Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Economic evaluations in the context of treatment recommendations in spondyloarthritis : analyses from… Harvard, Stephanie 2017

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

Item Metadata

Download

Media
24-ubc_2017_november_harvard_stephanie.pdf [ 4.92MB ]
Metadata
JSON: 24-1.0357139.json
JSON-LD: 24-1.0357139-ld.json
RDF/XML (Pretty): 24-1.0357139-rdf.xml
RDF/JSON: 24-1.0357139-rdf.json
Turtle: 24-1.0357139-turtle.txt
N-Triples: 24-1.0357139-rdf-ntriples.txt
Original Record: 24-1.0357139-source.json
Full Text
24-1.0357139-fulltext.txt
Citation
24-1.0357139.ris

Full Text

  ECONOMIC EVALUATIONS IN THE CONTEXT OF TREATMENT RECOMMENDATIONS IN SPONDYLOARTHRITIS:  ANALYSES FROM THE DESIR COHORT  by Stephanie Harvard  M.Sc., University of Manitoba, 2008 B.A.H., University of Winnipeg, 2005  A THESIS SUBMITTED IN PARTIAL FULLFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Population and Public Health)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  October 2017  © Stephanie Harvard, 2017    Université Pierre et Marie Curie University of British Columbia Ecole doctorale Pierre Louis de santé publique Groupe de recherche clinique 08 La Pitié Salpétrière rhumatologie Economic evaluations in the context of treatment recommendations in spondyloarthritis:  analyses from the DESIR cohort Par Stephanie Harvard Thèse de doctorat de santé publique Dirigée par Bruno Fautrel et Aslam Anis Présentée et soutenue publiquement le 4 octobre, 2017 Devant un jury composé de :  Aslam Anis, PhD, Co-directeur de thèse Bruno Fautrel, MD PhD, Co-directeur de thèse Nicholas Bansback, PhD, Membre du comité de thèse Cécile Gaujoux-Viala, MD PhD, Rapporteur Nigil Haroon, MD, PhD, Rapporteur Gilles Herjblum, PhD, Membre du jury Catherine Backman, MD PhD, Membre du jury   iii Abstract  Spondyloarthritis (SpA) is an inflammatory disease comprising both ankylosing spondylitis and non-radiographic axial SpA. This thesis conducted four studies using data from DESIR, a longitudinal cohort of 708 French SpA patients, focusing on economic questions in the context of SpA treatment recommendations. The objectives of study 1 were to value health resource use and productivity loss among DESIR patients and identify factors associated with costs. Cost valuation was done using public data and factors associated with costs were identified in multivariate regression models. This study showed that anti-TNF agents were the greatest cost driver in DESIR and generated the cost outcomes necessary to perform economic evaluations using DESIR data. The objective of study 2 was to collaborate with rheumatologists to develop measurable definitions of adherence to SpA treatment recommendations. A Delphi survey was conducted with 15 rheumatologists, who developed measurable definitions for 6/11 recommendations. The study uncovered differences of opinion between rheumatologists and generated the classification system necessary to explore adherence to recommendations among DESIR patients. The objective of study 3 was to examine the effect of adherence to anti-TNF use recommendations on outcomes in DESIR. Employing alternate definitions of adherence, patients were classed as adherent anti-TNF users, non-adherent anti-TNF users, adherent non-users, and non-adherent non-users. Following one potential definition, non-adherent anti-TNF users had significantly increased non-biologic costs compared to adherent users, while non-adherent non-users had significantly worse health outcomes than adherent anti-TNF users. This study showed that the impact of adherence to anti-TNF recommendations is sensitive to the definition of adherence and highlighted the need to validate methods to measure adherence. The objective of study 4 was to explore whether the French anti-TNF access restrictions are the most cost-effective in that setting relative to other potential restrictions. Five separate datasets were created comprising DESIR patients satisfying different sets of anti-TNF access criteria. Incremental cost-effectiveness ratios (ICERs) associated with anti-TNF use were calculated in each subset in basecase and sensitivity analyses. A sensitivity analysis simulating a 24-week stopping rule among anti-TNF non-responders demonstrated the effect of non-response on anti-TNF cost-effectiveness. The study underscored the need for evidence-informed anti-TNF access restrictions.   iv Lay Summary  SpA (short for 'spondyloarthritis') is a type of arthritis that affects the back, hips, and often other parts of the body. This disease is complex, and doctors and researchers are still learning about it, including what treatments to offer patients. Based on what they know, doctors have written 'SpA treatment recommendations' to help make sure all patients are treated equally and in the best way possible. But right now, it is hard to know the effect of SpA treatment recommendations, both on patients' health and on costs to the health system. In part, this is because treatment recommendations do not consider how much each treatment costs. This PhD thesis did four studies to help look at the effect of SpA treatment recommendations on patients' health and costs. Looking at the results, it then talks about how SpA treatment recommendations might change in the future to consider health and costs together.       v Preface  This statement is to certify that the work in this thesis was conceived, conducted and written by Stephanie Harvard (SH). The four research chapters of this dissertation (Chapters 2-5) were written as stand-alone manuscripts for publication in the peer-reviewed academic literature. Two of the chapters have already been published (Chapters 2,3), and one chapter has been accepted for publication (Chapter 4). As the primary author, SH led each of these chapters. The DESIR cohort was approved by the French Departmental Directorate of Health and Social Affairs (Directeur Départemental des Affaires Sanitaires et Sociales) Comittee for Protection of Persons (Comité de Protection des Personnes) (file reference number 2457). It was conducted in accord with the Declaration of Helsinki and the guidance for good clinical practice. All participants gave written informed consent to enter the study. Secondary data analysis in the costing study was reviewed and approved by the University of British Columbia Research Ethics Board (H13-01981). This dissertation is formatted in accordance with the regulations of X University and submitted in partial fulfillment of the requirements for a PhD degree awarded jointly by the Universitè Pierre et Marie Curie (Paris) and the University of British Columbia. Versions of this dissertation will exist in the institutional repositories of both institutions.  Chapter 1  SH conducted the literature review and wrote the first draft of the manuscript and all subsequent revisions. Drs. Anis, Fautrel, and Bansback edited the manuscript and suggested revisions.   Chapter 2 A version of Chapter 2 has been published: Harvard S, Guh D, Bansback N, Richette P, Dougados M, Anis A, Fautrel B. Costs of early spondyloarthritis: estimates from the first 3 years of the DESIR cohort. RMD Open. 2016 Apr 4;2(1). SH identified and selected the appropriate unit costs from publicly available data, performed the DESIR data cleaning, linkage to unit costs, and statistical analysis under the guidance of Daphne Guh (DG) and Huiying Su, and wrote the first draft of the manuscript and all subsequent revisions. Bruno Fautrel (BF) provided clinical expertise, collaborated with SH in designing the costing strategy, and helped guide the development of statistical models. Aslam Anis (AA) and Nick Bansback (NB) participated in   vi designing the costing strategy and helped guide the development of statistical models. Pascal Richette (PR), Maxime Dougados (MD) and all other authors (i.e., BF, AA, NB, DG) participated in the interpretation of results and suggested revisions to the manuscript.   Chapter 3 A version of Chapter 3 has been published: Harvard S, Gossec L, Pham T, Richette P, Dougados M, Anis A, Fautrel B. Measurable definitions of ankylosing spondylitis management recommendations are needed for use in observational studies. Joint Bone Spine. 2016 Jan;83(1):101-3. SH led the study design under BF's guidance, oversaw the working group and collected the resulting information, and developed the first draft of the online questionnaire and all subsequent revisions. SH was further responsible for the collection, management, analysis and interpretation of data from the online survey, and wrote the first draft of the manuscript and all subsequent revisions. BF helped guide the study design, participated in the working group, collaborated in the development and revision of the online questionnaire and interpretation of the data, and suggested revisions to the manuscript. Thao Pham (TP), Laure Gossec (LG), and MD participated in the working group and revision of the online questionnaire and suggested revisions to the manuscript. PR, AA, and NB participated in the revision of the online questionnaire, data interpretation, and suggested revisions to the manuscript.   Chapter 4 A version of Chapter 4 has been accepted for publication in the Journal of Rheumathology: Harvard S, Guh D, Bansback N, Richette P, Saraux A, Fautrel B, Anis A. Adherence to Anti-TNF Use Recommendations in Spondyloarthritis: Measurement and Impact in the DESIR Cohort. SH developed the hypotheses and analysis plan under the guidance of BF, AA, and NB, and performed the analyses under the guidance of DG. SH wrote the first draft of the manuscript and all subsequent revisions. BF, AA, NB, DG, PR, and Alain Saraux (AS) participated in the interpretation of results and suggested revisions to the manuscript.       vii Chapter 5 A version of Chapter 5 has been submitted for publication: Harvard S, Guh D, Bansback N, Richette P, Saraux A, Fautrel B, Anis A. Access Criteria for Anti-TNF Agents in Spondyloarthritis:  Influence on Comparative 1-Year Cost-Effectiveness Estimates.  SH developed the hypotheses and analysis plan under the guidance of BF, AA, and NB, and performed the analyses under the guidance of DG and Stanley Wong. SH wrote the first draft of the manuscript and all subsequent revisions. BF, AA, NB, DG, PR, and AS participated in the interpretation of results and suggested revisions to the manuscript.  Chapter 6 SH wrote the first draft of the manuscript and all subsequent revisions. AA, BF, and NB edited the manuscript and suggested revisions.                       viii Table of Contents  Abstract ......................................................................................................................................... iii	Lay Summary ............................................................................................................................... iv	Preface ............................................................................................................................................ v	Table of Contents ....................................................................................................................... viii	List of Tables ............................................................................................................................... xii	List of Figures ............................................................................................................................. xiv	List of Abbreviations .................................................................................................................. xv	Acknowledgements .................................................................................................................. xviii	Dedication .................................................................................................................................... xx	Chapter 1 Literature Review ....................................................................................................... 1	1.1 Overview ............................................................................................................................................ 1	1.2 Background ........................................................................................................................................ 1	1.2.1 Spondyloarthritis: an Evolving Disease Entity .......................................................................................... 1	1.3 Assessment of SpA Disease Activity and Outcomes ......................................................................... 14	1.3.1 Peripheral and Extra-Articular Manifestations of Disease ...................................................................... 15	1.3.2 Radiographic and MRI Findings .............................................................................................................. 15	1.3.3 Inflammatory Markers ('Acute Phase Reactants') .................................................................................... 16	1.3.4 Patient-Reported Disease Activity: BASDAI .......................................................................................... 16	1.3.5 Patient-Reported Functional Status and Disability: BASFI ..................................................................... 18	1.3.6. Patient's and Physician's Global Assessment of Disease Activity .......................................................... 18	1.3.7 Health-Related Quality of Life ................................................................................................................ 19	1.3.8 Work Productivity Loss ........................................................................................................................... 22	1.4 SpA Treatment Recommendations ................................................................................................... 24	1.4.1 General Treatment Recommendations ..................................................................................................... 24	1.4.1.1 Non-Pharmacological Treatment .................................................................................................... 25	1.4.1.2 Pharmacological Treatment ............................................................................................................. 26	1.4.1.2.1 Non-Steroidal Anti-Inflammatory Drugs .............................................................................. 26	1.4.1.2.2 Anti-TNF Agents ................................................................................................................... 27	1.4.1.2.3 Other Agents ......................................................................................................................... 28	1.4.1.3 Surgery ............................................................................................................................................ 29	1.4.1.4 Other Non-Specific Recommendations ........................................................................................... 29	1.4.2 Recommendations Specific to Anti-TNF Access .................................................................................... 30	1.4.3 SpA Treatment Recommendations: Considerations ................................................................................ 33	  ix 1.5 Overview of Economic Evaluation ................................................................................................... 35	1.5.1. Basic Costing Methods ........................................................................................................................... 36	1.5.2 Partial Economic Evaluations: Cost of Illness Studies ............................................................................ 39	1.5.3 Full Economic Evaluations ...................................................................................................................... 42	1.5.3.1 Types of CEA Using QALYs .......................................................................................................... 45	1.5.3.1.1 CEA Based on Observed Data From a Single RCT .............................................................. 46	1.5.3.1.2 CEA Based on Observed Data From a Single Observational Study ..................................... 48	1.5.3.1.3 Mathematical Modeling Studies ............................................................................................ 49	1.5.3.2 Presentation of CEA Results and Uncertainty ................................................................................ 52	1.5.1 Economic Issues in SpA .......................................................................................................................... 55	1.5.1.2 SpA Cost of Illness .......................................................................................................................... 56	1.5.2 Cost Effectiveness of Anti-TNF Therapy Among SpA Patients ............................................................. 61	1.6 Knowledge Gaps .............................................................................................................................. 65	1.7 Research Questions and Goals ........................................................................................................ 66	1.8 Outline of Subsequent Chapters ....................................................................................................... 67	Chapter 2   Costs of Early Spondyloarthritis: Estimates From the First 3 Years of the DESIR Cohort ............................................................................................................................. 69	2.1 Introduction ...................................................................................................................................... 69	2.2 Methods ............................................................................................................................................ 70	2.2.1 Participants ............................................................................................................................................... 70	2.2.2 Costing Methods ...................................................................................................................................... 70	2.2.2.1 Valuing Health Resource Use ......................................................................................................... 70	2.2.2.2 Valuing Work Productivity Loss ..................................................................................................... 72	2.2.2.3 Handling Missing Clinical and Cost Data ....................................................................................... 73	2.2.2.4 Statistical Analyses ......................................................................................................................... 73	2.3 Results .............................................................................................................................................. 75	2.3.1 Patient Characteristics .............................................................................................................................. 75	2.3.2 Costs ......................................................................................................................................................... 75	2.3.2.1 Health Resource Utilisation ............................................................................................................ 75	2.3.2.2 Productivity Losses ......................................................................................................................... 79	2.3.2.3 Total Costs ....................................................................................................................................... 80	2.3.3 Factors Associated With Costs ................................................................................................................ 80	2.3.3.1 Total Costs ....................................................................................................................................... 80	2.3.3.2 Cost Excluding Anti-TNF Agents ................................................................................................... 83	2.4 Discussion ........................................................................................................................................ 83	  x Chapter 3   Measurable Definitions of Ankylosing Spondylitis Management Recommendations Are Needed for Use in Observational Studies .......................................... 86	3.1 Introduction ...................................................................................................................................... 86	3.2 Materials and Methods .................................................................................................................... 88	3.2.1 Delphi Survey Development .................................................................................................................... 88	3.2.2 Participant Recruitment ........................................................................................................................... 88	3.2.3 Delphi Process ......................................................................................................................................... 89	3.3 Results .............................................................................................................................................. 89	3.3.1 Working Group Consensus and Definitions ............................................................................................ 89	3.3.2 Delphi Participant Feedback .................................................................................................................... 90	3.4 Discussion ........................................................................................................................................ 92	Chapter 4   Adherence to Anti-TNF Use Recommendations in Spondyloarthritis: Measurement and Impact in the DESIR Cohort ................................................................... 101	4.1 Introduction .................................................................................................................................... 101	4.2 Methods .......................................................................................................................................... 102	4.2.1 Data Source and Study Population ........................................................................................................ 102	4.2.2 Classification of Adherence ................................................................................................................... 102	4.2.3 Impact of Adherence Classifications ..................................................................................................... 104	4.3 Results ............................................................................................................................................ 105	4.3.1 Classification of Adherence ................................................................................................................... 105	4.3.2 Impact of Adherence Classifications ..................................................................................................... 107	4.3.2.1 Cost Outcomes .............................................................................................................................. 107	4.3.2.2 Health Outcomes ........................................................................................................................... 111	4.4 Discussion ...................................................................................................................................... 117	Chapter 5   Access Criteria for Anti-TNF Agents in Spondyloarthritis:  Influence on Comparative 1-Year Cost-Effectiveness Estimates ............................................................... 120	5.1 Introduction .................................................................................................................................... 120	5.2 Methods .......................................................................................................................................... 122	5.2.1 Study Setting and Data Source .............................................................................................................. 122	5.2.2 Selection of Anti-TNF Access Criteria .................................................................................................. 123	5.2.3 Creation of 'Study Population' Datasets ................................................................................................. 123	5.2.4 Descriptive Statistics .............................................................................................................................. 125	5.2.5 Adjustment of Costs and QALYs .......................................................................................................... 125	5.2.6 Cost-effectiveness Analysis Using Adjusted Costs and QALYs ........................................................... 126	5.3 Results ............................................................................................................................................ 