwent an emergent colectomy and were preoperatively tested for C. difficile. The median time from C. difficile diagnosis to surgery was 12 days (interquartile range 14 days). At least one postoperative complication was recorded in 30% of UC patients and 20% developed an infectious postoperative complication. Infectious postoperative complications experienced among UC patients with C. difficile included: sepsis (n = 3), abscess (n = 2), urinary tract infection (n = 1), pneumonia (n = 3) and infected central line (n = 1) (Appendix 2.C). C. difficile was diagnosed preoperatively in 24% of patients who experienced an infectious postoperative complication, compared with 7% of UC patients who did not develop an infectious postoperative complication ( P= 0.046) (Table 2.1). Preoperative diagnosis of C. difficile was not significantly associated with developing any postoperative complication (adjusted OR=3.16; 95% CI: 0.89-11.23) (Table 2.3). However, individuals diagnosed with C. difficile preoperatively had higher 32 odds of developing a new infectious postoperative complication (adjusted OR=4.76; 95% CI: 1.10-20.63) (Table 2.3). The sensitivity analysis that restricted the exposure definition to only patients tested for C. difficile in hospital or 14 days before admission resulted in similar associations between C. difficile infection and emergent colectomy (adjusted OR=3.70; 95% CI: 1.06-12.89) (Appendix 2.D), whereas the associations were strengthened for any complication (adjusted OR=6.11; 95% CI: 1.32-28.21) and infectious complications (adjusted OR=7.93; 95% CI: 1.54-40.75) (Appendix 2.E). The sensitivity analysis that included patients not tested for C. difficile resulted in a non-significant association between C. difficile infection and emergent colectomy (OR=2.80; 95% CI: 0.86-9.11) (Appendix 2.F). However, individuals not tested for C. difficile had higher odds of undergoing an emergent surgery when compared to those with a negative C. difficile test result (OR=1.69; 95% CI: 1.09-2.63) (Appendix 2.F). Individuals diagnosed with C. difficile preoperatively had higher odds of developing a new infectious postoperative complication (OR=4.67; 95% CI: 1.17-18.47) (Appendix 2.G). Similar to the author’s primary analysis, a preoperative diagnosis of C. difficile was not significantly associated with developing any postoperative complication (adjusted OR=3.55; 95% CI: 0.95-13.21) (Appendix 2.G). Patients with an unknown C. difficile status preoperatively were not at increased risk of developing any complication (OR=1.56; 95% CI: 0.80-3.07) or a new infectious postoperative complication (OR=1.49; 95% CI: 0.67-3.29) (Appendix 2.G). 33 2.5 Discussion UC patients diagnosed with C. difficile in hospital or 90 days before admission were more likely to undergo an emergent colectomy after controlling for factors such as age, disease extent, and corticosteroid use. Additionally, a preoperative diagnosis of C. difficile increased the risk of developing an infectious postoperative complication. Consequently, C. difficile infection is an important clinical outcome for UC patients admitted to hospital with a flare. UC patients have higher mortality and surgery risk at one and five years following a C. difficile diagnosis when compared to UC patients without C. difficile.19,20 However, reports describing the impact of C. difficile diagnosis on in-hospital and short-term risk of colectomy are inconsistent. Two single-center studies21,22 failed to find an association between C. difficile and the need of colectomy at index admission or three months following C. difficile diagnosis. In addition, a large, nationwide study using in-hospital data reported a negative association between C. difficile diagnosis and colectomy, even after taking into consideration patients who were admitted electively.11 The conflicting results between these previous studies and our study may be attributed to methodological differences from our study. Previous studies were potentially limited because of selection bias arising from data collected from tertiary care centers and/or misclassification bias from using administrative databases.23 In contrast, our study was population-based, and used chart reviews to confirm exposures and outcomes. Moreover, prior studies did not account for C. difficile testing that occurred before admission to hospital. Furthermore, surgery thresholds may differ between centers.23 Also, different C. difficile ribotypes8,25 and lack of consensus on the treatment of C. difficile infections among UC patients might 34 have influenced infection outcomes.26,27 Finally, over time, better surveillance and diagnostic tests along with more aggressive treatment of C. difficile infections may account for the differences observed.26-30 Several factors may explain the increased risk of postoperative infection among patients with a C. difficile infection. First, patients with a C. difficile infection may have greater UC disease severity, immunosuppression, or systemic toxicity.28 Second, antibiotic exposure prior to surgery may have increased susceptibility for acquisition of antibiotic-resistant nosocomial infections.29 Finally, C. difficile toxin worsens gut permeability30,31 and promotes bacterial migration, which may have increased the risk of septic complications.32 Our sensitivity analysis that restricted the study population to individuals tested within 14 days of hospital admission or during hospital admission suggested that the timing of C. difficile infection influenced postoperative morbidity. We studied a large population-based cohort of UC patients.16 The diagnosis of UC was confirmed in all cases, which improved the accuracy of our data. Administrative databases misclassify the diagnosis of UC, colectomy, and postoperative complications thereby influencing magnitude of risk estimates and the precision of CIs.16 Additionally, we were able to determine the reason for admission (flare versus elective colectomy) and only included individuals admitted as a flare in our analysis. Also, C. difficile test results were obtained from a population-based laboratory database that records both in-hospital and outpatient testing. Thus, we captured all tests performed on these patients during the study period. Using a centralized laboratory database that captures all C. difficile testing is essential because administrative coding of C. difficile includes a misclassification error and could miss C. difficile diagnosis before admission. Finally, patients were tested for C. 35 difficile using a sequential testing approach, which optimized the positive predictive value of detecting diseases with low prevalence. Several limitations to the present study should be considered. First, we excluded UC patients who were not tested for C. difficile in hospital or within 90 days of index admission. Consequently, some UC patients may have been C. difficile positive but were never detected. This gives rise to a possible differential misclassification bias where patients with a more severe form of colitis were more likely to be tested for C. difficile, reflecting a population with a more severe form of UC. Second, we were unable to report C. difficile incidence rates among UC patients.33 Third, patient information was obtained by reviewing patients’ charts, relying on various clinicians for their completeness and accuracy. Due to the retrospective nature of the data, we were not able to control disease severity by calculating a Mayo Score. Additionally, C-reactive protein levels were not reliably measured in all patients during the early study years. Fourth, we were unable to study other patient-related factors, such as antibiotic administration and response, because they were not reliably recorded on the patients’ charts. Antibiotic treatment of C. difficile (e.g., timing and first-line agent) may have influenced disease course. Fifth, approximately 42% of the patients were not tested for C. difficile, which could introduce a selection bias. Prior studies using administrative database have assumed that a negative code for C. difficile is a negative test for C. difficile.11 However, our sensitivity analysis demonstrated that individuals not tested for C. difficile were also at increased odds of colectomy when compared with those who tested negative. Finally, the number of patients with UC who tested positive for C. difficile was small, reducing the precision of our findings. Furthermore, the small number of outcomes in the postoperative 36 complications analysis may affect the generalizability of our results. Consequently, our findings should be independently replicated. C. difficile diagnosis was associated with undergoing an emergent colectomy and developing postoperative infectious complications. These findings have important clinical implications. First, physicians should carefully assess for C. difficile infection among all UC patients presenting with a flare of disease activity. Second, UC patients with a C. difficile infection who undergo a colectomy should be monitored closely and precautions should be taken to prevent infections. Third, increased surveillance for C. difficile infections with early identification and aggressive treatment provides potential avenues for improving outcomes among UC patients. Future studies are necessary to assess interventions that may reduce the morbidity of C. difficile infections among UC patients; and to determine whether timing of colectomy for UC patients with C. difficile infection should be adjusted to minimize postoperative infectious complications. At minimum, these patients require careful surveillance for infections postoperatively. 2.6 Acknowledgements The authors acknowledge the DIMR department for providing data from the CHZ. Dr. Kaplan is supported through a New Investigator Award from the Canadian Institute of Health Research and a Population Health Investigator Award from Alberta Innovates Health Solutions. 37 2.7 Financial Disclosures This project was funded by the Alberta Inflammatory Bowel Disease Consortium, which is funded by an AHFMR Interdisciplinary Team Grant. AHFMR is now Alberta Innovates - Health Solutions. 38 2.8 References 1. Molodecky NA, Soon IS, Rabi DM, et al. Increasing incidence and prevalence of the inflammatory bowel diseases with time, based on systematic review. Gastroenterol 2012;142:46-54. 2. Kaplan GG, Seow CH, Ghosh S, et al. Decreasing colectomy rates for ulcerative colitis: A population-based time trend study. Am J Gastroenterol 2012;12:187987. 3. Frolkis AD, Dykeman J, Negron ME, et al. Risk of surgery for inflammatory bowel diseases has decreased over time: a systematic review and meta-analysis of population-based studies. Gastroenterol 2013;145:996-1006. 4. de Silva S, Ma C, Proulx MC, et al. Postoperative complications and mortality following colectomy for ulcerative colitis. Clin Gastroenterol Hepatol 2011;11:972-80. 5. Soon IS, Wrobel I, deBruyn JC, et al. Postoperative complications following colectomy for ulcerative colitis in children. J Pediatr Gastroenterol Nutr 2012;54:763-8. 6. Ricciardi R, Ogilvie JW, Roberts PL, Marcello PW, Concannon TW, Baxter NN. Epidemiology of Clostridium difficile colitis in hospitalized patients with inflammatory bowel diseases. Dis Colon Rectum 2009;52:40-5. 7. Chandler RE, Hedberg K, Cieslak PR. Clostridium difficile associated disease in Oregon: increasing incidence and hospital-level risk factors. Infect Control Hosp Epidemiol 2007;28:116-122. 39 8. Hubert B, Loo VG, Bourgault AM, et al. A portrait of the geografic dissemination of the Clostridium difficile North American pulsed-field type 1 strain and the epidemiology of C. difficile-associated disease in Quebec. Clin Infect Dis 2007;44:238-44. 9. Ananthakrishnan AN. Clostridium difficile infection: epidemiology, risk factors and management. Nat Rev Gastroenterol Hepatol 2011;8:17-26. 10. Berg AM, Kelly CP, Farraye FA. Clostridium difficile infection in the inflammatory bowel disease patient. Inflamm Bowel Dis 2013;19:194-204. 11. Nguyen GC, Kaplan GG, Harris ML, Brant SR. A national survey of the prevalence and impact of Clostridium difficile infection among hospitalized inflammatory bowel disease patients. Am J Gastroenterol 2008;103:1443-50. 12. Issa M, Vijayapal A, Graham MB, et al. Impact of Clostridium difficile on inflammatory bowel disease. Clin Gastroenterol Hepatol 2007;5:345-51. 13. Ananthakrishnan AN, McGinley EL, Saeian K, Binion DG. Temporal trends in disease outcomes related to Clostridium difficile infection in patients with inflammatory bowel disease. Inflamm Bowel Dis 2011;17:976-83. 14. Ananthakrishnan AN, McGinley EL, Binion DG. Excess hospitalisation burden associated with Clostridium difficile in patients with inflammatory bowel disease. Gut 2008;57:205-10. 15. Alberta Health Services. Alberta Health Services Annual Report, April 1, 2009March 31, 2010. http://www.albertahealthservices.ca/Publications/ahs-pub- annual-rpt.pdf. (Version current at November 1, 2012). 40 16. Ma C, Crespin M, Proulx MC, et al. Postoperative complications following colectomy for ulcerative colitis: A validation study. BMC Gastroenterol 2012;12:39. 17. Dindo D, Demartines N, Clavien P-A. Classification of Surgical Complications. Ann Surg 2004;240:205-13. 18. Vandenbroucke JP, von Elm E, Altman DG, et al. Strengthening the reporting of observational studies in epidemiology (STROBE): explanation and elaboration. PLoS Med 2007;4:e297. 19. Murthy SK, Steinhart AH, Tinmouth J, et al. Impact of Clostridium difficile colitis on 5-year health outcomes in patients with ulcerative colitis. Aliment Pharmacol Ther 2012; 36:1032-9. 20. Navaneethan U, Mukewar S, Venkatesh PG, et al. Clostridium difficile infection is associated with worse long term outcome in patients with ulcerative colitis. J Crohns Colitis 2012;6:330-6. 21. Kariv R, Navaneethan U, Venkatesh PG, Lopez R, Shen B. Impact of Clostridium difficile infection in patients with ulcerative colitis. J Crohns Colitis 2011;5:3440. 22. Kaneko T, Matsuda R, Taguri M, et al. Clostridium difficile infection in patients with ulcerative colitis: investigations of risk factors and efficacy of antibiotics for steroid refractory patients. Clin Res Hepatol Gastroenterol 2011;35:315-20. 23. Molodecky NA, Panaccione R, Ghosh S, Barkema HW, Kaplan GG. Challenges associated with identifying the environmental determinants of the inflammatory bowel diseases. Inflamm Bowel Dis 2011;17:1792-9. 41 24. Kaplan GG, McCarthy EP, Ayanian JZ, Korzenik J, Hodin R, Sands BE. Impact of hospital volume on postoperative morbidity and mortality following a colectomy for ulcerative colitis. Gastroenterol 2008;134:680-7. 25. Sundram F, Guyot A, Carboo I, Green S, Lilaonitkul M, Scourfield A. Clostridium difficile ribotypes 027 and 106: clinical outcomes and risk factors. J Hospl Infect 2009;72:111-8. 26. Goodhand JR, Alazawi W, Rampton DS. Systematic review: Clostridium difficile and inflammatory bowel disease. Aliment Pharmacol Ther 2011;33:428-41. 27. Yanai H, Nguyen GC, Yun L, et al. Practice of gastroenterologists in treating flaring inflammatory bowel disease patients with clostridium difficile: Antibiotics alone or combined antibiotics/immunomodulators? Inflamm Bowel Dis 2011;17:1540-6. 28. Kaplan GG, Hubbard J, Panaccione R, et al. Risk of comorbidities on postoperative outcomes in patients with inflammatory bowel disease. Arch Surg 2011;146:959-64. 29. Tacconelli E, De Angelis G, Cataldo MA, et al. Antibiotic usage and risk of colonization and infection with antibiotic-resistant bacteria: a hospital populationbased study. Antimicrob Agents Chemother 2009;53:4264-9. 30. Moore R, Pothoulakis C, LaMont JT, Carlson S, Madara JL. C. difficile toxin A increases intestinal permeability and incduces Cl- secretion. Am J Physiol Gastrointest Liver Physiol 1990;259:G165-G72. 42 31. Hetch G, Charalabos P, LaMont JT, Madara JL. Clostridium difficile toxin A perturbes cytoskeletal sturcture and tight junction permeability of cultured human intestinal epithelial monolayers. J Clin Invest 1988;82:1516-24. 32. F Feltis BA, Kim AS, Kinneberg KM, et al. Clostridium difficile toxins may augment bacterial penetration of intestinal epithelium. Arch Surg 1999;134:123542. 33. Schulz KF, Grimes DA. Case-control studies: research in reverse The Lancet 2002;359:431-4. 43 Table 2.1. Characteristics of ulcerative colitis patients admitted to hospital stratified by flare (medically responsive) and emergent colectomy. Emergent colectomy patients were further stratified into development of any postoperative complication and development of infectious complication Characteristics Gender, %(n) Male Female All UC Patients (n = 278) Emergent Colectomy No Yes P-Value (n = 176) (n = 102) 52.3 (92) 47.7 (84) 62.8 (64) 37.2 (38) Emergent Colectomy Patients (n = 102) Any complication Infectious complication No Yes P-Value No Yes P-Value (n = 71) (n = 31) (n = 81) (n = 21) 0.090 54.9 (39) 45.1 (32) 80.6 (25) 19.4 (6) 0.015 56.8 (46) 43.2 (35) 85.7 (18) 14.3 (3) 0.021 43.0 50.0 65.0 0.028 31.0 43.0 55.0 46.0 50.0 72.4 0.022 0.085 92.6 (75) 7.4 (6) 76.2 (16) 23.8 (5) 0.046 0.474 0 2 7 0 1 4 n=3 Age at admission, y* 25th percentile 27.0 Median 36.0 75th percentile 47.0 31.0 47.0 58.0 < 0.001 30.0 40.0 55.0 C. difficile, %(n) No Yes 96.6 (170) 3.4 (6) 89.2 (91) 10.8 (11) 0.013 93.0 (66) 7.0 (5) 80.7 (25) 10.3 (6) Disease duration, y* 25th percentile Median 75th percentile Missing 0 1.0 6.0 n=5 0 2.0 6.0 n=3 0.401 0 2 7 0 1 4.5 n=3 44 0.549 Smoking, %(n) Current Ex-smoker Never Missing 10.8 (18) 28.9 (48) 60.3 (100) n = 10 9.8 (10) 35.3 (36) 54.9 (56) Flare duration, %(n) <2 wks 2-8 wks >8 wks Missing 18.3 (32) 49.1 (86) 32.6 (57) n=1 25.5 (26) 47.1 (48) 27.4 (28) n=1 Disease extent, %(n) Left-sided Pancolitis Missing 46.4 (71) 53.6 (82) n = 23 17.7 (17) 82.3 (79) n=6 0.550 11.3 (8) 33.8 (24) 54.9 (39) 6.4 (2) 38.7 (12) 54.8 (17) 0.332 25.4 (18) 50.7 (36) 23.9 (17) 25.8 (8) 38.7 (12) 35.5 (11) < 0.001 15.4 (10) 84.6 (55) 22.3 (7) 77.4 (24) n=6 51.6 (16) 48.4 (15) Comorbidity, %(n) No 68.8 (121) Yes 31.2 (55) 60.8 (62) 39.2 (40) 0.191 64.8 (46) 35.2 (25) Prednisone at admission, %(n) No Yes 20.6 (21) 79.4 (81) < 0.001 23.9 (17) 76.1 (54) 12.9 (4) 87.1 (27) 0.805 28.2 (20) 71.8 (51) 35.5 (11) 64.5 (20) 47.3 (84) 52.3 (92) Five-ASA†, %(n) No 31.8 (56) Yes 68.2 (120) 30.4 (31) 69.6 (71) 45 0.814 9.9 (8) 33.3 (27) 56.8 (46) 9.5 (2) 42.9 (9) 47.6 (10) 0.721 0.426 25.9 (21) 49.4 (40) 24.7 (20) 23.8 (5) 38.1 (8) 38.1 (8) 0.456 17.3 (13) 82.7 (62) n=6 19.1 (4) 80.9 (17) 1.0 0.210 64.2 (52) 35.8 (29) 47.6 (10) 52.4 (11) 0.166 0.205 23.5 (19) 76.5 (62) 9.5 (2) 90.5 (19) 0.229 0.460 28.4 (23) 71.6 (58) 28.1 (8) 69.9 (13) 0.389 0.403 Azathioprine†, %(n) No Yes 67.0 (118) 33.0 (58) 69.6 (71) 30.4 (31) Infliximab†, % (n) No 81.3 (143) Yes 18.7 (33) 86.3 (88) 13.7 (14) IV-steroids in hospital, % (n) No Yes 1.0 (1) 99.0(101) 12.5 (22) 87.5 (154) 0.659 70.4 (50) 29.6 (21) 67.7 (21) 32.6 (10) 0.281 88.7 (63) 11.3 (8) 80.7 (25) 19.3 (6) < 0.001 0 (0) 100 (71) 3.2 (1) 96.7 (30) * y- year; † medication taken at admission or in-hospital 46 0.787 70.4 (57) 29.6 (24) 66.7 (14) 33.3 (7) 0.742 0.275 86.4 (70) 13.6 (11) 85.7 (18) 14.3 (3) 1.0 0.304 1.2 (1) 98.8 (80) 0 (0) 100 (21) 1.0 Table 2.2. Logistic regression results of emergent colectomy with C. difficile diagnosis as primary exposure. All UC Patients (n = 278) Adjusted analysis OR 95% CI C. difficile diagnosis No Yes 1.00 3.39 Referent (1.02-11.23) Age 18-32 33-47 ≥ 48 1.00 1.21 5.05 Referent (0.06-2.44) (2.44-10.47) Disease distribution Left-side/Undetermined Pancolitis 1.00 4.41 Referent (2.33-8.33) Prednisone at admission No Yes 1.00 2.75 Referent (1.45-5.20) IV-steroids in hospital No Yes 1.00 15.61 Referent (1.77-137.18) 47 Table 2.3. Logistic regression results for emergent colectomy patients having any post-operative complication and infections post-operative complication with C. difficile diagnosis as primary exposure. Emergent Colectomy Patients (n = 102) Any complication Infectious complication Adjusted analysis Adjusted analysis OR 95% CI OR 95% CI C. difficile diagnosis No Yes 1.00 3.70 Referent (0.89-15.3) 1.00 4.76 Referent (1.10-20.63) Age 18-32 33-47 ≥ 48 1.00 1.35 2.77 Referent (0.24-5.30) (0.77-9.96) 1.00 1.56 3.43 Referent (0.31-7.93) (0.8-14.68) Gender Male Female 1.00 0.28 Referent (0.09-0.82) 1.00 0.23 Referent (0.6-0.88) Comorbidity No Yes 1.00 1.52 Referent (0.53-4.35) NS* *Not significant (NS): Was not significant in final model. 48 Figure 2.1. Flow diagram illustrating the inclusion and exclusion criteria for identifying patients admitted with a flare for ulcerative colitis and tested Clostridium difficile infection in hospital and up to 90 days prior to admission. 49 2.9 Appendix 2.A Post-operative complications classification Category Gastrointestinal Complications Small bowel obstruction, pouch leak or pouch failure, bowel perforation, ileus, ischemic bowel and gastrointestinal bleeding Wound Fistula, hematoma or seroma, wound dehiscence and delayed wound healing and iatrogenic injuries including foreign body accidentally left during procedure Infectious Sepsis and bacteremia, abscess, wound infection, urinary tract infection, pneumonia and empyema. Renal and Endocrine Acute renal failure, fluid and electrolyte disorders (e.g. hypokalemia) and adrenal disorders Cardiovascular Thrombosis or embolism, myocardial infarction, cardiac arrest, hypotension or shock, cardiac arrhythmias and congestive heart failure Pulmonary Acute respiratory failure, hypoxemia, pleural effusion and pulmonary edema, pneumothorax and atelectasis, asthma and COPD exacerbation Neurological Neurological disease, cerebrovascular disease, psychoses, delirium, seizures and neuropathies 50 2.10 Appendix 2.B Comorbidity classifications Category Coronary Artery Disease Cancer Comorbidity Coronary artery disease, ischemic heart disease, myocardial infarction, peripheral vascular disease Lymphoma, metastatic tumour, solid tumour without metastases Other cardiovascular Cardiac arrhythmia, valvular disorder CHF Congestive heart failure Diabetes Diabetes with complications, diabetes without complications Venous Thromboembolism GI Deep vein thrombosis CMV infection, pancreatitis, peptic ulcer disease HTN Blood loss anaemia, coagulopathy, cyclical neutropenia, deficiency anaemia Hypertension Liver Fatty liver, primary sclerosing cholangitis, liver disease Neurological Cerebrovascular disease, hemiplegia and paraplegia Pulmonary Renal Asthma, COPD, sarcoidosis Renal failure (acute or chronic) Rheumatoid Ankylosing spondylitis, episcleritis, uveitis and iritis, Gout, sacroiliitis, rheumatoid arthritis Haematological 51 2.11 Appendix 2.C Postoperative complications observed among patients tested for C. difficile. Complications categories* Gastrointestinal No complications At least one complication Pouch leak Bowel perforation Ileus Ischemic bowel GI bleed C. difficile-negative C. difficile-positive 82 9 2 1 4 0 3 8 3 0 0 0 1 2 Wound No complications At least one complication Hematoma/seroma Dehiscence 87 10 0 2 10 1 1 1 Infectious No complications At least one complication Sepsis Abscess Wound infection Urinary tract infection Pneumonia Infected central line Perianal infection 75 16 7 6 3 2 4 0 1 6 5 3 2 0 1 3 1 0 Renal and Endocrine No complications At least one complication Acute renal failure Fluid/electrolyte disorders Adrenal 87 4 1 2 1 9 2 2 1 0 52 Cardiovascular No complications At least one complication Thrombosis Myocardial infarction Cardiac arrest Hypotension/shock Cardiac arrhythmia 83 8 5 3 1 3 5 8 3 2 1 0 1 0 Pulmonary No complications At least one complication Acute respiratory failure Hypoxia Pleural effusion/Pulmonary edema Pneumothorax and atelectasis 84 9 4 2 4 1 7 2 2 1 1 0 91 0 11 0 Neurological No complications At least one complication * Patients had ≥ 1 postoperative complication 53 2.12 Appendix 2.D Sensitivity analysis: Logistic regression results of emergent colectomy with C. difficile diagnosis in hospital or 14 days prior to admission as primary exposure. All UC Patients (n = 278) Adjusted analysis OR 95% CI C. difficile diagnosis No Yes 1.00 3.70 Referent (1.06-12.89) Age 18-32 33-47 ≥ 48 1.00 1.14 5.06 Referent (0.56-2.34) (2.39-10.74) Disease distribution Left-sided/undetermined Pancolitis 1.00 5.04 Referent (2.39-10.74) Prednisone at admission No Yes 1.00 2.54 Referent (1.32-4.85) IV-steroids in hospital No Yes 1.00 11.51 Referent (1.36-97.46) 54 2.13 Appendix 2.E Sensitivity analyses: Logistic regression results of any postoperative complication and infectious post-operative complication with C. difficile diagnosis in hospital or 14 days prior to admission as primary exposure. Emergent Colectomy Patients (n = 102) Any complication Infectious complication Adjusted analysis Adjusted analysis OR 95% CI OR 95% CI C. difficile diagnosis No Yes 1.00 6.11 Referent (1.32-28.21) 1.00 7.93 Referent (1.54-40.75) Age 18-32 33-47 ≥ 48 1.00 1.86 3.61 Referent (0.42-8.29) (0.88-14.96) 1.00 2.98 5.18 