126	  xi 5.4 Discussion ...................................................................................................................................... 137	Chapter 6 Discussion: Informing Economic Questions and Treatment Recommendations with Observational Data .......................................................................................................... 141	6.1. Introduction ................................................................................................................................... 141	6.2 Contribution of This Thesis to Current Knowledge, Challenges Encountered, and Limitations .. 141	6.2.1 Objective 1: The Quantification of the Cost of Illness From the Societal Perspective Among DESIR Patients and Identification of Factors Associated With Costs in the First Three Years of Follow-Up .......... 141	6.3.2 Objective 2: The Development of a Method to Measure Adherence to Specific Elements of the ASAS Treatment Recommendations ......................................................................................................................... 142	6.3.3 Objective 3: The Evaluation of the Impact of Adherence to Measurable Elements of Recommended SpA Care Among DESIR Patients in Terms of Costs and Health Benefits ................................................... 144	6.3.4 Objective 4: The Assessment of Current French Anti-TNF Access Restrictions in Terms of the Cost-Effectiveness in That Setting Relative to Other Potential Restrictions .......................................................... 146	6.4 Significance of Findings and Implications ..................................................................................... 147	6.4.1 Research and Policy Directions ............................................................................................................. 153	6.4.1.1 Continue Work to Identify Which SpA Patients Are Likely to Respond to Anti-TNF Agents, Defining 'Response' in Terms of Meaningful Utility Gain ....................................................................... 153	6.4.1.2 Use Actionable Wording in Treatment Recommendations and Develop 'Quality Indicators' to Measure Adherence to Recommendations ................................................................................................ 155	6.4.1.3 Consider Innovative Policies Linking Anti-TNF Coverage to Outcome and/or Evidence Development; Emphasize Shared Goals of Clinical and Reimbursement Guidelines by Supporting Appropriate Anti-TNF Discontinuation .................................................................................................... 156	6.4.1.4 Invest in Observational Research Including Evaluation of Adherence to Recommendations ...... 157	6.5 Conclusion ..................................................................................................................................... 158	References .................................................................................................................................. 159	Appendix .................................................................................................................................... 178	     xii List of Tables  Table 1.1 Early Criteria Sets for AS ............................................................................................... 2	Table 1.2 Calin, Berlin, and ASAS Criteria for IBP ....................................................................... 3	Table 1.3 Modified New York Criteria for AS ............................................................................... 3	Table 1.4 Amor Criteria .................................................................................................................. 5	Table 1.5 Sensitivity and Specificity of the 25 Candidate Variables of the ESSG Criteria ........... 6	Table 1.6 ESSG Criteria ................................................................................................................. 6	Table 1.7 Comparison of Current AS and SpA Criteria Sets. ........................................................ 9	Table 1.8 Comparison of Current AS and SpA Criteria Sets ....................................................... 12	Table 1.9 SpA Disease Activity and Outcome Measures of Focus .............................................. 17	Table 1.10 2010 Update of the International ASAS Recommendations for the Use of Anti-TNF Agents in Patients With Axial Spondyloarthritis1 ................................................................ 28	Table 1.11 International Recommendations for Diagnosis, Disease Activity, and Treatment Failure Before Use of Anti-TNF Agents in SpA Described in 2011.2 .................................. 31	Table 1.12 Matrix of 19 Cost Domains3 ....................................................................................... 41	Table 1.13 SpA COI Studies in Biologic Era: Selected Characteristics ....................................... 58	Table 2.1 Baseline Characteristics of the DESIR Cohort by Anti-TNF Use ................................ 72	Table 2.2 Health Resource Use and Productivity Loss Costs by Year and Anti-TNF Use .......... 77	Table 2.3 Cost Component as Proportion of Total Costs by Year and Anti-TNF Use ................. 79	Table 2.4 Models of Total Cost and Non-Biologic Costs ............................................................. 82	Table 3.1 Summary of ASAS Recommendations 1 ...................................................................... 96	Table 3.2 Proposals Considered in Round 1 and Percent Agreement .......................................... 97	Table 3.3 Proposals Considered in Round 2 and Percent Agreement .......................................... 99	Table 3.4 Final Quantitative Definitions of 6 of 11 ASAS Recommendations .......................... 100	Table 4.1 Rheumatologist-Proposed Definitions of Adherence to ASAS Recommendations ... 108	Table 4.2 Patient Clinical Characteristics by Anti-TNF Use and Disease Activity Pre-Index ... 109	Table 4.3 Alternate Definitions Used to Classify Patients Based on Adherence to Anti-TNF Recommendations ............................................................................................................... 111	Table 4.4 Characteristics and Outcomes of Patients Defined as Adherent Users, Non-Adherent Users, and Non-Adherent Non-Users ................................................................................. 112	  xiii Table 4.5 Models of Impact of Adherence Classifications on Cost Outcomes .......................... 114	Table 4.6 Models of Impact of Adherence Classification on QALY Outcome .......................... 115	Table 4.7 Characteristics of Adherent Non-Users (n=255) ........................................................ 116	Table 5.1 Selected Criteria Sets: Satisfaction Among All DESIR Patients (n=708) .................. 127	Table 5.2 Characteristics of Anti-TNF Users and Non-Users by Criteria Set ............................ 129	Table 5.3 Characteristics of Anti-TNF Users not Satisfying Selected Criteria Sets .................. 130	Table 5.4 Incremental Costs, QALYs, and ICERS by Criteria Set ............................................ 131	Table 5.5 Comparative Estimates of Costs, QALYs, and ICERs: Basecase Analysis ............... 132	Table 5.6 Total Time on Anti-TNF Therapy Among Non-Responders (From Initiation Until End of Follow-Up) ..................................................................................................................... 134	Table 5.7 Comparative Estimates of Costs, QALYs, and ICERs: Sensitivity Analysis Excluding Non-Responder Anti-TNF Costs Past 24 Weeks ................................................................ 134	Table 5.8 Utility Gain Six and Twelve Months Post-Therapy Initiation in Anti-TNF Responders and Non-Responders ........................................................................................................... 135	Table 5.9 Comparative Estimates of Costs, QALYs, and ICERs in Sensitivity Analyses ......... 136	Table 6.1 Shared Characteristics of CER and Current Work ..................................................... 148	         xiv List of Figures  Figure 1.1 ASAS axSpA Criteria .................................................................................................... 8	Figure 1.2 ASAS Peripheral SpA Criteria ...................................................................................... 9	Figure 1.3 Evolution of SpA Concept ............................................................................................. 9	Figure 1.4 SpA: An Interconnected Family of Diseases ............................................................... 10	Figure 1.5 Cost-Effectiveness Plane ............................................................................................. 53	Figure 1.6 Bootstrap Estimates of Cost-Effectiveness ................................................................. 55	Figure 1.7 Confidence Ellipses ..................................................................................................... 55	Figure 2.1 Yearly Costs Among Non-Biologics (Left) and Biologics Users (Right) ................... 81	Figure 5.1 Confidence Intervals Around ICERs in Each of the Five Study Populations ........... 133	Figure 6.1 Redefined Relationships of Evidence Processes. Reproduced from Luce BR, Drummond M, Jönsson B, Neumann PJ, Schwartz JS, Siebert U, Sullivan SD. EBM, HTA, and CER: clearing the confusion. Milbank Q. 2010 Jun;88(2):256-76. ............................. 150	     xv List of Abbreviations  ACR American College of Rheumatology  Anti-TNF Anti-Tumour Necrosis Factor Alpha  AS Ankylosing Spondylitis ASDAS Ankylosing Spondylitis Disease Activity Score ASQoL Ankylosing Spondylitis Quality of Life ATIH Agence Technique de l’Information sur l’Hospitalisation axSpA Axial Spondyloarthritis BASDAI Bath Ankylosing Spondylitis Disease Activity Index  BASFI Bath Ankylosing Spondylitis Functional Index  BASGI Bath Ankylosing Spondylitis Disease Global Index  BASMI Bath Ankylosing Spondylitis Metrology Index  CBA Cost-Benefit Analysis CEA Cost-Effectiveness Analysis CER Cost-Effectiveness Research COI Cost-of-Illness  CRA Canadian Rheumatology Association  CRF Case Report Form CRP C-Reactive Protein  CUA Cost-Utility Analysis DC-ART Disease-Controlling Antirheumatic Treatments  DMARDs Disease-Modifying Antirheumatic Drugs DRG Diagnosis-Related Group EBM Evidence-Based Medicine EQ5D EuroQoL-5D ESR Erythrocyte Sedimentation Rate  ESSG European Spondyloarthropathy Study Group FC Friction Cost GDG Guideline Development Group GEE Generalized Estimating Equations    xvi HC Human Capital  HLA B-27 Human Leukocyte Antigen B-27 HRQoL Health-Related Quality of Life HTA Health Technology Assessment IBD Inflammatory Bowel Disease  IBP Inflammatory Back Pain ICER Incremental Cost-Effectiveness Ratio  IOM Institute of Medicine LOCF Last Observation Carried Forward  MCMC Monte Carlo Markov Chain  MCS  Mental Component Summary  MI Multiple Imputation  MID Minimally Important Difference mNY Modified New York Criteria MRI Magnetic Resonance Imaging MTX Methotrexate NICE National Institute for Health and Care Excellence  nr-axSpA Non-Radiographic Axial Spondyloarthritis NRS Numerical Rating Scale NSAIDs Non-Steroidal Inflammatory Drugs PASS Patient Acceptable Symptom State PCS Physical Component Summary  PhGA Physician’s Global Assessment  PRO Patient Reported Outcome PsA Psoriatic Arthritis PtGA Patient’s Global Assessment QALY Quality Adjusted Life Year QIC Quasi-Likelihood Statistic  r-axSpA Radiographic Axial Spondyloarthritis RCT Randomized Controlled Trials ReA Reactive Arthritis   xvii RR Rate Ratio  SD Standard Deviation SF-36 Medical Outcome Study Short Form 36 Health Survey SpA Spondyloarthritis SSZ Sulfasalazine UK United Kingdom US United States VAS Visual Analog Scale WTP Willingness-to-pay     xviii Acknowledgements  Who reads the Acknowledgements section of a PhD thesis? Perhaps only the very person who wrote the thesis- later, much later, when the trauma has subsided. Or perhaps future PhD students, curious as to the formatting requirements? ("Don't write anything that will become embarrassing with time!" said a friend, "I made an inside joke based on an episode of Top Chef which only my boyfriend would understand- we're no longer together.") Anyway, I think someone will read this, but I don't know who. I'll just write it in the style of a diary entry, imagining that the diary will fall out of my backpack and be lost on the bus, returned to me after untold adventures.  When I was a kid, I would watch the Oscars, and I formed the opinion that it's foolish to give speeches in which you systematically thank all the people who helped you achieve something. What if you forget someone? It would be an unspeakable disaster. Of course, it would seem unlikely that you would forget to thank, for example, the Casting Agent, the Producer, the Director, or your Co-Star (for taking a chance on you- for making this happen- for the guidance you needed to achieve your vision- for showing up on set with you every day, day after day, even when the going got tough). But it's not just that you might forget someone. By listing people to thank, one is obliged to impose an ORDER on the thanking, and you can't rank people who were all indispensable to you in accomplishing your goal. I think it's better just to speak of what you're grateful for, and to trust that the right people will recognize themselves as indispensable, and know the meaning of that.  This PhD thesis was very, very hard to accomplish. Terrible events unfolded relentlessly from the moment I enrolled until close to the end and my talents lay elsewhere. I'm grateful that, for good stretches of the worst times, I was in Paris living out a dream of mine. I'm grateful that as I did this PhD, over the years, I had the chance to do some work I was really good at and to know I was valued. I'm grateful that, for lots and lots of long days strung back to back, someone was there to help me and teach me and tolerate my shortcomings and model all manner of genius to me. I'm grateful for every big picture conversation that inspired me. I'm grateful to feel like there's the potential to do cool work in the future. If you recognize yourself as an indispensable part of any of this, I can only hope that you understand the significance of that.  I dedicated this thesis to my family, which comprises my friends. If you exchanged innumerable text messages with me between 2011 and 2017, if you came to my house for dinner, if you had me over to yours, if you talked with me and listened, know that I'm grateful, I thank you, you mean so much to me. Here is 1 page from an imaginary diary + 1 PhD thesis, to which I append an invisible IOU for 1+ bottle of champagne.     xix Remerciements particuliers  Cette thèse n'aurait pas été possible sans la base de données de DESIR, ce qui a été construit grâce aux contributions des 708 patients dans la cohorte. L'auteure se permet de remercier chacun de ces patients pour son temps et son engagement à l'objectif de recherche clinique en SpA.   Special Thanks  This thesis would not have been possible without the DESIR database, which was created thanks to the contributions of the 708 patients in the cohort. The author would like to extend a special thanks to each one of these patients for their time and their dedication to clinical research in SpA.    xx Dedication  To all my family, blood and chosen, here and elsewhere.  1 Chapter 1  Literature Review   1.1 Overview  This chapter is a review of SpA literature relevant to the subsequent chapters of this thesis, which focuses on economic evaluations in the context of SpA treatment recommendations. The goal of this chapter is to prepare any generalist in the health sciences, with or without a background in economic evaluation, to read the remaining chapters and interpret them. Specifically, this chapter describes the SpA disease entity, explaining that the concept of SpA has evolved over time and that the current conceptualization of the disease is relatively new. It then defines the SpA outcome measures that appear in the remaining sections of the thesis, describing the unique importance and relevance of each. Recommended SpA treatments, as outlined in clinical guidelines produced by major expert panels, are then discussed alongside the evidence base that supports their use in improving patient outcomes. Next, an overview of economic evaluation is provided, including a description of its purpose, basic methods, and relationship to evidence-based research paradigms, including comparative effectiveness research (CER). As all studies conducted as part of this thesis use observational data, the different types of data available for use in economic evaluations are reviewed, and the importance of observational studies to both economic evaluations and CER is highlighted. Finally, the economic impact of SpA is described, and current economic questions in the context of SpA treatment recommendations are discussed. The specific research questions that are addressed in the remaining chapters of this thesis are then introduced as well as the DESIR cohort, the source of the observational data used in this research.  1.2 Background  1.2.1 Spondyloarthritis: an Evolving Disease Entity  ‘Spondyloarthritis’ (sometimes 'spondyloarthropathy' or 'spondylitis') refers to a group of interrelated chronic rheumatic diseases, the most severe form of which is ankylosing spondylitis (AS). The first classification criteria for AS were proposed in Rome in 1961 at a symposium on   2 population studies in rheumatic diseases.4 This early classification system defined the disease by taking into account six criteria (Table 1.1). One important criterion in this set refers to changes in the sacroiliac joints (i.e., bone damage where the lower spine and pelvis connect), which can result from inflammation of the sacroiliac joints known as 'sacroiliitis'; sacroiliitis that is visible on X-Ray is referred to as 'radiographic sacroiliitis'. The Rome Criteria for AS were revised in 1966 to create the New York criteria for AS5 (Table 1.1). The major difference between the Rome and New York criteria is that radiographic sacroiliitis is considered in the latter as indispensable to the AS diagnosis.  Table 1.1 Early Criteria Sets for AS Rome Criteria (1961) 4 New York Criteria (1966) 5 1) Low back pain for over 3 months not relieved by rest;  2) Pain and stiffness in the thoracic region;  3) Limited motion in the lumbar spine;  4) Limited chest expansion;  5) History of iritis or its sequelae;  6) X-Ray evidence of bilateral changes in the sacroiliac joints characteristic of AS (eg excluding osteoarthrosis).  1) Limitation of motion in all three planes of the lumbar spine;   2) Pain at the dorsolumbar junction or in the lumbar spine;   3) Limitation of chest expansion to 2.5 cm or less Diagnosis Diagnosis Definite AS diagnosis if 4/5 of first five items satisfied, or item 6 (i.e., bilateral sacroiliitis) plus one other criterion. Definite AS diagnosis if 1 clinical criterion satisfied + grade 3-4 bilateral sacroiliitis  Definite AS diagnosis if clinical criteria 2 and 3 satisfied + grade 2 bilateral sacroiliitis   Definite AS diagnosis if clinical criteria 2 and 3 satisfied + grade 3-4 unilateral sacroiliitis   Probable AS diagnosis if grade 3-4 sacroiliitis present, even without additional clinical criteria   In 1977, a new, highly sensitive and specific marker of AS was discovered: inflammatory back pain (IBP). As demonstrated in Calin's seminal study,6 IBP could be defined by five factors (Table 1.2), where the presence of at least four had 95% sensitivity and 85% specificity in detecting AS. The IBP concept was to become influential in defining AS, and was reconceptualized by additional criteria sets7, 8 only twenty years after its introduction. The important concept of IBP prompted a modification to the New York criteria in 1984;9 the   3 Modified New York (mNY) criteria remain in wide use today (Table 1.3). In the mNY criteria, radiographic sacroiliitis remains essential to the definite AS diagnosis.9    Table 1.2 Calin, Berlin, and ASAS Criteria for IBP Calin Criteria (1977) 6 Berlin Criteria (2006) 7 ASAS Criteria (2009) 8 1) Insidious onset 2) Age at onset <40 years 3) Improved by exercise but not by rest 4) Associated with morning stiffness 5) Duration longer than 3 months 1) Morning stiffness >30 minutes 2) Improvement in back pain with exercise but not with rest 3) Alternating buttock pain 4) Waking during the second half of the night 1) Age at onset <40 years 2) Insidious onset 3) Improvement with exercise 4) No improvement with rest 5) Pain at night (with improvement upon getting up) Presence of 4 out of 5 of these factors had 95% sensitivity and 85% specificity in detecting AS in original study population Presence of 2 out of 4 factors had sensitivity of 70% and a specificity of 81% in original study population Presence of 4 out of 5 of these factors had 77% sensitivity and 92% specificity in original study population  Table 1.3 Modified New York Criteria for AS A. DIAGNOSIS 1. Clinical criteria a) Low back pain and stiffness for more than 3 months which improves with exercise, but is not relieved by rest. b) Limitation of motion of the lumbar spine in both the sagittal and frontal planes. c) Limitation of chest expansion relative to normal values corrected for age and sex  2. Radiologic criterion Sacroiliitis grade 2 bilaterally or sacroiliitis grade 3-4 unilaterally.  B. GRADING 1. Definite ankylosing spondylitis if the radiologic criterion is associated with at least 1 clinical criterion.  2. Probable ankylosing spondylitis if: a) Three clinical criteria are present. b) The radiologic criterion is present without any signs or symptoms satisfying the clinical criteria. (Other causes of sacroiliitis should be considered.)  Adapted from van der Linden et al. Evaluation of diagnostic criteria for ankylosing spondylitis. A proposal for modification of the New York criteria. Arthritis Rheum. 1984 Apr;27(4):361-8.   At the time that AS classification criteria were being evaluated in the 1970s and 80s, a discussion was taking place to suggest that AS is in fact just one manifestation of a larger family of diseases with overlapping symptoms.10 In a seminal 1974 article, Moll and colleagues described several   4 diseases with features in common with AS, including psoriatic arthritis (PsA), the 'intestinal arthropathies' such as ulcerative colitis and Crohn's disease, Reiter's and Whipple's disease (forms of arthritis thought to be caused by infection), and Behcet's syndrome. In all of these conditions, it was noted, certain symptoms tend to be common, including peripheral arthritis (i.e., inflammation of the joints of the arms and legs), psoriasis (i.e., inflammatory skin lesions), uveitis (i.e., inflammation of the eyes), and inflammatory bowel disease (IBD). The prevalence of radiographic sacroiliitis is also higher among patients with these conditions, yet it may not present as in classical AS. Importantly, patients with these conditions also test negative for rheumatoid factor, which led to Moll's proposal to call these conditions the 'seronegative spondarthritides', wherein the prefix 'spond' (i.e., spine) marks the association with spondylitis.   Following Moll's seminal article,10 the term SpA gradually came to be used to describe a family of diseases that are distinct from rheumatoid arthritis (RA) yet clearly linked to AS, and extensive research continued with the aim of describing the network of SpA symptoms.11-14, 14 No formal classification criteria for the disease were developed until 1990, when the Amor SpA criteria became available.15 The Amor criteria, published in a French language journal, took into account 12 clinical characteristics, including a positive test for human leukocyte antigen B-27 (HLA-B27), an immune-response mediating molecule within the major histocompatibility complex shown in various studies throughout the 1970s and 1980s to be associated with AS.16 The Amor criteria assign a unique number of points to each characteristic, with 6 or more points indicating the presence of disease (Table 1.4). In the original study population, the Amor criteria had 92% sensitivity and 98% specificity in detecting SpA.15           5 Table 1.4 Amor Criteria Clinical symptoms or history of scoring  Point Lumbar or dorsal pain at night or morning stiffness or lumbar or dorsal pain  1 Asymmetrical oligoarthritis  2 Buttock pain  1 If alternating buttock pain  2 Sausage like toe or digit  2 Heel pain or other well-defined enthesopathy  2 Iritis 1 Nongonococcal urethritis or cervicitis within one month before the onset of arthritis  1 Acute diarrhea within one month before the onset of arthritis 1 Psoriasis, balanitis, or inflammatory bowel disease (ulcerative colitis or Crohn’s disease)  2 Radiological findings  Sacroiliitis (bilateral grade 2 or unilateral grade 3)  3 Genetic background  Presence of HLA-B27 and/or family history of ankylosing  spondylitis, reactive arthritis, uveitis, psoriasis, or inflammatory bowel disease 2 Response to treatment  Clear-cut improvement within 48 hours after NSAIDs intake or rapid relapse of the pain after their discontinuation 2 2 A patient is considered as suffering from spondyloarthropathy if the sum is ≥6  Adapted from Akgul O, Ozgocmen S. Classification criteria for spondyloarthropathies. World J Orthop. 2011 Dec 18;2(12):107-15.  Just one year after the publication of the Amor criteria in France, the European Spondyloarthropathy Study Group (ESSG) published a new set of SpA classification criteria.17 Arguing that this family of diseases was broader than generally recognized, the ESSG described SpA in terms of five subtypes, including AS, PsA, 'reactive' arthritis (i.e., arthritis following infection, such as in Reiter's and Whipple's disease), SpA with IBD, and undifferentiated SpA (i.e., patients with features of spondyloarthropathy but who fail to meet diagnostic criteria). In the criteria development process, the ESSG calculated the sensitivity and specificity of 25 SpA-related characteristics among 403 SpA patients and 604 controls. This exhaustive list of SpA symptoms (Table 1.5) illustrates the complex, multi-faceted character of the disease; the final ESSG criteria are shown in Table 1.6. In the original study population,  sensitivity of the ESSG criteria set varied across the different subgroups from 100% for SpA with IBD, to 94% for AS, 82% for PsA, 81% for reactive arthritis, and 78% for undifferentiated SpA.17     6 Table 1.5 Sensitivity and Specificity of the 25 Candidate Variables of the ESSG Criteria Variable Sensitivity % Specificity% Spinal pain  84.9 28.8 Inflammatory spinal pain  74.6 82.5 Anterior chest wall pain  43.8 86.2 Buttock pain  52.9 74.0 Buttock pain alternating between the left and right gluteal areas  20.4 97.3 Buttock pain, unilateral, without radiation below the knee  13.1 93 .0 Chest expansion <2.5 cm  15.0 95.8 Reduction in spinal mobility  36.5 89.5 Synovitis, predominantly in the lower limbs  35.3 89.1 Asymmetric synovitis  41.3 87.3 Mono- or oligo- versus polyarticular involvement  14. I 74.5 Dactylitis  17.9 96.4 Enthesopathy at any site  56.4 77.6 Heel pain  36.5 88.9 Conjunctivitis  16.9 87.9 Uveitis (acute, anterior)  21.7 97.2 Psoriasis  22.7 95.2 Mucosal ulcerations  5.5 97.3 Acute diarrhea (1 month before arthritis)  12.3 98.1 Inflammatory bowel disease  9.6 97.3 Nongonococcal urethritis or cervicitis (1 month before arthritis)  6.8 96.7 HLA-B27  68.1 90.8 Family history of either ankylosing spondylitis, reactive arthritis, psoriasis, uveitis, or inflammatory bowel disease 32.2 94.5 Sacroiliitis (on radiography)  54.4 97.8 Positive effect of NSAIDs  65.4 49.0  Reproduced from Dougados et al. The European Spondylarthropathy Study Group preliminary criteria for the classification of spondylarthropathy. Arthritis Rheum. 1991 Oct;34(10):1218-27.  Table 1.6 ESSG Criteria INFLAMMATORY SPINAL PAIN OR SYNOVITIS*  (*alternating or predominantly in the lower limbs) Plus one or more of the following: Positive family history Psoriasis Inflammatory bowel disease Urethritis, cervicitis, or acute diarrhea within one month before arthritis Buttock pain alternating between right and left gluteal areas Enthesopathy Sacroiliitis  Adapted from Dougados et al. The European Spondylarthropathy Study Group preliminary criteria for the classification of spondylarthropathy. Arthritis Rheum. 1991 Oct;34(10):1218-27.  Throughout the 1990s, magnetic resonance imaging (MRI) became more widely available, and it was discovered that sacroiliitis, spondylitis, and inflammatory lesions of the spine are visible on   7 MRI. This discovery confirmed that the AS disease process begins before bone damage is visible on X-Ray18, 19 and underscored that less severe or undifferentiated forms of SpA without radiographic evidence of sacroiliitis may eventually progress to AS.20, 21 The distinction between 'non-radiographic' or 'pre-radiographic' SpA thus emerged in the literature and came to be of central importance in understanding SpA on a continuum.22, 23   In the late 1990s, an international working group, now known as the Assessment of Spondyloarthritis International Society (ASAS), was formed to select a core set of AS outcome measures for use in clinical trials.24-26 The work of ASAS surrounding outcome measures drew increased attention to the fact that the manifestation of SpA generally can be predominantly axial (i.e., with symptoms spine, pelvis, and thoracic cage being most noticeable) or predominantly peripheral (i.e., with symptoms in the extremities being most noticeable). This important distinction had clear implications for SpA treatment and outcome measurement, yet was not reflected in the Amor or ESSG criteria.   Subsequent to the developments throughout the 1990s in the understanding of SpA, and the increased importance of MRI, the need for updated classification criteria was voiced in 2005.22 ASAS began developing new candidate criteria,27 their work reflecting a new consensus that SpA should be distinguished as either 'axial' or 'peripheral', with the axial SpA (axSpA) diagnosis being independent of radiographic sacroiliitis. Axial disease without radiographic sacroiliitis was to be described as 'pre-radiographic' axSpA or 'non-radiographic' axSpA (nr-axSpA), acknowledging that non-radiographic disease may progress to radiographic disease.   Published in 2009 and 2011, respectively, ASAS's separate criteria sets for the classification of axSpA28 (Figure 1.1) and peripheral SpA29  (Figure 1.2) are in wide use today. In the original study populations, the ASAS axSpA criteria had 83% sensitivity and 84% specificity,28 the ASAS peripheral SpA criteria 78% sensitivity and 83% specificity, respectively.29 A comparison of the sensitivity and specificity of the ASAS, ESSG, and Amor criteria has been published by Zeidler,30 as shown in Table 1.7. The conceptual developments in the field that are represented by the ASAS classification criteria include the recognition of two distinct manifestations of the disease and the separation of 'imaging' versus 'clinical' criteria, reflecting the importance of new   8 MRI techniques. The ASAS criteria also incorporate elevated C-reactive protein (CRP), an inflammatory marker or 'acute phase reactant' shown to be associated with SpA disease activity31, 32 (erythrocyte sedimentation rate (ESR), another inflammatory marker associated with SpA disease activity,31, 32 was considered but not included in the final ASAS criteria set27).  Figure 1.1 ASAS axSpA Criteria   Reproduced from Rudwaleit et al. The development of Assessment of SpondyloArthritis international Society classification criteria for axial spondyloarthritis (part II): validation and final selection. Ann Rheum Dis. 2009 Jun;68(6):777-83.            9 Figure 1.2 ASAS Peripheral SpA Criteria  Reproduced from Rudwaleit et al. The Assessment of SpondyloArthritis International Society classification criteria for peripheral spondyloarthritis and for spondyloarthritis in general. Ann Rheum Dis. 2011 Jan;70(1):25-31.  Figure 1.3 Evolution of SpA Concept  Table 1.7 Comparison of Current AS and SpA Criteria Sets. Criteria Sensitivity (%) Specificity (%) Amor 77–91 92–98 ESSG 84–91 91–100 ASAS axial SpA 82.9 84.4  Modified ESSG (with MRI)* 85.1 65.1  Modified Amor (with MRI)* 82.9 77.5 ASAS peripheral SpA 77.8 82.9  Modified ESSG (with MRI)* 62.5 81.1  Modified Amor (with MRI)* 39.8 97.8 ASAS combined criteria 79.5 83.3  Modified ESSG (with MRI)* 79.1 68.8  Modified Amor (with MRI)* 67.5 86.7 * The ESSG criteria and the Amor criteria were modified in that active sacroiliitis on MRI was added.  Reproduced from Zeidler H, Amor B. The Assessment in Spondyloarthritis International Society (ASAS) classification criteria for peripheral spondyloarthritis and for spondyloarthritis in general: the spondyloarthritis concept in progress. Ann Rheum Dis. 2011 Jan;70(1):1-3. 30  Rome	AS	Criteria	1961	 New	York	AS	Criteria	1966	 Calin	IBP	Criteria	1977	 Modified	New	York	AS	Criteria	1984	 Amor	SpA	Criteria	1990	 ESSG	SpA	Criteria	1991	 Expansion	of	MRI	ASAS	axial	SpA	Criteria	2009	ASAS	peripheral	SpA	Criteria	2011	  10 From the evolution in SpA classification systems that has taken place since the 1960s (Figure 1.3), we see that what was once considered a distinct clinical entity defined by radiographic sacroiliitis is now understood to be a heterogeneous, yet interconnected family of diseases (Figure 1.4).   Figure 1.4 SpA: An Interconnected Family of Diseases   Reproduced from Zeidler H, Amor B. The Assessment in Spondyloarthritis International Society(ASAS) classification criteria for peripheral spondyloarthritis and for spondyloarthritis in general: the spondyloarthritis concept in progress. Ann Rheum Dis. 2011 Jan;70(1):1-3. 30  Developments in the SpA concept continue to be made: the term 'reactive arthritis' (ReA) is now used in place of the term Reiter's syndrome, with Whipple's disease and Behcet's syndrome no longer considered manifestations of SpA,30 and a juvenile form of SpA has been defined.30, 33 As SpA patients often present with different features over their life course,34 the understanding of SpA may also evolve at the patient level.  One important effect of the evolution in the SpA concept is on our understanding of the global burden of the disease. In a 2014 meta-analysis of 36 studies, Dean et al.35 estimated the prevalence of AS per 10,000 population to be 31.9 in North America, 23.8 in Europe, 16.7 in Asia, 10.2 in Latin America, and 7.4 in Africa. Additional estimates, weighted by study size, were calculated as 18.6, 18.0 and 12.2 for Europe, Asia and Latin America, respectively. Fewer studies have attempted to measure the prevalence of the broader category of axSpA and no meta-  11 analyses are as yet available to estimate global prevalence of axSpA.36 However, using data from the 2009-2010 United States (US) National Health and Nutrition Examination Survey, Reveille et al. estimated an axSpA prevalence in the US of 90 per 10,000 population using the Amor criteria and 140 per 10,000 population using the ESSG criteria.37 Thus, with changes to the conceptualization of SpA, we see a corresponding shift in the understanding of the burden of the disease.    12 Table 1.8 Comparison of Current AS and SpA Criteria Sets Features of criteria Modified New York criteria Amor criteria ESSG criteria ASAS axial SpA criteria ASAS peripheral SpA criteria Year of publication 1984 1990 1991 2009 2011 Inclusion or entry criteria Sacroiliitis on radiograph*plus ≥1 clinical criterion None, fulfillment of criteria requires a score of ≥6 points, assigned on the basis of clinical features that are considered from the list below. Weightings for each feature are shown in parentheses Either IBP or synovitis (asymmetric or predominantly of the lower limbs) plus at least 1 other SpA feature ≥3 months back pain before age 45 years and either sacroiliitis on imaging (radiographs or MRI) plus ≥1 other SpA feature (imaging arm) or HLA-B27 positive plus ≥2 other SpA features (clinical arm) Arthritis, enthesitis or dactylitis plus ≥1 SpA feature marked witha or ≥2 other SpA features marked withb SpA features to be considered IBP‡ √ × × √ √ (ever)b Alternating buttock pain × √ or gluteal pain (1 point) √ × × Pain at night or morning stiffness × √ (1 point) × × × Arthritis × √ asymmetrical oligoarthritis (2 points) × √ √b Dactylitis × √ (2 points) × √ √b Enthesitis (heel) × √ (2 points) √ √ √b Good response to NSAIDs × √ (2 points) × √ × Psoriasis × √ (2 points)§ √ √ √a Inflammatory bowel disease × √ (2 points)§ √ √ √a Balanitis × √ (2 points)§ × ×   Uveitis × √ (2 points) × √ √a Diarrhea <1 month before onset arthritis × √ (1 point) √# × × Urethritis/cervicitis <1 month before onset arthritis × √ (1 point) √# × × Preceding infection × – × × √a Positive family history for SpA|| × √ (2 points)¶ √ √ √b   13 Features of criteria Modified New York criteria Amor criteria ESSG criteria ASAS axial SpA criteria ASAS peripheral SpA criteria HLA-B27 × √ (2 points)¶ × √ √a Elevated CRP × × × √ × Sacroiliitis × √ (radiographic*; 3 points) √ (radiographic*) × √a (radiographic* or MRI-detected) Limitation in mobility of lumbar spine √ × × × × Limitation in chest expansion √ × × × × *Radiographic sacroiliitis is considered present when at least grade 2 bilaterally or grade 3–4 unilaterally. ‡Different definitions for IBP exist for the different criteria sets. §Presence of psoriasis, balanitis or inflammatory bowel disease is considered as 1 item, and 2 points are given in total if at least one is present. ||Different definitions for a positive family history exist for the different criteria sets. ¶Presence of either HLA-B27 positivity or a positive family history is sufficient to obtain the score of 2 points. #Presence of either urethritis/cervicitis or acute diarrhea within 1 month before onset of arthritis is sufficient to fulfill this SpA feature in the ESSG criteria. Abbreviations: AS, ankylosing spondylitis; ASAS, Assessment of Spondyloarthritis International Society; CRP, C-reactive protein; ESSG, European Spondyloarthropathy Study Group; IBP, Inflammatory back pain; SpA, spondyloarthritis.  Reproduced from: van Tubergen A, Weber U. Diagnosis and classification in spondyloarthritis: identifying a chameleon. Nat. Rev. Rheumatol. 2012 Mar 27;8(5):253-61.    14 Finally, changes to SpA classification criteria have implications for the results of clinical studies in SpA, which can be expected to vary based on the entry criteria for included patients. In the following sections discussing SpA research, participants' clinical characteristics are described in order to better contextualize the study findings. As it is a relatively new clinical concept, many questions remain surrounding SpA outcomes and treatment, particularly with respect to differences between nr-axSpA and AS populations. The following section describes a number of common assessments of disease activity and outcomes that are used in studies of SpA patients, specifically those that appear in the remainder of the thesis. Selected outcomes among untreated SpA patients are described in order to illustrate the disease burden and to introduce the subsequent section on SpA treatment, which aims to improve patient outcomes.  1.3 Assessment of SpA Disease Activity and Outcomes   Regular assessment of SpA patients is needed in order to track and predict disease progression, determine the need for therapy, and evaluate subsequent treatment response.38 In ASAS's published guide to SpA assessments,39 a comprehensive list of measurements is organized by clinical domain and instrument, which may be used to collect information in accordance with different purposes in the clinical context and research context, respectively. Some assessments are objective measures of disease activity (e.g., laboratory analyses), others reflect the physician's perspective on the patient's health status, while others are patient-reported outcomes (PROs). Defined as "any aspect of a patient's health status that comes directly from the patient (i.e., without the interpretation of the patient's responses by a physician or anyone else)",40 PROs are increasingly recommended for measuring quality of care in rheumatology.41 In the context of the current thesis, health-related quality of life (HRQoL) and work productivity loss are two PROs of particular importance, as these are central outcomes in economic studies. When incorporated into economic studies, these outcomes are assigned a value, as described in section 1.5. However, it should be noted that both of these outcomes may also be useful in non-economic studies, in which they are analyzed without attaching a value to them. In the subsequent chapters of this thesis, HRQoL and work productivity loss are outcomes of focus, while other assessments of disease activity are used to understand important differences between patients and to help measure their independent effects on outcomes.    15 1.3.1 Peripheral and Extra-Articular Manifestations of Disease  As shown in Figure 1.3, ASAS's criteria for peripheral SpA take into consideration many of the patient's clinical features, including peripheral arthritis, enthesitis, dactylitis, and extra-articular manifesations of the disease, such as psoriasis, IBD, and uveitis.29 Vital to characterizing the patient population under study in terms of axial versus peripheral SpA, and describing disease activity in the individual, information on these clinical features is routinely collected in the context of SpA research studies. Furthermore, the same clinical features that are considered in distinguishing axial from peripheral SpA are associated with unique characteristics of disease progression and treatment response, about which there is still much to be learned.42 Information in terms of peripheral and extra-articular manifestations of disease activity is thus central to research surrounding SpA treatment and outcomes.   1.3.2 Radiographic and MRI Findings  As shown in Table 1.3, the definite diagnosis of AS according to the mNY criteria is dependent on radiographic evidence of sacroiliitis. In addition to radiographic damage of the sacroiliac joints specifically, axSpA patients may develop structural damage to the spine, known as syndesmophytes.43 Prior to being visible on X-Ray, inflammation in the sacroiliac joints or spine may be present that is visible on MRI,44 and this inflammation is associated with progression to AS.45 Recent research has begun to characterize the differences in the clinical characteristics of patients with and without radiographic damage and/or inflammation visible on MRI, the rate at which patients may progress from non-radiographic to radiographic disease, and how treatment response may differ between AS and nr-axSpA patients.46 Currently, there are important outstanding questions regarding how patient outcomes, including economic outcomes, may be predicted by radiographic and MRI findings, independently and in interaction with other clinical variables.     16 1.3.3 Inflammatory Markers ('Acute Phase Reactants')  As noted, the inflammatory markers CRP and ESR have been shown to be associated with SpA disease activity.31, 32 Both CRP and ESR are widely used biomarkers in SpA clinical trials and practice, although CRP may be the most common.47 The cut-off value for normal CRP is < 6 mg/L and the cut-off value for normal ESR < 10 mm/h.48 Elevated CRP or ESR is present in only approximately 40–50 % of patients with AS, meaning these measures have limitations in terms of detecting disease activity.47 However, elevated inflammatory markers have been shown to be associated with radiographic sacroiliitis, 49 sacroiliiac inflammation on MRI48, and with response to certain treatments.50 In a comparison of AS and nr-axSpA patients, significantly higher CRP values were found in AS patients.51 Inflammatory markers are in some cases used to guide treatment recommendations, as discussed further in section 1.4.  1.3.4 Patient-Reported Disease Activity: BASDAI  Disease activity among SpA patients is commonly measured using the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), part of the wider Bath outcome measurement system in AS.52 The BASDAI is a self-reported measure and is therefore considered a PRO.53 The BASDAI includes six questions regarding the symptoms fatigue, spinal pain, peripheral arthritis, enthesitis, intensity of morning stiffness, and duration of morning stiffness (Table 1.9).39             17 Table 1.9 SpA Disease Activity and Outcome Measures of Focus Domain Measure of focus Peripheral arthritis and enthesitis Presence of peripheral arthritis yes vs. no Presence of enthesitis yes vs. no Extra-articular manifestations of disease Presence of extra-articular manifestations yes vs. no If yes, type: Uveitis yes vs. no;  Psoriasis yes vs.no; Pustulosis yes vs. no; Inflammatory bowel disease yes vs. no Radiographic findings Presence of definite sacroiliitis yes vs. no (left sacroiliac joint) Presence of definite sacroiliitis yes vs. no (right sacroiliac joint) MRI findings Presence of inflammatory lesions visible on MRI yes vs. no (spine) Presence of inflammatory lesions visible on MRI yes vs. no (left sacroiliac joint) Presence of inflammatory lesions visible on MRI yes vs. no (right sacroiliac joint) Inflammatory Markers CRP, where normal <10 mm/h Patient-Reported Disease Activity  BASDAI39: 1) How would you describe the overall level of fatigue/tiredness you have  experienced? 0='none' and 10= 'very severe'   2) How would you describe the overall level of AS neck, back or hip pain you have had?  3) How would you describe the overall level of pain/swelling in joints other than neck, back, hips you have had? 4) How would you describe the overall level of discomfort you have had from any areas tender to touch or pressure? 5) How would you describe the overall level of morning stiffness you have had from the time you wake up? (from 0 hours to 2 or more hours) 6) How long does your morning stiffness last from the time you wake up?   Q1-5 scored on scale 0='none' and 10= 'very severe', Q6 scored on scale from 0 hours to 2 or more hours. Total BASDAI= Q1+Q2+Q3+Q4+(Q5+Q6/2)/5, where active disease ≥4. Patient-Reported Functional Status and Disability BASFI39: "Please indicate your level of ability with each of the following activities during the past week",  where 0 indicates no problems performing the task and 10 indicates the task is impossible.  1) Putting on your socks or tights without help or aids (e.g sock aid). 2) Bending from the waist to pick up a pen from the floor without aid. 3) Reaching up to a high shelf without help or aids (e.g helping hand). 4) Getting up from an armless chair without your hands or any other help. 5) Getting up off the floor without help from lying on your back. 6) Standing unsupported for 10 minutes without discomfort       7) Climbing 12-15 steps without using a handrail or walking aid.      8) Looking over your shoulder without turning your body. 9) Doing physically demanding activities (e.g physiotherapy exercises, gardening or sports). 10) Doing a full day’s activities whether it be at home or at work. Patient's Global Assessment of Disease Activity BASG54:  "How have you been over the last week?", where very good=0, very bad=10     18 Domain Measure of focus Physician's Global Assessment of Disease Activity According to physician, where 1=inactive disease, 10=active disease Health-Related Quality of Life SF-36 questionnaire Work Productivity Loss "How many days of work did you lose as a result of your spondyloarthritis over the last year?"   Active disease is defined as a BASDAI total score ≥4. Active disease on the BASDAI is often used as a criterion for treatment and for patient entry into clinical trials, although it has been noted that a BASDAI score of ≥4 may not be meaningful at the patient level.55 Research conducted to evaluate what BASDAI score represents the 'patient acceptable symptom state' (PASS), defined as the highest level of symptoms that a patient considers acceptable, suggests that a BASDAI score of 4.1 (95% confidence interval (CI) 3.8–4.4) is acceptable to 75% of patients.55  1.3.5 Patient-Reported Functional Status and Disability: BASFI  Self-reported functional status and disability among SpA patients is commonly measured using the Bath Ankylosing Spondylitis Functional Index (BASFI)56 considered to be the most widely-validated PRO in use in SpA assessment.53The BASFI measures patients' ability to perform ten daily tasks on a scale of 0 to 10, where 0 indicates no problems performing the task and 10 indicates the task is impossible (Table 1.9). The total BASFI score is the mean of the 10 item scores completed on a numerical rating scale. Functional impairment is defined as a BASFI total score ≥4.39 Research surrounding the PASS level on the BASFI scale indicates that a score of 3.8 (95% CI 3.5–4.1) is acceptable to 75% of patients.55   1.3.6. Patient's and Physician's Global Assessment of Disease Activity  Included in the ASAS core set of outcomes is that referred to as the Patient's Global Assessment of Disease Activity (PtGA).39  This is a PRO measured with a single question- 'how active was your spondylitis on average during the last week?'- on which the patient is asked to provide a response on a numerical rating scale (NRS) or visual analog scale (VAS) (a horizontal scale on which health is represented visually in a linear fashion from worse to better). The Bath   19 Ankylosing Spondylitis Global Index or BASG also contains a question representing a PtGA over the last week ('How have you been over the last week?' where 0=very good, 10=very bad).54 In addition to the PtGA it is also common for physicians to complete a Physician's Global Assessment (PhGA) of the patient's disease activity from the provider perspective.53 On both the PtGA and PhGA scales, the threshold indicating high disease activity has been defined as >6/10 on a 10 cm VAS scale, while low disease activity has been defined as <4/10 on a 10 cm VAS scale.57 Patient and physician perspectives on global disease activity have been shown to differ, with each being informed by different variables.58, 59 Specifically, the PtGA appears to be primarily informed by function, largely captured by the BASFI, while the PhGA appears to reflect objective clinical measures, such as CRP and ESR, number of swollen joins, and physical examinations of mobility.58 Consistent with these findings, the PtGA and PhGA have been found to be only moderately correlated (0.47), with the PtGA correlating highly with the BASDAI (r=0.71) and the PhGA correlating only moderately with the BASDAI (r=0.44).57 As the PtGA and PhGA reflect different perspectives, it is common for these instruments to be used in conjunction to describe SpA patients' disease activity.  