Referent (0.43-20.60) (0.8-33.77) Gender Male Female 1.00 0.29 Referent (0.1-0.89) 1.00 0.21 Referent (0.57-0.86) Comorbidity No Yes 1.00 1.75 Referent (0.58-5.26) 1.00 2.05 Referent (0.57-7.45) 55 2.14 Appendix 2.F Sensitivity analysis: Logistic regression results of emergent colectomy including those not tested for C. difficile. All UC Patients (n = 481) Adjusted analysis OR C. difficile diagnosis No Yes Not tested 1.00 2.80 1.69 95% CI Referent (0.86-9.11) (1.09-2.63) Age 18-32 33-47 ≥ 48 1.00 1.27 3.52 Referent (0.75-2.15) (2.05-6.04) Disease distribution Left-sided/Undetermined Pancolitis 1.00 4.51 Referent (2.83-6.04) Prednisone at admission No Yes 1.00 3.11 Referent (1.95-4.99) IV-steroids in hospital No Yes 1.00 6.21 Referent (2.33-16.53) 56 2.15 Appendix 2.G Sensitivity analyses: Logistic regression results of any postoperative complication and infectious post-operative complication including those not tested for C. difficile. Emergent Colectomy Patients (n = 186) Any complication Infectious complication Adjusted analysis Adjusted analysis OR 95% CI OR 95% CI C. difficile diagnosis No Yes Not tested 1.00 3.55 1.56 Referent (0.95-13.21) (0.80-3.07) 1.00 4.67 1.49 Referent (1.17-18.47) (0.67-3.29) Age 18-32 33-47 ≥ 48 1.00 1.16 2.81 Referent (0.47-2.89) (1.24-6.41) 1.00 1.03 3.61 Referent (0.32-3.37) (2.04-9.81) 57 Chapter Three: Clostridium difficile infection diagnosed early in ulcerative colitis patients impacts disease related outcomes: A population-based inception cohort study 58 3.1 Abstract Ulcerative colitis (UC) patients diagnosed with Clostridium difficile infection have a higher overall risk of colectomy. Still, previous research has been performed on prevalent UC cases rather than UC incident cases. The objectives of this study were to: 1) validate the ICD-10 code for C. difficile diagnosis among UC patients; 2) determine the 1, 3, and 5-year cumulative risk of a C. difficile diagnosis after their UC diagnosis; 3) determine the effect of a C. difficile diagnosis on the risk of colectomy and mortality; and 4) determine the association between a C. difficile diagnosis and developing postoperative complications in a population-based inception cohort of patients with UC. ICD-10 codes for C. difficile were compared to toxin tests from Calgary Laboratory Services and sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) with 95% CI were calculated. The 1, 3 and 5-year cumulative risk of being diagnosed with C. difficile infection following UC diagnosis date was calculated with a life table. Next, the effect of a C. difficile diagnosis on colectomy for UC and mortality was modelled using an adjusted competing risk survival regression model and a Cox proportional hazard model, respectively. Finally, the effect of a C. difficile diagnosis on postoperative complications was assessed using a mixed effects logistic regression model. The sensitivity, specificity, PPV and NPV of the ICD-10 diagnostic code of C. difficile diagnosis were 82.1% (95% CI: 71.7-89.9%), 99.4% (95% CI: 99.1-99.7%), 88.4% (95% CI: 82.9-92.3%) and 99.1% (95% CI: 98.5-99.4%), respectively. The cumulative risk of C. difficile diagnosis at 1, 3, and 5 years after UC diagnosis was 1.6% (95% CI: 1.1-2.3), 2.5% (95% CI: 1.9-3.4), and 3.2% (95% CI: 2.2-4.2), respectively. Patients with a C. difficile diagnosis were more likely to have colectomy compared to 59 those who had no prior C. difficile diagnosis (SHR=2.36; (95% CI: 1.47-3.80) after adjusting for age, sex, and comorbidities. The multilevel logistic regression model shows that C. difficile diagnosis in hospital ≤ 90 days prior to colectomy admission was associated with developing at least one postoperative complication (adjusted OR=4.84; 95% CI: 1.28-18.35). C. difficile diagnosis worsened the prognosis of UC by increasing the risk of colectomy and postoperative infectious complications following colectomy. Future studies are needed to explore whether early detection and aggressive management of C. difficile infection will improve UC outcomes. 60 3.2 Introduction Ulcerative colitis (UC) is an inflammatory bowel disease that results in chronic inflammation of the large intestine. The clinical course of UC is highly variable ranging from mild disease severity treated effectively with mesalamine to an aggressive disease course that requires treatment with immunosuppressants and/or biologics. While advances in medical management have led to decreased rates of colectomy in patients with UC, 16% of patients with UC require a colectomy within 10 years of their diagnosis.1 Colectomy for UC is associated with significant postoperative mortality (3.2%) and morbidity (33.3%).2,3 Thus, identifying modifiable risk factors that influence disease course is a clinical priority. Over the last 15 years, special attention has been given to the effect of Clostridium difficile infections on the clinical course of UC. Patients with UC have a high risk of acquiring C. difficile when compared to Crohn’s patients or the general population.4 Specifically, Nguyen et al. demonstrated that the prevalence rate of C. difficile among UC patients was approximately 8 and 3 times greater when compared to the non-IBD and CD patients, respectively.5 In addition, among UC patients C. difficile infections are associated with an increased risk of colectomy after C. difficile diagnosis.512 As demonstrated in Chapter 2, C. difficile infection prior to colectomy among UC patients admitted to hospital for a flare increased the risk of postoperative complications; however, this study was limited by small sample size.13 Most of the studies that assessed the risk of developing C. difficile infection and its effect on colectomy among UC patients used prevalent cohorts admitted to hospital. 61 These prevalent cohorts are mainly patients followed by gastroenterologists, most likely with severe form of UC.14 Thus, patients with an indolent course of UC may be missed. Further, many of these studies used administrative databases without validating the code to identify C. difficile infection.5,6,12,15,16 Thus, we cannot estimate the impact of misclassifying patients based on C. difficile exposure status. The objectives of this study were to: 1) validate the International Classification of Diseases-10 (ICD-10) diagnostic code for C. difficile diagnosis among UC patients; 2) determine the 1, 3, and 5-year cumulative risk of a C. difficile diagnosis after their UC diagnosis; 3) determine the effect of a C. difficile diagnosis on the risk of colectomy and mortality; and 4) to determine the association between a C. difficile diagnosis and developing postoperative complications using a population-based inception cohort of UC patients. 3.3 Materials and Methods 3.3.1 Study population 3.3.1.1 Validation cohort The study population for the validation of the ICD-10 code for C. difficile consisted of all adult UC patients (≥ 18 years of age) who were admitted to Calgary Health Zone (CHZ) hospitals with a diagnosis of UC between 2002 and 2009. The CHZ in Alberta, Canada, is a population-based health authority providing health care to residents in Calgary and over 20 nearby cities under a public single-payer system.17 62 3.3.1.2 Inception cohort of UC patients The study population consisted of all adults (≥ 18 years of age) diagnosed with UC and living in Alberta, Canada, between 2003 and 2010 while registered under the Alberta Health Care Insurance Plan (AHCIP). The AHCIP is publicly funded and provides hospital and health care services to >99% of Alberta’s population (~3 million residents). Every Alberta resident registered under the AHCIP has a unique personal health identification number, which can be used to track the patient through time and to link the patient to different health data sources in Alberta.18 3.3.2 Data sources 3.3.2.1 Validation of the ICD-10 code for C. difficile diagnosis among UC patients The Data Integration, Measurement and Reporting (DIMR) Hospital Discharge Abstract Database is a population-based database that includes patient’s demographical data, admission and discharge date, 42 diagnostic coding fields using the International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-10CM), and 25 procedural codes using the Canadian Classification of Health Intervention (CCI). The DIMR was used to capture all adults (≥ 18 years) admitted to hospital with a diagnosis of UC (ICD-10-CA K51.X) from 2002 to 2009. Charts were reviewed on all of these patients to confirm the diagnosis of UC and C. difficile infection. The ICD-10 codes for UC hospitalization and for colectomy were previously validated.19 Calgary Laboratory Services (CLS) is a medical diagnostic laboratory that provides services to the Calgary region and Southern Alberta. CLS is responsible for the majority of laboratory tests (microbiology, anatomic pathology and cytopathology) performed in hospital settings in Calgary. Calgary Laboratory Services provided 63 information on all C. difficile test results (date, type of test, result) performed on all UC patients admitted to hospitals in the CHZ. A patient was considered C. difficile-positive if at least one test result was positive. All stool toxin test for C. difficile captured by CLS were linked by a unique identifying variable to patients with UC who were identified in the DIMR Hospital Discharge Abstract Database. 3.3.2.2 Inception cohort of patients with UC The Alberta-wide UC inception cohort was derived from four linked data sources from AHCIP: 1) the Registry Database, contains participants’ demographic and geographic information for each year; 2) the Alberta Physicians’ Claims (PC) Database, contains outpatient claims for all people seen by a practitioner in Alberta. Information consists of demographic data, service date, 3 diagnostic codes (ICD-9) and one procedure code (CCI); 3) the Alberta Inpatient Hospital Discharge Abstract Database (DAD) contains demographic information, fiscal year, service dates, 25 diagnostic codes (ICD-9: 1995 to 2001; ICD-10-CA: 2002 to present) and 20 procedure codes (CCI); and 4) the Ambulatory Care Classification System (ACCS), contains information on facility-based ambulatory care ranging from emergency visits to outpatient procedures (e.g. colonoscopy). Data recorded in ACCS consists of demographic data, fiscal year, service dates, 10 diagnostic codes (ICD-9: 1998 to 2001; ICD-10-CA: 2002-present), and 10 procedure codes (CCI). Information from all databases was available from April 1, 1995 to March 31, 2010 with the exception of ACCS that was available from April 1, 1997 to March 31, 2010. Information provided to investigators is de-identifiable. 64 3.3.3 Development of the UC inception cohort We used the four linked data sources from AHCIP to identify all IBD patients living in Alberta using a previously validated IBD case definition.20 This validated IBD algorithm involves the identification of diagnostic codes for IBD (i.e. ICD-9: 555; 556; ICD-10: K50; K51) in the DAD database (≥ 2 hospitalizations within a 2-year period), or ACCS database (≥ 2 ambulatory services contacts within a 2-year period), or PC database (≥ 4 physician contacts within a 2-year period). Date of the first contact with an IBD code was recorded on those patients who fulfilled the case definition. Next, a previously validated IBD scoring system was used to distinguish UC from Crohn’s disease (CD).20 A +1 score was assigned to any related admission/visit with an ICD-9/10 code for UC and a -1 score for any related admission/visit with an ICD-9/10 code for CD.20 A cumulative score was calculated and those with a score ≥ 2 were considered UC patients and those with a score ≤ -2 were considered CD patients. Patients with a score between -2 and 2 were classified as IBD unknown. Patients with CD and IBD unknown were excluded from the UC inception cohort. Validation studies have shown that an 8-year washout period is necessary to differentiate prevalent from incident cases.20 Thus, a UC patient was considered to be an incident case only if the patient was an Alberta resident for at least 8 years prior to the date of their first UC contact. Patients who were < 18 years of age at UC diagnosis were excluded. In order to minimize misclassification errors patients who had a diagnosis of C. difficile prior to the diagnosis of UC were excluded (n = 32) (Figure 3.1). Though, patients who underwent a colectomy prior to their diagnosis of UC (n = 24) were 65 excluded. Patients that had a C. difficile diagnosis and/or colectomy 14 days prior to their diagnosis date were included in the study. 3.3.4 Outcome and predictor variables The primary outcome was having a colectomy. A patient was defined as a colectomy case if a colectomy procedure code (CCI: 1.NM.87, 1.NM.89, 1.NM.91, 1.NQ.89, 1.NQ.90) was identified in the DAD. The first colectomy was recorded in situations where more than one admission had a colectomy procedure code. Patients that were identified as having a colectomy up to 14 days prior to their diagnosis date were included in the analysis as some of these patients were diagnosed with UC based on histopathology of the resected colon. Secondary outcomes included mortality and the development of in hospital postoperative complications. A patient was recorded as having a postoperative complication if any of the pre-defined ICD-10 codes for postoperative complications occurred within the same admission as the colectomy procedure (Appendix 3.A).2,12,21,22 These postoperative complications were further classified into seven categories including infectious complications (Appendix 3.A).2,12,21,22 Postoperative complications codes following colectomy for UC have been previously validated. A diagnosis of C. difficile within the same colectomy admission was not considered to be an infectious complication. The primary exposure was the diagnosis of C. difficile in either the DAD or ACCS database. The Physician Claims database cannot identify C. difficile because the diagnostic code is limited to 3 characters and the ICD-9 code for C. difficile (008.45). The date of the first positive test was recorded as the index date of C. difficile infection. 66 Additional demographic and clinical data was extracted from the Registry, PC, DAD and ACCS datasets and included: age, sex, death, disease duration (defined as the interval between UC diagnosis and colectomy or death), type of surgery procedure (defined as open or laparoscopic approach), type of admission on colectomy admission (defined as elective or emergent) and a de-identifiable hospital identification number. The Charlson comorbidity index was used to identify patients’ comorbidities (i.e. health conditions occurring before hospitalization).23 The diagnostics codes used for differentiating between types of surgical procedures and the Charlson comorbidities are in Appendices 3.B and 3.C. 3.3.5 Analyses All analyses was done using STATA statistical software version 11 (STATA Corp, College Station, TX, USA). The study was approved by the Conjoint Health Research Ethics Board of the University of Calgary 3.3.5.1 Validation of the ICD-10 code for C. difficile diagnosis among UC patients Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) with 95% CI were calculated for the ICD-10 code for C. difficile during admission to hospital. This was done by comparing the presence of an ICD-10 code for C. difficile during hospital admission to the CLS provided stool toxin test result for C. difficile performed during hospital admission and 7 days prior to hospital admission (gold standard). A dichotomous variable was created for the presence of an ICD-10 code for C. difficile for the hospital discharge abstract database. 67 3.3.5.2 Effect of C. difficile on UC incident cases A life table was constructed to calculate the 1-, 3-, and 5-year cumulative risk of being diagnosed with their first C. difficile infection following UC diagnosis. For the primary outcome, we modelled the cumulative incidence of colectomy in the presence of competing mortality using competing risk regression. The primary exposure was being diagnosed with C. difficile any time between diagnosis of UC and colectomy. The following variables were included a priori in to the model: sex, age at diagnosis of UC, and comorbidities (0 = no, 1 = one comorbidity, 2 = ≥ 2 comorbidities). Patients were censored if they moved out of Alberta, Canada. Interactions between C. difficile and the presence of comorbidities, sex and age (continuous) were explored and included in the model if the likelihood ratio test was statistically significant (P < 0.05). Statistical significance was defined as a two-sided P-value < 0.05. Point estimates were presented as sub hazard ratio (SHR) with 95% confidence interval. A secondary analysis using a Cox proportional hazard model was used to assess the effect of a C. difficile infection on the overall mortality. The following variables were included a priori in to the model: sex, age at diagnosis of UC, and comorbidities (0 = no, 1 = one comorbidity, 2 = ≥ 2 comorbidities). Patients were censored if they moved out of Alberta, Canada. Interactions between C. difficile and the presence of comorbidities, sex and age (continuous) were explored and included in the model if they were tested and included in the model if the likelihood ratio test was statistically significant (P < 0.05). Statistical Significance was defined as a two-sided P-value < 0.05. Point estimates were presented as a hazard ratio (HR) with 95% confidence interval. 68 The association between a C. difficile diagnoses on postoperative complications was evaluated in a subset of the UC population who underwent colectomy. For this analysis, C. difficile was stratified into 3 categories: no C. difficile diagnosis prior to colectomy admission, C. difficile diagnosis captured > 90 d prior to colectomy admission and C. difficile diagnosis < 90 d prior to colectomy admission. A mixed effects logistic regression model was used to evaluate the risk of C. difficile on postoperative complication including hospital as a random effect (to account for clustering within hospital). The following covariates were a priori forced in the model because they have been previously shown to influence the risk of postoperative complications following colectomy for UC: age, sex, comorbidities, type of colectomy admission (i.e. emergent versus elective) and type of surgery procedure (i.e. open versus laparoscopic).2 Interactions between C. difficile and variables independently associated with colectomy were tested and included in the model if the likelihood ratio test was statistically significant (P < 0.05). Point estimates were presented as adjusted odds ratios (OR) with 95% confidence interval (CI). A sensitivity analyses was performed to determine whether C. difficile had an effect on developing a postoperative infectious complication using the same procedure as described above. 3.4 Results 3.4.1 Validation of the ICD-10 code for C. difficile diagnosis among UC patients Between 2002 and 2009, 1,686 UC patients with 3,603 admissions to CHZ hospitals were identified. A C. difficile test was performed in 37.0% (n = 1,333) of the 69 admissions. Among these, a C. difficile positive test result was identified in 2.3% (n = 84) of the admissions. An ICD-10 code for C. difficile diagnosis was recorded in 2.2% (n = 78) of the admissions. Sensitivity, specificity, PPV and NPV the ICD-10 code for C. difficile diagnosis was 82.1% (95% CI: 71.7-89.9%), 99.4% (95% CI: 99.1-99.7%), 88.4% (95% CI: 82.9-92.3%) and 99.1% (95% CI: 98.5-99.4%), respectively. 3.4.2 Effect of C. difficile on UC incident cases Between April 1, 1994 and March 31, 2010, 8,011 people with a UC diagnosis received medical services in Alberta (Figure 3.1). From these, between April 1, 2003 and March 31, 2010, 1,754 patients met our inclusion criteria of an incident case. Table 1 shows the characteristics of the UC incident cohort stratified by whether they had a colectomy or not. The median time to follow-up for the inception cohort was 3.8 years. A C. difficile diagnosis was recorded in 4.6% of the patients. The cumulative risk of the C. difficile diagnosis at 1-, 3-, and 5-years after UC diagnosis was 1.8% (95% CI: 1.2-2.5%), 2.8% (95% CI: 2.0-3.7%), and 3.4% (95% CI: 2.5-4.6%), respectively. The 1-, 3-, and 5-year cumulative colectomy risk on the entire inception cohort was 5.7% (95% CI: 4.6-6.9%), 12.0% (95% CI: 10.4-13.8%), and 14.0% (95% CI: 12.3-16.1%), respectively. The 1-, 3- and 5-year cumulative colectomy risk among those with no prior C. difficile diagnosis was 5.0% (95% CI: 4.1-6.2%), 11.3% (95% CI: 9.7-13.1%), and 13.3% (95% CI: 12.3-17.2%), respectively. The 1-, 3- and 5-year cumulative colectomy risk among UC patients with a C. difficile infection was 18.7% (95% CI: 11.5-29.6%), 27.2% (95% CI: 18.1-39.5%), and 29.3% (95% CI: 19.7-42.1%), respectively. The results of the competing risk regression demonstrate that C. difficile increases the incidence of colectomy or death (Figure 3.2). Patients with a C. difficile diagnosis were more likely to 70 have colectomy compared to those who had no prior C. difficile diagnosis (SHR=2.36; (95% CI: 1.47-3.80) after adjusting for age, sex, and comorbidities (Table 3.2 & Figure 3.2). We further explored the timing of C. difficile and colectomy among these patients and found that those who had a C. difficile diagnosis ≤ 90 days prior to colectomy had an earlier disease course (median time: 32 days, 25th percentile: 2.5 days, 75th percentile: 180 days) compared to those who had a C. difficile diagnosis > 90 prior to (median time: 441 days, 25th percentile: 273.5 days, 75th percentile: 1004 days) and those without a C. difficile diagnosis (median time: 287 days, 25th percentile: 25 days, 75th percentile: 667 days). Overall 5-year mortality risk was higher among patients diagnosed with C. difficle infection (5-year mortality= 17.5%; 95% CI: 9.2-31.7) compared to those with no C. difficile diagnosis (5-year mortality=3.3%; 95% CI: 2.4-4.5). The results of the Cox proportional hazard model shows that a C. difficile diagnosis increases the hazard of dying by 2.56 times (95% CI: 1.28-5.10) after adjusting for age, sex, and comorbidities (Table 3.2). Table 3.3 shows patients characteristics among those who had a colectomy stratified by having any postoperative complication. At least one postoperative complication was recorded in 41% of UC patients (see Appendix 3.D for further information on the types of postoperative complications observed). The multilevel logistic regression model shows that C. difficile diagnosis in hospital ≤ 90 days prior to colectomy admission was associated with developing at least one postoperative complication (adjusted OR=4.84; 95% CI: 1.28-18.35) (Table 3.4). 71 The sensitivity analysis shows that a diagnosis of C. difficile in hospital ≤ 90 days prior to colectomy admission was not associated with developing an infectious postoperative complication (adjusted OR=1.53; 95% CI: 0.46-5.09) (Appendix 3.E and 3.F). 