1.3.7 Health-Related Quality of Life   One of the most important PROs assessed in SpA research studies is HRQoL; as it also comes into focus in economic evaluations, this outcome is described here in some detail.   The term HRQoL is used to refer to multiple aspects of patients' subjective experience, including both physical and emotional function, that result specifically from health status.60 It is now understood that many of the clinical outcomes traditionally measured in SpA research fall short of capturing the multi-dimensional impact of disease on a patient's HRQoL.61 In fact, HRQoL researchers have shown that clinical outcomes that are of central interest to clinicians may be of limited interest to patients themselves; furthermore, clinical outcomes may affect individuals in different ways, meaning that patients with the same clinical values may have different experiences in terms of overall well-being.62 As a result, it is now argued that researchers should evaluate outcomes that patients report as important, rather than just those considered to be important by clinicians.63-65  Instruments to measure HRQoL aim to help accomplish this goal,   20 and can be used to detect differences in HRQoL between patients or to detect changes in HRQoL in a single patient over time.62    An overview of HRQoL measurement has been provided by Guyatt,60, 62 who notes that measurement of HRQoL may be done using either specific or generic instruments. Specific instruments focus on differences or changes in HRQoL that result from a particular disease, population, or aspect of health status that is of interest in the research context.60, 62 A number of disease-specific instruments for measuring HRQoL among patients with AS have been developed,66 including the Ankylosing Spondylitis Quality of Life questionnaire (ASQoL),67 the Evaluation of Ankylosing Spondylitis Quality of Life,65 the Ankylosing Spondylitis Arthritis Measurement Scales,68 and the Patient Generated Index- Ankylosing Spondylitis.69 Each of these validated instruments focuses on aspects of HRQoL that are most relevant to patients with AS specifically. As an example, the ASQoL questionnaire includes questions on sleep, mood, motivation, coping, activities of daily living, independence, relationships, and social life; proponents of this questionnaire note that it captures important information on limitation of activities and participation that are not measured by other instruments.70 By focusing on aspects of HRQoL that are most relevant to patients with the disease in question, disease-specific HRQoL instruments  have the advantage of greater responsiveness, i.e., greater ability to detect changes in patients over time.71 However, a disadvantage of disease-specific instruments is that they do not allow for comparison of HRQoL across different patient populations, or for  comparison of a patient group to the general population.  In addition to specific instruments, HRQoL can be measured using generic instruments, designed for use with any population. There are two types of generic HRQoL instruments. The first type, referred to as 'health profiles', aims to measure HRQoL comprehensively, covering multiple domains of health. Two common health profiles are the EuroQoL-5D (EQ5D) and the Medical Outcome Study Short Form 36 Health Survey (SF-36). The EQ5D measures five domains of health, including mobility, usual activities, self-care, pain/discomfort, and anxiety/depression. A person taking the EQ5D is given a score in each of these five domains, as well a score rating their general health on a VAS.72 By comparison, the SF-36 measures eight domains of health, including physical functioning, role limitations due to physical functioning, bodily pain, general   21 health, mental health, role limitations due to emotional health, social functioning, and vitality. The SF-36 gives respondents a score in each domain, as well as a physical component summary (PCS) score and a mental component summary (MCS) score (both on scales of 0 to 100 where higher values represent better health).   An important advantage of using generic health profiles to measure HRQoL is that these instruments allow for a comparison of HRQoL across different populations. To make these comparisons, it is useful to know the instrument's standardized values among the general population, as well as the Minimally Important Difference (MID) on the instrument, defined as "the smallest difference in score in the domain of interest which patients perceive as beneficial and which would mandate, in the absence of troublesome side effects and excessive cost, a change in the patient's management".73 In the case of the SF-36, for example, mean PCS and MCS scores among the general population are 50 with a standard deviation of 1074, and the MID on the SF-36 PCS and MCS scales is between 3 and 5 points.75 With those values in mind, a meta-analysis of 38 studies evaluating HRQoL in AS patients using the SF-36 questionnaire76 demonstrates the impact of AS on HRQoL relative to the general population, showing that pooled mean PCS scores from all included studies of HRQoL among AS patients were 37.5 (95% CI 35.5-39.7), i.e., 12.5 units or 2.5-4 MID lower than the population mean; pooled mean MCS scores were 44.7 (95% CI 41.9-47.5), i.e., 5.3 units or 1-1.75 MID lower than the population mean.   The second type of generic HRQoL instrument is referred to as 'utility measures', which were developed in the fields of economics and decision theory.60, 62 The unique quality of utility measures is that they directly incorporate people's preferences for different health states, meaning utility scores represent not only a health state but how individuals tend to value that health state. Utility scores are on a scale from 0 to 1, where 0 represents dead and 1 represents perfect health, dead being the least preferred state on the scale and perfect health being the most preferred. While these scores can be derived using a number of methods (discussed further in section 1.5), one popular method is to obtain utility scores from the results of health profiles, such as the EQ5D and SF-36, by converting the results of those questionnaires into utility scores using validated algorithms.77 A number of studies have used this method to report health state   22 utility values among patients with SpA, most often using the EQ5D.78-88 In a review of such studies, Bansback et al. concluded that mean utility among AS patients is between -0.1 and -0.3 points below that of the general population.89   Utility measures do not provide information about a patient's status across the different domains of health, which is a distinct disadvantage over both specific HRQoL instruments and generic health profiles. However, utility scores have the advantage of representing overall HRQoL in the context of preferences, weighing both any positive effects of treatment minus any side-effects.60, 62 Most importantly, their unique properties, i.e., incorporation of preferences and scaling between 0-1, allow utility measures to be used as the quality adjustment weight in estimating quality-adjusted life years (QALYs) using mortality data.90 QALYs, which incorporate information about both quality of life and quantity of life, are a standard outcome measure used in economic evaluations, as discussed in detail in section 1.5.  1.3.8 Work Productivity Loss   Increasingly, studies focused on quantifying the burden of rheumatic disease include a measurement of work productivity loss, defined as reduced labour output due to poor health in the context of paid or unpaid work.91, 92 According to Zhang et al., the specific components of reduced labour output include presenteeism, absenteeism, and employment status changes such as reduced regular working time, job loss, and early retirement.91 As defined by Zhang et al., the term presenteeism refers to "reduced intensity and/or quality of labour input due to health problems while working", while the term absenteeism "commonly refers to the number of missed workdays for employed people", although it may also refer to missed work days that result from employment status changes.91 Specific concepts related to work productivity loss that have previously been measured among SpA patients include work participation,93, 94 work disability status,95 and sick leave.96 Work productivity outcomes, including their definitions and means of measurement, routinely vary across studies, and these concepts require close attention when comparing studies of work productivity loss.91 The use of a validated questionnaire that measures both absenteeism and presenteeism is recommended as the most rigorous method to evaluate work productivity loss and several such questionnaires exist.97 Like any instrument, work   23 productivity loss questionnaires require time and resources to administer, both on the part of participants and researchers. As a result, clinical research studies may ask participants to report only their employment status and number of work days lost, information which allows for a more limited, yet still useful, analysis of work productivity loss among patients.  Work productivity loss be described either in monetary or non-monetary terms. The latter approach, taken in non-economic studies, helps describe the disease burden on patients, while leaving it to interpreted from more than one perspective. For example, a recent systematic review described work productivity loss in non-monetary terms including 'unfavourable work status', sick leave, and presenteeism.98 Unfavourable work status (defined as full/partial work disability, permanent work disability, unemployment, withdrawal from the work force, or inability to work) ranged from 13% of AS patients reporting permanent work disability status99 to 61% of AS patients reporting receiving full or partial disability pension.100 Sick leave ranged from a mean of 6 days a year101 to 69 days a year,102 with the proportion of patients with at least one sick leave episode per year ranging from 40% to 69%.102 Presenteeism, i.e., reduction in at-work productivity relative to normal, ranged from a 6% reduction 103 to a 19% reduction 104; a separate study using a different measurement method found a 41% impairment in at-work productivity.105   Studies describing work productivity loss among SpA patients in non-monetary terms give a general indication of the burden of the disease on a patient's work life, supporting the view that improving the patient's ability to work is an important goal of SpA treatment. In economic studies, a monetary value is attached to work productivity loss outcomes, with the goal of interpreting the impact of work productivity loss from an economic perspective specifically.  Such studies often go on to compare work productivity loss among patients receiving different treatments, taking into account the effect of treatment on patients' ability to work. Economic studies of work productivity loss whose aim is to describe outcomes in monetary terms are discussed in section 1.5.       24 1.4 SpA Treatment Recommendations  The above sections have reviewed common disease activity and outcome measures in SpA research, including clinical assessments, HRQoL, and work productivity loss. In general, it is agreed that to improve these outcomes is the goal of SpA treatment.1, 106, 107 SpA treatment recommendations fall under the definition of clinical practice guidelines, which are defined as ''systematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances".108 In this section, which discusses SpA treatment recommendations, the terms 'recommendations' and 'guidelines' are used interchangeably.  Treatment recommendations are created by panels of experts based on the relevant health outcomes data available to guide decision-making.108 This process is a reflection of broader initiatives to use research evidence to improve the quality and efficiency of health care, promoted within the overlapping paradigms of evidence-based medicine (EBM),109 health technology assessment (HTA),110 and, more recently, CER.111  The latter  paradigm is defined by the US Institute of Medicine (IOM) as "the generation and synthesis of evidence that compares the benefits and harms of alternative methods to prevent, diagnose, treat, and monitor a clinical condition or to improve the delivery of care",111 and is distinguished by its emphasis on generating evidence in real-world settings and on identifying individual patient characteristics that influence outcomes.  In addition to health outcomes data, expert panels responsible for developing treatment recommendations may also consider economic evidence, although this is not currently routine practice.112 In fact, the current SpA treatment recommendations consider only clinical evidence and do not incorporate findings related to the economic impact of treatments. The implications of this are discussed in the concluding portion of this section, which serves to introduce the topic of economic evaluation and its relationship to CER.   1.4.1 General Treatment Recommendations  International AS treatment recommendations, developed based on an extensive literature review113 and applicable to all patients with axSpA, were originally published by ASAS in   25 2006114 and first updated in 2010;1  the 2010 ASAS recommendations specifically are discussed in the current thesis. 1  In addition to the ASAS recommendations, SpA treatment recommendations based on extensive literature reviews have been published by the American College of Rheumatology (ACR)106 and the Canadian Rheumatology Association (CRA), respectively.107, 115 As noted in a recent review,116 there is substantial agreement between SpA treatment recommendations. All sets of recommendations describe the SpA treatment paradigm as consisting of non-pharmacological treatment, pharmacological treatment including both symptom-modifying and disease-modifying agents, and sometimes surgery. One difference between published sets of recommendations is the system used to rate the quality of evidence in support of or against specific treatments, and to rate the strength of the recommendation given. For its part, ASAS does not rate the quality of evidence directly but rather rates only the strength of the recommendation, using a continuous scale from 0 to 10 along with a confidence interval. The ACR rates the quality of evidence using a scale from high, moderate, low, to very low, as well as the strength of the recommendation using the percentage of voting rheumatologists in favour of the recommendation. The CRA rates the quality of evidence using a custom scale based on the type of study from which the evidence was derived, but not taking into account the quality of the study, and rates the strength of the recommendation from strong to weak based on the quality of evidence. The specific SpA treatments described below are supported by all three sets of recommendations.106, 107, 115, 116 For ease of interpretation, just the ACR's quality of evidence and strength of recommendation ratings are noted unless there is clear disagreement between ASAS, the ACR, and CRA.   1.4.1.1 Non-Pharmacological Treatment  The main non-pharmacological treatment for SpA is exercise. This includes both home exercise and supervised physiotherapy, although all recommendations favour the latter. The ACR rates the quality of evidence in favour of physiotherapy as moderate, however given the low risk of                                                 1 ASAS published updated treatment recommendations in 2017,{{6967 van der Heijde,D. 2017;}}  which contain notable differences from the 2010 recommendations discussed in this thesis. These differences are discussed briefly in the final chapter of the thesis.   26 harm, the treatment is strongly recommended (100% voter agreement). A Cochrane review on the effects of physiotherapy in AS that was taken into account by recommendations concluded that moderate, positive effects of physiotherapy have been demonstrated only with respect to spinal mobility and self-reported physical function specifically, with no effect on pain, stiffness, or PtGA.117 Beyond physiotherapy, other recommended non-pharmacological interventions for SpA include patient education and self-management programs. The ACR rates the quality of evidence in favour of formal group or individual self-management education as moderate, given some indication of small, positive effects on BASDAI, BASFI, pain, fatigue, stiffness, and selected SF-36 scales (91% agreement).  1.4.1.2 Pharmacological Treatment  1.4.1.2.1 Non-Steroidal Anti-Inflammatory Drugs   The first-line pharmacological treatment for SpA are non-steroidal anti-inflammatory drugs (NSAIDs), including both Cox-1 inhibitors (e.g., ibuprofen, naproxen, diclofenac), and Cox-2 inhibitors (e.g., celecoxib, etoricoxib) that are effective in reducing back pain and stiffness.118 The ACR's quality of evidence rating with respect to NSAID treatment is low, however the strength of the recommendation is high (100% voter agreement). A 2010 meta-analysis taken into account by recommendations measured NSAID effect sizes reported in AS trials conducted from 1966 to 2009, with effect sizes calculated as the standardized mean difference, equal to the mean effect of NSAIDs minus the mean effect of placebo divided by the pooled standard deviation.119 With effect sizes between 0.5 and 0.8 and effect sizes being classed as medium and large, respectively, a large effect size was found for pain and for PtGA, while a medium effect size was found for physical function; NSAIDs had an insignificant effect on mobility. For patients with active disease, recommendations support continuous rather than episodic treatment with NSAIDs, although note that the cardiovascular, renal, and gastrointestinal risks of NSAIDs are to be taken into consideration.1 No particular NSAID is recommended over another by any of the recommendations. Where pain persists despite the use of NSAIDs, the ASAS and CRA recommendations support the use of analgesics, including acetaminophen and opioids, as a second-line symptom-modifying therapy; the ACR does not address the use of analgesics.   27 1.4.1.2.2 Anti-TNF Agents  All recommendations agree that SpA patients with persistently active disease should receive therapy with an anti-tumor necrosis factor (anti-TNF) agent. Anti-TNF agents are a class of biological drugs (i.e., made from living cells, not synthetic chemicals) that block TNF-alpha, a cell signaling protein known to mediate the inflammatory process in arthritis.120 Anti-TNF agents, including infliximab, adalimumab, etanercept, certolizumab pegol, and golimumab, emerged as treatment for SpA around the year 2000 and since then have been proven clinically effective in several trials.121, 122 A 2015 Cochrane meta-analysis of anti-TNF therapy in AS patients,122 which included 21 randomized controlled trials of maximum 24 weeks duration encompassing 3308 patients, reported that patients receiving an anti-TNF agent were 3 to 4 times more likely to achieve an ASAS 40 response123  than those receiving placebo, indicating significant positive effects on PtGA of disease activity, pain, function, and inflammation. The ASAS recommends anti-TNF therapy specifically for patients who fulfill either the mNY criteria for AS or the ASAS criteria for axSpA, who have had active disease for 4 or more weeks, an adequate therapeutic trial of NSAIDs (defined as at least two NSAIDs over a 4 week period in total at the maximum recommended or tolerated dose), and a positive expert opinion.1  It should be noted that ASAS does not describe in quantitative terms the criteria that should inform the expert opinion, stating only that "expert opinion should consider clinical features (history and examination) as well as either serum acute phase reactant levels or imaging results, such as radiographs demonstrating rapid progression or MRI scans indicating inflammation."124 Like ASAS, the ACR recommends anti-TNF agents for patients with active disease (which it defines as "disease causing symptoms at an unacceptably bothersome level as reported by the patient, and judged by the examining clinician to be due to SpA") who have a lack of response or intolerance to a minimum of two different NSAIDs over one month or incomplete responses to a minimum of two NSAIDs over two months.106 In contrast, the CRA recommends anti-TNF therapy for patients with active disease (which it defines by two of three clinical features: BASDAI ≥4, elevated acute phase reactants, inflammatory lesions in the sacroiliac joint and/or spine on MRI) despite NSAID treatment.115 The ACR rates the quality of evidence in favour of anti-TNF therapy as being moderate, with 80% of voters agreeing with the recommendation. Notably, these ratings could be interpreted as being lower than those given by the CRA, which   28 gives top ratings to both quality of evidence and strength of recommendation. The ACR performed its own review of evidence supporting the use of anti-TNF agents, which included 13 clinical trials of 24 weeks duration. The CRA cites a previous metaanalysis of data from 2005-2009125 (led by ASAS) which concluded that evidence was "high" and which found that anti-TNF were similarly effective in AS and nr-axSpA.   Table 1.10 2010 Update of the International ASAS Recommendations for the Use of Anti-TNF Agents in Patients With Axial Spondyloarthritis1    Diagnosis Patients fulfilling modified New York criteria for definitive ankylosing spondylitis or the ASAS criteria for axial SpA Active disease Active disease for ≥4 weeks BASDAI ≥4  (0–10) and positive expert opinion Treatment failure All patients: should have had adequate therapeutic trial of at least two NSAIDs; defined as at least two NSAIDs over a 4-week period in total at maximum recommended dose unless contraindicated; Axial disease: no pretreatment with DMARDs required Peripheral arthritis: one local corticosteroid injection if appropriate; should normally have had a therapeutic trial of a DMARD, preferably sulfasalazine Enthesitis: appropriate local treatment Contraindications Refer to annually updated consensus statement on biological agents Assessment of disease ASAS core set for daily practice and BASDAI Assessment of response 50% Improvement in BASDAI or absolute change of 2 (0–10) and positive expert opinion in favour of continuation Assessment after at least 12 weeks   1.4.1.2.3 Other Agents  Pharmacologic agents other than NSAIDs and anti-TNF agents are generally not recommended for treating SpA. Traditional disease-modifying anti-rheumatic drugs (DMARDs), for example, such as methotrexate (MTX) and sulfasalazine (SSZ), appear to have a limited role in axSpA treatment. These agents have been shown to be effective in treating peripheral arthritis among SpA patients but there is little evidence to suggest they are effective in treating axial manifestations of the disease.106 The ACR conditionally recommends against the use of DMARDs in treating SpA, rating the quality of evidence against their use as very-low to   29 moderate, depending on the specific DMARD, with 90% voter agreement in recommending against DMARDs generally. The recommendation that ASAS gives against the use of DMARDs could be interpreted as even firmer, with the strength of the recommendation rated as 9.4 ± 0.2. Beyond DMARDs, both ASAS and the ACR recommendations also advise against the systemic use of glucocorticoids such as prednisone and prednisolone. Although both ASAS and the ACR agree that local glucocorticoid injections may be considered for patients with peripheral disease, both note that the use of systemic glucocorticoids is not supported by evidence (ACR quality of evidence rating very-low, 100% voter agreement). In contrast, the CRA supports the use of local glucocorticoid injections and short courses of systemic glucorticoids for some manifestations of SpA.   1.4.1.3 Surgery  Potential surgical interventions for SpA patients include total hip arthroplasty and spinal corrective osteotomy. The statements in the ASAS recommendations on the topic of surgery are vague: it asserts that total hip arthroplasty "should be considered" in patients with refractory pain and disability and radiographic evidence of structural damage, while spinal corrective osteotomy "may be considered" in patients with severe disabling deformity. The strength of the ASAS recommendation is 9.2 ± 0.3. In comparison, the ACR strongly recommends total hip arthroplasty for adult AS patients with advanced hip arthritis (very-low quality evidence, 100% voter agreement). For adult AS patients with severe kyphosis, the ACR conditionally recommends against spinal osteotomy. The CRA notes that recommendations with respect to spinal surgery are derived from expert opinion based on little evidence.  1.4.1.4 Other Non-Specific Recommendations  In addition to specific treatments, the ASAS recommendations include a statement emphasizing that the extra-articular manifestations of SpA, such as psoriasis, uveitis, and IBD, should be managed in collaboration with relevant specialists (i.e., dermatologists, opthalmologists, gastroenterologists). The ACR mentions that iritis should be managed by an opthalmologist, while the CRA does not include a statement regarding the use of specialist care.   30 1.4.2 Recommendations Specific to Anti-TNF Access  In addition to the general SpA treatment recommendations made by ASAS, ACR, and CRA, numerous patient organizations and health systems worldwide have issued recommendations to guide anti-TNF use specifically.2 Among the 23 sets of recommendations reviewed by Van den Berg et al. (Table 1.11),2 some are clinical recommendations, while others are reimbursement criteria, which designate for which patients the cost of anti-TNF therapy will be covered by the government. Each of these sets of anti-TNF recommendations differs in terms of the diagnosis, disease activity level, and history of treatment failure that is advised or required before anti-TNF therapy is considered appropriate. In terms of diagnosis, for example, some guidelines approve of anti-TNF therapy for AS patients only, i.e., requiring the mNY criteria to be met, while others allow for their use in nr-axSpA patients. Where guidelines allow for anti-TNF use among nr-axSpA patients, the definition of nr-axSpA often differs, with some bodies demanding satisfaction of either the ASAS, Amor, or ESSG classification criteria (e.g., Colombia), others the ASAS criteria specifically (e.g., Korea, Mexico), and still others allowing for various adaptations to these criteria, for example permitting the use of MRI or computerized tomography (CT) evidence instead of X-Ray (e.g., France) (Table 1.11). With respect to disease activity, all recommendations use the BASDAI cut-off of ≥4 to indicate high disease activity, however some demand BASDAI alone (e.g., Korea), others demand BASDAI plus some combination of additional markers (e.g., Spain), while others leave space for BASDAI to be excluded (e.g., Poland, which requires two of three criteria be satisfied: BASDAI ≥4, pain on VAS ≥4, CRP >10 mg/dl over 12 weeks while on treatment).           31 Table 1.11 International Recommendations for Diagnosis, Disease Activity, and Treatment Failure Before Use of Anti-TNF Agents in SpA Described in 2011.2 Region Country Diagnosis Disease activity Treatment failure Asia Australiaa Sacroiliitis (X-ray) Grade II bi- or Grade III unilateral BASDAI ≥ 4 and abnormal lab tests (ESR > 25, CRP > 10) In the preceding 3 months: two different NSAIDs and a specified exercise programme (both stretching and daily aerobic exercise) Asia Hong Konga Modified NY criteria or ESSG Persistent active disease (BASDAI and patient and physician GH) and ESR ≥ 50 mm/h and CRP ≥ 50 mg/l Three NSAIDs (different chemical classes), ≥4 weeks each and ≥2 DMARDs (SSZ/MTX/Arava) ≥ 3 months (peripheral joint) Asia Koreaa According to ASAS According to ASAS Two DMARDS or NSAIDS, 3 months Americas Canadab Expert opinion and unequivocal evidence of sacroiliitis or spinal inflammation on X-ray/CT/MRI Two out of the 3 following: BASDAI ≥ 4 and ↑ CRP and/or ESR and inflammatory lesions SI joint and/or spine on MRI ≥2 NSAIDs 2 weeks and CS injections may be considered and SSZ ≥ 3 months in peripheral arthritis and MTX ≥ 3 months in peripheral arthritis Americas Colombiaa According to ASAS, or Amor or ESSG According to ASAS, duration not specified According to ASAS and >2 infiltration intra-steroids in peripheral arthritis and >2 CSs injections in enthesitis Americas Mexicob According to ASAS According to ASAS According to ASAS Europe Belgiuma Modified NY criteria and expert opinion BASDAI > 4 and elevated CRP Insufficient response on at least ≥2 NSAIDs, optimum dosage ≥3 months or contraindication for NSAIDs Europe Czech Repa,b According to ASAS, or MRI instead of X-ray BASDAI ≥ 4 and CRP ≥ 10 at two consecutive FU visits separated by ≥4 weeks According to ASAS Europe Finlanda Not mentioned Active disease not specified Two NSAIDs and MTX and SSZ 6–12 months and intolerability/lack of efficacy DMARDs Europe Franceb According to ASAS, or MRI/CT instead of X-ray or cervical syndesmophytes without any sacroiliac structure lesion According to ASAS and peripheral: TJC and SJC (≥3 of 76–78 joints) According to ASAS and ≥3 NSAIDs instead of two NSAIDs and enthesitis not specified Europe Germanya,b Secured diagnosis of  AS  According to ASAS and disease symptomatic ≥6 months  According to ASAS and enthesitis not specified  Europe Greecea Clinical and laboratory and radiological findings According to ASAS According to ASAS and MTX ≥ 2 months in peripheral arthritis and ≥2 topical infusions of CSs in enthesitis Europe Hungarya,b According to ASAS According to ASAS According to ASAS and ≥2 IA steroid injections, SSZ or other DMARD ≥4 months in peripheral arthritis  Europe Italyb According to ASAS According to ASAS According to ASAS Europe Netherlandsa,b According to ASAS According to ASAS According to ASAS                  32 Region Country Diagnosis Disease activity Treatment failure Europe Norwaya Conventional diagnosis Prescription from department with >2 specialists (rheumatologists) Approval based on disease history, previous treatment and current status - ESR/CRP, clinical status and imaging - Peripheral arthritis: joint counts and disease activity indices - Axial: BASDAI and BASFI - Both axial and peripheral: pain, fatique and globals on VAS (not specified) Axial: 2 NSAIDs and peripheral arthritis: DMARD (pref. SSZ) If relevant: IA steroidal injections Europe Polandb Conventional diagnosis Prescription from department with >2 specialists (rheumatologists) Approval based on disease history, previous treatment and current status - Two of three following parameters: BASDAI ≥ 4, pain VAS ≥ 4, CRP > 10 mg/dl in 12 weeks interval on stable treatment - One or more parameters limited in 1-month interval: chest expansion, occiput-to-wall distance, Schober test According to ASAS and >2 IA glucocorticosteroids injections in peripheral arthritis Europe Portugalb According to ASAS, or MRI/CT instead of X-ray  According to ASAS and in case of BASDAI < 4: expert opinion  According to ASAS and 4 weeks instead of 3 months  Europe Slovakiaa,b According to ASAS, or MRI instead of X-ray  BASDAI ≥ 4 and CRP ≥ 10 at two consecutive FU visits separated by ≥4 weeks  According to ASAS Europe Spainb Expert opinion BASDAI > 4 and one out of three: spinal pain or patient global assessment (VAS >4) or ↑ ESR/CRP, for > 3 months  According to ASAS and any DMARD (pref. SSZ) instead of only SSZ  Europe Swedenb According to ASAS According to ASAS According to ASAS Europe Switzerlanda Expert opinion Expert opinion Expert opinion and mandatory consent of the consultant physician of the health insurance company  Europe UKa,b According to ASAS BASDAI > 4 and spinal pain (VAS > 4) on two occasions >4 weeks apart and no change in treatment  >2 NSAIDs for 4 weeks   Reproduced from van den Berg R, Stanislawska-Biernat E, van der Heijde DM. Comparison of recommendations for the use of anti-tumour necrosis factor therapy in ankylosing  spondylitis in 23 countries worldwide. Rheumatology. 2011 Dec;50(12):2270-7. Note: some inconsistencies are apparent between the criteria cited in this published table and criteria included in original published recommendations (R. Van den Berg, personal communication by email, May 22, 2015). The study in Chapter 5 incorporates criteria from original published recommendations and cites the primary reference.  Another key difference between the recommendations in terms of disease activity is whether or not they require elevated acute-phase reactants, and if so whether only CRP is accepted (e.g.,   33 Czech Republic) or whether CRP or ESR may be considered (e.g., Australia). Finally, all anti-TNF recommendations reviewed by Van den Berg2 include a statement as to what evidence of treatment failure should be in place prior to anti-TNF use, with both the recommended number of NSAIDs and the period of use varying between sets. Several countries require that the patient try 2 NSAIDs for a total period of 4 weeks, following the ASAS recommendations; others increase the number of NSAIDs to 3 (e.g., France), while others increase the period of use to 3 months (e.g., Belgium).  1.4.3 SpA Treatment Recommendations: Considerations  As shown, there are many similarities between the ASAS, ACR, and CRA general SpA treatment recommendations, just as there are overlapping criteria across different anti-TNF use recommendations in place worldwide. This reflects some degree of consensus as to the mainstays of SpA treatment, and to principles guiding anti-TNF use. Nonetheless, there are some important differences in how evidence has been interpreted by ASAS, ACR, and CRA, respectively. In general, differences between treatment recommendations for the same condition reflect limitations in the quality of their underlying evidence.126 Indeed, it stands to reason that the more robust the data on a particular treatment, the more consistent the interpretation of those data and the more uniform their transformation into treatment recommendations. Where data are lacking or of inferior quality, their interpretation can be expected to be more variable, and their implications for treatment recommendations less obvious.   As is made clear by the ACR recommendations- in which the highest quality evidence rating granted is 'moderate'- there are important data limitations surrounding every aspect of SpA treatment.  Particularly in the case of anti-TNF agents- in use in SpA for little over fifteen years at time of writing- the body of evidence to inform recommendations is incomplete. Currently, various bodies worldwide recommend the use of anti-TNF agents in patients with high disease activity, but use alternate definitions of high disease activity, meaning patients in different settings do not have equal access to anti-TNF agents. Furthermore, there are differences in how experts have assessed the quality of evidence in favour of anti-TNF agents. For example, the CRA suggests that anti-TNF agents have similar efficacy in both AS and nr-axSpA patients,107   34 yet this differs from the opinion of the United Kingdom's (UK) National Institute for Health and Care Excellence (NICE), which concluded that on data on anti-TNF efficacy in nr-axSpA are too heterogeneous to compare to AS.121   There is a strong argument to suggest that treatment recommendations require their own evaluation in terms of their impact on patient health outcomes.127 For example, treatment recommendations often incorporate 'expert opinion' to fill the gaps left by incomplete evidence,127 and they can be shown to reflect the values of those involved in the development process.128 As a result, treatment recommendations should be seen as an entity unto themselves, a distinct part of evidence-based health care. To date, a number of studies have focused on the impact of treatment recommendations on patient health outcomes129-132 but none in the field of SpA. Importantly, to evaluate the impact of recommended care, methods are needed to measure adherence to treatment recommendations. It should be noted that, in this thesis, adherence to clinical guidelines refers to the concept of following the recommended practices described therein. This concept of adherence is distinct from other commonly studied forms of adherence, such as patient-level adherence to prescribed drug regimens. Throughout this thesis, adherence to SpA clinical guidelines is a matter of focus, and the term 'adherence' refers exclusively to the concept of following treatment recommendations. The phenomenon of patient-level adherence to prescribed drug regimens is not discussed in this thesis.  Adherence to clinical guidelines may vary as a result of practices on the part of the physician, the patient, or both. Studies that measure adherence to clinical guidelines may measure the reasons for adherence, or they may measure only adherence itself without addressing underlying causes; the latter applies to this thesis. To measure adherence itself, the process may begin with the development 'quality indicators', based directly on clinical guidelines or on the evidence used to inform the guidelines.132-134 Quality indicators allow for the measurement of adherence by defining the eligible patient and the care process that the patient should receive, often in the form of an 'if, then' statement.133 Although quality indicators have been developed in the areas of RA134 and osteoarthritis,133 there appears to be a lack of methods developed to measure adherence to clinical guidelines in SpA.     35 Not only is it possible to evaluate the impact of SpA treatment recommendations on health outcomes, it is possible to evaluate their impact on economic outcomes. Importantly, although some clinical guidelines are beginning to take costs into account,112 the current SpA treatment recommendations weigh only the clinical evidence and neglect to consider the economic impact of recommended treatments. Increasingly, it is argued that economic evidence should be taken into consideration alongside clinical evidence when developing treatment recommendations,135, 136  a view that acknowledges that treatment choices will have an economic impact and potentially divert resources away from other activities. The incorporation of economic data into the development of SpA treatment recommendations is thus a distinct goal, to be furthered through economic evaluation of SpA treatments on an individual basis. A related, yet unique goal is to evaluate the economic impact of treatment recommendations themselves, examining patient outcomes while taking into consideration to what extent the recommendations were adhered to. In all economic evaluations, the purpose to inform decision-making, a purpose it shares with CER.137 As described further below, CER and economic evaluation have not only shared goals, but shared methods and shared imperatives, including the use of data from observational studies in order to describe real-world outcomes.  1.5 Overview of Economic Evaluation   In the healthcare context, there are more ways to spend money than there are resources available. The concept of an 'opportunity cost' suggests that the true value of a treatment is not simply its monetary value, but rather the value of whatever program will not be made available to society because the treatment is funded instead. The concept of the opportunity cost is central to economic evaluation in health, a form of analysis that compares the costs and effects of two or more alternatives. As economic evaluations must collect information on comparative effects, there is an important methodological overlap between economic evaluation and CER, although economic evaluation may or may not be considered a part of CER.138  Numerous sources have published methodological guidelines for the conduct of economic evaluations in healthcare, focusing on the handling of cost data.139 In this section, unless otherwise specified, the source of the information on economic evaluation is a definitive, often-cited textbook by Drummond that reflects a general consensus on the topic.140    36 In economic evaluations, the alternatives for comparison may be defined as a discrete treatment or intervention (e.g., a drug, a diagnostic procedure), a complex health program involving more than one element of care (e.g., a comprehensive recommended management strategy), or even the absence of treatment. In fact, 'no treatment' is a common comparator in economic evaluations, as when the costs and effects of treating a disease with a drug are compared to the costs and effects of not treating the disease. In all of these cases, economic evaluations assume that policymakers must choose one alternative over another, and that this choice should be informed by knowledge of the opportunity cost. Importantly, although the term economic evaluation implies a comparative analysis of both the costs and effects of two or more alternatives, by virture of their methodological pathway, economic evaluations often produce preliminary results that are purely descriptive, focusing only on the costs of a given treatment. In the health economics literature, descriptive cost studies (sometimes called 'partial' economic evaluations, in contrast to 'full', i.e., comparative, evaluations, as described by Drummond) have their own place and purpose, such as in the case of cost-of-illness (COI) studies that aim to describe the economic burden of a disease. The following section describes the basic methods of economic evaluation that underlie the studies undertaken as part of this thesis, describing the necessary data and corresponding potential data sources, the steps involved in a descriptive cost study, and the methods of comparative analysis. The section concludes by reviewing selected findings from existing economic evaluations in SpA, which serves to contextualize the thesis and the economic questions which it aims to address.   1.5.1. Basic Costing Methods   The same basic methods of cost analysis underlie all economic evaluations. A first requirement is to obtain data on the costs associated with each of the treatment alternatives under consideration. One fundamental methodological issue in the conduct of economic evaluations, therefore, concerns the study's perspective, as this determines which costs are to be identified and measured. One option is to conduct the evaluation from the perspective of a specific entity, such as the public healthcare system or a private insurer, in which case only the costs covered by the entity in question are included.141 The other option is to conduct the evaluation from the societal perspective, and thereby aim to identify and include all costs associated with the   37 treatment under consideration. This includes all health resource use costs, patient out-of-pocket costs (such as adaptive devices and informal caregiving), 'non-medical' but disease-related costs (such as transportation), as well as work productivity loss. Because they attempt to include all sources of costs associated with a disease, including work productivity loss (which may be substantial in the context of chronic disease91) economic evaluations conducted from the societal perspective will, as a rule, produce higher cost estimates compared to those conducted from the perspective of a single entity.   Once all relevant sources of costs have been identified in accordance with a study's chosen perspective, the next step is to measure them in appropriate units. Examples of units include a specific blood test, an MRI procedure for a specific body part, taxi fare over a specific distance to and from a medical appointment, a single day of work lost, etc.. Data on the frequency/quantity of such units may come from a multitude of sources, including population-based health administrative databases, patient charts, randomized controlled trials (RCTs) and observational studies. Some of these sources, such as health administrative databases, represent routinely collected data, and are not affected by the problems associated with self-report data, such as recall bias. However, routinely collected data will seldom include all variables of interest to researchers; for example, health administrative databases will generally not contain information on patient out-of-pocket costs or on work productivity loss. Conversely, other data sources, such as observational studies, provide self-report data that may be affected by factors such as recall bias, but support the use of specially designed questionnaires and/or patient diaries, facilitating the collection of comprehensive information on sources of costs.  Following the quantification of all sources of relevant costs in unit form, the next step in a cost analysis is to assign each unit a monetary value, which involves choosing an appropriate source of information on 'unit costs'. For practical reasons, it is common to use publicly available market prices for unit costs. In cases where market prices are not available, not applicable (e.g., unpaid caregiver time), or otherwise inappropriate (e.g., market prices are subsidized by a third party), researchers are advised choose a value they believe is justified from a theoretical perspective and to present the consequences of their choice. If the choice of unit cost has the potential to substantially affect the results, researchers should conduct a 'sensitivity analysis',   38 which is the term for any form of analysis designed to estimate the impact on overall study results of varying the value of a given parameter.142  With respect to assigning unit costs to work productivity loss specifically, some special theoretical considerations apply. The simplest method to value a single day of lost work is to use the individual's daily wage rate, or an average wage rate, multiplied by the number of days the individual is absent from work. This method, which corresponds to one of two main methods to measure the value of work productivity loss, is referred to as the Human Capital (HC) approach.143 The HC approach is a common, simple method of valuation, but it has been criticized for misrepresenting the true cost of work productivity loss; the argument is that a worker's absence may not generate added costs if another person, be it a colleague or an unemployed person, replaces them. In response to this criticism of the HC approach, an alternative was developed called the Friction Cost (FC) method, which assumes that when a person is absent from work, productivity is reduced only until they are replaced.144 Following the FC approach, work productivity costs are estimated by multiplying the chosen wage rate for an individual worker by the estimated number of days that it will take to replace that worker. By definition, the HC approach to valuing work productivity loss results in higher estimates than the FC approach. In fact, in a seminal study, Koopmanschap showed that productivity costs in the Netherlands were nearly nine times higher when estimated by the HC approach compared to the FC approach.144   Whether to use the HC or FC approach is but one of many special methodological issues that pertain to the valuation of work productivity loss, the appropriate methods for which are still debated.145 Some other important outstanding questions on this topic relate to whether to value presenteeism in addition to absenteeism, and whether to include the value of unpaid labour. Overall, the inclusion of work productivity loss in economic evaluations is supported by many, but not all, experts, and the arguments for and against their inclusion rest on theoretical foundations that are outside the scope of this thesis. Many current guidelines for conducting economic evaluations suggest that researchers include estimates of work productivity loss costs but report them separately,140, 145 so that decision-makers can consider the costs of a treatment both including and excluding work productivity. For as long as the best methods to value work   39 productivity loss are still being contested, researchers should describe their methodological choices in a transparent manner and discuss their meaning and implications on study results. Again, if the method used to value work productivity loss has the potential to substantially affect the results, a sensitivity analysis may be conducted to explore the implications of this.  1.5.2 Partial Economic Evaluations: Cost of Illness Studies   As noted, studies whose only goal is to describe the costs of a given treatment are not, strictly speaking, full economic evaluations. Nonetheless, purely descriptive, 'partial' economic evaluations are often performed, sometimes as precursors to more sophisticated economic analyses, other times to achieve their own limited, yet specific objectives. A valuable type of  partial economic evaluation is COI studies, whose objective is to describe the economic impact of a disease by calculating its associated costs. In general, COI studies can be useful in demonstrating which cost components are most important in a given disease, helping to predict costs in light of changing epidemiology or technological advances, and identifying priority areas for full economic evaluations.146 The methodological framework for COI studies has been described by numerous authors, including the World Health Organization,141, 147, 148 and involves the same basic methods of cost analysis common to all economic evaluations. In terms of more sophisticated methodological choices, there is a degree of overlap between COI studies and full economic evaluations, which stems from the choice of data source.  A fundamental methodological choice in COI studies is the study's epidemiological framework, which informs the timeframe for the analysis.141, 148 In COI studies taking an 'incidence-based' approach, the costs of new cases of a disease detected within a specified period are estimated over a lifetime, while in those taking a 'prevalence-based' approach, the costs of existing cases at a given point are estimated over a specified period. Prevalence-based studies are recommended for estimating the total current burden of a disease and often take the timeframe of one year.141 In prevalence-based studies, the specific the calendar year (or years) in which the costing process is carried out can have a large impact on cost estimates. This is due to the fact that evolution takes place in the development and use of health technologies and treatments. A vivid example of this comes from Poland, where the national payer began covering the cost of anti-TNF agents for AS   40 treatment in 2009. While in 2008 the total annual cost of AS treatment in Poland was approximately 6.3 million US dollars (USD), this figure rose to 10.5 million in 2009 and reached 37.6 million in 2013, highlighting the importance of the year of costing.149 As a result of its potential large effect on cost estimates, the calendar year of costing is an important methodological choice to report in the results of a COI study.   