3.5 Discussion This is the first study confirming the validity of ICD-10 code for C. difficile among UC patients. Using this validated code, we showed that within 5 years of being diagnosed with UC 3.2% of UC patients were diagnosed with C. difficile infection. Patients with UC who were diagnosed with C. difficile were significantly more likely to undergo a colectomy or die. Further, C. difficile infection reduced the interval from diagnosis to colectomy when compared to UC patients without a C. difficile diagnosis. Finally, patients with UC who were diagnosed with a C. difficile infection were at increased risk of developing a postoperative complication. Collectively, these data highlight that a diagnosis of C. difficile has a major impact on the course of UC. Our population-based UC inception cohort demonstrated that within 5 years of diagnosis 3.4% of UC patients are diagnosed with C. difficile infection and nearly 1 in 3 of these patients undergo colectomy within 5 years of their diagnosis of C. difficile infection. These data suggest that either C. difficile infection directly worsens the prognosis of UC or that patients with an aggressive severe phenotype of UC are more likely to be diagnosed with C. difficile infection. While we controlled for comorbidities and disease duration, our study could not account for disease severity or extent, and we did not control for different C. difficile ribotypes that may have influenced the 72 relationship between C. difficile infection and UC prognosis. Further, we adjusted for clustering within hospitals, but could not directly account for variation in treatment for C. difficile infection between hospitals and practitioners.11,24-26 Future prospective studies should be designed to identify a risk-profile at diagnosis of UC that predicts acquisition of C. difficile infection after adjusting for disease severity. The increased risk of acquiring C. difficile among UC patients has been consistently shown in epidemiological studies.11 Further, our study and others, demonstrated that UC patients with a C. difficile diagnosis had an increased risk of colectomy, postoperative complications and mortality.5-10,12,13,15,16,27,28 The reason why C. difficile is more predominant in UC is likely multifactorial. First, the altered gut microbiota, observed among UC patients provides an optimal environment for C. difficile to proliferate.29 Second, patient related factors such as the ability to develop an adequate innate and adaptive immune response have been shown to influence the sequelae following C. difficile colonization. For example, the innate immune response, which minimizes mucosal damage and inflammation, is dependent on commensal bacteria.29 Thus, UC patients might have increased mucosal damage as the result the observed gut dysbiosis characteristic of UC. Also, the lack of an adequate adaptive immune response following C. difficile colonization has been associated with developing C. difficileassociated diarrhoea among the non-IBD population.30 In UC patients, genetic polymorphisms, which are involved in the regulation of innate or adaptive immunity, have been found to be associated with an increased risk of acquiring C. difficile infection.31 While these genetic polymorphisms were not associated with a more aggressive phenotype of UC,31 the presence of multiple risk loci was associated with 73 developing C. difficile infection early in the UC disease course.31 This could explain our findings that patients who had a colectomy within 90 days of a C. difficile diagnosis were earlier in their UC disease course. Our data demonstrated that a C. difficile diagnosis prior to colectomy increased the risk of postoperative complications, but not specifically postoperative infection. This finding was in contrast to a recent study that demonstrated a diagnosis of C. difficile infection was associated with the development of infectious postoperative complications.13 The previous study confirmed postoperative complications through medical chart review, whereas the current study relied on the accuracy of administrative codes. A prior validation study demonstrated that the sensitivity of identifying any major postoperative complication in administrative data was 85%, but dropped to 66% for postoperative infectious complications.19 The lower sensitivity likely led to a higher number of unidentified patients with an infectious postoperative complication, which decreased the power and precision of our estimates. We studied a large population-based incident cohort of UC patients using administrative data. However, there are several limitations that should be considered with the use of administrative databases.14 First, misclassification of UC and C. difficile diagnosis and the development of postoperative complications might have influenced the magnitude of the risk estimates and the precision of the confidence intervals. However, the IBD case definition and the UC scoring were validated within the data obtained from AHCIP with a sensitivity and specificity of 83.2% (95% CI: 82.1-84.3) and 86.3% (+/SD: 1.9), respectively. Additionally, we validated the diagnostic validity of the ICD-10 code for C. difficile among UC patients using a subset of UC patients living in Alberta, 74 Canada. However, during the C. difficile validation phase we found that only one third of the patients had a test done for C. difficile. This could give rise to a possible differential misclassification bias where patients with a severe form of UC were more likely tested for C. difficile, inflating the risk estimate. On the other hand, assuming that patients with no C. difficile code were in fact negative may have shifted the risk estimates towards the null. Additionally, we were unable to control for disease severity and other markers for C. difficile outcomes, disease extent and the use of IBD medication and antibiotics. Thus, our findings should be independently replicated. These findings have important clinical implications. First, physicians should carefully assess for C. difficile infection among all UC patients. However, special attention should be given to those presenting with C. difficile infection early on the UC disease course and carefully monitor them as they are at increased risk of colectomy and postoperative complications. Future studies are necessary to: 1) determine the impact of host immunity on C. difficile outcomes among UC patients and 2) clinical trials on which treatment course improves the outcomes among those with C. difficile infections. 75 3.6 References 1. Frolkis AD, Dykeman J, Negron ME, et al. Risk of surgery for inflammatory bowel diseases has decreased over time: a systematic review and meta-analysis of population-based studies. Gastroenterology 2013;145:996-1006. 2. Frolkis AD, Kaplan GG, Patel A, et al. Short-term emergent readmission and postoperative complications in children and adults with inflammatory bowel disease who undergo intestinal resection – A population-based study. Accepted: Inflamm Bowel Dis 2014. 3. de Silva S, Ma C, Proulx MC, et al. Postoperative Complications and Mortality Following Colectomy for Ulcerative Colitis. Clin Gastroenterol Hepatol 2011;9:972-80. 4. Nguyen GC, Tuskey A, Dassopoulos T, et al. Rising hospitalization rates for inflammatory bowel disease in the United States between 1998 and 2004. Inflamm Bowel Dis 2007;13:1529-35. 5. Nguyen GC, Kaplan GG, Harris ML, et al. A national survey of the prevalence and impact of Clostridium difficile infection among hospitalized inflammatory bowel disease patients. Am J Gastroenterol 2008;103:1443-50. 6. Murthy SK, Steinhart AH, Tinmouth J, et al. Impact of Clostridium difficile colitis on 5-year health outcomes in patients with ulcerative colitis. Aliment Pharm Ther 2012;36:1032-9. 7. Ananthakrishnan AN, McGinley EL, Saeian K, et al. Temporal trends in disease outcomes related to Clostridium difficile infection in patients with inflammatory bowel disease. Inflamm Bowel Dis 2011;17:976-83. 76 8. Issa M, Vijayapal A, Graham MB, et al. Impact of Clostridium difficile on inflammatory bowel disease. Clin Gastroenterol Hepatol 2007;5:345-51. 9. Navaneethan U, Mukewar S, Venkatesh PG, et al. Clostridium difficile infection is associated with worse long term outcome in patients with ulcerative colitis. J Crohn's Colitis 2012;6:330-6. 10. Jen MH, Saxena S, Bottle A, et al. Increased health burden associated with Clostridium difficile diarrhoea in patients with inflammatory bowel disease. Aliment Pharm Ther 2011;33:1322-31. 11. Goodhand JR, Alazawi W, Rampton DS. Systematic review: Clostridium difficile and inflammatory bowel disease. Aliment Pharm Ther 2011;33:428-41. 12. Khan NA, Quan H, Bugar JM, et al. Association of postoperative complications with hospital costs and length of stay in a tertiary care center. J Gen Intern Med 2006;21:177-80. 13. Negron ME, Barkema HW, Rioux K, et al. Clostridium difficile infection in ulcerative colitis: Increased risk of colectomy and postoperative infectious complications. Am J Gastroenterol 2011;106:S481-2. 14. Molodecky NA, Pannacione R, Ghosh S, et al. Challenges associated with identifying the environmental determinants of the inflammatory bowel diseases. Inflamm Bowel Dis 2011;17:1792-9. 15. Ananthakrishnan AN, McGinley EL, Binion DG. Excess hospitalisation burden associated with Clostridium difficile in patients with inflammatory bowel disease. Gut 2008;57:205-10. 77 16. Ricciardi R, Ogilvie JW, Roberts PL, et al. Epidemiology of Clostridium difficile colitis in hospitalized patients with inflammatory bowel diseases. Dis Colon Rectum 2009;52:40-5. 17. Alberta Health Services. Alberta Health Services Annual Report, April 1,2009March 31, 2010. http://www.albertahealthservices.ca/Publications/ahs-pub-annualrpt.pdf. Accessed: June 1, 2014. 18. Alberta Health. Alberta health care insurance plan. 2014. Alberta Health Services. http://www.health.alberta.ca/health-care-insurance-plan.html. Accessed: June 1, 2014. 19. Ma C, Crespin M, Proulx MC, et al. Postoperative complications following colectomy for ulcerative colitis: A validation study. BMC Gastroenterology 2012;12:39. 20. Rezaie A, Quan H, Fedorak R, et al. Development and validation of an administrative case definition for inflammatory bowel diseases. Can J Gastroenterol 2012;26:711-8. 21. Ma C, Crespin M, Proulx MC, et al. Accuracy of administrative data in identyfying ulcerative colitis patients presenting with acute flare and undergoing colectomy. Ame J Gastroenterol 2010;105:A1270. 22. Kaplan GG, McCarthy EP, Ayanian JZ, et al. Impact of hospital volume on postoperative morbidity and mortality following a colectomy for ulcerative colitis. Gastroenterology 2008;134:680-7. 78 23. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130-9. 24. Hubert B, Loo VG, Bourgault AM, et al. A portrait of the geografic dissemination of the Clostridium difficile North American pulsed-field type 1 strain and the epidemiology of C. difficile-associated disease in Quebec. Clin Infect Dis 2007;44:238-244. 25. Sundram F, Guyot A, Carboo I, et al. Clostridium difficile ribotypes 027 and 106: clinical outcomes and risk factors. J Hosp Infect 2009;72:111-118. 26. Yanai H, Nguyen GC, Yun L, et al. Practice of gastroenterologists in treating flaring inflammatory bowel disease patients with Clostridium difficile: Antibiotics alone or combined antibiotics/immunomodulators? Inflamm Bowel Dis 2011;17:1540-6. 27. Jodorkovsky D, Young Y, Abreu MT. Clinical outcomes of patients with ulcerative colitis and co-existing Clostridium difficile infection. Dig Dis Sci 2010;55:415-20. 28. Murthy SK, Steinhart AH, Tinmouth J, et al. Impact of Clostridium difficile colitis on 5-year health outcomes in patients with ulcerative colitis. Aliment Pharm Ther 2012;36:1032-9. 29. Buffie CG, Pamer EG. Microbiota-mediated colonization resistance against intestinal pathogens. Nat Rev Immunol 2013;12:790-801. 30. Kyne L, Warny M, Qamar A, et al. Asymptomatic carriage of Clostridium difficile and serum levels of IgG antibody against Toxin A. New Engl J Med 2000;342:3907. 79 31. Ananthakrishnan AN, Oxford EC, Nguyen DD, et al. Genetic risk factors for Clostridium difficile infection in ulcerative colitis. Aliment Pharmacol Ther 2013;38:522-3 80 Table 3.1. Characteristics of patients diagnosed with ulcerative colitis in Alberta, Canada, between April 1st, 2003 and March 31st, 2010 stratified by colectomy. Characteristics All (n = 1,754) No Colectomy (n = 1,518) Colectomy (n = 234) P-Value Gender, % (n) Male Female 50.4 (884) 49.6 (870) 49.1 (746) 50.9 (774) 59.0 (138) 41.0 (96) 0.005 Age at diagnosis, y* 25th percentile Median 75th percentile 28 41 55 28 41 55 30 46 57 N/A N/A 24 275.5 648 N/A Comorbidities, % (n) 0 1 ≥2 40.8 (715) 29.2 (515) 30.0 (527) 42.2 (642) 29.9 (454) 27.9 (424) 31.2 (73) 24.8 (58) 44.0 (103) < 0.001 C. difficile, % (n) No Yes 95.4 (1,673) 4.6 (81) 96.25 (1,463) 3.75 (57) 89.7 (210) 10.3 (24) < 0.001 Time to first C. difficile, d† 25th percentile 0 0 0 0.588 Time to surgery, d† 25th percentile Median 75th percentile 0.051 81 Median 75th percentile 14 274 16 286 13 228.5 Death No Yes 96.9 (1,699) 3.1 (55) 97.1 (1,476) 2.9 (44) 95.3 (223) 4.7 (11) * years, † days 82 0.156 Table 3.2. Results of the competing risk regression and the Cox proportional hazard models for time to colectomy and time to death since UC diagnosis with prior C. difficile diagnosis as the primary predictor Variables Time to Colectomy SHazard ratio (95% CI) Time to death Hazard ratio (95% CI) C. difficile No Yes Baseline 2.36 (1.47-3.80) Baseline 2.56 (1.28-5.10) Gender Female Men Baseline 1.65 (1.24-2.19) Baseline 0.89 (0.52-1.52) Age at diagnosis, y* 0.99 (0.98-1.0) 1.09 (1.07-1.12) Comorbidities 0 1 ≥2 Baseline 1.10 (0.75-1.61) 2.30 (1.60-3.32) Baseline 0.68 (0.09-4.94) 5.96 (1.36-26.04) * years 83 Table 3.3. Characteristics of colectomy patients stratified by developing any postoperative complication in hospital Characteristics No (n = 138) Any complication Yes (n = 96) P-Value Gender, % (n) Male Female 57.25 (79) 42.75 (59) 61.5 (59) 38.5 (37) 0.589 Age at admission, y* 25th percentile Median 75th percentile 28 40 56 30 50 64.5 0.001 Time to surgery, d† 25th percentile Median 75th percentile 0 52 358 5 193 648.5 0.038 Comorbidities 0 1 ≥2 65.9 (91) 18.2 (25) 15.9 (22) 59.4 (57) 24.0 (23) 16.7 (16) 0.511 Time between first C. difficile and colectomy, No C. difficile 94.9 (131) † > 90 d 2.2 (3) ≤ 90 d† 2.9 (4) 82.3 (79) 5.2 (5) 12.5 (12) 0.005 Type of Surgery Endoscopic Open 18.1 (25) 81.9 (113) 10.4 (10) 89.6 (86) 0.136 Type of admission Elective Emergent 43.5 (60) 56.5 (78) 20.8 (20) 79.2 (76) < 0.001 * years † days 84 Table 3.4. Results of multilevel regression models for the development of any postoperative complication and a postoperative infectious complication with C. difficile as primary predictor (facility was included as a random effect in the models) Variables Any Complication OR (95%CI) Time between first C. difficile and colectomy, No C. difficile > 90 d† ≤ 90 d† Baseline 2.48 4.84 (0.42-14.65) (1.28-18.35) Age at admission, y* 1.03 (1.01-1.05) Gender Female Male Baseline 1.58 (0.83-2.99) Colectomy admission Elective Emergent Baseline 3.36 (1.68-6.71) Type of procedure Endoscopic Open Baseline 2.90 (1.13-7.45) Comorbidities 0 1 ≥2 Baseline 1.19 0.63 (0.54-2.61) (0.26-1.54) * years, † days 85 Figure 3.1. Flow diagram of patient selection criteria Alberta Health Care Insurance Plan Registry DAD1 ACCS2 PC3 IBD case definition4: 2-4-2-2 23,681 IBD patients identified Crohn’s disease patients (n=12,153) & IBDU5 patients (n=3,517) 8,011 UC patients Not registered as an Alberta resident (n=38) 7,973 UC patients Prevalent cases(n=5,965) 2,008 UC incident6 patients A colectomy code prior to UC diagnosis (n=24) Diagnosis of C. difficile prior to UC diagnosis (n=32) Age at UC diagnosis <18 years (n=203) 1,754 UC incident patients 1 Alberta Inpatient Hospital Discharge Abstract Database Ambulatory Care Classification System Database 3 Alberta Physicians’ Claims Database 4 A diagnostic code for IBD in ≥2 hospitalizations within a 2-year period (DAD) or ≥4 ambulatory contact within a 2-year period (ACCS) or ≥2 physician contacts within a 2-year period (PC) 5 IBD unknown based on scoring system 6 Incident cases where those who were Alberta residents for at least 8 years prior to the date of the first UC contact 2 86 30 25 20 15 10 5 0 Cumulative incidence of colectomy (%) Figure 3.2. Survival curves for the proportion of colectomy free UC patients (a) and survival of UC patients (b) stratified by having C. difficile diagnosis (red line) and no C. difficile diagnosis (blue line) 0 365 730 1095 1460 1825 Time since UC diagnosis (days) cdif_1=0 87 cdif_1=1 2190 2555 3.7 Appendix 3.A List of postoperative complications and the respective ICD-10 code by category Category Gastrointestinal Code K22.8 K25 K26 K27 K29.0 K55.0 K56.0 K56.6 K56.7 K62.5 K63.1 K63.8 K66.0 K72.0 K72.9 K85 K92 K56.6 K91.3 K91.4 K91.8 K91.9 T76.2 T81.0 Renal and Endocrine E15 E27.2 E86 E87 N13.9 Complication Other specified diseases of esophagus or haemorrhage of esophagus NOS Gastric ulcer Duodenal ulcer Peptic ulcer, site unspecified Acute haemorrhagic gastritis Acute vascular disorders of intestine, unspecified Paralytic ileus Other and unspecified intestinal obstructions Ileus, unspecified Haemorrhage of anus and rectum Perforation of intestine (nontraumatic) Other specified diseases of intestine Peritoneal adhesions Acute and subacute hepatic failure Hepatic failure, unspecified Acute pancreatitis Other diseases of digestive system Intestinal adhesion [bands] with obstruction Postoperative intestinal obstruction Colostomy and enterostomy malfunction Other postprocedural disorders of digestive system, not elsewhere classified Postprocedural disorder of digestive system, unspecified Traumatic secondary and recurrent haemorrhage Haemorrhage and haematoma complicating a procedure, not elsewhere classified Non diabetic hypoglaecemic coma Addisonian crisis; adrenal crisis; adrenocortical crisis Volume depletion Other disorders of fluid, electrolyte and acid-base balance Obstructive and reflux uropathy, unspecified: urinary 88 N17 N19 N31.2 R33 N10 N12 N99.0 N99.9 Cardiovascular I21 I26 I46 I48 I49 I50 I74 I80 I81 I92 I95.0 I95.2 I95.9 R57 I97.8 I97.9 T79.0 T80.0 T80.1 T81.1 T81.7 tract obstruction NOS Acute renal failure Unspecified kidney failure Flaccid neuropathic bladder, not elsewhere classified Retention of urine Acute tubulo-interstitial nephritis Tubulo-interstitial nephritis, not specified as acute or chronic Postprocedural renal failure Postprocedural disorder of genitourinary system, unspecified Acute myocardial infarction Pulmonary embolism Cardiac arrest Atrial fibrillation and flutter Other cardiac Heart failure Arterial embolism and thrombosis Phlebitis and thrombophlebitis Portal vein thrombosis Other venous embolism and thrombosis Idiopathic hypotension Hypotension due to drugs Hypotension, unspecified Shock, not elsewhere specified Other postprocedural disorders of the circulatory system, not elsewhere classified Post procedural disorder of circulatory system, unspecified Air embolism (traumatic) Air embolism following infusion, transfusion and therapeutic injection Vascular complications following infusion, transfusion and therapeutic injection Shock during or resulting from a procedure, not elsewhere classified Vascular complications following a procedure, not 89 Pulmonary Neurological T88.2 J80 J81 J90 J91 J93 J96.0 J96.9 J98.1 R09 J95.5 J95.8 F05 F13 F15 F19 G45 G46 G81 G82 G83 G93.1 G93.6 G97.0 G97.1 163 165 G97.8 elsewhere classified Shock due to anaesthesia Adult respiratory distress syndrome Pulmonary oedema Pleural effusion, not elsewhere classified Pleural effusion in conditions classified elsewhere Pneumothorax Acute respiratory failure Respiratory failure, unspecified Pulmonary collapse Other symptoms and signs involving the circulatory and respiratory systems Postprocedural subglottic stenosis Other postprocedural respiratory disorders Delirium, not induced by alcohol or other psychoactive substances Mental and behavioural disorders due to use of sedatives or hypnotics Mental and behavioural disorders due to use of other stimulants, including caffeine Mental and behavioural disorders due to multiple drug use and use of other psychoactive substances Transient cerebral and ischaemic attacks and related syndromes Vascular syndromes of brain in cerebrovascular diseases Hemiplegia Paraplegia and tetraplegia Other paralytic syndromes Anoxic brain damage, not elsewhere classified Cerebral oedema Cerebrospinal fluid leak from spinal puncture Other reaction to spinal and lumbar puncture Cerebral infarction Occlusion and stenosis of precerebral arteries, not resulting in cerebral infarction Other postprocedural disorders of nervous system 90 Wounds G97.9 K60.3 K60.4 K60.5 K63.2 K82.9 K83.2 L89 N36.0 N82.4 T81.2 T81.3 T81.5 T81.8 Infections T81.9 A40 A41 B95 B96 J10.0 J11.0 J12 J13 J14 J15 J16 J17 J18 J69.0 J85 J86 Postprocedural disorder of nervous system, unspecified Anal fistula Rectal fistula Anorectal fistula Fistula of intestine Disease of gallbladder, unspecified Perforation of bile duct Decubitus ulcer and pressure Urethral fistula Other female intestinal-genital tract fistulae Accidental puncture and laceration during a procedure, not elsewhere classified Disruption of operation wound, not elsewhere classified Foreign body accidentally left in body cavity or operation wound following a procedure Other complications of procedures, not elsewhere classified Unspecified complication of procedure Streptococcal sepsis Other sepsis Streptococcus and staphylococcus as the cause of disease classified to the other chapters Influenza with pneumonia, other influenza virus identified Influenza with pneumonia, virus not identified Viral pneumonia, not elsewhere classified Pneumonia due to Streptococcus pneumoniae Pneumonia due to Haemophilus influenzae Bacterial pneumonia, not elsewhere classified Pneumonia due to other infectious organisms, not elsewhere classified Pneumonia in diseases classified elsewhere Pneumonia, organism unspecified Pneumonitis due to food and vomit Abscess of lung and mediastinum Pyothorax Abscess of anal and rectal regions 91 K61 K63 K65 L03 L04 N15.1 N15.9 N30.9 N39.0 T80.2 A49 T79.3 T81.4 T81.6 T82.7 T83.6 T85.7 Other T88.3 T88.5 T88.6 T88.7 R78.8 T88.8 T88.9 Abscess of intestine Peritonitis Cellulitis Acute lymphadenitis Renal and perinephric abscess Renal tubule-interstitial disease, unspecified Acute cystitis Cystitis, unspecified Urinary tract infection, site not specified Infections following infusion, transfusion and therapeutic injection Bacterial infection of unspecified site Post-traumatic wound infection, not elsewhere classified Infection following a procedure, not elsewhere classified Acute reaction to foreign substance accidentally left during a procedure Infection and inflammatory reaction due to other cardiac and vascular devices, implants and grafts Infection and inflammatory reaction due to prosthetic device, implant and graft in genital tract Infection and inflammatory reaction due to other internal prosthetic devices, implants, and grafts Malignant hyperthermia due to anaesthesia Other complications of anaesthesia Anaphylactic shock due to adverse effect of correct drug or medicament properly administered Unspecified adverse effect of drug or medicament Finding of other specified substances, not normally found in blood Complication of surgical and medical care, unspecified Other specified complications of surgical and medical care, not elsewhere classified 92 3.8 Appendix 3.B Charlson comorbidities Comorbidity Codes Acute myocardial infarction 410, 412, I21, I22, I252 Congestive heart failure 428, I43, I50, 39891, 40201, 40211, 40291, 40401, 4254, 4255, 4257,4258, 4259, I099, I110, I130, I426, I427, I428, I429, P290, I132, I255, I420, I425, 40403, 40411, 40413, 40491, 40493 4439, V434, 7854, 5571, 5579, V434, 0930, 4373, 4431, 4432, 4438, 4439, 4471, 441, I70, I71, I731, I738, I739, I771, I790, I792, K551, K558, K559, Z958, Z959 430, 431, 432, 433, 434, 435, 436, 437, 438, 36234, 430, 431, 432, 433, 434, 435, 436, 437, 438, G45, G46, I60, I61, I62, I63, I64, I65, I66, I67, I68, I69, H340 290, 2941, 3312, F00, F01, F02, F03, G30, F051, G311 Peripheral vascular disease Cerebrovascular disease Dementia Chronic pulmonary disease 490, 491, 492, 493, 494, 495, 496, 500, 501, 502, 503, 504, 505, 5064, 4168, 4169, 5064, 5081, 5088, 490, 491, 492, 493, 494, 495, 496, 500, 501, 502, 503, 504, 505, J40, J41, J42, J43, J44, J45, J46, J47, J60, J61, J62, J63, J64, J65, J66, J67, I278, I279, J684, J701, J703 Rheumatologic disease (connective 7100, 7101, 7104, 7140, 7141, 7142, 725, 71481, 4465, 7100, 7101, 7102, 7103, 7104, tissue disease) 7140, 7141, 7142, 7148, 725, M05, M32, M33, M34, M06, M315, M351, M353, M360 Peptic ulcer disease 531, 532, 533, 534, 531, 532, 533, 534, K25, K26, K27, K28 Mild liver disease 5712, 5714, 5715, 5716, 57140, 57141, 57149, 07022, 07023, 07032, 07033, 07044, 07054, 0706, 0709, 5733, 5734, 5738, 5739, V427, 570, 571, B18, K73, K74, K700, K701, K702, K703, K709, K713, K714, K715, K717, K760, K762, K763, K764, K768, K769, Z944 93 Moderate or severe liver disease Metastatic carcinoma 5722, 5723, 5724, 5728, 4560, 4561, 4562, 45620, 45621, 4560, 4561, 4562, 5722, 5723, 5724, 5728, K704, K711, K721, K729, K765, K766, K767, I850, I859, I864, I982 2500, 2501, 2502, 2503, 2507, 2500, 2501, 2502, 2503, 2508, 2509, E100, E101, E106, E108, E109, E110, E111, E116, E118, E119, E120, E121, E126, E128, E129, E130, E131, E136, E138, E139, E140, E141, E146, E148, E149 2504, 2505, 2506, 2504, 2505, 2506, 2507, E102, E103, E104, E105, E107, E112, E113, E114, E115, E117, E122, E123, E124, E125, E127, E132, E133, E134, E135, E137, E142, E143, E144, E145, E147 342, G81, G82, 3341, 3440, 3441, 3442, 3443, 3444, 3445, 3446, 3449, G041, G114, G801, G802, G830, G831, G832, G833, G834, G839 5830, 5831, 5832, 5834, 5836, 5837, 582, 585, 586, 588, 582, 585, 586, V56, 5830, 5831, 5832, 5834, 5836, 5837, 5880, V420, V451, 40301, 40311, 40391, 40402, 40403, 40412, 40413, 40492, 40493, N052, N053, N054, N055, N056, N057, N250, I120, I131, N032, N033, N034, N035, N036, N037, Z490, Z491, Z492, Z940, Z992, N18, N19 C00, C01, C02, C03, C04, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 170, 171, 172, 174, 175, 176, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 200, 201, 202, 203, 204, 205, 206, 207, 208, C05, C06, C07, 2386, C08, C09, C10, C11, C12, C13, C14, C15, C16, C17, C18, C19, C20, C21, C22, C23, C24, C25, C26, C30, C31, C32, C33, C34, C37, C38, C39, C40, C41, C43, C45, C46, C47, C48, C49, C50, C51, C52, C53, C54, C55, C56, C57, C58, C60, C61, C62, C63, C64, C65, C66, C67, C68, C69, C70, C71, C72, C73, C74, C75, C76, C81, C82, C83, C84, C85, C88, C90, C91, C92, C93, C94, C95, C96, C97, 174, 195, 200, 208, 140, 172 196, 197, 198, 199, 196, 197, 198, 199, C77, C78, C79, C80 HIV 042, 043, 044, 042, 043, 044, B20, B21, B22, B24 Diabetes without complication Diabetes with chronic complications Hemiplegia or paraplegia Renal disease Cancer 94 3.9 Appendix 3.C Classification of colectomy approach and ICD-10 codes Approach Code Technique Open Code Endoscopic Code 1.NM.87.DA 1.NM.87.BA 1.NM.87.LA 1.NM.87.