Another basic distinction between COI studies is the costing method, defined as either a 'top-down', i.e., population-based, or 'bottom-up', i.e., person-based approach to costing. Importantly, this methodological choice relates intrinsically to the choice of data source for costs, which, as noted, includes both routinely collected, administrative health data and data from specially-designed health research studies. Top-down COI studies use population-based, aggregate cost data, and use additional data linking those costs to specific diseases and/or procedures in order to attribute a fraction of total costs to a particular disease of interest.141 A limitation of top down COI studies is that the accuracy of their estimates will depend in part on the accuracy of data linking health resource use to a specific disease. Furthermore, in general, top-down COI studies do not include work productivity loss costs, as these data are not typically available in population databases; the top up approach can therefore seldom be used to capture all the societal costs of a disease. In comparison, bottom-up COI studies use individual-level data, in which patient case-report forms or specially designed questionnaires are used to collect data on patients' costs. These forms may be used to inquire as to any number of sources of costs, including work productivity loss, meaning COI studies conducted from the societal perspective generally use the bottom-up approach to costing.140 For the purpose of including the comprehensive costs associated with a disease and correctly attributing costs to a specific disease (rather than to, for example, a co-morbidity), the individual-level costing strategies of the bottom-up approach are considered to be superior.148  Another distinction between COI studies relates to the timing of data collection relative to the study itself.148 In a retrospective COI study, all costs have occurred and all data are available by the time the study is undertaken. In contrast, a prospective COI study is characterized by longitudinal follow-up of patients and the collection of cost data over time. Prospective COI studies, by nature, tend to be bottom-up studies and are thus characterized by the same   41 advantages, including the potential to collect a wide breadth of data and to focus on items of interest using specially designed questionnaires. One limitation of prospective COI studies is potential recall bias, as patients are asked to remember their experiences with health resource use and work productivity loss over a specified timeframe. However, this limitation of prospective studies is to be weighed against their advantages, including collecting highly specific, individual-level data that allows for changes in health resource use and work productivity loss over time.  In general, the costing methods for a COI study are selected in light of the study's context, purpose, and the availability of unit cost data. A common criticism of COI studies is that even those conducted from the same perspective on the same disease often include different health resource use components, use different methods to derive unit costs, and value different aspects of work productivity loss using different approaches.150 In an effort to address problems of comparability of COI studies in rheumatology, Merkesdal et al. 3 suggested a standardized list of 19 cost domains to be included in all such studies (Table 1.12). However, such standardized lists have not been widely adopted, and COI studies remain difficult to compare, limiting their usefulness in the view of many critics. Even more importantly, COI studies do not provide guidance as to how to optimize resource allocation, which is the most crucial shortcoming of COI studies.151  Table 1.12 Matrix of 19 Cost Domains3 Cost Domain 1. Health care costs (direct) 1.1 Outpatient costs 1.1.1 Visits to physicians (specialists and other) 1.1.2 Outpatient surgery 1.1.3 Emergency room visits 1.1.4 Non-physician service utilization (physiotherapist, occupational therapist, social worker, psychologist) 1.1.5 Medication 1.1.6 Diagnostic/therapeutic procedures and tests (radiological examinations, laboratory tests) 1.1.7 Devices and aids 1.2 Inpatient costs 1.2.1 Acute hospital facilities (without surgery) 1.2.2 Acute hosptial facilities (surgery) 1.2.3 Non-acute hospital facilities (rehabilitation, nursing homes) 2. Other disease related costs (direct) 2.1 Transportation 2.2` Home health care services 2.3 Home remodeling 2.4 Medical equipment (nonprescription)   42 Cost Domain 2.5 Non-medical practitioner, alternative therapy 2.6 Patient time 3. Productivity costs 3.1 Loss of productivity in employed patients (sick leave, work disability) 3.2 Opportunity costs (loss of productivity due to time spent by nursing family members, disabilities leading to impaired housekeeping or activities of daily life) 3.3 Lost wages  Reproduced from Merkesdal S, Ruof J, Huelsemann JL, Schoeffski O, Maetzel A, Mau W, Zeidler H. Development of a matrix of cost domains in economic evaluation of rheumatoid arthritis. J Rheumatol. 2001 Mar;28(3):657-61.  1.5.3 Full Economic Evaluations  In their standard text, Drummond et al. refer to three types of full, i.e., comparative, economic evaluations: cost-benefit analysis (CBA), cost-effectiveness analysis (CEA), and cost-utility analysis (CUA). All of these can be used to compare the costs and effects of two or more treatments, with each one producing results to indicate each treatment's cost per unit of effect, as well as its incremental cost per unit of effect relative to its comparator. However, each type of evaluation differs in the method it uses to value effects. It should be noted that, although Drummond, and other experts, speak of CUA as a distinct form of economic evaluation, other authors view CUA as a subtype of CEA and avoid using the term CUA; the reasons for this discrepancy are explained below.   In a CBA, effects are valued specifically in monetary form, meaning that both costs and effects are communicated using the same unit. The advantage of CBA is that the net benefit of a treatment can be easily calculated; in the case a treatment will 'pay for itself', this will be clearly demonstrated by a CBA. The disadvantage of a CBA lies in the inherent difficulty in assigning monetary values to health outcomes. Several methods have been developed in order to achieve this, including the HC approach (i.e., valuing healthy time using a wage rate), the revealed preferences approach (i.e., valuing healthy time based on observations of individuals' willingness to risk their health for money, or accepted monetary compensation for injury), and the willingness-to-pay (WTP) approach (i.e., valuing treatments based on the maximum price that individuals say they are willing to pay for them). However, each of these methods has been subject to criticism, including problems using wage rates to value healthy time given wage inequities, limitations in observed data on health risk tradeoffs, and the influence of individual   43 income on self-reported WTP values. While CBA can be a useful form of analysis, CEA and CUA may be used in order to avoid the problems that arise in assigning monetary values directly to health outcomes.  In a CEA, effects are measured in units relevant to the circumstances of the study. For example, in a study of SpA patients, a CEA could measure the benefit of a treatment in terms of its effects on function using the BASFI scale; alternatively, it could examine the benefit of treatment in terms of its effects on disease activity using the BASDAI scale, or on the PtGA or PhGA using a ten-point VAS. Each of these would allow for an estimate of each treatment's cost per unit of effect, as well as a direct comparison of the costs and effects of two treatments using the Incremental Cost-Effectiveness Ratio (ICER). The ICER is calculated as the cost of one treatment minus the cost of the other, divided by the effect of one treatment minus the effect of the other (i.e., CostTreatment1-CostTreatment2/EffectTreatment1-EffectTreatment2). The ICER is discussed in further detail below.  A CEA is an appropriate and useful metric when the purpose of a study is to compare two treatments that target the same outcome and when the value of a unit of improvement on the selected scale is implicitly understood by those using the results of the CEA. For example, in the case of studies using the BASFI as the outcome, the ICER would help inform a funding choice between two SpA treatments to be made by policymakers who understand the clinical significance of BASFI improvement. However, the disadvantage of CEA arises when comparisons are to be made between treatments that target different outcomes for improvement, or when the significance of the selected unit of effect is not intuitive to those using the CEA. For example, if one treatment aims to improve function on the BASFI, and another treatment aims to improve PhGA ratings, it would not be possible to compare the value of the two treatments using an ICER. Furthermore, not all policymakers understand the significance of the BASFI or of the PhGA. As a result, researchers have sought to develop a standardized outcome measure interpretable by policymakers that allows for a comparison of treatments developed for different purposes.    44 In a particular form of CEA, sometimes called a CUA, the benefits of a treatment are examined in terms of its effects on QALYs, which incorporate information about both the quality of life and quantity of life associated with a treatment strategy. In calculating QALYs, utility scores (on a scale of 0-1, from death to perfect health, as described) are multiplied by data on length of life in order to produce an adjusted measure that reflects both people's preferences for living longer and in better health states. For example, a person who receives a treatment that extends their life for one year, but whose health state following the treatment is associated with utility value of 0.5, is said to have gained 0.5 QALY (1 year x 0.5 utility). By comparison, another person who receives a treatment that extends their life for six months, but whose health state following the treatment is associated with a utility value of 0.9, is said to have gained 0.45 QALY (0.5 year x 0.9 utility). As a result of the utility weight of reflecting the preference for a better quality of life, two such people have nearly the same QALY gain post-treatment, despite the fact that the first person will live six months longer than the second. One of the distinct advantages of using QALYs as an outcome measure in a CEA is that the metric incorporates both people's preferences for living in specific health states and their preferences for living longer. An additional advantage of using QALYs as the outcome measure in economic evaluations is that QALYs can be calculated for any patient group following any treatment. This allows for economic evaluations to compare the value of any two treatments, including those targeting different intermediate outcomes within the same patient group, as well as those targeting different patient groups altogether.   Utility values used to calculate QALYs are often derived from HRQoL questionnaires, using validated algorithms to convert HRQoL data into utility scores. In fact, HRQoL questionnaires are called 'indirect methods' of measuring utility; the algorithms used to convert HRQoL data into utility scores are informed by previous studies that applied one or more 'direct methods' to elicit information on utility. For example, the algorithm that allows HRQoL data from the EQ5D to be converted into utility scores is informed by a study of approximately 3000 people, who were shown a subset of the various EQ5D health states and asked to describe the amount of life time they would be willing to sacrifice to live in perfect health rather than in a given health state (the 'Time Trade-Off' method). By comparison, the algorithm that allows data from the SF-36 to be converted into utility scores is informed by a study of over 800 people who were shown a   45 subset of the various SF-36 health states (called the SF-6D). These people were asked to choose between accepting life in a given health state or avoiding it by taking a gamble between perfect health and immediate death (the 'Standard Gamble' method, in which the risk level for the gamble between perfect health and death is varied as a means of eliciting information on utility). Despite the fact that different methods of eliciting utility underlie their scoring algorithms, the EQ5D and SF-36 questionnaires provide equal means of calculating QALYs. Because direct methods of eliciting utility involve comparatively complex tasks, it is most common in health research studies to use HRQoL questionnaires and link them to utility values.  As noted, some experts avoid using the term CUA. One reason for this is that not all experts agree on which methods of elicitation result in true utility values. Another reason is that QALYs assign a utility to the quality of life component, but not the length of life component, which is simply a multiplication of years of life in each health state. This ignores that people may value different times of their life in different ways. As a result of this, QALYs represent a unique outcome that is not perfectly represented by the term utility, making the term CUA somewhat misleading. The subsequent chapters of this thesis report the results of full economic evaluations in which QALYs are used as an outcome measure. In accordance with the view that QALYs are a unique outcome not equal to utility, these economic evaluations are referred to in this thesis as CEA using QALYs.   1.5.3.1 Types of CEA Using QALYs  In a CEA using QALYs, some source of utility data is required, such as HRQoL questionnaires. Utility data are, generally speaking, not contained in population-based data sources, meaning that CEA using QALYs require individual-level data derived from specially designed, primary health research studies. Two types of research studies are common sources for individual-level data for use in economic evaluations: RCTs and observational studies, particularly longitudinal cohort studies. In both RCTs and observational studies, cost data can be collected alongside clinical information using a case-report form (CRF) or patient diary.152 Using a CRF, data on costs are elicited by asking patients to report their experience with health resource use, other specified disease-related costs, and work productivity loss over a chosen recall period; using a patient   46 diary, patients are asked to record costs as they occur. Data on HRQoL, as well as data for the valuation of work productivity loss, can also be collected using standardized questionnaires administered at regular follow-up intervals in either an RCT or observational study.   A CEA may be conducted based exclusively on observed data from a single RCT, exclusively on observed data from a single observational study, or by combining data from multiple sources. In general (although not as a rule), CEA that combine data sources also apply special techniques in order to extrapolate outcomes beyond those observed empirically; these are referred to here as 'mathematical modeling' studies. This section briefly reviews these three types of CEA.  1.5.3.1.1 CEA Based on Observed Data From a Single RCT   Often, RCTs are referred to as the ‘gold standard’ research design, 153 as only in an RCT are patients assigned to treatment conditions using a randomized process. Generally speaking, this randomized process ensures that treatment groups are balanced in terms of patient characteristics, which allows for differences in treatment effects to be observed independently of other factors. While this is a strong advantage of the RCT design, RCTs have their own limitations, which arise in part from their goal of demonstrating a treatment's potential for efficacy under ideal conditions, rather than its effectiveness under real-world circumstances. As a result of this specific objective, RCTs tend to enroll the subset of patients that is most likely to benefit from the treatment under study, rather than a representative sample of the total patient population. RCTs also often use a primary outcome measure (i.e., the outcome measure specified when estimating the study's power to detect differences, which informs the number of patients to be enrolled in the trial) that is designed to be sensitive to the treatment effect, but which may be less relevant to patients; for example, RCTs examining SpA treatments have used disease activity measures as primary outcomes rather than HRQoL,122 though HRQoL is a higher-level PRO. Furthermore, RCTs often administer treatment under conditions that do not fully reflect those in clinical practice (e.g., increased safety monitoring). All of these attributes of an RCT mean that its results may not be generalizable to all relevant patients in the real world.     47 An additional limitation of RCTs is that they tend to have relatively short follow-up periods, which means that adverse events or other phenomena expected to occur only rarely or over the long-term may not be observed during the trial. For example, some SpA treatments may have a latent effect on aspects of the disease that progress slowly, such as bone damage. As excess costs may accumulate with bone damage, it is important to incorporate these costs into CEA, yet they may not occur until years after an RCT is completed. In addition to their limited follow-up period, it is problematic that RCTs tend to compare a treatment under study to just one or two comparators, rather than all potential alternative treatments. Often, the comparator chosen in an RCT is not the most relevant to clinical practice (e.g., the comparator is often placebo, rather than the standard of care). Yet another limitation is that the amount of data collected in an RCT may be subject to limitations in order to reduce costs as well as participant burden; thus an RCT may neglect to collect some information that is potentially relevant to a CEA. Finally, irrespective of costs, all treatments to be investigated in an RCT must be ethical to administer to participants. This means that certain factors that may influence costs and outcomes in clinical practice, such as the use of non-recommended treatments, are not possible to observe in an RCT.  In accordance with the unique strengths and weaknesses of RCTs, economic evaluations based exclusively on observed data from a single RCT have their own advantages and disadvantages.140, 154 Such CEA may not examine all potentially cost-effective treatments, and the costs and effects experienced by the RCT patient population may not be generalizeable to other patients. Furthermore, the 'time horizon' (i.e., the period over which treatment alternatives are compared) of such CEA may be inadequate to fully capture the downstream costs and effects of each treatment. Finally, as they are derived from single sample, the data informing the CEA will necessarily carry with them a substantial measure of uncertainty (i.e., the possibility that different results would be obtained with a different sample). Despite these concerns, CEA based on observed data from a single RCT have advantages. They generate high-quality data on treatment efficacy, which results from the reduction of bias resulting from randomization as well as from blinded and rigorous assessment of outcomes,155, and they serve an important role in generating parameter estimates (i.e., estimates of costs and effects) specific to an observed population. Finally, CEA based on observed data from a single RCT tend to be transparent and easily interpretable.   48 1.5.3.1.2 CEA Based on Observed Data From a Single Observational Study  Unlike RCTs, observational studies do not confer the benefit of balanced treatment groups achieved by random treatment allocation. On the contrary, participants in an observational study, such as a longitudinal cohort study, usually receive treatment and follow-up that reflects usual care. As a feature of usual care, patients whose disease is severe enough to warrant intensive therapy will receive it, whereas those without severe disease will not be treated as intensively. This results in a source of bias referred to as 'confounding by indication', whereby systematic differences between patients who receive and do not receive a certain treatment can give the illusion that treatment is associated with more severe disease. Confounding by indication is a well-known limitation of observational studies. Recent advances in statistical methods (e.g., multivariate regression) allow researchers to adjust for differences between treated and untreated patients and thus isolate the independent effect of treatment. In a recent Cochrane review of meta-analyses describing the impact of study design on reported effect sizes, there was no significant difference in effect sizes reported by RCTs versus observational studies, challenging the notion that observational studies are inherently less valuable than RCTs.156 However, statistical methods used to isolate treatment effects are not foolproof, and the potential for residual confounding by indication is a concern in any observational study. In addition to confounding by indication, observational studies may be susceptible to misclassification bias, if the data collected do not allow for the perfect identification of exposures.157  Despite their potential risk of bias, observational studies offer a number of distinct advantages over RCTs that make them a valuable source of data for economic evaluations. Importantly, these are the same advantages that observational studies have in the context of CER, in which the incorporation of observational data is considered imperative for numerous reasons.137, 158 For one, observational studies such as longitudinal cohort studies collect data from participants over a much longer follow-up period than RCTs, allowing for the detection of rare events or outcomes that occur only over the long-term. As they tend to be lower cost than RCTs, observational studies also allow for the collection of more exhaustive information on patient history, sociodemographic and clinical characteristics, and other potentially influential variables. Perhaps most importantly, observational studies are more likely to collect data on a representative sample   49 of patients, rather than a specially-selected subset that is anticipated to respond best to treatment. Because observational studies enroll a more heterogeneous population of patients, they allow for more meaningful analyses of patient subsets, giving them the potential to identify patient characteristics associated with outcomes. Furthermore, the data are derived from a real-world setting, one that is more likely to reflect true clinical practice than an RCT. Observational studies thus give researchers the opportunity to study the effect of variables that may be carefully eliminated from the RCT setting, but remain present in clinical practice. An example of this is the receipt of non-recommended care, which only observational studies offer researchers the chance to examine and thus describe in terms of its impact on patients. In general, observational studies present a better opportunity to collect comprehensive information and to observe relationships between multiple variables in real-world clinical practice, which is why they are considered indispensable to CER.   Ultimately, CEA that use data from a single observational study ultimately take on the same strengths and weaknesses of observational studies themselves.140, 154 Their results are subject to the risk of confounding by indication and bias due to misclassification error, for which it may not be possible to perfectly control.157 Yet due to their use of data derived from a real-world patient population, these CEA may be more generalizeable than others, and they may be able to estimate the cost-effectiveness of alternative treatments or practices not studied elsewhere. Furthermore, as their time horizon may be up to several years when using data from a longitudinal cohort study, CEA based on observational studies may be able to provide better estimates of costs and effects as they accumulate over time. Finally, CEA based on observational studies have the important strength of transparency and easy interpretability.  1.5.3.1.3 Mathematical Modeling Studies   A form of CEA broadly known as mathematical modeling (often 'decision analytic modeling') studies attempts to address the limitations of CEA based on data from a single RCT or observational study. Two things are important to note about the mathematical modeling form of CEA. First, other CEA not in this category also employ statistical models, but in general these are regression models, which are used to describe relationships between variables in observed   50 data. Mathematical modeling studies, on the other hand, are those that apply specialized techniques in order to extrapolate outcomes beyond the observed data.159, 160 Second, one feature of mathematical modeling studies is that they tend to use multiple sources of data, combining inputs from one or more RCTs, observational studies, or other sources. For example, CEA may take data on treatment effect sizes from RCTs, but data on resource use from observational studies, drawing on the strengths of each unique study design. While it is possible for mathematical modeling studies to perform extrapolation using data from a single source, this section focuses on mathematical modeling studies that combine multiple data sources.  As noted, the follow-up periods in RCTs and observational studies are limited, and even the longest of cohort studies may be inadequate to capture all the relevant latent costs and effects of a treatment. It has been argued that the most relevant time horizon to evaluate a treatment's cost-effectiveness is the patient's lifetime, and that CEA based on RCTs and observational studies necessarily fail to incorporate downstream sequelae that would otherwise substantially change cost-effectiveness estimates.154 To address this concern, mathematical modeling studies use specialized techniques to extrapolate outcomes over longer periods than have been observed empirically and thereby make projections of cost-effectiveness over the patient's lifetime.159, 160 Some examples of outcomes among SpA patients that require extrapolation over the lifetime horizon include rates of radiographic progression and continued response to anti-TNF therapy, respectively. In attempting to estimate the cost-effectiveness of anti-TNF agents over a patient's lifetime, mathematical modeling studies have made assumptions regarding at what rate bone damage will occur over the long-term, and to what extent anti-TNF response will attenuate or plateau.122  Numerous forms of mathematical models exist, of which just one common form used in CEA using QALY is the 'Markov' model, which is a complex form of decision-tree.159, 160 In a Markov model, patients are represented as independent from another (i.e., patients do not interact) and they transition into different states in 'cycles' in accordance with probabilities fed into the model using data from external sources (e.g., RCTs or observational studies). The states in a Markov model may be characterized by a certain level of disease activity or function, by the use of a certain treatment, by the experience of a certain adverse event, and so on, and patients   51 accumulate costs and QALYs in accordance with their time spent in each state throughout various cycles. Again, the information on the outcomes associated with different states is usually derived from primary research studies, such as RCTs and observational studies, or other sources of population-based data; sometimes, an assumption must be made in the model for which there is no empirical evidence. Notably, in a Markov model, the probability of transitioning from one state to another depends only on the state the patient is in at the beginning of the cycle- not on the amount of time the patient has been in the state or on any aspect of the patient's history- which can be a limitation of Markov models relative to other more sophisticated models, including those allowing interaction between patients. It is well understood that the structure of the model can have substantial influence on cost-effectiveness estimates and the choice of model must be tailored to the question that is being addressed in the CEA.159, 160   With any mathematical model, including Markov models, the quality of the results obtained rests on the veracity of the assumptions made within it. Each individual assumption made in a mathematical model introduces uncertainty into its results, and most models make numerous assumptions, as multiple parameters influence a treatment's cost-effectiveness. In general, researchers must use the best available data to inform each of these parameters and then systematically test the impact of their input choices through sensitivity analyses. However, the veracity of assumptions by definition cannot be proven; this is but one of many limitations of mathematical models, which have been previously summarized by Sheldon.161 Importantly, mathematical modeling studies often lack the very transparency and interpretability that other forms of CEA possess.  A common view is that CEA conducted using data from a single primary research study and those conducted using mathematical models are complementary, each having its own limitations.154 The subsequent chapters of this thesis report the results of CEA using QALYs that are based on data from a single observational study, without employing techniques of mathematical modeling. The strengths and weaknesses of this form of CEA relative to other forms are to be taken into account when interpreting these results.    52 1.5.3.2 Presentation of CEA Results and Uncertainty   All CEA carry with them some measure of uncertainty, including uncertainty around costs and effects and corresponding ICER values. In many research studies, uncertainty surrounding outcomes is presented using 95% CI, which display the range of values that an outcome of interest would be expected to take on in 95% of future samples, based on patterns in the observed data from the current sample. However, in a CEA, uncertainty surrounds both costs and effects, and ICERs are not amenable to the calculation of a 95% CI. For this reason, as well as other unique characteristics of the ICER, the presentation of CEA results and surrounding uncertainty follows a unique format.  Commonly, the results of a CEA are shown by plotting the incremental costs and effects of a treatment relative to its comparator (e.g., the standard of care) on a two-dimensional plot called a cost-effectiveness plane. As shown in Figure 1.5, the origin of the plot represents the costs and effects of the standard of care; the incremental costs and incremental effects, respectively, of the new treatment may be greater (i.e., to the right of the origin), lesser (i.e., to the left of the origin), or equal (i.e., equal to the origin) to the standard of care.155 As a point of terminology, in the case that the new treatment has both lower costs and greater effect than the standard, the former is referred to as 'dominant' (located in the bottom right quadrant of the plane) and the latter as 'dominated' (located in the top right quadrant of the plane); when one treatment is dominant and the other is dominated it is not necessary to calculate an ICER. When both the costs and effects of one treatment are higher than the costs and effects of another, or both the costs and effects of one treatment are lower, the ICER is calculated to assist decision-making.     53 Figure 1.5 Cost-Effectiveness Plane  Reproduced from Chete et al. Cost-effectiveness Assessment of Cardiac Interventions: Determining a Socially Acceptable Cost Threshold. Interv Cardiol. 2014; 6(1): 45–55. 162  In order to express the uncertainty that surrounds the results of a CEA, a number of methods may employed.163 One common method is to use non-parametric bootstrapping 163-165 in order to explore the range of plausible values for both costs and effects and thereby create a distribution of plausible ICERs. To do so, numerous simulated samples are created by sampling with replacement from observed data on costs and effects. For example, if a dataset contains cost and QALY data on a sample of 100 patients who received a certain treatment, that dataset will produce a mean cost and a mean QALY for those 100 patients. By sampling with replacement from those 100 patients, additional samples will be created (often 10,000 or more), each of which will produce its own value for mean cost and mean QALY among patients who received that treatment. If a second dataset contains cost and QALY data on a sample of another 100 patients who received an alternative treatment, the same process can be followed, resulting again in multiple mean cost and QALY estimates for patients who received the alternative treatment. Using the mean cost and QALY values from the two original datasets, an ICER point estimate can be produced on the basis of the observed costs and QALYs among patients receiving alternate treatments; using the bootstrap samples generated from those original datasets, a distribution of ICERs is created. Once the bootstrap estimate of the ICER distribution has been   54 created, there are several approaches to create confidence limits around the ICER point estimate, although a simple common method is to use a specified upper and lower percentile of the distribution (e.g., for 95% confidence, the range of values excluding the bottom and top 2.5% of the distribution of ICERs is taken as the confidence interval).   A common way to present the results of a CEA is to plot not only the ICER point estimate on the cost-effectiveness plane, but the range of ICER estimates obtained using the bootstrap method, as shown in Figure 1.6. Instead of plotting each individual ICER estimate, it is also possible to draw a 95% 'confidence ellipse', which is effectively the smallest circle that will encompass 95% of the data points representing the ICER distribution (Figure 1.7).166 In reviewing the results of a CEA presented on a cost-effectiveness plane, decision-makers must take into account both the ICER point estimate and the range of uncertainty surrounding it. They must then determine whether they are willing to pay the estimated added cost of the new treatment in exchange for the estimated added benefit.  This determination may be made in a number of ways, one of which is by setting a cost-effectiveness threshold, an arbitrary cut-off that separates treatments that are considered to be good value for money versus poor value for money.167 It has been remarked that the most common threshold is USD $50,000 per QALY,168 roughly €47,000 or £40,000 at time of writing.                55 Figure 1.6 Bootstrap Estimates of Cost-Effectiveness    Reproduced from MacNeil Vroomen et al. The Cost-Effectiveness of Two Forms of Case Management Compared to a Control Group for Persons with Dementia and Their Informal Caregivers from a Societal Perspective. PLoS One. 2016; 11(9): e0160908. 169  Figure 1.7 Confidence Ellipses  Reproduced from Noyes et al. Cancer Care Economics of Bladder Cancer: Cost-Enhancing Factors and Possible Cost Reduction. Curr Opin Urol. 2008 Sep; 18(5): 533–539 170   1.5.1 Economic Issues in SpA  The goal of this section is two-fold. One, it aims to summarize the knowledge surrounding the economic impact of SpA that has been gained through COI studies. This serves to explain the importance of addressing outstanding economic questions in the field of SpA and also to prepare the reader to interpret a novel COI study performed as part of this thesis. Two, this section aims   56 to introduce the issue of the cost-effectiveness of anti-TNF therapy for SpA, which is the economic question that has attracted the most attention in the SpA field in recent years. In the subsequent chapters of this thesis, which focus on economic evaluations in the context of SpA treatment recommendations, anti-TNF therapy is the recommended SpA treatment that receives the most scrutiny.   1.5.1.2 SpA Cost of Illness   Given the substantial heterogeneity between SpA patients, the estimated economic impact of the disease is likely to vary based on the classification criteria used to define the patient population. Furthermore, given the high cost of anti-TNF therapy, COI studies conducted in the 'pre-biologic' versus 'biologic' era will inevitably vary in their cost estimates. As a formal systematic review of COI studies in SpA is outside the scope of this thesis, this section discusses selected findings from COI studies in SpA, briefly reviewing estimates from the pre-biologic era (chiefly derived from studies of AS patients) before describing findings obtained in the biologic era among various SpA populations.   In a review of 4 COI studies in AS conducted in the pre-biologic era, Boonen et al.171 found that mean total annual per patient costs in 2002 USD were between US$7,243 and $11,840, with work productivity costs accounting for 53–73% of total costs (HC approach), or between US$3,353 and $3,903, with productivity costs accounting for 15–26% of the total costs (FC method). A systematic review by Palla et al.172 identified 15 COI studies conducted among AS patients published between 2002 and 2012, of which 12 were conducted in the pre-biologic era.95, 173-183 In their review, Palla et al. did not analyze the mean total costs across included studies due to heterogeneity, but concluded that work productivity costs were the largest component of total costs among AS patients.   At least 12 COI studies of SpA patients have been conducted in the biologic era, including nine studies of AS patients,184-192 two studies of both AS and PsA patients,193, 194 and one study of PsA patients.195 Table 1.13 shows selected characteristics of these studies and how they differ. A number of studies have produced estimates of annual per-patient costs in the range of $25,000   57 USD (e.g., Azevedo187($23,184, 2012 USD), Cooksey188 (£19,016, 2010 Great Britain Pound (GBP)), and Kobelt184 (€20,328 2005 Euros), while others have produced estimates in the range of $10,000 USD per year, such as that by Tsifetaki194 (€8,680, 2014 Euros), Tu 192($10,856, 2012 USD), or Olivieri195(€11,844, 2011 Euros). Importantly, the proportion of patients treated with anti-TNF agents varied widely across these studies, from "hardly any"192 to 63%187, 194 to 100%195. As shown in Table 1.13, 5 of the 12 selected studies found that anti-TNF agents represented the largest source of costs, while the rest identified work productivity loss as the largest source. Among all studies that identified anti-TNF agents as the largest source of costs, the proportion of total costs consumed by anti-TNF agents varied from 53%190 to 93%.195    58 Table 1.13 SpA COI Studies in Biologic Era: Selected Characteristics First author Year of costing Population (place of recruitment) Data source: a=questionnaire b=administrative c=registry d=hospital % on anti-TNF Health resource measurements Work productivity measurements (valuation method) Largest cost component Mean annual cost per patient Olivieri (2016) 2005-2010 PsA (hospital clinic) a 100% Therapies, laboratory and other diagnostic examinations, hospitalisations, surgery, rehabilitation procedures, physician visits, and 'any other possible cost'; patient assistance. Caregiver and patient absenteeism (HC, national salaries by work category) Anti-TNFs (96%) 11,844 (2011 Euros) Sliwczynski (2015) 2013 AS (national payer database) b 1% Healthcare expenditures related to AS. NA Anti-TNFs (75%) 760 (USD, year ns) Azevedo (2015) 2011-2012 AS + ASAS axSPa (hospital clinic) a 63% Anti-TNFs; other drugs; medical consultations; physical therapy; 'complementary examinations'; adaptations; equipment. Retirement due to AS; sick leave due to AS; unemployed (HC, national minimum wage) Anti-TNFs (96%) 23,184 (2012 USD) Tsifetaki (2015) 2012-2013 AS or PsA (hospital clinic) d 63% Monitoring visits, inpatient hospitalizations, laboratory, imaging and other tests (scheduled or emergency), in-hospital biologic administration; only costs attributable to SpA. NA Medications (89%), of which 95% are anti-TNF costs 8,680 (2014 Euros) Cooksey (2015) 2009-2010 AS (rheumatologist office) a, b Not specified Visits to health professionals; transport; investigations performed; Rx and OTC medication; adaptations to home/car; carer assistance; self-funded visits to health professionals; unpaid assistance. Early retirement; absenteeism; presenteeism (HC, age and sex specific daily wage rates) Productivity loss (63%) 19,016 (2010 GBP)   59 First author Year of costing Population (place of recruitment) Data source: a=questionnaire b=administrative c=registry d=hospital % on anti-TNF Health resource measurements Work productivity measurements (valuation method) Largest cost component Mean annual cost per patient Tu (2014) 2012-2013 AS (hospital clinic) a "Hardly any" Medications, hospital outpatient appointments, hospital inpatient durations, physical therapy, traditional Chinese therapy; transportation fees, paid help. Unemployment, sick leave, reduced work productivity, early retirement due to AS (HC, wage rate not specified) Productivity loss (65%) 10,856 (2012 USD) Lee (2014) 2010 AS (hospital clinic) a, d 26% In-hospital, Rx drugs, physiotherapy, oriental/alternative medicine, travel, auxiliary devices, dietary supplements, home assistance, physician visits, treatments, hospitalization, monitoring. Job loss and sick leave HC, self-reported income Productivity loss (40%) 11,646,180 (Korean Won, year ns) Petrikova (2013) 2008 AS (patient organisation) a 6% Medical services or treatment for the disease, e.g., outpatient and inpatient treatment, treatment and assistive devices, medicines, etc. Long- and short term absence from work, sick leave, work time reduced due to AS, early retirement (FC, gross income in 2008, 6 months) Anti-TNFs (53%) 5,806 (2008 Euros) Kvamme (2012) Various up to 2010 AS or PsA (DMARD registry) c 74% AS; 17% PsA Drugs; infusions; GP visits; in-hospital general; in-hospital rheumatology; imaging; outpatient general; outpatient rheumatology; physiotherapy; rehabilitation Disability; rehabilitation; sick leave; work time reduced HC, median income (FC, median income, 5 months) PL if measured by HC (AS 69%; PsA 74.5%); anti-TNF if PL measured by FC HC: 44,277; (2011 Euros)   60 First author Year of costing Population (place of recruitment) Data source: a=questionnaire b=administrative c=registry d=hospital % on anti-TNF Health resource measurements Work productivity measurements (valuation method) Largest cost component Mean annual cost per patient Rafia (2012) NS AS (hospital clinic) a 27% Medications (DMARDs, anti-TNF agents), hospital inpatient duration, hospital outpatient appointments (physiotherapy, hydrotherapy), GP appointments Unemployment or early retirement due to AS/unrelated to AS, absenteeism and presenteeism (HC, self-reported income or age/sex specific mean hourly wage) Productivity loss (84%) 11,204 (GBP, year ns) Strombeck (2010) 2005-2007 AS (national payer database) b 22% Merkesdal cost domains3: inpatient care, outpatient care, drugs. Sickness benefit, work disability Productivity loss (53%) $18,548 (USD, year ns) Kobelt  (2008) 2005 AS (hospital clinic) a 26% Health care and community services related to AS; Rx and OTC medication; medical devices; investments; informal care Sick leave; work time reduced; early retirement (HC, sex specific hourly wage) Investments and informal care (43%) 20,328 (2005 Euros)     61 The proportion of patients using anti-TNF agents is just one factor that stands to influence COI estimates in SpA. As shown in Table 1.13, the cost components included in COI studies and corresponding terminology has varied widely. The valuation of work productivity loss has also varied, with the inclusion or exclusion of presenteeism having a strong influence on estimated value of work productivity loss. In a study by Kvamme et al.,193 the greatest source of costs depended on the method used to value work productivity loss (i.e., work productivity loss was the greatest source of costs when using the HC approach, anti-TNF agents when using the FC approach). As demonstrated by Cooksey et al.,188 the data source for a COI study can also influence cost estimates. In their study, patient questionnaires overestimated the number of healthcare visits compared to administrative data on GP visits and hospitalizations, while administrative data did not capture drug costs; the mean estimated total cost to the National Health Service per AS patient per year was £3,320 based on questionnaire data and £2,343 based on administrative data.188   Given multiple sources of variation in costing methods and differences across geographical settings, it is difficult to compare the results of COI studies conducted among SpA patients. In the event that COI studies serve to provide cost data for full economic evaluations, such as CEA, the impact of the costing methods on the results of the analysis should be considered. Comparing COI studies among SpA patients in the pre-biologic versus biologic era clearly shows that the economic burden of the disease has increased. To date, no COI studies have been conducted among a large sample of patients with early SpA.   1.5.2 Cost Effectiveness of Anti-TNF Therapy Among SpA Patients  To date, the vast majority of CEA of SpA treatments have focused on anti-TNF agents.196 This is consistent with the view that policy decisions around these agents are the most important in the SpA field, as their use may lead to significant increases in drug budgets.197 In 2016, the UK's NICE published a systematic review and mathematical models of anti-TNF cost-effectiveness for the treatment of AS and nr-axSpA, respectively.198 Here, the NICE results are discussed as a means of presenting the latest knowledge of anti-TNF cost-effectiveness, and to demonstrate that    62 this knowledge is derived mostly from RCT-based economic evaluations and mathematical models, with little contribution from observational studies.  The NICE models of anti-TNF cost-effectiveness incorporated evidence from 24 RCTs, including 19 trials of AS patients, 4 of nr-axSpA patients, and 1 of AS and nr-axSpA patients. Twenty-two trials were placebo-controlled, with observation periods under placebo control generally lasting 16 weeks or less; extension phases without placebo control took place in 17 trials and lasted from 40 weeks to 8 years. In addition to short follow-up periods, the included RCTs had other limitations, including patient populations that may not be generalizable.  For example, patients included in these RCTs had relatively high BASDAI scores, ranging from 5.3 to 7, despite the fact that BASDAI ≥4 was the standard trial entry criterion and is the most commonly cited threshold in anti-TNF use recommendations. Furthermore, the nr-axSpA patients included in five separate RCTs were clinically heterogeneous (e.g., variation in baseline CRP, proportion of patients with positive MRI) and there was uncertainty surrounding nr-axSpA diagnoses in two key trials.199, 200 Finally, due to variation in tools used to collect HRQoL data in RCTs, the NICE authors were required to predict utility using BASDAI and BASFI scores in order to calculate QALYs.  In addition to RCT data, NICE incorporated data from 14 UK-specific mathematical models of anti-TNF agents, including eight conducted by anti-TNF manufacturers. Importantly, all of these models used RCT data to define rates and magnitude of response to anti-TNF agents, and none incorporated data on anti-TNF effectiveness derived from real-world populations. Furthermore, these models made assumptions regarding what happens to patients who lose their response to anti-TNF agents. In some models, these patients were assumed to revert to the level of function they had prior to responding to anti-TNF therapy, a scenario referred to as 'rebound equal to gain.' In other models, these patients were assumed to take on the level of function they would be predicted to have at that time point had they never responded to anti-TNF therapy, a scenario referred to as 'rebound equal to natural history'. As the rebound to natural history assumption places patients at a lower level of function than the rebound equal to gain assumption, and because lower function is associated with higher cost and lower benefit, ICERs will appear more favourable under the rebound equal to gain assumption.    63 Given the influence of the rebound assumption, the NICE authors created a series of mathematical models in which they varied key assumptions, making separate estimates of anti-TNF cost-effectiveness for AS and nr-axSpA patients. In all models, a lifetime, i.e., 60-year, time horizon is adopted. Assuming rebound equal to gain, the ICERs for anti-TNF use in AS patients ranged from £19,240 per QALY for certolizumab (assuming it could be obtained by the government at a reduced price) to £40,467 QALY for infliximab in AS patients. Assuming rebound to natural history, the ICERs for anti-TNF use in AS patients ranged from £33,762 per QALY for certolizumab (again assuming reduced cost) to £66,529 per QALY for infliximab. With respect to patients with nr-axSpA, assuming rebound equal to gain, ICERs associated with anti-TNFs use ranged from £28,247 per QALY for reduced-price certolizumab to £29,784 per QALY for etanercept. Assuming rebound to natural history, the ICERs associated with anti-TNFs for nr-axSpA ranged from £32,528 per QALY for reduced-price certolizumab to £34,232 per QALY for etanercept.   