^^ 1.NM.89.^^ 1.NM.91.^^ Partial excision of large intestine Total excision of large intestine Radical excision of large intestine 1.NM.87.RN 1.NM.87.DF 1.NM.87.RD 1.NM.87.DE 1.NM.87.RE 1.NM.87.DN 1.NM.87.TF 1.NM.87.DX 1.NM.87.TG 1.NM.87. 1.NM.89.RN 1.NM.89.DF 1.NM.89.TF 1.NM.89.DX 1.NM.91.RN 1.NM.91.DF 1.NM.91.RD 1.NM.91.DE 1.NM.91.RE 1.NM.91.DN 95 Definition simple excisional technique colocolostomy anastomosis technique colorectal anastomosis technique enterocolostomy anastomosis technique stoma formation and distal closure stoma formation with creation of mucous fistula ileorectal anastomosis technique stoma formation with distal closure colocolostomy anastomosis technique colorectal anastomosis technique enterocolostomy anastomosis technique 1.NQ.89.^^ 1.NQ.90^^ Total excision of the rectum Total excision with reconstruction of rectum 1.NM.91.TF 1.NM.91.DX 1.NM.91.TG 1.NM.91.DY 1.NQ.89.SF 1.NQ.89.KZ 1.NQ.89.SF-XX-G 1.NQ.89.KZ-XX-G 1.NQ.89.RS 1.NQ.89.LH 1.NQ.89.RS-XX-G 1.NQ.89.LH-XX-G 1.NQ90.LA-XX-G 96 1.NQ.89.GV stoma formation and distal closure stoma formation with creation of mucous fistula coloanal anastomosis technique pouch formation 1.NQ.89.AB stoma formation with distal closure continent ileostomy formation pouch formation 3.10 Appendix 3.D Types of postoperative complications Timing of C. difficile Postoperative complication Gastrointestinal, % (n) No C. difficile (n=210) Over 90 days prior to colectomy (n=8) Up to 90 days prior to colectomy (n=16) 19.1 (40) 37.5 (3) 37.5 (6) 0 (0) 18.8 (3) Renal and Endocrine, % 5.7 (12) (n) Cardiovascular, % (n) 9.5 (20) 0 (0) 31.3 (5) Pulmonary, % (n) 5.2 (11) 0 (0) 37.5 (6) Neurological, % (n) 1.4 (3) 0 (0) 6.3 (1) Wound, % (n) 9.05 (19) 25.0 (2) 31.3 (5) Infectious, % (n) 17.1 (36) 25.0 (2) 31.3 (5) Other, % (n) 1. (2) 0 (0) 0 (0) 97 3.11 Appendix 3.E Characteristics of colectomy patients stratified by developing infectious complications in hospital Characteristics No (n = 191) Infectious complication Yes P-Value (n = 43) Gender, % (n) Male Female 39.8 (76) 60.2 (115) 46.5 (20) 53.5 (23) 0.493 Age at admission, y* 25th percentile Median 75th percentile 29 44 57 46 52 68 0.004 Time to surgery, days 25th percentile Median 75th percentile 36 351 680 0 54 627 0.011 Comorbidities 0 1 ≥2 63.9 (122) 19.4 (37) 16.7 (32) 69.5 (26) 25.6 (11) 13.9 (6) 0.651 Time between first C. difficile and colectomy, No C. difficile 91.1 (174) > 90 days 3.1 (6) ≤ 90 days 5.8 (11) 83.7 (36) 4.7 (2) 11.6 (5) 0.244 Type of Surgery Endoscopic Open 16.7 (32) 83.3 (159) 6.7 (3) 93.0 (40) 0.153 Type of admission Elective Emergent 38.7 (74) 61.3 (117) 14.0 (6) 86.0 (37) 0.002 98 3.12 Appendix 3.F Results of multilevel logistic regression model for the development of infectious complications with C. difficile as primary predictor (hospital was included as a random effect in the model) Variables Infectious Complication OR (95% CI) Time between first C. difficile and colectomy, No C. difficile > 90 days ≤ 90 days Baseline 1.67 1.53 (0.26-10.59) (0.46-5.09) Age 1.03 (1.01-1.05) Gender Female Male Baseline 0.80 (0.39-1.65) Colectomy admission Elective Emergent Baseline 3.88 (1.50-10.00) Type of procedure Endoscopic Open Baseline 3.02 (0.83-11.01) Comorbidities 0 1 2 Baseline 1.27 0.59 (0.52-3.08) (0.20-1.73) 99 Chapter Four: Changes in the annual incidence of colectomy for colorectal neoplasia among patients with ulcerative colitis: A population-based cohort 100 4.1 Abstract Incidence of surgery among ulcerative colitis (UC) patients has decreased over time. This decrease in incidence has been attributed to medications (i.e. immunomodulators and biologics) that have improved disease management. However, the reduced risk of colectomy could also be attributed to a lower incidence of colectomy for colorectal dysplasia or cancer. Thus, a multi-method approach was performed to determine whether the incidence of colectomy for colorectal dysplasia or cancer has changed over time. Adult UC patients who had a colectomy between 1997 and 2009 were identified. First, a logistic regression analysis examined the association between year and indication for colectomy (dysplasia or cancer vs. medically refractory disease) after controlling for potential confounders. Second, the annual incidence of colectomy for colorectal dysplasia or cancer and medically refractory disease was calculated for each year. The referent population was all patients with UC living in the Calgary Health Zone and at risk for undergoing a colectomy. A generalized linear model was built to assess changes in colectomy rates during the study period. Of the 638 patients admitted to hospital who underwent a colectomy, 10% (n = 58) had a colectomy due to colorectal dysplasia or cancer. The incidence of colectomy for colorectal neoplasia was stable in both analyses, whereas colectomy rates for medically refractory disease decreased from 1997 to 2009. Future studies are needed to evaluate modifiable factors that may reduce the risk of dysplasia or cancer among patients with UC. 101 4.2 Introduction Ulcerative colitis (UC) is a chronic disease characterized by inflammation of the large intestine. Most patients with UC require medications to induce and maintain remission. Colectomy is warranted among those with medically refractory disease and/or complications, and for those with colorectal dysplasia or cancer. Colectomy is an important outcome among patients with UC due to the morbidity and mortality associated with this procedure.1 A meta-analysis demonstrated that 16% of patients with UC underwent a colectomy within 10 years of diagnosis.2 Furthermore, the risk of colectomy within 10 years of diagnosis decreased significantly across time. Advances in medical management of UC including the advent of mesalamine, immunomodulators, and biologics have led to a decrease in the risk of colectomy for medically refractory disease. UC patients have a higher risk of developing colorectal cancer when compared to the non-IBD population. A meta-analysis reported that the 10 and 20-year risk of developing dysplasia or cancer was 2 and 8%, respectively.3 However, advances in the medical management of UC may have affected the risk of dysplasia and colorectal cancer among patients with UC. More recent estimates from a meta-analysis of population-based studies reported a 20-year risk of dysplasia and cancer between 1.1-5.3%.4 Some of the factors that increase the risk of colorectal cancer are: age and gender, greater disease extent, long disease duration, primary sclerosing cholangitis (PSC), familial history of colorectal cancer.4-7 Experimental studies have shown that both 5-ASA and biologics could be used as chemoprotectants.8,9 However, a recent meta-analysis concluded that 5ASA does not confers a protective effect against the development of colorectal cancer.10 102 Studies that have evaluated whether the risk of dysplasia or cancer has decreased over time are contradictory. Only one of these studies reported that the incidence of colorectal cancer is increasing with an overall incidence rate of 14 cases per 1,000 patient years (95%CI: 5-34 per 1,000 patient years).11 Two studies reported a decrease in the incidence of colorectal cancer with only one performing a meta-regression but the results were not statistically significant.12,13 These differences may be explained by selection biases of the target population (i.e. tertiary referral centers vs. population-based studies), the methodology used to assess changes over time, and geographic differences in clinical practice.14 Thus, we evaluated the risk profile for the development of dysplasia and colorectal cancer, and determined if the incidence of colectomy for colorectal dysplasia or cancer has change over time. 4.3 Materials and Methods 4.3.1 Data sources and study population 4.3.1.1 Data Integration, Measurement, and Reporting Hospital Discharge Abstract The Calgary Health Zone (CHZ) in Alberta, Canada is a population-based health authority providing health care to residents in Calgary and over 20 nearby cities under a public single-payer system.15 In 2009 the estimated population of the CHZ in 2009 was 1,326,115.16 The DIMR database was used to identify all adult UC patients (≥ 18 years of age) who were admitted to CHZ hospitals with a diagnosis of UC who underwent a colectomy between 1 January 1997 and 31 December 2009. Medical chart review confirmed all patients with UC who underwent a colectomy. 103 4.3.1.2 Alberta Health Care Insurance Plan (AHCIP) The AHCIP provides universal health coverage to more than 99% of Alberta’s population (~3 million residents).17 Each Alberta resident has a unique personal health identification number, which can be used to track the patient through time and to link the patient to different health data sources in Alberta. Information provided to investigators is de-identifiable. Four data sources from AHCIP were used in this study to identify UC patients at risk for a colectomy. The AHCIP Registry database contains participants’ demographic and geographic information for each fiscal year. The Alberta Inpatient Hospital Discharge Abstract Database (DAD) contains demographic information, year, service dates, 25 diagnostic codes (ICD-9: from 1995 to 2002; ICD-10-CA: from 2003 to 2010), 20 procedure codes. The Alberta Physicians’ Claims (PC) Database contains claims for all people seen by a practitioner in Alberta. Information consists of demographic data, service date, 3 diagnostic codes (ICD-9) and one procedure code. The Ambulatory Care Classification System (ACCS) Database collects information on facility-based ambulatory care ranging from emergency visits and outpatient procedures to education treatment and rehabilitation programs. Recorded data consists of demographic data, year, service dates, 10 diagnostic codes (ICD-9: from 1997 to 2002; ICD-10-CA: from 2003 to 2010), and 10 procedure codes. The AHCIP databases were used to identify all adult UC patients (≥ 18 years of age) who lived in the CHZ between 1 January 1997 and 31 December 2009. 4.3.2 Identification of UC colectomy patients in CHZ hospitals and data collection The DIMR database was used to identify hospitalizations of UC patients who had a colectomy in CHZ hospitals between 1997 and 2009. The approach of identifying the 104 study population was previously validated.18 Briefly, the DIMR was used to identify: 1) patients admitted to hospital with a UC diagnosis (ICD-9-CM 556.X or ICD-10-CA K51.X) in any diagnostic position and a procedural code for colectomy (ICD-9-CM 45.7, 45.8 or CCI 1.NM.87, 1.NM.89, 1.NM.91, 1.NQ.89, 1.NQ.90). The primary outcome was the indication for colectomy: colorectal dysplasia or cancer versus medically refractory. Colorectal dysplasia or cancer was defined by review of the pathology report for the colectomy specimen. Medically refractory disease was defined as patients who underwent colectomy because of failed medical management or complication (e.g. toxic megacolon) requiring surgery. Data extracted during chart review for the primary variables of interests was based on a priori defined risk factors4-7 for colorectal cancer and included: age, (defined as age at colectomy admission, stratified as: 18-34, 35-64, and ≥ 65 years); sex; disease extent (left-sided versus pancolitis); disease duration (defined as the interval between UC diagnosis and colectomy date), and smoking (current, ex-smoker, never) at time of colectomy. IBD medications (5-ASA or sulfasalazine, azathioprine/6-mercaptopurine) taken at time of admission and/or administered in-hospital were recorded. Patients were classified as having comorbidities (i.e. health conditions occurring before hospitalization) if they had at least one of the comorbidities listed in Appendix 4.A (stratified as no comorbidities, 1-2 comorbidities or > 2 comorbidities). Of note, PSC was not included as comorbidity as it was included as a separate variable on the analysis. The definition of comorbidities have been previously validated for UC.18 105 4.3.3 Identification of UC patients at risk for colorectal cancer in CHZ using AHCIP The AHCIP’s Registry, PC, DAD, and ACCS databases were used to identify those UC patients at risk of having a colectomy in the CHZ. First, we identified all IBD patients living in Alberta using a previously validated case definition of IBD.19 The validated Alberta algorithm for identifying a prevalent case of IBD involves the identification of diagnostic codes for IBD (i.e. ICD-9: 555; 556; ICD-10: K50; K51) in the DAD database (≥ 2 hospitalizations within a 2-year period), or the PC database (≥ 4 physician contacts within a 2-year period), or the ACCS database (≥ 2 physician contacts within a 2-year period). This algorithm has a sensitivity of 83.4% and a specificity of 99.8%.19 Next, we used a previously validated IBD scoring system to distinguish UC from CD on those patients that fulfilled the prevalent case definition. A +1 score was assigned to any hospital admission or physician contact with an ICD code for UC (556 or K51) and a -1 score for an ICD code for CD (555 or K51).19 A cumulative score was calculated and those with a score > 2 were considered UC patients and those with a score < -2 were considered CD patients. Patients with a score between -2 and 2 were classified as IBD unknown. This scoring system has a sensitivity of 86.3% and a specificity of 99.7%.19 Prevalent cases of CD and IBD unknown were excluded. Next, prevalent UC cases were identified for each fiscal year between 1997 and 2009. UC patients with a procedural code for colectomy (ICD-9-CM 45.7, 45.8 or CCI 1.NM.87, 1.NM.89, 1.NM.91, 1.NQ.89, 1.NQ.90) prior to the fiscal year being assessed were excluded. 106 4.3.4 Analyses 4.3.4.1 Risk factors for colectomy for colorectal dysplasia or cancer Descriptive statistics were performed using the Fisher’s exact test or the X2 test. Continuous variables were expressed as medians with interquartile ranges (IQR) and compared using the Wilcoxon rank sum test. We performed a multivariate logistic regression model to determine differences in patients’ characteristics between UC patients who had a colectomy because of colorectal dysplasia or cancer as compared to those who had a colectomy because of medically refractory disease with year of admission as the primary exposure. Known risk factors associated with colorectal dysplasia or cancer among UC patients that were a priori forced into the regression model included: age, sex, disease duration, disease extent, PSC, smoking, and the presence of comorbidities. A temporal analysis was conducted by entering year into the model as a continuous variable. For medications, a backwards elimination approach was used to examine independent effects of additional variables on the need of surgery with an entry P-value of < 0.20 and were kept in the model if the two-sided P-value was < 0.05 or there was evidence of confounding because their removal resulted in a 30% change in the estimate of the primary exposure. Statistical Significance was defined as a two-sided P-value < 0.05. Point estimates were presented as adjusted odds ratios (OR) with 95% confidence interval (CI). 4.3.4.2 Incidence of colectomy for colorectal dysplasia or cancer and medically refractory disease The annual incidence of colectomy for colorectal dysplasia or cancer and the annual incidence of colectomy for medically refractory disease was calculated by 107 dividing the total number of colectomies by the estimated number of UC patients at risk of having a colectomy each year within the CHZ. A generalized linear model (GLM) was created to test whether colectomy rates for colorectal dysplasia or cancer have changed over time using year of colectomy admission (continuous) as the primary predictor while adjusting for age and sex. An additional variable was included in the model to account for annual differences on the size of the population at risk living in CHZ. A scale parameter was included in the model to account for overdispersion. Significance was defined as a two-sided P-value < 0.05. Point estimates were presented as adjusted risk ratios (RR) with 95% CI. A second GLM was created to determine whether colectomy rates for medically refractory disease have changed over time using the same procedure as before. Statistical analyses were conducted using STATA statistical software version 11 (STATA Corp, College Station, TX, USA). The study was approved by the Conjoint Health Research Ethics Board of the University of Calgary. Our study was conducted in accordance with the strengthening the reporting of observational studies in Epidemiology (STROBE).20 4.4 Results 4.4.1 Risk factors for colectomy for colorectal dysplasia or cancer Between 1997 and 2009 we identified a population-based cohort of 638 UC patients who had a colectomy at a CHZ hospital. From these, 9.1% (n = 58) had surgery because of colorectal dysplasia or cancer. The median age at surgery among those who had surgery for colorectal dysplasia or cancer was 53.9 years (range: 18.1-86.6 years), while 77.6% were males (Table 4.1). The median age at surgery among those who had 108 surgery because of medically refractory UC was 39.1 years (range: 21.8-86 years), while 58.45% were males (Table 4.1). Year of admission was not significantly associated with having a colectomy for colorectal dysplasia or cancer (adjusted OR=1.04; 95% CI: 0.96-1.14) after adjusting for clinical risk factors of colorectal dysplasia or cancer (Table 4.2). Patients with UC who underwent a colectomy due to colorectal dysplasia or cancer were more likely to be men (OR=2.55; 95%: 1.20-5.41) with longer disease duration (OR= 1.12; 95% CI: 1.09-1.15). Prescription of mesalamine (OR=0.43; 95% CI: 0.22-0.86) and azathioprine/6mercaptopurine (OR=0.33; 95% CI: 0.12-0.91) were associated with a decreased risk of colectomy due to dysplasia or cancer when compared to colectomy for medically refractory disease. 4.4.2 Incidence of colectomy for colorectal neoplasia and medically refractory disease Among patients with UC who were at risk for a colectomy between 1997 and 2009, the risk of having a colectomy due to medically refractory disease decreased (Year: RR=0.97; 95% CI: 0.94-0.99), whereas the risk of having a colectomy for dysplasia or cancer was stable (Year: RR=1.03; 95% CI: 0.94-1.11) (Figure 4.1). Additionally, the risk of colectomy for colorectal dysplasia or cancer was stable after adjusting for age (3564 versus 18-34 years: RR=9.77, 95% CI: 1.55-61.54; >65 years versus 18-34 years: RR=11.62, 95% CI: 1.79-75.59) and sex (males versus females: RR 3.42; 95% CI: 1.667.05). 109 4.5 Discussion In this population-based study, we used a multi-method approach to determine whether the incidence of colectomy for colorectal dysplasia or cancer has changed over time. In both approaches the incidence of colectomy for neoplasia remained stable. In contrast, the risk of colectomy for medically refractory patients (i.e. excluding patients with colectomy for dysplasia or cancer) dropped significantly by 3% per year between 1997 and 2009. This finding is consistent with previously conducted population-based studies demonstrating that the risk of colectomy has been decreasing across time.2,21 In addition, those who had a colectomy for colorectal dysplasia and cancer were older males with longer disease duration as compared to patients who underwent a colectomy due to medically refractory UC. Colorectal cancer is an established complication of chronic UC.3,4,11-13,22 In this study, the risk of colectomy for dysplasia or colorectal cancer remained stable between 1997 and 2009. The annual incidence of colectomy for dysplasia in Calgary varied between 0.2 and 1.1 cases per 1,000 UC patients. We used two control populations to evaluate the risk of colectomy for dysplasia of cancer. The referent population in the first analysis was patients with UC who underwent a colectomy due to failure of medical management. The advantage of using this referent population was that the control population had a complete chart review, which extracted clinically relevant information that may have influenced the risk of colectomy. This analysis demonstrated that the risk of colectomy for dysplasia and colorectal cancer was stable after adjusting for age, sex, disease duration, disease extent, primary sclerosing cholangitis and other comorbidities, smoking, and the use of medications. However, the conclusion that colectomy rates were 110 stable overtime is relative to a control population that only included patients with severe UC requiring colectomy. This analysis did not account for patients with milder UC or those with worse prognosis, but responded to medical management (e.g. biologics). In response, we conducted a second analysis whereby the denominator was defined as all patients with UC at risk for colectomy. This referent population was derived from administrative health databases in Alberta using a previously validated case-definition.19 This temporal analysis included all UC patients at risk for colectomy and accounted for the increasing prevalence of UC overtime.23 This analysis also demonstrated a stable rate of colectomy for dysplasia and cancer; however, the temporal analysis was only adjusted for age and sex because these databases do not include clinically relevant information on risk factors for colectomy. Despite the differences in study design between both of these analyses, the risk estimates for temporal analyses were nearly identical (year as continuous variable: OR=1.03 and RR=1.04) and nonsignificant. Thus, this multi-method study suggests that despite the advances to medical management during this time period, colectomy for dysplasia or cancer remained constant between 1997 and 2009. We found that patients who had a colectomy for colorectal dysplasia and cancer were more likely to be men with longer disease duration compared to those who had a colectomy due to medically refractory disease. The role of IBD medications as chemoprotectants remains controversial with differences observed between published meta-analyses attributed to study inclusion criteria.10,24 A meta-analysis by Velayos et al., using population-based and tertiary center studies, demonstrated a protective association between 5-ASA and colorectal dysplasia or cancer (OR=0.51; 95% CI: 0.38-0.69).24 On 111 the other hand, Nguyen et al., using only population-based studies, demonstrated that 5ASA did not protect against developing colorectal cancer (OR=0.95; 95% CI: 0.541.26).10 In our study, both mesalamine and azathioprine/6-mercaptopurine decreased the risk of colectomy for dysplasia or cancer. However, because the control population was UC patients with colectomy for medically refractory disease, the negative associations may reflect risk adjustment for disease severity rather than a protective effect of these medications. Additionally, we did not evaluate the effect of anti-TNF therapies on the risk of neoplasia because the time between its introduction (2005)25 and our study end point (2009) was insufficient to evaluate a reduction in the risk of dysplasia or cancer (i.e. >10 years of follow-up required26,27). Future studies are needed to assess the effect of anti-TNF therapies on the risk of colorectal dysplasia or cancer. Both methodological approaches have distinct limitations.28 The first approach extracted patient information through medical chart review that relied on clinicians for the accuracy and completion of patients’ medical history. In addition, because of the retrospective nature of the chart review, we did not include some important disease characteristics like severity of inflammation.29 Also, the prevalence of dysplasia or cancer was rare in our study population (n=58 over 13 years). For example, a diagnosis of PSC doubled the odds of colectomy for colorectal neoplasia; however, the study lacked power to demonstrate a significant effect. In addition, the study was limited to patients with UC who underwent colectomy for dysplasia or cancer. UC patients with dysplasia that was not treated with colectomy (e.g. unifocal low-grade dysplasia)30 and patients with cancer not treated with colectomy (e.g. died) would not have been captured. For our second approach, we used administrative data to identify patients with UC who were at risk for 112 colectomy for dysplasia or cancer. Despite using a validated case-definition for UC, misclassification bias is inherent (e.g. 3% of UC cases are false-positive)19 and we were not able to adjust for important confounders like medication and disease duration. Finally, we could not adjust for PSC in the control population due to the low accuracy of algorithms for PSC patient identification in administrative databases.31 Both methodological approaches with different referent populations consistently demonstrate that the incidence of colectomy for colorectal dysplasia or cancer remained stable between 1997 and 2009. Thus, the reported reduction in the overall rates of colectomy for UC21 is most likely driven by a decrease in the incidence of colectomy due to medically refractory disease. Our findings suggest that those at risk for a colectomy for colorectal dysplasia and cancer are males with longer disease duration. Future studies are needed to evaluate modifiable factors that may reduce the risk of dysplasia or cancer among patients with UC. 113 4.6 References 1. Kaplan GG, McCarthy EP, Ayanian JZ, et al. Impact of hospital volume on postoperative morbidity and mortality following a colectomy for ulcerative colitis. Gastroenterology 2008;134:680-7. 2. Frolkis AD, Dykeman J, Negron ME, et al. Risk of surgery for inflammatory bowel diseases has decreased over time: a systematic review and meta-analysis of population-based studies. Gastroenterology 2013;145:996-1006. 3. Eaden JA, Abrams KR, Mayberry JF. The risk of colorectal cancer in ulcerative colitis: a meta-analysis. Gut 2001;48:526-35. 4. Jess T, Rungoe C, Peyrin-Biroulet L. Risk of colorectal cancer in patients with ulcerative colitis: a meta-analysis of population-based cohort studies. Clin Gastroenterol Hepatol 2012;10:639-45. 5. Triantafillidis JK, Nasioulas G, Kosmidis PA. Colorectal cancer and inflammatory bowel disease: Epidemiology, risk factors, mechanisms of carcinogenesis and prevention strategies. Anticancer Res 2009;29:2727-38. 6. Velayos FS, Loftus EV, Jr., Jess T, et al. Predictive and protective factors associated with colorectal cancer in ulcerative colitis: A case-control study. Gastroenterology 2006;130:1941-9. 7. Collins PD, Mpofu C, Watson AJ, et al. Strategies for detecting colon cancer and/or dysplasia in patiens with inflammatory bowel disease. Cochrane Database Syst Rev 2006;CD000279. 114 8. Onizawa M, Nagaishi T, Kanai T, et al. Signaling pathway via TNF-alpha/NFkappaB in intestinal epithelial cells may be directly involved in colitis-associated carcinogenesis. Am J Physiol Gastrointest Liver Physiol 2009;296:G850-9. 9. Yan F, Polk DB. Aminosalicylic acid inhibits IkB kinase a phosphorylation of IkBa in mouse intestinal epithelial cells. Gastroenterology 1999;116:602-9. 10. Nguyen GC, Gulamhusein A, Bernstein CN. 5-aminosalicylic acid is not protective against colorectal cancer in inflammatory bowel disease: a metaanalysis of non-referral populations. Am J Gastroenterol 2012;107:1298-304. 11. Thomas T, Abrams KA, Robinson RJ, et al. Meta-analysis: cancer risk of lowgrade dysplasia in chronic ulcerative colitis. Aliment Pharmacol Ther 2007;25:657-68. 12. Lutgens MW, van Oijen MG, van der Heijden GJ, et al. Declining risk of colorectal cancer in inflammatory bowel disease: an updated meta-analysis of population-based cohort studies. Inflamm Bowel Dis 2013;19:789-99. 13. Castano-Milla C, Chaparro M, Gisbert JP. Systematic review with meta-analysis: the declining risk of colorectal cancer in ulcerative colitis. Aliment Pharmacol Ther 2014;39:645-59. 14. Andersen NN, Jess T. Has the risk of colorectal cancer in inflammatory bowel disease decreased? World J Gastroenterol 2013;19:7561-8. 15. Alberta Health Services. Alberta Health Services: Calgary Zone. 2014. http://www.albertahealthservices.ca/calgary-zone.asp. Accessed: June 1, 2014. 115 16. Alberta Health Services. Alberta Health Services Annual Report, April 1,2009March 31, 2010. http://www.albertahealthservices.ca/Publications/ahs-pub- annual-rpt.pdf. Accessed: June 1, 2014. 17. Alberta Health. Alberta health care insurance plan. 2014. Alberta Health Services. http://www.health.alberta.ca/health-care-insurance-plan.html. Accessed: June 1, 2014. 18. Ma C, Crespin M, Proulx MC, et al. Postoperative complications following colectomy for ulcerative colitis: A validation study. BMC Gastroenterology 2012;12:39. 19. Rezaie A, Quan H, Fedorak R, et al. Development and validation of an administrative case definition for inflammatory bowel diseases. Can J Gastroenterol 2012;26:711-8. 20. Vandenbroucke JP, von Elm E, Altman DG, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): Explanation and elaboration. PLoS Med 2007;4:e297. 21. Kaplan GG, Seow CH, Ghosh S, et al. Decreasing colectomy rates for ulcerative colitis: a population-based time trend study. Am J Gastroenterol 2012;107:187987. 22. Crohn BB, Rosenberg H. The sigmoidoscopic picture of chronic ulcerative colitis (non-specific). Am J Med Sci 1925;170:220-7. 23. Molodecky NA, Soon IS, Rabi DM, et al. Increasing incidence and prevalence of the inflammatory bowel diseases with time, based on systematic seview. Gastroenterology 2012;142:46-54. 116 24. Velayos FS, Terdiman JP, Walsh JM. Effect of 5-aminosalicylate use on colorectal cancer and dysplasia risk: a systematic review and metaanalysis of observational studies. Am J Gastroenterol 2005;100:1345-53. 25. Stidham RW, Lee TC, Higgins PD, et al. Systematic review with network metaanalysis: the efficacy of anti-tumour necrosis factor-alpha agents for the treatment of ulcerative colitis. Aliment Pharmacol Ther 2014;39:660-71. 26. Kornbluth A, Sachar DB. Ulcerative colitis practice guidelines in adults: American College Of Gastroenterology, Practice Parameters Committee. Am J Gastroenterol 2010;105:501-23. 27. Farraye FA, Odze RD, Eaden J, et al. AGA medical position statement on the diagnosis and management of colorectal neoplasia in inflammatory bowel disease. Gastroenterology 2010;138:738-45. 28. Molodecky NA, Kaplan G. Environmental risk factors for inflammatory bowel disease. Gastroenterol Hepatol 2010;6:339-46. 29. Rubin DT, Huo D, Kinnucan JA, et al. Inflammation is an independent risk factor for colonic neoplasia in patients with ulcerative colitis: a case-control study. Clin Gastroenterol Hepatol 2013;11:1601-8. 30. Siegel CA, Schwartz LM, Woloshin S, et al. When should ulcerative colitis patients undergo colectomy for dysplasia? Mismatch between patient preferences and physician recommendations. Inflamm Bowel Dis 2010;16:1658-62. 31. Molodecky NA, Myers RP, Barkema HW, et al. Validity of administrative data for the diagnosis of primary sclerosing cholangitis: a population-based study. Liver Int 2011;31:712-20. 117 Table 4.1. Characteristics of ulcerative colitis patients admitted to Calgary Health Zone hospitals that had a colectomy for colorectal dysplasia or cancer versus colectomy for medically refractory UC. Colectomy indication Characteristics Colorectal neoplasia (n = 58) Medical refractory (n = 580) Age at operation, y* 25th Percentile Median 75th Percentile 46.5 53.9 63.3 28.8 39.1 51.7 Gender, % (n) Male Female 77.6 (45) 22.4 (13) 58.45 (339) 41.55 (241) Disease duration, y* 25th Percentile Median 75th Percentile Missing (n) 10 18 30 3 1 3 9 24 Comorbidity, % (n) 0 1-2 >2 39.7 (23) 48.3 (28) 12.1 (7) 51.4 (298) 40.5 (235) 8.1 (47) Primary sclerosing cholangitis, % (n) No Yes 91.4 (53) 8.6 (5) 97.9 (568) 2.1 (12) Smoking status, % (n) Current Ex-smokers Never Missing (n) 3.7 (2) 33.3 (18) 63.0 (34) 4 8.2 (46) 34.8 (195) 57.0 (320) 19 Disease Extent, % (n) 118 P-Value <0.001 0.005 <0.001 0.178 0.014 0.513 Left-sided Pancolitis Missing (n) 27.8 (15) 72.2 (39) 4 21.7 (121) 78.3 (437) 22 Mesalamine/sulfasalazine†, % (n) No Yes Missing, (n) 36.4 (20) 63.6 (35) 3 26.25 (152) 73.75 (427) 1 Azathioprine/6mercaptopurine†, % (n) No Yes Missing, (n) 90.9 (50) 9.1 (5) 3 72.9 (422) 27.1 (157) 1 y*- year, † Medication taken at admission or in-hospital 119 0.306 0.114 0.002 Table 4.2. Logistic regression results comparing colectomy patients’ characteristics (colorectal dysplasia or cancer vs. medically refractory) with year of colectomy as primary exposure. Variables Final model Odds Ratio (95% CI) Year 1.04 (0.96-1.14) Age 18-34 35-64 ≥ 65 Baseline 2.14 (0.96-4.77) 2.91 (0.99-8.59) Gender Female Male Baseline 2.55 (1.20-5.41) Disease duration (years) 1.12 (1.09-1.15) Disease extent Pancolitis Left sided Baseline 1.85 (0.88-3.87) Comorbidities 0 1-2 >2 Baseline 1.47 (0.73-2.97) 1.58 (0.50-4.96) Primary sclerosing cholangitis No Yes Baseline 2.03 (0.46-8.88) Mesalamine/sulfasalazine No Yes Baseline 0.43 (0.22-0.86) Azathioprine/6-mercaptopurine No Yes Baseline 0.33 (0.12-0.91) 120 Smoking Current Ex-smokers Never Baseline 0.96 (0.17-4.98) 1.36 (0.28-6.61) 121 13 11 9 .3 .6 .9 1. 2 7 5 Colectomies per 1,000 UC patients at risk 15 Figure 4.1. Annual incidence of colectomy for colorectal dysplasia or cancer (green line), and medically refractory disease (red line) among UC patients living in the Calgary Health Zone between 1997 and 2009. The graph presents annual adjusted incidence of colectomy per 1,000 UC patients at risk* 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Year Colorectal neoplasia Medically refractory *Adjusted for the UC prevalent population at risk of colectomy in Calgary, Canada. 122 4.7 Appendix 4.A Comorbidity classifications Category Coronary Artery Disease Comorbidity Coronary artery disease, Ischemic heart disease, Myocardial infarction, Peripheral vascular disease Cancer Lymphoma, Metastatic tumor, Solid tumor without metastases Other cardiovascular Cardiac arrhythmia, Valvular disorder CHF Congestive heart failure Diabetes Diabetes with complications, Diabetes without complications Venous Thromboembolism Deep vein thrombosis GI CMV infection, Pancreatitis, Peptic ulcer disease Haematological Blood loss anaemia, Coagulopathy, Cyclical neutropenia, Deficiency anaemia HTN Liver Hypertension Fatty liver Neurological Cerebrovascular disease, Hemiplegia and paraplegia Pulmonary Renal Asthma, COPD, Sarcoidosis Renal failure (acute or chronic) Rheumatoid Ankylosing spondylitis, Episcleritis, Uveitis and iritis, Gout, Sacroiliitis, Rheumatoid arthritis 123 Chapter Five: Colorectal cancer surveillance in patients with inflammatory bowel disease and primary sclerosing cholangitis: An economic evaluation 124 5.1 Abstract The cost-effectiveness of annual colonoscopy for detection of colorectal neoplasia among patients with inflammatory bowel disease (IBD) and primary sclerosing cholangitis (PSC) is uncertain. The aim of this study was to determine whether annual colonoscopy among patients with IBD-PSC is cost-effective compared to less frequent intervals from a publicly funded health care system perspective. A cost-utility analysis using a Markov model was used to simulate a 35 year-old patient with a 10-year history of well-controlled IBD and a recent diagnosis of concomitant PSC. The following strategies were compared: no surveillance, colonoscopy every 5 years, biennial colonoscopy and annual colonoscopy. Outcome measures included: costs, number of cases of dysplasia found, number of cancers found and missed, deaths, quality-adjusted life-years (QALYs) gained and the incremental cost per QALY gained. In the base-case analysis, no surveillance was the least expensive and least effective strategy. Compared to no surveillance, the cost per QALY of surveillance every 5 years was CAN $15,021. The cost per QALY of biennial surveillance compared to surveillance every 5 years was CAN $37,522. Annual surveillance was more effective than biennial surveillance, but at an incremental cost of CAN $174,650 per QALY gained compared to biennial surveillance. More frequent colonoscopy screening intervals improves effectiveness (i.e. detects more cancers and prevents additional deaths), but at higher cost. Health systems must consider the opportunity costs associated with different surveillance colonoscopy intervals when deciding which strategy to implement among patients with IBD-PSC. 125 5.2 Introduction Colorectal cancer (CRC) is the third most common type of cancer and accounts for 12% of cancer related deaths among the general population.1 Compared to the general population, patients with inflammatory bowel disease (IBD) are at increased risk of developing CRC. Risk factors associated with development of CRC among IBD patients include longer duration of disease, greater disease extent, familial history of CRC, chronic active inflammation and being concurrently diagnosed with primary sclerosing cholangitis (PSC).2 Primary sclerosing cholangitis is a chronic cholestatic liver disease, characterized by chronic inflammation, and destruction and fibrosis of the intrahepatic and/or extrahepatic biliary tree.3 Approximately 68% of PSC patients will also have a diagnosis of IBD.4 A recent population-based study concluded that the risk of developing CRC among IBD-PSC patients is 10 times higher and their diagnosis of cancer/dysplasia occurs at a younger age when compared to the general IBD population.5 In addition, cancers diagnosed among IBD-PSC patients are more often located in the right colon with an overall worse prognosis.6 Contemporary guidelines for the management of IBD patients recommend surveillance colonoscopy every 1-2 years, 8-10 years after diagnosis.7,8 In contrast, annual colonoscopy is recommended for IBD-PSC patients immediately after both diseases are concurrently recognized.7,8 However, the implications of annual surveillance among IBD-PSC patients on the number of cancer cases detected, CRC-related mortality as well as the cost-effectiveness of this strategy is unknown. Thus, the objective of this 126 study was to perform a cost-utility analysis to determine whether annual colonoscopy surveillance is cost-effective compared to less frequent surveillance intervals in IBD patients with new concurrent PSC. 5.3 Materials and Methods 5.3.1 Model overview An incremental cost-utility analysis was performed from the perspective of the publicly funded health care system according to contemporary guidelines.9 Our hypothetical population consisted of young to middle-aged patients (35 years on average) with a 10-year history of well-controlled IBD, recently diagnosed with PSC. Both IBD and PSC are lifelong diseases with liver transplant being the only curative option for advanced PSC. Furthermore, the risk of CRC persists following liver transplantation among patients with an intact colon.10 Therefore, we considered the following CRC screening strategies: no screening, annual colonoscopy, biennial colonoscopy, and colonoscopy performed every 5 years over a lifetime horizon. The Markov model was constructed using decision analysis software (TreeAge Pro 2014, USA). A concise version depicting the flow of the model is presented in figure 5.1. All patients had compensated liver disease at the start of the model. As the model progressed, patients were at risk of developing decompensated disease requiring liver transplantation. During each year of the model (one-year cycle length), patients with IBD-PSC with or without dysplasia or CRC could either remain in the same health state, progress to another health state or die. Cancers were modeled to arise through both a dysplasia-carcinoma sequence and through a faster pathway without any prior identifiable 127 dysplastic lesion. Dysplasia or CRC could be diagnosed during surveillance colonoscopy, as part of the liver transplant workup when they developed decompensated liver disease or when CRCs became clinically manifest in those missed by surveillance colonoscopy. Patients who were noted during surveillance to have dysplasia or CRC were treated with colectomy. Consistent with liver transplant guidelines, a history of any extra-hepatic malignancy within the previous 5 years was considered a contraindication to transplantation. Please refer to Appendix 5.A for a list of the health states included in our model. 5.3.2 Model assumptions The model structure assumed the following: 1) All patients had well-controlled colitis in order to focus on the impact of surveillance rather than diagnosing and managing flares. 2) We assumed 100% compliance in the base case across screening strategies.11,12 3) Consistent with the literature, those diagnosed with CRC during a surveillance colonoscopy had improved survival compared to patients presenting with a missed cancer during colonoscopy or in the context of no program. This differential survival was on the basis of a more favourable CRC stage distribution (i.e., more early stage cancers) at diagnosis during surveillance with lower treatments costs.13-18 4) Based on expert opinion, and consistent with other decision analyses, we assumed that cancers missed during colonoscopy (or due to no colonoscopy performed in the natural history arm) became clinically manifest and were diagnosed within a median time of two years.19 5) Consistent with clinical practice, some patients who underwent colectomy were subsequently noted to not have dysplasia (i.e. false-positive test results). 6) Postcolectomy CRC patients alive after 5 years were considered to be cured of their cancer. 7) 128 All patients that had surgery because of dysplasia or CRC underwent a total colectomy because of the high risk of pouchitis among IBD-PSC patients.20 8) We assumed a lower risk of recurrent PSC and graft loss among those with a pre-/peri-liver transplant colectomy compared to those with an intact colon after liver transplant.21 9) We assumed that there was a one-year delay on average between the indication of liver transplant and receipt of the transplant. Based on expert opinion, only one-third of patients requiring a liver transplant survived and successfully had their procedure. 10) The mortality risk among IBD-PSC patients is 3 times higher than the general IBD population,22 the latter of which is equivalent to the general population without IBD.23 These assumptions were applied across all strategies in the model and were varied in sensitivity analyses. 5.3.3 Model validation A previously validated and published model was adapted for this study.16 First, clinical specialists in IBD and PSC carefully reviewed the structure and flow of the decision tree to ensure the face validity of the model. Following this, a series of “debugging” exercises and Markov cohort analyses were performed to identify and eliminate all syntactical errors. Lastly, a Monte Carlo simulation was performed to calculate the median time to liver transplant or death, the number of subjects who developed dysplasia and CRC, and the number of colectomy and CRC-related deaths. These numbers were used to calibrate the model against a recent IBD-PSC populationbased study to ensure that the number of cancers developed in the model closely approximated those reported in the study.5 129 5.3.4 Parameters MEDLINE was used to retrieve all relevant studies published on CRC, IBD and PSC and additional articles were identified from originally retrieved studies. Expert opinion was used in circumstances were data was not available. A list of the base case model inputs and ranges considered is presented in Table 1. 5.3.4.1 Dysplasia and cancer related risk The cumulative 10-year risks of dysplasia and CRC from the general IBD population were used to calculate the baseline prevalence of dysplasia and CRC at the time of concurrent IBD-PSC diagnosis (year 0 of the model).24,25 Thereafter, the incidence of colonic neoplasia was derived by pooling data from previous IBD-PSC studies that enabled us to calculate the incidence rate upon coexistence of both IBD and PSC.11,26-28 A weighted average of the annual probability of developing neoplasia and the proportion of dysplasia and CRC cases was then calculated. The rate of transition from dysplasia to CRC is uncertain among IBD-PSC patients. As such, the transition rate from dysplasia to CRC was estimated from follow-up studies in the general IBD population.29 5.3.4.2 Median time to liver transplant and graft loss In the base case the median time to liver transplant was considered to be 13 years.5 The annual risk of graft loss was calculated directly from studies describing the rate of graft loss among IBD-PSC patients.30 5.3.4.3 Mortality The annual risk of death varied according to the clinical state. Life tables and published literature on mortality of IBD-PSC patients were used to obtain the annual background mortality risk.31 The risk of death due to colectomy was derived from a 130 general IBD population-based study.32,33 The 5-year CRC mortality according to cancer stage was obtained from the literature and a weighted average was calculated based on cancer stage distribution among IBD patients under a CRC surveillance program versus no surveillance program.15,17,18 Consistent with the literature, those diagnosed with CRC under a surveillance program had improved survival over patients presenting with symptomatic cancer, on the basis of a more favourable CRC stage distribution (i.e., more early stage cancers) at diagnosis.15,17,18 Similar to other economic analyses, mortality from colonoscopy was not considered, as this represents a negligible risk.19 Patient survival following liver transplant was derived from the IBD-PSC literature.21 5.3.4.4 Colonoscopy performance characteristics. Colonoscopy is the most widely used modality for diagnosing dysplasia and cancer. Colonoscopy test performance characteristics were based on 33 biopsy samples as recommended by CRC surveillance guidelines.7,8,34-36 5.3.4.5 Costs All costs are reported in 2012 CAN$. Direct costs of colonoscopy and colectomy were derived from published estimates within the Calgary Health Zone.11,16,37,38 Colonoscopy costs included: physician and non-physician costs (i.e. capital, nursing, drugs) as well as biopsy costs and pathology fees.16 Nonmedical costs (patient +/caregiver time and travel costs) were also included according to guidelines and were obtained from the literature.9,16 The costs of managing CRC included surgery and chemotherapy treatment costs and were obtained from published studies.16 The average cancer stage costs were used to 131 calculate weighted average costs for overall cancer management based on cancer stage distribution.16-18 The total cost of liver transplantation was taken from a report that estimated the medical costs during the 30-day period prior to the transplant, graft procurement, hospital admission for the transplant including physician costs, post-discharge facility and nonfacility services, and the mean costs resulting from hospital readmissions and other evaluations and treatments during the first year post-transplant.39 All future costs and health benefits were discounted at 5% annually.40 5.3.4.6 Utilities Utilities for most of the health states were obtained from studies among the general IBD population because utilities for patients with IBD-PSC are unknown.41,42 When available, studies reporting utilities using a standard gamble exercise were used. Weighted averages for cancer utilities were calculated based on cancer stage distribution among IBD patients under a CRC surveillance program versus patients under no surveillance program.17,18,43 Utilities post liver transplant were obtained from studies looking at post liver transplant health-related quality of life among all liver transplant patients.44,45 5.3.5 Analyses Incremental analyses (expressed as the cost per QALY gained) were performed by rank ordering all competing strategies by increasing cost after eliminating strategies that were more costly and less effective (i.e. dominated). Outcome measures included: costs, number of cases of dysplasia found, number of cancers found and missed, deaths, quality-adjusted life-years (QALYs) gained and the incremental cost per QALY gained. 132 Multiple one-way sensitivity analyses were performed in order to assess the extent to which the model results were affected by each parameter independently. We performed a number of scenario analyses in order to evaluate the impact of adherence to screening, median time to liver transplant, and adherence with biopsy protocol guidelines during colonoscopy. We performed 3 scenario analyses taking into consideration a surveillance program adherence rate of 25%, 50% and 75%, recognizing that compliance rates to CRC screening programs can vary. Taking into consideration a potential selection bias from median time to liver transplant reports, a scenario analysis evaluated the potential impact of extending the median time to liver transplant to 21 years based on a recent population based study.5. Finally, a scenario analysis was performed to assess the impact of taking only 18 biopsy samples during surveillance colonoscopies. For this we lowered the sensitivity for dysplasia and CRC to 60% and 85%, respectively. A probabilistic sensitivity analysis (PSA) was performed to take into account the overall model uncertainty by simultaneously varying all the model parameter estimates within their range of uncertainty. The PSA incorporated 100,000 trials (iterations), and statistical distributions were created around all parameters. The type of distribution and values used are presented in Table 5.1. When possible, base case values were taken from the median or mean of published data and the 95% confidence intervals were used as the lower and upper values. Beta distributions were used for probabilities and utilities. Gamma or normal distributions were used for costs. Triangular distributions were used in cases where expert opinion was used. The sensitivity and specificity of colonoscopy was not considered for the PSA because these values are correlated and we did not have access to the receiver operator curves. PSA results are presented using incremental cost133 effectiveness scatterplots with 95% confidence ellipses. For this, each surveillance strategy was compared to the next most expensive strategy, and the incremental cost and effectiveness for each of the iterations was plotted in the cost-effectiveness plane. Also, the probability that each intervention was cost-effective at multiple willingness-to-pay (WTP) thresholds was calculated and presented on a cost-effectiveness acceptability curve graph. 5.4 Results 5.4.1 Base-case analysis For every 1,000 IBD-PSC patients followed over a lifetime, 255 cases of CRC and 58 CRC related deaths would be expected with no surveillance at a cost of $101,663 per patient (Tables 5.2 & 5.3). When compared to no surveillance, the number of CRC deaths could be reduced by 50% with surveillance every 5 years, 69% with biennial colonoscopy and 81% with annual colonoscopy (Table 5.2). Under no surveillance, 255 dysplasia cases would be expected and 178 of them would have progressed to cancer (Table 5.2). In contrast, with surveillance every 5 years, biennial and annual colonoscopy surveillance the number of dysplasia cases that progressed to cancer could be reduced by 64%, 78% and 95%, respectively (Table 5.2). Considering a lifetime horizon, the cost per patient was estimated to be $101,663, $104,517, $107,894 and $114,880 for no surveillance, surveillance every 5 years, biennial surveillance and annual surveillance respectively (Table 5.3). As expected, no surveillance was the least costly, but also the least effective strategy. Colonoscopy every 5 years was the least costly screening strategy, but also generated the fewest QALYs at a 134 cost of $15,021 per QALY gained compared to no screening (Table 5.3). Compared to surveillance every 5 years, biennial surveillance was more costly and generated more QALYs at a cost of $37,522 per QALY gained (Table 5.3). Annual colonoscopy produced additional QALYs, but at a cost of $174,650 per QALY gained compared to biennial colonoscopy (Table 5.3). 5.4.2 Sensitivity analyses All the one-way sensitivity analysis yielded similar rakings as the base case. No surveillance remained the least expensive and effective strategy followed by surveillance every 5 years, biennial and annual surveillance. As expected, as the annual incidence of neoplasia increases, the cost per QALY of all the surveillance strategies decreases. For example, when the annual incidence of neoplasia was 3.3% (the upper range of our estimate) the cost per QALY of biennial and annual surveillance decreased to $28,499 and $125,985, respectively. However, under all circumstances, the ICER of annual surveillance remained well above $100,000/QALY. Adherence to surveillance influenced the cost per QALY of all surveillance strategies. When we assumed a 25% adherence to the surveillance program, biennial surveillance was eliminated due to extended dominance and annual surveillance was the most effective strategy at a cost per QALY of $19,950 (Table 5.4). Assuming a 50% adherence to the surveillance program, the cost per QALY for the biennial surveillance program was similar to the base case but the cost per QALY for the annual surveillance program decreased to $80,616 (Table 5.4). Compared to the base case, the cost per QALY for biennial surveillance decreased to $21,836 assuming a 75% adherence rate. 135 However, with 75% adherence, the cost per QALY of annual surveillance remained similar to the base case results ($180,300). The cost per QALY for biennial surveillance was similar to the base case when time to liver transplant was extended to 21 years ($34,025 per QALY compared to colonoscopy every 5 years). However, the cost per QALY of annual surveillance increased significantly to $403,300 compared to biennial surveillance (Table 5.4). Lowering colonoscopy performance by decreasing the number of tissue biopsies obtained led to an increase of the cost per QALY of biennial surveillance ($59,871) and to a decrease of the cost per QALY of annual surveillance ($101,114) (Table 5.4). 5.4.3 Probabilistic sensitivity analysis Figure 5.2 presents the incremental cost-effectiveness scatterplot comparing annual surveillance to biennial surveillance. When we compared annual surveillance to biennial surveillance, 94% of the iteration results were located in the top right quadrant indicating that annual surveillance is more effective, but also more costly than biennial surveillance. However, in 98% of the iterations, annual surveillance was not recommended because the incremental cost-effectiveness ratio was above the WTP threshold of $50,000 per QALY. When we compared biennial surveillance to surveillance every 5 years, 98% of the iteration results were located on the top right quadrant meaning that biennial surveillance is more costly and more effective than surveillance every 5 years. In 57% of the iterations, the incremental cost-effectiveness ratio for biennial surveillance was above the generally accepted WTP threshold of $50,000 per QALY. 136 The cost-effectiveness acceptability curve shows that at a WTP threshold of $50,000, biennial surveillance was likely to be cost-effective ~40% of the time while annual surveillance was likely to be cost-effective only ~2% of the time. Surveillance every 5 years is the most likely strategy to be cost-effective with a probability of ~55% (Figure 5.3). Only at a WTP of $175,000, annual surveillance is likely to be cost-effective 50% of the time. 5.5 Discussion This is the first cost-utility analysis to evaluate different CRC surveillance intervals among IBD-PSC patients. Current CRC screening guidelines recommend annual surveillance colonoscopy for IBD-PSC patients based only on existing observational data that highlights the elevated CRC risk among these patients.7,8 Our study demonstrates that, for a typical IBD-PSC patient with stable colitis and non-decompensated liver disease, biennial surveillance was associated with a cost per QALY of $37,522 compared to surveillance every 5 years. Biennial colonoscopy would reduce the number of cancer cases and cancer-related deaths by 61% and 69%, respectively, compared to no surveillance. Annual surveillance could prevent an additional 19% of cancer cases and 29% cancer-related deaths compared to biennial screening, but at an additional average cost of $6,986 per patient or $174,650 per QALY. This information can be used by clinicians and healthcare administrators to plan surveillance programs; particularly, in recognition of the evolving incidence and risk of surgery for IBD in the 21st Century.46,47 Cost-effectiveness studies are often presented using a generally accepted WTP threshold of $50,000/QALY (in the USA) or between £20,000-£30,000 (in the UK). 137 These WTP thresholds have been used to facilitate decision-making for the costeffectiveness of medical technologies and health interventions despite of not being endorsed by the Panel on Cost-Effectiveness in Health and Medicine or NICE (UK).48,49 Some authors suggest that the use of these thresholds may have limited generalizability for severe/chronic diseases and propose a WTP of $100,000/QALY as a less conservative approach reflecting inflation and increased health care costs.50 It should be noted that health care interventions are still being implemented despite being well above a WTP of $50,000/QALY; for example, the use of biologics in IBD patients.51,52 In our study, biennial surveillance was a reasonable strategy using both WTP thresholds of $50,000/QALY and $100,000/QALY. This model takes into consideration the progression of liver disease among this patient population. The development of end stage liver disease and subsequent high mortality limits the benefits of CRC prevention through surveillance. In addition, our model suggests that approximately 30% of dysplasia cases missed during surveillance colonoscopy are diagnosed during the liver transplant workup. Consequently, we observed little difference between the effectiveness of biennial and annual surveillance under the base case and under most of the scenario analyses with the exception of 25% compliance rate and suboptimal surveillance biopsy protocol practices. This study has several important strengths. First, we used a recently published population-based study that described the natural history of CRC among IBD-PSC patients to validate our model.5 This external validation process increases the confidence that our model resembles real life natural history with respect to the number of CRC cases developed and the median time to liver transplant or death. Second, our model 138 represents the average IBD-PSC patient with a median time to liver transplant of 13 years. However, we performed a scenario analysis extending time to liver transplant to 21 years. This provides further insight on the potential impact of surveillance among the IBD-PSC population outside of large referral or liver transplantation centers.5 Fourth, this study highlights the importance of standardizing the follow-up period and reporting of dysplasia and CRC incidence. Finally, our study has identified important knowledge gaps in the field that should be addressed to improve patient care. However, several limitations to this study should be considered. First, this study was limited by the data available for IBD-PSC patients. Some parameters of our model are not well characterized (e.g. incidence of neoplasia/dysplasia/CRC, distinguishing between indefinite, low or high grade dysplasia) or not described at all (e.g. probability of progression from dysplasia to CRC and utilities) in the IBD-PSC population. Heterogeneity was observed between studies that report dysplasia and cancer incidence rates among IBD-PSC patients. Some studies report incidence rates of neoplasia while others report incidence rates that distinguish between dysplasia and cancer. Furthermore, the start of the follow-up period varies among studies. Some studies use age at diagnosis of IBD as the beginning of the follow up period, while others use the age at diagnosis of PSC or age at coexistence of both diseases. Also, lack of adherence to not only the recommended screening intervals, but also the recommended 33 tissue biopsies per colonoscopy would likely lead to underestimating the incidence rate of dysplasia and/or CRC.11 Further, the diagnosis of indefinite dysplasia, low or high grade dysplasia will lead to different clinical outcomes. However, the data differentiating practice between grades of dysplasia is limited and thus, we only modeled dysplasia as a single entity. In 139 addition, the probability of progression from dysplasia to cancer was extrapolated from IBD studies due to the small sample size and short duration of patient follow-up in IBDPSC studies.53 Health related utilities were extrapolated from the IBD population. Thus, extrapolating some of the probabilities, utilities and costs from the general IBD population most likely underestimated the benefit of the surveillance strategies being evaluated. However, we presented a series of sensitivity and scenario analyses demonstrating the impact of the uncertainty around these parameters. Second, this costutility model was based on the traditional random biopsy driven colonoscopy surveillance practice that most community gastroenterologists currently apply. Novel techniques that may improve the sensitivity of identifying colonic dysplasia are available in certain academic centers including mucosal dye spray during colonoscopy (i.e. chromoendoscopy) and confocal endomicroscopy.2 While these techniques may improve the detection rates of dysplasia and early CRC, they are more time consuming, require additional training, and the costs to integrate these modalities into an endoscopy unit are higher. Third, our model assumed that the IBD disease status was stable. We did not factor in additional colonoscopies or the added IBD related costs (e.g. medications and surgery) during disease flares. Thus, CRC diagnosed during a flare would not have been accounted for in our model.11 However, this bias is minimized because surveillance colonoscopies are predominantly completed when IBD is in remission. Further, the impact of additional colonoscopies would be the same across strategies and thus would not have affected our conclusions. Finally, our model, did not account for the rare occurrence of neoplasia in patients with an ileal pouch-anal anastomosis.54,55 140 Based on the best available data, this cost-utility model demonstrated that more frequent surveillance improves effectiveness (detects additional cancers earlier and improves survival) at a higher cost. Healthcare systems might choose different intervals depending on what is affordable. Recognizing that the model was influenced by uncertainty in key parameters, future studies that address these knowledge gaps would aid in providing more robust estimates of cost-effectiveness for decision makers. 141 5.6 References 1. Canadian Cancer Society's Advisory Committee on Cancer Statistics. Canadian Cancer Statistics 2012. Toronto, ON: Canadian Cancer Society; 2012. 2. Collins PD, Mpofu C, Watson AJ, et al. Strategies for detecting colon cancer and/or dysplasia in patiens with inflammatory bowel disease. Cochrane Database Syst Rev 2006:CD000279. 3. Kaplan GG, Laupland KB, Butzner D, et al. The burden of large and small duct primary sclerosing cholangitis in adults and children: a population-based analysis. Am J Gastroenterol 2007;102:1042-9. 4. Molodecky NA, Kareemi H, Parab R, et al. Incidence of primary sclerosing cholangitis: a systematic review and meta-analysis. Hepatology 2011;53:1590-9. 5. Boonstra K, Weersma RK, van Erpecum KJ, et al. Population-based epidemiology, malignancy risk and outcome of primary sclerosing cholangitis. Hepatology 2013;58:2045-55. 6. Claessen MM, Lutgens MW, van Buuren HR, et al. More right-sided IBDassociated colorectal cancer in patients with primary sclerosing cholangitis. Inflamm Bowel Dis 2009;15:1331-6. 7. Kornbluth A, Sachar DB. Ulcerative colitis practice guidelines in adults: American College Of Gastroenterology, Practice Parameters Committee. Am J Gastroenterol 2010;105:501-23. 8. Farraye FA, Odze RD, Eaden J, et al. AGA medical position statement on the diagnosis and management of colorectal neoplasia in inflammatory bowel disease. Gastroenterology 2010;138:738-45. 142 9. Guidelines for the economic evaluation of health technologies: Canada [3rd edition ed] Ottawa, Canadian Agency for Drugs and Technologies in Health; 2006. 10. Singh S, Varayil JE, Loftus EV Jr., et al. Incidence of colorectal cancer after liver transplantation for primary sclerosing cholangitis: a systematic review and metaanalysis. Liver Transpl 2013;19:1361-9. 11. Kaplan GG, Heitman SJ, Hilsden RJ, et al. Population-based analysis of practices and costs of surveillance for colonic dysplasia in patients with primary sclerosing cholangitis and colitis. Inflamm Bowel Dis 2007;13:1401-7. 12. Kottachchi D, Yung D, Marshall JK. Adherence to guidelines for surveillance colonoscopy in patients with ulcerative colitis at a Canadian quaternary care hospital. Can J Gastroenterol 2009;23:613-6. 13. McArdle CS, Hole DJ. Emergency presentation of colorectal cancer is associated with poor 5-year survival. Br J Surg 2004;91:605-9. 14. Amri R, Bordeianou LG, Sylla P, et al. Impact of screening colonoscopy on outcomes in colon cancer surgery. JAMA Surg 2013:1-7. 15. O'Connell JB, Maggard MA, Ko CY. Colon cancer survival rates with the new American Joint Committee on Cancer sixth edition staging. J Nat Cancer Inst 2004;96:1420-5. 16. Heitman SJ, Hilsden RJ, Au F, et al. Colorectal cancer screening for average-risk North Americans: an economic evaluation. PLoS One 2010;7:e1000370. 143 17. Choi PM, Nugent FW, Schoetz DJ, et al. Colonoscopic surveillance reduces mortality from colorectal cancer in ulcerative colitis. Gastroenterology 1993;105:418-24. 18. Lutgens MW, Oldenburg B, Siersema PD, et al. Colonoscopic surveillance improves survival after colorectal cancer diagnosis in inflammatory bowel disease. Br J Cancer 2009;101:1671-5. 19. Nguyen GC, Frick KD, Dassopoulos T. Medical decision analysis for the management of unifocal, flat, low-grade dysplasia in ulcerative colitis. Gastrointest Endosc 2009;69:1299-310. 20. Gorgun E, Remzi FH, Manilich E, et al. Surgical outcome in patients with primary sclerosing cholangitis undergoing ileal pouch-anal anastomosis: a casecontrol study. Surgery 2005;138:631-9. 21. Alabraba E, Nightingale P, Gunson B, et al. A re-evaluation of the risk factors for the recurrence of primary sclerosing cholangitis in liver allografts. Liver Transpl 2009;15:330-40. 22. Ananthakrishnan AN, Cagan A, Gainer VS, et al. Mortality and extraintestinal cancers in patients with primary sclerosing cholangitis and inflammatory bowel disease. J Crohns Colitis 2014; http://dx.doi.org/10.1016/j.crohns.2014.01.019. 23. Jess T, Loftus EV Jr., Harmsen WS, et al. Survival and cause specific mortality in patients with inflammatory bowel disease: a long term outcome study in Olmsted County, Minnesota, 1940-2004. Gut 2006;55:1248-54. 24. Delvin J, O'Grady J. Indications for referral and assessment in adult liver transplantation: a clinical guideline. Gut 1999;45:Supp6:VII-VI22. 144 25. Carrithers RL. Liver transplantation: AASLD practice guidelines. Liver Transpl 2000;6:122-35. 26. Weinstein MC, O'Brien B, Hornberger J, et al. Principles of good practice for decision analytic modeling in health-care evaluation: report of the ISPOR Task Force on good research practices-modeling studies. Value Health 2003;6:9-17. 27. Eaden JA, Abrams KR, Mayberry JF. The risk of colorectal cancer in ulcerative colitis: a meta-analysis. Gut 2001;48:526-535. 28. Thomas T, Abrams KA, Robinson RJ, et al. Meta-analysis: cancer risk of lowgrade dysplasia in chronic ulcerative colitis. Aliment Pharmacol Ther 2007;25:657-68. 29. Gurbuz AK, Giardiello FM, Bayless TM. Colorectal neoplasia in patients with ulcerative colitis and primary sclerosing cholangitis. Dis Colon Rectum 1995;38:37-41. 30. Navaneethan U, Kochhar G, Venkatesh PG, et al. Duration and severity of primary sclerosing cholangitis is not associated with risk of neoplastic changes in the colon in patients with ulcerative colitis. Gastrointest Endosc 2012;75:104554. 31. Broome U, Lofberg R, Veress B, et al. Primary sclerosing cholangitis and ulcerative colitis: Evidence for increased neoplastic potential. Hepatology 1995;22:1401-08. 32. Ullman T, Croog V, Harpaz N, et al. Progression of flat low-grade dysplasia to advanced neoplasia in patients with ulcerative colitis. Gastroenterology 2003;125:1311-29. 145 33. Rowe IA, Webb K, Gunson BK, et al. The impact of disease recurrence on graft survival following liver transplantation: a single centre experience. Transpl Int 2008;21:459-65. 34. Complete life table, Canada, 2000-2002: males. Ottawa, Canada: Statistics Canada: 2006. 35. de Silva S, Ma C, Proulx MC, et al. Postoperative complications and mortality following colectomy for ulcerative colitis. Clin Gastroenterol Hepatol 2011;9:972-80. 36. Kaplan GG, McCarthy EP, Ayanian JZ, et al. Impact of hospital volume on postoperative morbidity and mortality following a colectomy for ulcerative colitis. Gastroenterology 2008;134:680-7. 37. Rubenstein JH, Waljee AK, Jeter JM, et al. Cost-effectiveness of ulcerative colitis surveillance in the setting of 5-aminosalicylates. Am J Gastroenterol 2009;104:2222-32. 38. Pickhard PJ, Hassan C, Halligan S, et al. Colorectal cancer: CT colonography and colonoscopy for detection: Systematic review and meta-analysis. Radiology 2011;259:393-405. 39. Rubin CE, Haggitt RC, Burmer GC, et al. DNA aneuploidy in colonic biopsies predicts future development of dysplasia in ulcerative colitis. Gastroenterology 1992;103:1611-20. 40. Heitman SJ, Au F, Manns BJ, et al. Nonmedical costs of colorectal cancer screening with the fecal occult blood test and colonoscopy. Clin Gastroenterol Hepatol 2008;6:912-17. 146 41. Coward S, Heitman S, Hubbard J, et al. Ulcerative colitis associated hospitalization costs: A population-based study, In: Canadian Digestive Diseases Week, 2012; Victoria, British Columbia. 42. Milliman research report: 2008 U.S. organ and tissue transplant cost estimates and discussion.Milliman Inc., Brookfield, WI, USA. 2008 43. Poole CD, Connolly MP, Nielsen SK, et al. A comparison of physician-rated disease severity and patient reported outcomes in mild to moderately active ulcerative colitis. J Crohn's Colitis 2010;4:275-82. 44. Muir AJ, Edwards LJ, Sanders LL, et al. A prospective evaluation of healthrelated quality of life after ileal pouch anal anastomosis for ulcerative colitis. Am J Gastroenterol 2001;96:1480-85. 45. Ness RM, Holmes AM, Klein R, et al. Utility valuations for outcome states fo colorectal cancer. Am J Gastroenterol 1999;94:1650-1657 46. Longworth L, Bryan S. An empirical comparison of EQ-5D and SF-6D in liver transplant patients. Health Econ 2003;12:1061-7. 47. Younossi ZM, Boparai N, McCormick M, et al. Assessment of utilities and health-related quality of life in patients with chronic liver disease. Am J Gastroenterol 2001;96:579-83. 48. Molodecky NA, Soon IS, Rabi DM, et al. Increasing incidence and prevalence of the inflammatory bowel diseases with time, based on systematic review. Gastroenterology 2012;142:46-54. 147 49. Frolkis AD, Dykeman J, Negron ME, et al. Risk of surgery for inflammatory bowel diseases has decreased over time: a systematic review and meta-analysis of population-based studies. Gastroenterology 2013;145:996-1006. 50. Rawlins MD, Culyer AJ. National Institute for Clinical Excellence and its value judgements. BMJ 2004;329:224-7. 51. Weinstein MC, Siegel JE, Gold MR, et al. Recommendations of the Panel on Cost-Effectiveness in Health and Medicine. JAMA 1996;276:1253-8. 52. Lee JK, Tang DH, Mollon L, et al. Cost-effectiveness of biological agents used in ulcerative colitis. Best Pract Res Clin Gastroenterol 2013;27:949-60. 53. Park KT, Bass D. Inflammatory bowel disease-attributable costs and costeffective strategies in the United States: a review. Inflamm Bowel Dis 2011;17:1603-9. 54. Dretzke J, Edlin R, Round J, et al. A systematic review and economic evaluation fo the use of tumour necrosis factor-alpha (TNF-α) inhibitors, adalimumab and infliximab, for Crohn's disease. Health Technol Assess 2011;15(6) doi: 10.3310/hta15060. 55. Eaton JE, Smyrk TC, Imam M, et al. The fate of indefinite and low-grade dysplasia in ulcerative colitis and primary sclerosing cholangitis colitis before and after liver transplantation. Aliment Pharmacol Ther 2013;38:977-987 56. Borjersson L, Willen R, Haboubi N, et al. The risk of dysplasia and cancer in the ileal pouch mucosa after restorative proctocolectomy for ulcerative proctocolitis is low: A long-term follow-up study. Colorectal Dis 2004;6:494-8. 148 57. Thompson–Fawcett MW, Marcus V, Redston M, et al. Risk of dysplasia in longterm ileal pouches and pouches with chronic pouchitis. Gastroenterology 2001;121:275-81. 149 Table 5.1. Base-case parameter estimates and distributions and ranges used in the probabilistic sensitivity analysis Variables Colonoscopy Sensitivity for detecting dysplasia Sensitivity for detecting cancer Specificity Baseline prevalence of dysplasia Baseline prevalence of cancer Incidence of neoplasia Proportion of dysplasia cases Transition from dysplasia to cancer Probability of cancer becoming symptomatic Probability of developing decompensated liver disease Probability of graft loss (pre-/peritransplant colectomy) Probability of graft loss (intact colon) Baseline mortality Mortality associated with cancer Missed cancer‡ Early diagnosed cancer§ Mortality from colectomy Mortality of end stage liver disease° Mortality during first year of liver transplant Mortality after first year of liver transplant Base-case analysis Distribution Range Reference 0.90 0.95 0.999 0.021 0.002 0.023/yr 0.75 0.14/yr 0.6/yr N/A N/A N/A Beta Beta Triangular Triangular Beta Triangular (0.58-1.0) (0.81-1.0) (0.992-1.0) (0.015-0.03) (0.001-0.002) (0.013-0.033) (0.5-1.0) (0.06-0.32) (0.2-1.0) 37,39 0.0385/yr Beta (0.024-0.0385) 5 0 Beta (0-0.15) 21 0.18/yr Beta (0.15-0.22) (0.0011-0.00654) 21 0.09/yr 0.05/yr 0.015 0.70 0.20 Beta Beta Beta Triangular Beta (0.02-0.4) (0.02-0.4) (0.005-0.023) 15,17,18 (0.159-0.2) 33 0.03/yr Beta (0.02-0.03) 33 150 38 38 28 27 11,29-31 11,29-31 32 EO* 22,34 15, 17,18 35,36 EO* Costs (CAD)† Cost of colonoscopy Cost colonoscopy (medical) Cost colonoscopy (non-medical) Biopsy cost (per biopsy) Pathologist fee Cost of surgery Cost of cancer Missed cancer‡ Early diagnosed cancer§ Cost of liver transplant Cost of pre-liver transplant workup Utilities Baseline Following surgery no cancer Dysplasia Cancer Missed cancer‡ Early diagnosed cancer§ First year of liver transplant After first year of liver transplant Discount rate Time horizon Mean (Range) 11,16,40 913 308 9 186 25,000 Gamma Gamma Gamma Gamma Gamma 913 (532 – 1,278) 328 (212 -479) 9 (7 - 18) 186 (175 - 372) 25,000 (15,000 - 50,000) 73,933 51,617 585,774 25,781 Gamma Gamma Normal Normal 73,933 (26,688 – 142,784) 51,617 (26,688 – 142,784) 585,774 (+/- SD 73,222) 25,781 (+/- SD 3,222) 0.94 0.93 0.74 Beta Beta Beta Range (0.89-1.0) (0.76-1.0) (0.69-0.78) 0.58 0.67 0.62 0.84 5% lifetime Beta Beta Beta Beta (0.25-0.74) (0.25-0.74) (0.57-0.65) (0.70-0.90) 41 ,EO* 16-18,41 42 42 43 44 45 17,18,45 47 46 *EO- expert opinion ‡ Cancers missed during colonoscopy or because no colonoscopy was performed (in the case of the natural history arm, surveillance every 5 years and biennial surveillance) § Cancers diagnosed at colonoscopy within a year ° Patients who developed decompensated disease and died prior undergoing the liver transplant † Cost were rounded to the nearest CAN$ 151 Table 5.2. The number of patients who developed dysplasia and cancer, as well as the number of colonoscopies performed over a lifetime horizon for a hypothetical 1,000 average IBD-PSC patient cohort No surveillance Dysplasia cases overall Diagnosed During surveillance During LT* workup Progressed to cancer Other† 255 Surveillance every 5 years 255 Biennial surveillance 253 Annual surveillance 252 0 56 178 21 158 28 63 6 212 17 22 2 236 13 3 0 Cancer cases overall Diagnosed During surveillance During liver transplant Symptomatic Other‡ 256 141 101 82 0 26 223 7 48 12 79 2 66 27 7 1 74 5 3 0 Cancer related deaths Number of surveillance colonoscopies 58 0 29 2,349 18 6,289 11 12,738 * Liver transplant † Dysplasia cases that died due to background mortality or decompensated liver disease by the end of the model remained undiagnosed ‡ Cancer cases that died due to background mortality or decompensated liver disease by the end of the model remained undiagnosed 152 Table 5.3. Base case-analysis results (in CAN$) for cost-utility analysis Strategy Average cost* (95% CI) † Average effectiveness (95% CI) † Incremental cost* Incremental Incremental cost/effectiveness effectiveness* ratio* No surveillance 101,663 (100,477102,849) 9.84 (9.81-9.88) Surveillance every 5 years 104,517 (103,336105,687) 2,854 10.03 (10.00-10.06) 0.19 15,021 Biennial surveillance‡ 107,894 (106,713109,074) 3,377 10.12 (10.09-10.15) 0.09 37,522 Annual surveillance 114,880 (113,698116,061) 6,986 10.16 (10.13-10.19) 0.04 174,650 * Compared to the next most costly strategy † 95% confidence interval 153 Table 5.4. Scenario-analysis results (in CAN$) for cost-utility analysis Strategy cost (CAN$) Effectiveness Incremental cost/effectiveness ratio* Compliance 25% No surveillance Surveillance every 5 years Biennial surveillance Annual surveillance 101,123 102,327 103,276 104,322 9.83 9.94 9.98 10.04 10,945 extended dominance† 19,950 Compliance 50% No surveillance Surveillance every 5 years Biennial surveillance Annual surveillance 99,733 103,074 105,048 109,885 9.83 9.98 10.03 10.09 22,273 39,480 80,616 Compliance 75% No surveillance Surveillance every 5 years Biennial surveillance Annual surveillance 101,278 104,287 106,689 112,098 9.83 9.99 10.10 10.13 18,806 21,836 180,300 Time to liver transplant 21 years No Surveillance Surveillance every 5 years Biennial surveillance Annual surveillance 78,750 80,521 84,604 92,670 11.28 11.48 11.60 11.62 8,855 34,025 403,300 Suboptimal surveillance‡ No surveillance Surveillance every 5 years Biennial surveillance Annual surveillance 101,714 104,342 108,533 115,611 9.84 10.00 10.07 10.14 16,425 59,871 101,114 *Compared to the next most costly strategy † The dominated strategy is both less effective and less costly compared to a mix of two other strategies ‡ Suboptimal surveillance was defined as taking only 18 biopsy samples during surveillance colonoscopy. The sensitivity of dysplasia and cancer was assumed to be 60 and 85% respectively 154 Figure 5.1. Bubble diagram summarizing the structure of the Markov model, flow of patients through the health states, and the outcomes considered in this cost-utility analysis. Colonoscopy screening strategy Target population No surveillance Annual surveillance Biennial surveillance Surveillance every 5 years 35 year-old male diagnosed with inflammatory bowel disease and primary sclerosing cholangitis No dysplasia/Cancer Dysplasia Outcomes Colectomy Death 155 Invasive cancer Figure 5.2. Incremental cost-effectiveness scatterplots with 95% confidence ellipses for annual surveillance compared to biennial surveillance. The diagonal line represents a willingness to pay threshold of CAN$50,000 per quality-adjusted lifeyear. 156 Figure 5.3. Cost-effectiveness acceptability curve. 157 5.7 Appendix 5.A Summary of health states included in the model Health states Prior to developing decompensated liver disease/undergoing liver transplant 1-No dysplasia 2- Missed dysplasia 3- Missed CRC 4- Post-colectomy dysplasia (true-positive/false-positive dysplasia) 5- Post-colectomy CRC (diagnosed during surveillance) 6- Post-colectomy CRC (missed during surveillance) After first liver transplant 1- Post-liver transplant, intact colon/no dysplasia 2- Post-liver transplant, intact colon/missed dysplasia 3- Post-liver transplant, intact colon/ missed CRC 4- Post-liver transplant, pre/peri-transplant colectomy 5- Post-liver transplant, no colon (previous colectomy for CRC) 6- Post-colectomy CRC (missed during surveillance) Other 1- Death 158 Chapter Six: Discussion of work presented 159 6.1 Overview of main findings The primary focus of this thesis was to understand the epidemiology of Clostridium difficile infection and colorectal dysplasia/cancer in order to identify ways of decreasing the risk of colectomy among UC patients. Chapters 2 and 3 established the effect of C. difficile infections on short- and long-term risk of colectomy. In particular, the cumulative risk of acquiring C. difficile infection after the diagnosis of UC was established. Patients with UC who were diagnosed with C. difficile were shown to have an increased risk of colectomy. Further, both Chapters 2 and 3 demonstrated that C. difficile infections increased the risk of postoperative complications following colectomy for UC. Chapter 4 demonstrated that the incidence of colectomy for colorectal dysplasia and cancer has remained stable over time, and showed that the observed decrease in colectomy rates is most likely due to a decrease in colectomies due to medically refractory disease. Finally, Chapter 5 determined the cost-effectiveness of different surveillance colonoscopy intervals for colorectal dysplasia among a subgroup of inflammatory bowel disease (IBD) patients with a higher risk of developing colorectal dysplasia and cancer; i.e., patients with IBD and primary sclerosing cholangitis (PSC). The aims of the study presented in Chapter 2 were to determine if C. difficile infection was associated with having an emergent colectomy and developing postoperative complications in a population-based cohort of UC patients who were admitted to Calgary Health Zone (CHZ) hospitals between 2000 and 2009. The medical charts of these patients were reviewed to extract clinical data and Calgary Laboratory Services (CLS) provided patients’ C. difficile tests results done in hospital and 90 days 160 prior to hospital admission. This study demonstrated that UC patients diagnosed with C. difficile infection in hospital or 90 days prior to admission were more likely to be unresponsive to rescue therapy in hospital and need an emergent colectomy (adjusted odds ratio (OR)=3.39; 95% confidence interval (CI): 1.02-11.23). Additionally, a preoperative diagnosis of C. difficile infection increased the odds of developing an infectious postoperative complication (adjusted OR=4.76; 95% CI: 1.10-20.63). However, a preoperative diagnosis of C. difficile infection was not significantly associated with developing any postoperative complication (adjusted OR=3.16; 95% CI: 0.89-11.23). The study presented in Chapter 3 was a population-based inception cohort of patients from Alberta who were diagnosed with UC between 2003 and 2010. This cohort was identified using Alberta Health administrative databases. The objectives of this study were to 1) validate the International Classifications of Diseases (ICD-10) code for C. difficile diagnosis among UC patients; 2) determine the 1-, 3-, and 5-year cumulative risk of a C. difficile diagnosis after UC diagnosis; 3) determine the effect of a C. difficile diagnosis on the risk of colectomy and mortality; and 4) determine the association between a C. difficile diagnosis and developing postoperative complications. The sensitivity, specificity, PPV and NPV of the ICD-10 diagnostic code of C. difficile diagnosis were 82.1% (95% CI: 71.7-89.9%), 99.4% (95%CI: 99.1-99.7%), 88.4% (95% CI: 82.9-92.3%), and 99.1% (95% CI: 98.5-99.4%), respectively. The cumulative risk of C. difficile diagnosis at 1-, 3-, and 5-years after UC diagnosis was 1.6%, 2.5%, and 3.2%, respectively. Patients with a C. difficile diagnosis were more likely to have colectomy compared to those who had no prior C. difficile diagnosis (sub-hazard ratio (SHR)= 2.36; 161 (95% CI: 1.47-3.80) after adjusting for age, sex, and comorbidities. Additionally, patients with a C. difficile diagnosis had an overall increased risk of death compared to those with no prior C. difficile diagnosis (hazard ratio (HR)= 2.56; 95% CI: 1.28-5.10). Finally, a C. difficile diagnosis was associated with the development of post-operative complications (adjusted OR=4.84; 95% CI: 1.28-18.35). Chapter 4 used two different methodological approaches to determine whether the incidence of colectomy for colorectal dysplasia or cancer has changed over time. This study demonstrates that the incidence of colectomy for colorectal dysplasia and cancer remained stable in the last decade. Thus, the observed reduced colectomy risk previously reported by Kaplan et al. is most likely driven by a decrease in the incidence of colectomy due to medically refractory disease and not colorectal dysplasia or cancer.7 Finally, the objective of Chapter 5 was to determine the cost-effectiveness of the current recommendations for surveillance of colorectal dysplasia and cancer among patients with IBD and PSC. A cost-utility analysis using Markov modelling was used to simulate a cohort of IBD-PSC patients in order to compare the following strategies: no surveillance, colonoscopy every 5 years, biennial surveillance and annual surveillance (currently recommended). Annual surveillance was more effective than biennial surveillance, but at an incremental cost of CAN $174,650 per QALY gained compared to biennial surveillance. This means that more frequent colonoscopies improve effectiveness (i.e. detect more dysplasia and cancer cases and prevents additional deaths), but at a higher cost. Understanding that the willingness to pay will vary among different health systems, this study made no recommendations on which strategy to choose rather, we reported the costs and outcomes associated with each strategy. This information can be 162 used by clinicians and healthcare administrators to plan surveillance programs; particularly, in recognition of the evolving incidence and risk of surgery for IBD in the 21st Century. 6.2 Limitations Epidemiological research often encounters situations where the data used for research is not perfect. Thus, it is important to acknowledge and discuss how these limitations impact the study results and the generalizability of the findings. This section expands on the limitations discussed on each chapter, and presents additional methodological challenges. Because of the similarities in methods and limitations, Chapters 2-4 are discussed together, and Chapter 5 is discussed separately. 6.2.1 Limitations of observational studies The studies presented in Chapters 2-4 used secondary data (medical charts, CLS for C. difficile test results, and administrative health databases). Secondary data are defined as “data generated for different purposes that may overlap with the objective of the study”.108 Some of the benefits of using secondary data are that they are readily available and inexpensive. Since the information was not collected for research purposes, secondary data have inherent limitations that researchers need to take into consideration. For example, some of the limitations of the secondary data used in the studies presented in this dissertation include accuracy of the data and completeness of the information.109 Another challenge for Chapters 2 to 4 was the potential for misclassification of the disease (i.e. UC), the outcomes (i.e. colectomy and post-operative complications) and the exposures (i.e. C. difficile infection). For example, in Chapter 3 the misclassification 163 of postoperative complications (i.e. outcome) in the Discharge Abstract Database (DAD) most likely contributed to the differences in the results between Chapter 2 and Chapter 3. In Chapter 2 a preoperative diagnosis of C. difficile increased the risk of postoperative infectious complications. In contrast, in Chapter 3 a preoperative diagnosis of C. difficile increased the risk of having at least one postoperative complication, but this effect was not specific to infectious complications. In Chapter 2 postoperative complications were confirmed by chart review, whereas in Chapter 3 postoperative complications were derived from ICD codes for complications. Thus, in Chapter 3 the definition of postoperative complication was subject to a misclassification bias. Ma et al. validated the accuracy of postoperative complications following colectomy for UC.10 The sensitivity and specificity of “any complications” (i.e. at least one complication) was more accurate as compared to a group of complications (e.g. infectious complications).10 Further, the accuracy of ICD coding for postoperative complications was correlated to the severity of the complication.10 As a result, in Chapter 3, patients with less severe postoperative complications were more often classified as controls when compared to those with severe postoperative complications. This likely resulted in a non-differential misclassification error that biased the risk estimates for postoperative infectious complications towards the null. Another example is the misclassification of the UC population in Chapter 3 and Chapter 4 with the use of administrative databases. A validated case-definition to identify UC patients (incident cases in Chapter 3 and prevalent cases in Chapter 4) in Alberta’s administrative databases was used to identify the patient population.110 This algorithm has a sensitivity of 83.4%, a specificity of 99.8%, a positive predictive value of 97.4% 164 and negative predictive value of 98.5%.110 The case-definition was shown to optimize the balance between the specificity (few false-positives) and sensitivity (few false-negatives). Still, a sensitivity of 83% means that ~17% of the IBD population in Alberta was not captured. Misclassification is also introduced when stratifying IBD patients by UC versus Crohn’s disease. A validated scoring system was used to differentiate UC from CD, which had a sensitivity of 86.3% and a specificity of 99.7%.110 Patients who were not classifiable to UC or CD by this scoring system were excluded. Finally, Chapter 3 required the development of an inception cohort, which has the potential of mixing prevalent with incident cases. A previous validation study demonstrated that an 8-year washout period is required to exclude prevalent cases; however, this long washout period would have resulted in the exclusion of some true incident cases. Completeness of the data is another limitation of secondary data sources. For example, in Chapter 2, one third of hospitalized UC patients were not tested for C. difficile infection. Surveillance of C. difficile infection among flaring UC patients has improved overtime and thus, patients admitted to hospital in earlier years were less likely to be tested. Another explanation for the low test rates is that diagnosing C. difficile infection when an UC patients is flaring can be challenging due to the similarity of the clinical symptoms of the two disorders. Among UC patients, the presence of microscopic pseudomembranes, characteristic of C. difficile infection among the non-IBD population, cannot be used to diagnose C. difficile due to low sensitivity and specificity.111 Additionally, a C. difficile diagnosis is further complicated by asymptomatic carriage, which has been detected in 9% of UC patients.112 Because discriminating C. difficile infection from UC flare is difficult, in addition to high numbers of asymptomatic 165 carriage, some UC patients with milder disease severity may have not been tested.113 Not testing for C. difficile may have introduced a differential misclassification bias where patients with a severe form of colitis are more likely to be tested for C. difficile infection, which may overestimate the association between C. difficile infection and colectomy. Additionally, the type of data that was collected was limited because of the retrospective nature of the chart- reviews and the limited number of clinically relevant variables in administrative health databases. As a result, there is information missing that could be clinically relevant to these studies. For example, indicators of the disease severity (e.g. C-reactive protein) and disease activity (e.g. Mayo Score) would have been useful to better discern between a mild flare and severe colitis. In addition, medication history did not reliably capture duration of use or dosage in the case of the chapters using the medical chart-review data. Further, administrative databases did not provide any information on patient medication. This is of importance in both Chapter 2 and 3 because there is no clear consensus on treatment course for C. difficile infection in the setting of UC and it has been demonstrated that the choice of treatment influences C. difficile infection outcomes among UC patients.114, 115 In the case of Chapter 4, it was impossible to determine for how long these patients were on those medications, which could potentially influence the results. Finally, the comparison groups used in the studies described in both Chapters 2 and 4 limited the interpretation and generalizability of the study findings. In the case of Chapter 2, the study was limited to patients admitted to hospital due to a flare. Thus, we cannot extrapolate the study results to those UC patients with a mild course of UC who have never been in hospital. In Chapter 4, we identified patients who had a colectomy for 166 medically refractory disease as our comparison group. This patient group represents a much more severe diseased population, potentially biasing some of the results. Like in the case of the observed “protective effect” of 5-ASA and azathioprine/6-mercaptopurine. The fact that our control group consisted of UC patients that were sicker than the general UC population hinders any conclusion that the observed association was due to a true protective effect on these medications. 6.2.2 Limitations of decision analysis study A decision-analytic model was used in Chapter 5 in order to compare different colonoscopy surveillance strategies in terms of outcomes and costs to determine the costeffectiveness of each strategy. In this model, mathematical relationships were used to estimate a series of outcomes that would result from the strategies being evaluated.116 Thus, modelling was the best option in this situation given that the current evidence needed for this analysis was incomplete, integrating information from primary and secondary sources like clinical trials, meta-analyses, observational studies and/or expert opinion. This decision tree was constructed based on what has been observed in the population to represent a simplified version of real life situations. A key component of this simplification process was deciding which events that are part of the disease process in the real population should be included or excluded and understanding the impact these assumptions had on the results. An example of an event excluded from the decision tree was the occurrence of colonoscopies performed due to flares. In real life, some cancers are diagnosed by chance at the time of a colonoscopy that is performed because of a flare. Also, one of the main problems encountered during the design phase in this study was the uncertainty around the model parameters. The lack of data available and the 167 heterogeneity between the studies hindered an attempt to conduct a meta-analysis. Furthermore, most of the studies used to populate this model were of short duration and assumptions were made in order to fit the lifetime horizon of this model. In addition, assumptions were made based on expert opinion in circumstances were no information was available. For example, the proportion of patients that survive the period between being diagnosed with end-stage liver disease and having a liver available for transplant. However, we performed a thorough external validation process and multiple scenario analyses were performed to ensure the model outputs mimicked real population outcomes reported in the literature in order to minimize biases introduced by assumptions and less than ideal estimates. In addition, the impact on the model’s results of the uncertainty around the parameters was assessed by probabilistic sensitivity analysis. 6.3 Implications on Clinical Management and Public Health Epidemiology is crucial in medicine as it attempts to describe the variation in the outcomes of illness and identify risk factors that could explain the outcomes observed.117 This is done in order to identify potential ways to modify risk factors that could result in a reduction of the burden of disease. The results presented in this thesis will have an impact on prevention of adverse outcomes (i.e. colectomy, post-operative complications, mortality), and may influence guidelines on patient management and shape future research development. The research presented in this dissertation demonstrates that C. difficile infection increases the risk of colectomy, post-operative complications and mortality. Thus, reducing the risk of developing a C. difficile infection, by identifying UC patients at high 168 risk of acquiring C. difficile infection and managing them aggressively may result in a reduction in risk of colectomy and postoperative complications and reducing overall mortality. Additionally, C. difficile-positive patients who had a colectomy should be monitored closely for early detection of post-operative complications. Also, this work highlights the importance of testing patients who were admitted to hospital for a flare for C. difficile, so that the appropriate treatment course can be given to these patients. The findings of Chapter 5 will guide health care systems on choosing between colorectal cancer surveillance strategies; based on what they are willing to pay. Adherence to surveillance guidelines is a physician and patient issue.118 Thus, the results of this study can be used to inform and educate both physicians and patients on the benefits associated with adherence to surveillance recommendations. Finally, this study paved the road for future directions of colorectal cancer research among IBD-PSC patients. 6.4 Future Research These studies answer knowledge gaps in the literature and also served generate additional hypotheses. Further, the acquisition of the Alberta Health administrative databases along with the validation of the case definitions provide numerous opportunities for studies in the areas of C. difficile infection and colorectal cancer in UC. For example, the Alberta health databases could be used to compare and contrast clusters in space and time of C. difficile cases in IBD and non-IBD patients. These findings could lead to studies that compare C. difficile strains affecting IBD and non-IBD patients. Additionally, the Alberta Health databases could be used to determine the annual 169 incidence, mortality and geographic location of colorectal cancers among IBD patients and compared it to that of the non-IBD population. We demonstrated that a C. difficile diagnosis increases the risk of colectomy. Clinical trials are needed to identify which treatment course (e.g. antibiotics and/or UC flare treatment or fecal microbiota transplant) improves outcomes among these patients. These findings could further help identify which of these treatments is the most costeffective way of managing these patients. Another area that needs further investigation is the development of patient related factors (clinical profile, genetic susceptibility, serology, etc.) that may predict the acquisition of C. difficile infection among UC patients. Finally, we mentioned that C. difficile diagnosis is complicated because it mimics the symptoms of a flare and because of asymptomatic carriage.112 Similarly asymptomatic carriage is observed in children, and thus, future studies should evaluate pediatric-onset UC. The presence of elevated fecal inflammatory cytokines has been used to differentiate children with asymptomatic colonization from those with the disease.119 Thus, future studies should identify inflammatory biomarkers that distinguish C. difficile colonization from disease among UC patients. Based on Chapter 4 results, the incidence of colorectal dysplasia and cancer has not changed in the last decade. The next step would be to investigate factors that may influence the risk of colectomy for dysplasia or cancer; for example, lack of adherence to surveillance practices or the use of chemoprotective IBD medications. While a recent meta-analysis concluded that 5-ASA does not confer a protective effect against the development of colorectal cancer, recent experimental colitis models and case-control studies suggest that biologics may decrease the risk of colorectal cancer.104-106 170 Prospective studies to evaluate the effect of biologics on cancer risk should be conducted on newly diagnosed patients to allow for enough lag time between the introduction of these medications and the risk of colorectal cancer. In Chapter 5 we identified several knowledge gaps in the area of colorectal dysplasia and cancer among IBD-PSC patients. For example, more accurate information is needed for the cumulative incidence of colorectal dysplasia and cancer among IBDPSC patients. Also, IBD-PSC patients’ utilities are missing from the literature. Thus, a study should be conducted to determine patients’ utilities for colorectal cancer, liver transplant, and colectomy. Also, newer surveillance techniques (e.g. chromoendoscopy or confocal microscopy) have better detection rates of colorectal dysplasia than colonoscopy with random biopsies. 98 These techniques were not incorporated into the model because they are not readily used in practice. However, the cost-utility model in Chapter 5 may be expanded to incorporate these surveillance techniques when wider utilization occurs in the community. 6.5 Conclusions Collectively, the 4 studies in this dissertation evaluate the impact of C. difficile infection and colorectal neoplasia on colectomy for UC. In Chapters 2 and 3 we demonstrated that C. difficile impacts the short- and long-term risk of colectomy. Further, Chapters 2 and 3 are the first studies to demonstrate that C. difficile infection increases the risk of postoperative complications among UC patients. Chapter 3 is the first study to establish the 1-, 3- and 5-year risk of developing a C. difficile infection after UC diagnosis. Chapter 4 demonstrated that the colectomy incidence for colorectal dysplasia 171 and cancer remained stable between 1997-2009. Thus, the previously reported decrease in colectomy rates is most likely due to a decrease in colectomies for medically refractory UC. Finally, Chapter 5 was the first cost-utility analysis to establish the cost-effectiveness of the current guidelines for colorectal neoplasia surveillance among IBD-PSC patients. While these studies answered many knowledge gaps in the literature, it also served as the foundation for future research. So in the future, we will have a better understanding of these modifiable risk factors in order to reduce the burden of UC. 1! ! 172 References 1. Molodecky NA, Soon IS, Rabi DM, et al. Increasing incidence and prevalence of the inflammatory bowel diseases with time, based on systematic review. Gastroenterology 2012;142:46-54. 2. Crohn’s & Colitis Foundation of Canada. The impact of inflammatory bowel disease in Canada: 2012 final report and recommendations, 2012. 3. Jess T, Gamborg M, Munkholm P, et al. Overall and cause-specific mortality in ulcerative colitis: meta-analysis of population-based inception cohort studies. Am J Gastroenterol 2007;102:609-17. 4. Bewtra M, Kilambi V, Fairchild AO, et al. Patient preferences for surgical versus medical therapy for ulcerative colitis. Inflamm Bowel Dis 2014;20:103-14. 5. Ghosh S, Mitchell R. Impact of inflammatory bowel disease on quality of life: Results of the European Federation of Crohn's and Ulcerative Colitis Associations (EFCCA) patient survey. J Crohns Colitis 2007;1:10-20. 6. Frolkis AD, Dykeman J, Negron ME, et al. Risk of surgery for inflammatory bowel diseases has decreased over time: a systematic review and meta-analysis of population-based studies. Gastroenterology 2013;145:996-1006. 7. Kaplan GG, Seow CH, Ghosh S, et al. Decreasing colectomy rates for ulcerative colitis: a population-based time trend study. Am J Gastroenterol 2012;107:187987. 8. de Silva S, Ma C, Proulx MC, et al. Postoperative complications and mortality following colectomy for ulcerative colitis. Clin Gastroenterol Hepatol 2011;11:972-80. 173 9. Frolkis AD, Kaplan GG, Patel A, et al. Short-term emergent readmission and postoperative complications in children and adults with inflammatory bowel disease who undergo intestinal resection – A population-based study. Accepted: Inflamm Bowel Dis 2014. 10. Ma C, Crespin M, Proulx MC, et al. Postoperative complications following colectomy for ulcerative colitis: A validation study. BMC Gastroenterol 2012;12:39. 11. Nguyen GC, Kaplan GG, Harris ML, et al. A national survey of the prevalence and impact of Clostridium difficile infection among hospitalized inflammatory bowel disease patients. Am J Gastroenterol 2008;103:1443-50. 12. Rodemann JF, Dubberke ER, Reske KA, et al. Incidence of Clostridium difficile infection in inflammatory bowel disease. Clin Gastroenterol Hepatol 2007;5:33944. 13. Issa M, Vijayapal A, Graham MB, et al. Impact of Clostridium difficile on inflammatory bowel disease. Clin Gastroenterol Hepatol 2007;5:345-51. 14. Bossuyt P, Verhaegen J, Van Assche G, et al. Increasing incidence of Clostridium difficile-associated diarrhea in inflammatory bowel disease. J Crohn's Colitis 2009;3:4-7. 15. Goodhand JR, Alazawi W, Rampton DS. Systematic review: Clostridium difficile and inflammatory bowel disease. Aliment Pharmacol Ther 2011;33:428-41. 16. Jen MH, Saxena S, Bottle A, et al. Increased health burden associated with Clostridium difficile diarrhoea in patients with inflammatory bowel disease. Aliment Pharmacol Ther 2011;33:1322-31. 174 17. Jodorkovsky D, Young Y, Abreu MT. Clinical outcomes of patients with ulcerative colitis and co-existing Clostridium difficile infection. Dig Dis Sci 2010;55:415-20. 18. Murthy SK, Steinhart AH, Tinmouth J, et al. Impact of Clostridium difficile colitis on 5-year health outcomes in patients with ulcerative colitis. Aliment Pharmacol Ther 2012;36:1032-9. 19. Ananthakrishnan AN, McGinley EL, Binion DG. Excess hospitalisation burden associated with Clostridium difficile in patients with inflammatory bowel disease. Gut 2008;57:205-10. 20. Ricciardi R, Ogilvie JW, Roberts PL, et al. Epidemiology of Clostridium difficile colitis in hospitalized patients with inflammatory bowel diseases. Dis Colon Rectum 2009;52:40-5. 21. Khan NA, Quan H, Bugar JM, et al. Association of postoperative complications with hospital costs and length of stay in a tertiary care center. J Gen Intern Med 2006;21:177-80. 22. Rubin DT, Huo D, Kinnucan JA, et al. Inflammation is an independent risk factor for colonic neoplasia in patients with ulcerative colitis: a case-control study. Clin Gastroenterol Hepatol 2013;11:1601-8. 23. Rutter M, Saunders B, Wilkinson K, et al. Severity of inflammation is a risk factor for colorectal neoplasia in 2004;126:451-9. 175 ulcerative colitis. Gastroenterology 24. Nieminen U, Jussila A, Nordling S, et al. Inflammation and disease duration have a cumulative effect on the risk of dysplasia and carcinoma in IBD: a case-control observational study based on registry data. Int J Cancer 2014;134:189-96. 25. Jess T, Rungoe C, Peyrin-Biroulet L. Risk of colorectal cancer in patients with ulcerative colitis: a meta-analysis of population-based cohort studies. Clin Gastroenterol Hepatol 2012;10:639-45. 26. Boonstra K, Weersma RK, van Erpecum KJ, et al. Population-based epidemiology, malignancy risk and outcome of primary sclerosing cholangitis. Hepatology 2013;58:2045-55. 27. Kornbluth A, Sachar DB. Ulcerative colitis practice guidelines in adults: American College Of Gastroenterology, Practice Parameters Committee. Am J Gastroenterol 2010;105:501-23. 28. Farraye FA, Odze RD, Eaden J, et al. AGA medical position statement on the diagnosis and management of colorectal neoplasia in inflammatory bowel disease. Gastroenterology 2010;138:738-45. 29. Targan SR, Shanahan F, Karp LC. Inflammatory Bowel Disease: Translating Basic Science into Clinical Practice. West Sussex, UK: Blackwell Publishing, 2010. 30. Odze R. Diagnostic problems and advances in inflammatory bowel disease. Mod Pathol 2003;16:347-58. 31. Cohen RD, Yu AP, Wu EQ, et al. Systematic review: the costs of ulcerative colitis in Western countries. Aliment Pharmacol Ther 2010;31:693-707. 176 32. Orholm M, Munkholm P, Lanhholz E, et al. Familial occurrence of inflammatory bowel disease. N Engl J Med 1991;324:84-8. 33. Jostins L, Ripke S, Weersma RK, et al. Host-microbe interactions have shaped the genetic architecture of inflammatory bowel disease. Nature 2012;491:119-24. 34. Thompson AI, Lees CW. Genetics of ulcerative colitis. Inflamm Bowel Dis 2011;17:831-48. 35. Gent AE, Hellier MD, Grace RH, et al. Inflammatory bowel disease and domestic hygiene in infancy. Lancet 1994;343:766-7. 36. Stallmach A, Carstens O. Role of infections in the manifestation or reactivation of inflammatory bowel diseases. Inflamm Bowel Dis 2002;8:213-18. 37. Klement E, Cohen RV, Boxman J, et al. Breastfeeding and risk of inflammatory bowel disease: a systematic review with meta-analysis. Am J Clin Nutr 2004;80:1342-52. 38. Kaplan GG, Hubbard J, Korzenik J, et al. The inflammatory bowel diseases and ambient air pollution: a novel association. Am J Gastroenterol 2010;105:2412-9. 39. Hou JK, Abraham B, El-Serag H. Dietary intake and risk of developing inflammatory bowel disease: a systematic review of the literature. Am J Gastroenterol 2011;106:563-73. 40. Salim SY, Kaplan GG, Madsen KL. Air pollution effects on the gut microbiota: A link between exposure and inflammatory bowel disease. Gut Microbes 2014;5:215-19. 41. Ng SC, Bernstein CN, Vatn MH, et al. Geographical variability and environmental risk factors in inflammatory bowel disease. Gut 2013;62:630-49. 177 42. Langholz E, Munkholm P, Davidsen M, et al. Course of ulcerative colitis: analysis of changes in disease activity over years. Gastroenterology 1994;107:311. 43. Vavricka SR, Brun L, Ballabeni P, et al. Frequency and risk factors for extraintestinal manifestations in the Swiss inflammatory bowel disease cohort. Am J Gastroenterol 2011;106:110-9. 44. Kaplan GG, Laupland KB, Butzner D, et al. The burden of large and small duct primary sclerosing cholangitis in adults and children: a population-based analysis. Am J Gastroenterol 2007;102:1042-9. 45. Molodecky NA, Kareemi H, Parab R, et al. Incidence of primary sclerosing cholangitis: a systematic review and meta-analysis. Hepatology 2011;53:1590-9. 46. Ananthakrishnan AN, Cagan A, Gainer VS, et al. Mortality and extraintestinal cancers in patients with primary sclerosing cholangitis and inflammatory bowel disease. J Crohns Colitis 2014; doi: 10.1016/j.crohns.2014.01.019. [Epub ahead of print]. 47. Broome U, Olsson R, Loof L, et al. Natural history and prognostic factors in 305 Swedish patients with primary sclerosing cholangitis. Gut 1996;38:610-615. 48. Silverberg MS, Satsangi J, Ahmad T, et al. Toward an integrated clinical, molecular and serological classification of inflammatory bowel disease: report of a Working Party of the 2005 Montreal World Congress of Gastroenterology. Can J Gastroenterol 2005;19 Suppl A:5a-36a. 49. Quezada SM, Cross RK. Association of age at diagnosis and ulcerative colitis phenotype. Dig Dis Sci 2012;57:2402-7. 178 50. Regueiro M, Loftus EV, Steinhart AH, et al. Medical management of Left-sided ulcerative colitis and ulcerative proctitis: Critical evaluation of therapeutic trials. Inflamm Bowel Dis 2006;12:979-94. 51. Gionchetti P, Amadini C, Rizzello F, et al. Review article: treatment of mild to moderate ulcerative colitis and pouchitis. Aliment Pharmacol Ther 2002;16:13-9. 52. Feagan BG, MacDonald JK. Oral 5-aminosalicylic acid for induction of remission in ulcerative colitis. Cochrane Database Syst Rev 2012; CD000543. 53. Kane S, Huo D, Aikens J, et al. Medication nonadherence and the outcomes of patients with quiescent ulcerative colitis. Am J Med 2003;114:39-43. 54. Faubion WA, Loftus EV, Harmsen WS, et al. The natural history of corticosteroid therapy for inflammatory bowel disease: A population-based study. Gastroenterology 2001;121:255-60. 55. Lopez-Sanroman A, Bermejo F, Carrera E, et al. Efficacy and safety of thiopurinic immunomodulators (azathioprine and mercaptopurine) in steroiddependent ulcerative colitis. Aliment Pharmacol Ther 2004;20:161-6. 56. Stidham RW, Lee TC, Higgins PD, et al. Systematic review with network metaanalysis: the efficacy of anti-tumour necrosis factor-alpha agents for the treatment of ulcerative colitis. Aliment Pharmacol Ther 2014;39:660-71. 57. Costa J, Magro F, Caldeira D, et al. Infliximab reduces hospitalizations and surgery interventions in patients with inflammatory bowel disease: a systematic review and meta-analysis. Inflamm Bowel Dis 2013;19:2098-110. 179 58. Kaplan GG, McCarthy EP, Ayanian JZ, et al. Impact of hospital volume on postoperative morbidity and mortality following a colectomy for ulcerative colitis. Gastroenterology 2008;134:680-7. 59. Tulchinsky H, Averboukh F, Horowitz N, et al. Restorative proctocolectomy impairs fertility and pregnancy outcomes in women with ulcerative colitis. Colorectal Dis 2013;15:842-7. 60. Cornish JA, Tan E, Teare J, et al. The effect of restorative proctocolectomy on sexual function, urinary function, fertility, pregnancy and delivery: a systematic review. Dis Colon Rectum 2007;50:1128-38. 61. Solberg IC, Lyrgen I, Jahnsen J, et al. Clinical course during the first 10 years of ulcerative colitis: Results from a population-based inception cohort (IBSEN Study). Scand J Gastroenterol 2009;44:431-40. 62. Chandler RE, Hedberg K, Cieslak PR. Clostridium difficile associated disease in Oregon: increasing incidence and hospital-level risk factors. Infect Control Hosp Epidemiol 2007;28:116-22. 63. Hubert B, Loo VG, Bourgault AM, et al. A portrait of the geografic dissemination of the Clostridium difficile North American Pulsed-Field Type 1 strain and the epidemiology of C. difficile-associated disease in Quebec. Clin Infect Dis 2007;44:238-44. 64. Musa S, Thomson S, Cowan M, et al. Clostridium difficile infection and inflammatory bowel disease. Scand J Gastroenterol 2010;45:261-72. 65. Greenfield C, Aguilar-Ramirez JR, Pounder RE, et al. Clostridium difficle and inflammatory bowel disease. Gut 1983;24:713-7. 180 66. Rolny P, Jarnerot G, Mollby R. Ocurrence of Clostridium difficile toxin in inflammatory bowel disease. Scand J Gastroenterol 1983;18:61-4. 67. Keighley MRB, Young D, Johnson M, et al. Clostridium difficile toxin in acute diarrhoea complicating inflammatory bowel disease. Gut 1982;23:410-4. 68. Burke DA, Axon AT. Clostridium difficile, sulphasalazine, and ulcerative colitis. Postgrad Med J 1987;63:955-7. 69. Weber P, Koch M, Heizmann WR, et al. Microbic superinfection in relapse of inflammatory bowel disease. J Clin Gastroenterol 1991;14:301-8. 70. Ott C, Girlich C, Klebl F, et al. Low risk of Clostridium difficile infections in hospitalized patients with inflammatory bowel disease in a German tertiary referral center. Digestion 2011;84:187-92. 71. Masclee GM, Penders J, Jonkers DM, et al. Is Clostridium difficile associated with relapse of inflammatory bowel disease? results from a retrospective and prospective cohort study in the Netherlands. Inflamm Bowel Dis 2013;19:212531. 72. Bignardi GE. Risk factors for Clostridium difficile infection. J Hosp Infect 1998;40:1-15. 73. Kelly CP, LaMonte JT. Clostridium difficile infection. Ann Rev Med 1998;49:375-90. 74. Bien J, Palagani V, Bozko P. The intestinal microbiota dysbiosis and Clostridium difficile infection: is there a relationship with inflammatory bowel disease? Ther Adv Gastroenterol 2013;6:53-68. 181 75. Schneeweiss S, Korzenik J, Solomon DH, et al. Infliximab and other immunomodulating drugs in patients with inflammatory bowel disease and the risk of serious bacterial infections. Aliment Pharmacol Ther 2009;30:253-64. 76. Ananthakrishnan AN, Oxford EC, Nguyen DD, et al. Genetic risk factors for Clostridium difficile infection in ulcerative colitis. Aliment Pharmacol Ther 2013;38:522-30. 77. Powell N, Jung SE, Krishnan B. Clostridium difficile infection and inflammatory bowel disease: a marker for disease extent? Gut 2008;57:1183-4. 78. Ananthakrishnan AN. Clostridium difficile infection: epidemiology, risk factors and management. Nat Rev Gastroenterol Hepatol 2011;8:17-26. 79. Croabach MJT, Dekkersm OM, Wilcox MH, et al. European Society of Clinical Microbiology and Infectious Diseases (ESCMID): data review and recommendations for diagnosing Clostridium difficile-infection (CDI). Clin Microbiol Infect 2009;15:1053-66. 80. Guo B, Harstall C, Louie T, et al. Systematic review: faecal transplantation for the treatment of Clostridium difficile-associated disease. Aliment Pharmacol Ther 2012;35:865-75. 81. Cammarota G, Ianiro G, Gasvarrini A. Fecal Microbiota transplantation for the treatment of Clostridium difficile infection: A systematic review. J Clin Gastroenterol 2014; In Press. doi: 10.1097/MCG.0000000000000046 82. Surawicz CM, Brandt LJ, Binion DG, et al. Guidelines for diagnosis, treatment, and prevention of Clostridium difficile infections. Am J Gastroenterol 2013;108:478-98. 182 83. Ananthakrishnan AN, McGinley EL. Infection-related hospitalizations are associated with increased mortality in patients with inflammatory bowel diseases. J Crohns Colitis 2013;7:107-12. 84. Ullman T, Loftus EV, Kakar S, et al. The fate of low grade dysplasia in ulcerative colitis. Am J Gastroenterol 2002;97:922-7. 85. Itzkowitz SH, Yio X. Inflammation and cancer: IV. Colorectal cancer in inflammatory bowel disease: the role of inflammation. Am J Physiol Gastrointest Liver Physiol 2004;281:7-17. 86. Eaden JA, Abrams KR, Mayberry JF. The risk of colorectal cancer in ulcerative colitis: a meta-analysis. Gut 2001;48:526-35. 87. Thomas T, Abrams KA, Robinson RJ, et al. Meta-analysis: cancer risk of lowgrade dysplasia in chronic ulcerative colitis. Aliment Pharmacol Ther 2007;25:657-68. 88. Lutgens MW, van Oijen MG, van der Heijden GJ, et al. Declining risk of colorectal cancer in inflammatory bowel disease: an updated meta-analysis of population-based cohort studies. Inflamm Bowel Dis 2013;19:789-99. 89. Castano-Milla C, Chaparro M, Gisbert JP. Systematic review with meta-analysis: the declining risk of colorectal cancer in ulcerative colitis. Aliment Pharmacol Ther 2014;39:645-59. 90. Triantafillidis JK, Nasioulas G, Kosmidis PA. Colorectal cancer and inflammatory bowel disease: Epidemiology, risk factors, mechanisms of carcinogenesis and prevention strategies. Anticancer Res 2009;29:2727-38. 183 91. Velayos FS, Loftus EV, Jr., Jess T, et al. Predictive and protective factors associated with colorectal cancer in ulcerative colitis: A case-control study. Gastroenterology 2006;130:1941-9. 92. Grivennikov SI. Inflammation and colorectal cancer: colitis-associated neoplasia. Semin Immunopathol 2013;35:229-44. 93. Jess T, Loftus EV, Velayos FS, et al. Incidence and prognosis of colorectal dysplasia in inflammatory bowel disease: A population-based study from Olmstead county, Minnesota. Inflamm Bowel Dis 2006;12:669-76. 94. Soetikno RM, Lin OS, Heidenreich PA, et al. Increased risk of colorectal neoplasia in patients with primary sclerosing cholangitis and ulcerative colitis: A meta-analysis. Gastrointest Endosc 2002;56:48-54. 95. Claessen MM, Lutgens MW, van Buuren HR, et al. More right-sided IBDassociated colorectal cancer in patients with primary sclerosing cholangitis. Inflamm Bowel Dis 2009;15:1331-6. 96. Singh S, Varayil JE, Loftus EV, Jr., et al. Incidence of colorectal cancer after liver transplantation for primary sclerosing cholangitis: a systematic review and metaanalysis. Liver Transpl 2013;19:1361-9. 97. Riddell RH, Goldman H, Ransohoff DF, et al. Dysplasia in inflammatory bowel disease: Standardized classification with provisional clinical applications. Hum Pathol 1983;14:931-68. 98. Collins PD, Mpofu C, Watson AJ, et al. Strategies for detecting colon cancer and/or dysplasia in patiens with inflammatory bowel disease. Cochrane Database Syst Rev 2006; CD000279. 184 99. Wu L, Li P, Wu J, et al. The diagnostic accuracy of chromoendoscopy for dysplasia in ulcerative colitis: meta-analysis of six randomized controlled trials. Colorectal Dis 2012;14:416-20. 100. Dong YY, Li YQ, Yu YB, et al. Meta-analysis of confocal laser endomicroscopy for the detection of colorectal neoplasia. Colorectal Dis 2013;15:488-95. 101. Ullman T, Croog V, Harpaz N, et al. Progression of flat low-grade dysplasia to Advanced neoplasia in patients with ulcerative colitis. Gastroenterology 2003;125:1311-29. 102. Bernstein CN, Shanahan F, Weinstein WM. Are we telling patients the truth about surveillance colonosocpy in ulcerative colitis? Lancet 1994;343:71-4. 103. Siegel CA, Schwartz LM, Woloshin S, et al. When should ulcerative colitis patients undergo colectomy for dysplasia? Mismatch between patient preferences and physician recommendations. Inflamm Bowel Dis 2010;16:1658-62. 104. Nguyen GC, Gulamhusein A, Bernstein CN. 5-aminosalicylic acid is not protective against colorectal cancer in inflammatory bowel disease: a metaanalysis of non-referral populations. Am J Gastroenterol 2012;107:1298-304. 105. Baars JE, Looman CW, Steyerberg EW, et al. The risk of inflammatory bowel disease-related colorectal carcinoma is limited: results from a nationwide nested case-control study. Am J Gastroenterol 2011;106:319-28. 106. Onizawa M, Nagaishi T, Kanai T, et al. Signaling pathway via TNF-alpha/NFkappaB in intestinal epithelial cells may be directly involved in colitis-associated carcinogenesis. Am J Physiol Gastrointest Liver Physiol 2009;296:G850-9. 185 107. Hansen JD, Kumar S, Lo WK, et al. Ursodiol and colorectal cancer or dysplasia risk in primary sclerosing cholangitis and inflammatory bowel disease: a metaanalysis. Dig Dis Sci 2013;58:3079-87. 108. Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. Philadelphia, PA, USA: Lippincott Williams &Wilkins, 2008. 109. Molodecky NA, Kaplan G. Environmental risk factors for inflammatory bowel disease. Gastroenterol Hepatol 2010;6:339-46. 110. Rezaie A, Quan H, Fedorak R, et al. Development and validation of an administrative case definition for inflammatory bowel diseases. Can J Gastroenterol 2012;26:711-8. 111. Wang T, Matukas L, Streutker CJ. Histologic findings and clinical characteristics in acutely symptomatic ulcerative colitis patients with superimposed Clostridium difficile infection. Am J Clin Pathol 2013;140:831-7. 112. Clayton EM, Rea MC, Shanahan F, et al. The vexed relationship between Clostridium difficile and inflammatory bowel disease: an assessment of carriage in an outpatient setting among patients in remission. Am J Gastroenterol 2009;104:1162-9. 113. Burnham CA, Carroll KC. Diagnosis of Clostridium difficile infection: an ongoing conundrum for clinicians and for clinical laboratories. Clin Microbiol Rev 2013;26:604-30. 114. Yanai H, Nguyen GC, Yun L, et al. Practice of gastroenterologists in treating flaring inflammatory bowel disease patients with Clostridium difficile: Antibiotics 186 alone or combined antibiotics/immunomodulators? Inflamm Bowel Dis 2011;17:1540-6. 115. Ben-Horin S, Margalit M, Bossuyt P, et al. Combination immunomodulator and antibiotic treatment in patients with inflammatory bowel disease and Clostridium difficile infection. Clin Gastroenterol Hepatol 2009;7:981-7. 116. Briggs A, Claxton K, Sculpher M. Decision modelling for health economic evaluation: Oxford University Press, 2008. 117. Oleckno WA. Epidmeiology: Concepts and methods. Long Grove, IL, USA: Waveland Press, Inc., 2008. 118. Kaplan GG, Heitman SJ, Hilsden RJ, et al. Population-based analysis of practices and costs of surveillance for colonic dysplasia in patients with primary sclerosing cholangitis and colitis. Inflamm Bowel Dis 2007;13:1401-7. 119. El Feghaly RE, Stauber JL, Tarr PI, et al. Intestinal inflammatory biomarkers and outcome in pediatric Clostridium difficile infections. J Pediatr 2013;163:1697704. 187