Arguably, the NICE models of anti-TNF cost-effectiveness represent the most cutting-edge economic evidence on this topic to date. Nonetheless, some shortcomings in this evidence should be noted. For one, the NICE models are built largely on the basis of RCT data, and legitimate questions surround the generalizability of anti-TNF efficacy based on RCT findings. It should not be overlooked that RCT populations have generally had higher BASDAI scores than the threshold currently cited in anti-TNF use recommendations, which means that anti-TNF users in the real world will likely have less active disease than trial populations. Only five RCTs have been conducted among nr-axSpA populations and it is difficult to synthesize their results due to heterogeneity in the clinical characteristics of participants. As well, NICE used BASDAI and BASFI scores to predict utility, a form of extrapolation that could introduce uncertainty surrounding the benefits of anti-TNF agents.201 In addition, the majority of RCTs have lasted 16 weeks or less; extension phases have taken place but by definition have included highly selected samples, i.e., patients included in the original trial and choosing to both continue anti-TNF therapy and participate in long-term follow-up.   Several significant sources of uncertainty in the NICE models arise from attempting to model the effect of anti-TNF therapy over 60 years – far longer than any empirical study of SpA patients.    64 Just one of these sources of uncertainty is the assumption of what occurs to patients who lose response to anti-TNF therapy. Importantly, the 'rebound to natural history' scenario used in some of the NICE models requires assuming the rate at which untreated patients will naturally experience functional decline. As acknowledged by the NICE authors, it is in fact highly uncertain at what rate function will decline without treatment; there are only limited long-term data on AS patients to indicate this and no long-term data on nr-axSpA patients. To model functional decline without treatment, the authors used data on the relationship between radiographic progression and BASFI scores,202, 203 assuming an increase of 0.056 BASFI units per year for patients with radiographic disease and no increase in BASFI for patients without. The authors were thus required to make additional assumptions regarding the proportion of nr-axSpA patients who progress to AS. Separate assumptions were then made regarding the rate of functional decline among patients responding to anti-TNF therapy. All of the necessary assumptions were made using the best available data and sensitivity analyses were employed to examine their implications. Nonetheless, all assumptions were based on data from samples of SpA patients with specific characteristics and it is unknown whether these assumptions are generalizable to all SpA patients in the real world. Furthermore, all assumptions were based on data from studies with observation periods far shorter than the 60-year time horizon employed in the models. As significant uncertainty surrounds the extrapolation of outcomes over unobserved periods, the cost-effectiveness of anti-TNF agents over the lifetime horizon is necessarily uncertain.  In summary, the best available evidence to-date on anti-TNF effectiveness is derived largely from RCTs and mathematical models that combine many sources of data, apply numerous assumptions, and employ sophisticated methods that are difficult to interpret. Studies are needed to describe outcomes among large numbers of SpA patients in real-world settings, including outcomes that are not always collected in RCTs, such as HRQoL. Observational studies of SpA patients treated and not treated with anti-TNF agents should provide complementary, transparent, and easily interpretable information to decision-makers on the cost-effectiveness of anti-TNF agents.      65 1.6 Knowledge Gaps  The concept of SpA has evolved over recent years and is now understood to comprise a heterogeneous patient population, including patients with nr-axSpA. Current SpA treatment recommendations are based largely on evidence derived from studies of AS patients, and significant knowledge gaps remain as to the effect of treatment on other SpA patients. Furthermore, no studies to date have examined the effect of adherence to SpA treatment recommendations as a whole. This results in part from the lack of validated methods to measure adherence to SpA treatment recommendations. Studies are needed to define adherence to each element of recommended care, and to measure the impact of adherence to all recommendations on patients' health and economic outcomes.   Economic outcomes among SpA patients are of rising importance, as the introduction of anti-TNF agents has substantially increased the economic burden of SpA. Worldwide, most health systems restrict access to anti-TNF agents to patients meeting specific criteria in order to reduce costs; the cost-effectiveness of anti-TNF agents is of great interest to decision-makers. However, it is unknown which restrictions around access to anti-TNF agents are most cost-effective. Treatment recommendations specific to anti-TNF agents vary across geographical settings, citing different access criteria without corresponding cost-effectiveness evidence. Many anti-TNF recommendations endorse the use of the drugs among nr-axSpA patients, despite the incomplete evidence base as to their costs and benefits in this population.   To date, estimates of the cost-effectiveness of anti-TNF use among SpA patients has been almost exclusively derived from RCTs and mathematical models. This leaves room for error, as RCT evidence may not be generalizable to real-world SpA populations and mathematical models make assumptions that could render their estimates inaccurate. Observational cohort studies provide comprehensive data on representative samples of SpA patients, treated under conditions most likely to reflect true clinical practice. Accordingly, observational cohort studies could be a valuable means of estimating the costs and benefits of anti-TNF therapy, both on its own and alongside other recommended treatments. While observational studies are also an important source of data for eventual use in mathematical models, there is an outstanding need to describe    66 observed outcomes among SpA patients without making assumptions. Describing observed costs and benefits among a heterogeneous population of SpA patients, while taking into account adherence to anti-TNF use and other treatment recommendations, will help generate transparent evidence to inform the development of clinical and economic guidelines.  Since the mid-nineties, numerous observational cohort studies have been established with the goal of addressing the knowledge gaps in the field of SpA research.204-210 In France, the DESIR cohort was established in 2007, enrolled 708 patients aged >18 and <50 with IBP fulfilling the Calin or Berlin criteria, with symptoms lasting >3 months and <3 years and suggestive of SpA according to the local investigator’s assessment (i.e., a score ≥5 on a 0 to 10 numerical rating scale in which 0 = no suggestive and 10 = very suggestive of SpA). Patients with a definite diagnosis of non-SpA spinal disease, history of anti-TNF use or corticoid use exceeding 10 mg/day, or conditions affecting informed consent or compliance were excluded. The DESIR cohort was the data source for all studies included in the current thesis, which aimed to address outstanding economic questions in the context of SpA treatment recommendations.  1.7 Research Questions and Goals  The specific research questions and goals addressed in the thesis are:   1. What was the cost of illness from the societal perspective among DESIR patients in the first three years of follow-up and what factors were associated with costs? Calculating these costs would allow for further research questions to be addressed using DESIR data, including economic research questions in the context of SpA treatment recommendations.  2. Can a method be developed to allow for the measurement of adherence to each element of the ASAS SpA treatment recommendations using observational data in DESIR? Developing a means of measuring adherence to each element of recommended care would allow for an analysis of the impact of such adherence on costs and health benefits. 3. What was the independent effect of adherence to measurable elements of recommended SpA care among patients in the DESIR cohort in terms of costs and health benefits?    67 4. Are current French anti-TNF access restrictions are the most cost-effective in that setting relative to other potential restrictions?  1.8 Outline of Subsequent Chapters  The subsequent chapters in this thesis, with the exception of Chapter 6, were written with the goal of publication in peer-reviewed journals. Specifically, Chapters 2-5 each represent a separate research article, while Chapter 6 aims to summarize and contextualize them. Consequently, some facts are stated more than once throughout the subsequent chapters of this thesis.   Chapter 2 describes the cost-of-illness study that was foundational to the thesis in that it allowed for further economic research questions to be addressed. Specifically, this chapter describes the costing methods employed in order to quantify the cost-of-illness in the DESIR cohort over the first three years of follow-up, as well as the results of a descriptive cost analysis and model to elucidate predictors of various cost components and total costs. Chapter 2 should be seen as the first step toward the realization of Chapters 4 and 5.  Chapter 3 describes a study aimed at developing and describing a means of measuring adherence to the various elements of the ASAS SpA treatment recommendations. While the focus of Chapter 3 may not appear at first to follow sequentially from a cost-of-illness study, it in fact constitutes the second step towards the realization of Chapter 4. To quantify the cost of illness in DESIR (Chapter 2) and to develop a means of measuring adherence to recommended care (Chapter 3) were necessary in order to estimate the costs and benefits of adherence to the ASAS recommendations for SpA care (Chapter 4).  Chapter 5 describes a study to evaluate the costs and benefits of anti-TNF use and non-use among patients who satisfied different sets of anti-TNF use recommendations, in an effort to determine whether the current French anti-TNF access restrictions are the most cost-effective in that setting relative to other potential restrictions     68 The objective of Chapter 6 is to comment on the totality of the four distinct research studies that were undertaken as part of this thesis. This chapter raises a discussion as to the meaning of this research, including its strengths, limitations, and implications. The nature of the contribution to SpA research field is interpreted, and ideas for future directions for SpA research are given from the perspective of the author.       69 Chapter 2ii  Costs of Early Spondyloarthritis: Estimates From the First 3 Years of the DESIR Cohort  2.1 Introduction  SpA is a family of chronic rheumatic diseases that includes AS, PsA, reactive SpA, SpA with IBD, and undifferentiated SpA.211 In its early phases, SpA is often undifferentiated, and corresponds to the concept of IBP, the key feature of the disease.212 To date, most economic studies of SpA have examined outcomes among patients with AS or PsA specifically, and have not included patients with other SpA subtypes or with early SpA symptoms that are still evolving. A severe form of SpA, AS has been shown to cause significant pain, functional disability, and loss of mobility.213 In turn, these physical outcomes are associated with substantial health resource utilisation and work productivity loss among AS patients.214, 215  In 2012, a systematic review identified 15 economic studies of SpA conducted worldwide since 2002, nine of which evaluated both health resource utilisation and work productivity loss. The median cost of health resource utilisation in 2012 US dollars was $3,764 per patient per year, compared to a median cost of productivity loss of $4,999.216 More recently, biologic anti-TNF agents have been made available for the treatment of moderate to severe SpA, increasing the cost of illness compared to previous years in which NSAIDs were the only available drug therapy.217  The objective of this study was to describe health resource utilisation and productivity loss costs among patients in the DESIR cohort,210 a longitudinal, multi-centre study of early SpA in France. We further aimed to determine factors associated with total costs and with costs excluding anti-TNF agents ('non-biologic costs') among DESIR patients.                                                    ii A version of Chapter 2 has been published: Harvard S, Guh D, Bansback N, Richette P, Dougados M, Anis A, Fautrel B. Costs of early spondyloarthritis: estimates from the first 3 years of the DESIR cohort. RMD Open. 2016 Apr 4;2(1).    70 2.2 Methods  2.2.1 Participants  The DESIR cohort218 includes 708 patients aged 18-50 with early IBP lasting more than 3 months but less than 3 years, and suggestive of SpA according to the rheumatologists' assessment (score ≥5 on a 0-10 Numerical Rating Scale where 0=not suggestive and 10=very suggestive). Patients were required to fulfil the Calin6 or Berlin219 IBP criteria. Patients with a definitive diagnosis of non-SpA back pain, a history of anti-TNF use, or conditions that could affect informed consent and/or compliance (e.g., alcoholism, psychiatric disorders) were excluded. All patients attended their baseline study visit between October 2007 and May 2010. Follow-up visits collecting self-report questionnaire data occurred every six months in the first two years and every year thereafter, and covered the health resource use since the previous visit. Our analyses included the first three years of follow-up, i.e., baseline plus follow-up visits at months 6, 12, 18, 24, 36. Data up to month 12 were from the DESIR database locked in February 2014, data from months 18 to 36 from that locked in April, 2015.  2.2.2 Costing Methods  Our analysis estimated the total cost from the societal perspective (all payers combined) of all-cause health resource utilisation, specifically health practitioner visits, hospitalizations (including emergency room visits and surgeries), medical workups, and medications, and work productivity losses. Although we aimed to include as many societal costs as possible, other patient out-of-pocket costs (e.g., transportation, devices, caregiver expenses, and other costs sometimes called 'non-medical') were not included.  2.2.2.1 Valuing Health Resource Use   Detailed costing methods are described in Supplementary Tables 1-4 in the Appendix. Unit costs for health practitioner visits were estimated by adding base tariffs from the French National Health Insurance220 to the average cost of extra-billings based on data from the IRDES (Institut    71 de Recherche et de Documentation en Economie de la Santé)221 Eco-Santé database.222 Data on medical workups included laboratory analyses on blood and urine and functional/diagnostic tests. Based on clinical expertise, the cost of blood tests was estimated assuming the following standard laboratory analyses among SpA patients: ESR; CRP; hemogram including platelets; thrombocytes; transaminases; creatinine; and creatinine clearance. Blood collection was valued at fifteen minutes of nurse care; the unit cost of blood tests was the cost of analysis plus the cost of collection as per the French public insurance fee schedule. Urine tests were assumed to include urine protein only and no collection fee was applied. The following tests were attributed a single standard cost: mammography, MRI, bone densitometry, scintigraphy, respirometry, colonoscopy, and fibroscopy. For X-rays, ultrasounds, scans, and 'other exams', free-text data were reviewed individually and assigned specific exam codes which were linked to unit costs.  Data on medications included NSAIDs, DMARDs (conventional synthetic or biologic anti-TNF agents), corticosteroids (oral, intramuscular, intraarticular and intraveinous), and analgesics. For abatacept, infliximab, and certolizumab, prices-per-milligram were derived from list prices in France using the Vidal dictionary.223 For all other drugs, cost data from the French public drug program224 were used to derive a price-per-milligram for each drug. Drugs were valued by multiplying price-per-milligram by number of milligrams per day by number of days of use (except injectable corticoids, which were valued by multiplying price-per-milligram by number of milligrams in the reported number of injections). Where dose data were missing, standard dosages were imputed taking into account the patient's weight where applicable. For infliximab only, a standard cost for administration in day hospital was applied every 6 weeks for the duration of the therapy.   Data on hospitalizations were assigned a diagnosis or act code and linked to Diagnosis-Related Group (DRG) codes using 2012 data from the French national agency for hospital information (ATIH, 'Agence Technique de l’Information sur l’Hospitalisation').225 Selected DRG were then linked to costs using 2012 ATIH data.226 The base-case analysis used the 6-digit DRG selected as the best-fit; sensitivity analyses were performed using DRG sharing the same first 5 digits as the best-fit DRG (Supplementary Table 3). All unit costs were adjusted to 2013 Euros based on the Consumer Price Index227 and multiplied by frequency of use.    72 2.2.2.2 Valuing Work Productivity Loss  Data on patients’ profession were collected in 8 categories (Table 2.1). Average daily wage data were obtained from the French Ministry of Agriculture228 and from the French National Institute of Statistics and Economic Studies ('INSEE').228  Productivity loss was valued by multiplying the number of self-reported work days lost per period by the average daily wage by professional category over the entire population of French workers as estimated by these public data sources. The age and sex distribution of DESIR the cohort was compared to that of the population of French workers from which average daily wages were obtained and wages were not further adjusted for age and sex.  Table 2.1 Baseline Characteristics of the DESIR Cohort by Anti-TNF Use   All (n=708) No Anti-TNF=483 Anti-TNF=225 Age, Sex, Ethnicity Count (%) Count (%) Count (%) Age group  <25 133 (18.0) 94 (19) 39 (17) Age group >=25  and <55 572 (81.0) 388 (80) 184 (82) Age group >= 50 3 (0.4) 1 (0) 2 (1) Male  327 (46.2) 230 (48) 97 (43) Caucasian 634 (89.6) 435 (90) 199 (88) Highest Level of Education        High school or less  290 (40.4) 183 (52) 107 (128) Post-secondary school 418 (59.0) 300 (38) 118 (107) Profession        Agriculturer 6 (0.9) 4 (1) 2 (1) Artisan, retailer or small business owner 35 (4.9) 25 (5) 10 (4) Executive or academic 151 (21.3) 114 (24) 37 (16) Intermediate professional 54 (7.6) 44 (9) 10 (4) Employee 299 (42.2) 192 (40) 107 (48) Tradesperson 64 (9.0) 40 (8) 24 (11) Without professional occupation 93 (13.1) 60 (12) 33 (15) Clinical characteristics        Early onset (<17 years) 13 (1.8) 8 (2) 5 (2) Late onset (>44 years) 85 (12.0) 54 (11) 31 (14) Presence of peripheral arthritis 405 (57.2) 248 (51) 157 (70) Presence of extra-articular manifestations          Uveitis 60 (8.5) 38 (8) 22 (10)   Psoriasis 112 (15.8) 66 (14) 46 (20)   Crohn's disease 18 (2.5) 10 (2) 8 (4)    73   All (n=708) No Anti-TNF=483 Anti-TNF=225 Age, Sex, Ethnicity Count (%) Count (%) Count (%) Treated with conventional DMARD  100 (14.1) 50 (10.4) 50 (22.2) HLA B27 Positive 410 (57.9) 284 (59) 126 (56) X-Ray and MRI damage 132 (18.6) 76 (15.7) 56 (24.9) X-Ray damage only 51 (7.2) 34 (7.0) 17 (7.6) MRI damage only 98 (13.8) 65 (13.5) 33 (14.7) Neither 403 (56.9) 289 (59.8) 114 (50.7) Missing 24 (3.3) 19 (3.9) 5 (2.2) ASAS criteria (total): Positive 486 (68.6) 322 (67) 164 (73) ASAS clinical criteria: Positive 404 (57.1) 279 (58) 125 (56) ASAS imaging criteria: Positive 286 (40.4) 178 (37) 108 (48) Clinical Values Mean SD Mean SD Mean SD Disease duration in years  1.5 0.9 1.5 0.9 1.5 0.9 Physician's assessment of disease activity (0-10) 4.3 2.2 3.8 2.0 5.6 2.0 BASDAI Score (0-100)  44.7 20.0 40.0 19.9 54.7 16.1 BASFI Score (0-100) 30.4 22.8 25.3 21.7 41.3 21.2  2.2.2.3 Handling Missing Clinical and Cost Data  Missing cost and clinical data were imputed using the Monte Carlo Markov Chain (MCMC) multiple imputation (MI) procedure, the last observation carried forward (LOCF) method, probabilistic imputation, or with negative values based on clinical expertise. Specifically, for clinical variables in which fluctuations are normal and commonly observed clinically (e.g., CRP, BASDAI, BASFI), missing data was handled with MCMC imputation, which was considered the best strategy for representing this variability. For clinical variables observed to have greater stability, either the LOCF method or probabilistic imputation was used in lieu of MCMC imputation, as these methods provide better control over the variability of values imputed; LOCF was used if all patients had baseline data, while probabilistic imputation was used if baseline data were missing. The imputation model is described in detail in the Appendix.  2.2.2.4 Statistical Analyses  Patient characteristics were described by the mean ± SD or number (%) at baseline. For descriptive purposes, patients were divided into four subsets based on imaging status at baseline: X-Ray damage only; MRI inflammation only; both X-Ray damage and MRI inflammation;    74 neither. Costs of health resource utilisation and productivity losses were expressed as yearly costs. For health resource utilisation, year 1 costs were the sum of costs reported at months 6 and 12; year 2 costs were the sum of costs reported at months 18 and 24; year 3 costs were those reported at month 36. For productivity losses, yearly costs were those reported at the 12, 24, and 36 month visits. Yearly costs were described by the median and mean ± SD.   To describe factors associated with annual total costs and non-biologic costs over the three years we used generalized linear models with a gamma distribution of log link and a generalized estimating equations (GEE) algorithm to account for repeated measures within subjects. The independent variables year, baseline age, and sex were chosen to be included in all models throughout development. Other independent variables of interest included DESIR study centre; baseline education, profession, smoking, presence of peripheral arthritis, HLA-B27, imaging status and BASDAI and BASFI values, as well as mean BASDAI and BASFI values (where year 1 mean= mean of months 0, 6, 12; year 2 mean= mean of months 12, 18, 24; and year 3 mean= mean of month 24, 36) number of months on anti-TNF therapy. At the outset, baseline and mean BASDAI and BASFI values were tested for collinearity and examined alone and together in models. Goodness-of-fit assessed by quasi-likelihood statistic (QIC) was used to compare the model fit among models with various BASDAI and BASFI values (baseline, mean, or baseline and mean together).229-231 The model with the best fit was chosen for inclusion in subsequent models.  All other independent variables of interest were examined separately in models adjusting for year, age, and sex, and variables associated with the outcome at p<0.20 were selected for inclusion in subsequent models. Model selection was then done in a backward stepwise manner, beginning with all variables and removing those that were not associated with the outcome at p<0.05 to increase goodness-of-fit based on the QIC. All analyses were performed using SAS 9.4.          75 2.3 Results  2.3.1 Patient Characteristics  Table 2.1 shows the baseline characteristics of all DESIR patients and of anti-TNF users and non-users. At baseline, the mean age of patients was 33.7 ± 8.6 years and mean disease duration was 1.5 ± 0.8 years. Just under half of patients were male (46.2%) and the majority was Caucasian (89.6%) (Table 2.1). Most patients were HLA-B27-positive (57.9%). The mean BASDAI score (on 100) at baseline was 44.7 ± 20.0 and the mean BASFI score was 30.5 ± 22.8. The most common extra-rheumathological manifestations were psoriasis (15.8%) and uveitis (8.5%). At baseline, 286 patients satisfied the ASAS imaging criteria (40.4%) and 404 the ASAS clinical criteria (57.1%); 486 patients satisfied one or the other (68.6%). At baseline, the majority of patients had neither visible X-Ray damage nor MRI inflammation (n=403; 56.9%), while approximately a fifth of patients had both X-Ray damage and MRI inflammation (n=132; 18.6%); a minority of patients had X-Ray damage alone (n=51; 7.2%) or MRI inflammation alone (n=98; 13.8%).   2.3.2 Costs  2.3.2.1 Health Resource Utilisation  Table 2.3 shows the estimated costs of health resource utilisation and productivity losses among all DESIR patients over three years. Medication was the largest cost component in all years, representing over 50% of all costs in all years. The absolute mean cost of medication increased from €2680.2 ± 5339.6 in year 1 to €3339.2 ± 6224.1 in year 2 to €3396.3 ± 6476.5 in year 3. The proportion of patients incurring costs from anti-TNF agents was approximately a quarter each year. The cost of anti-TNF agents accounted for approximately 95% of medication costs each year.  Health practitioner visits, the second-largest cost component of health resource utilisation in all years, declined from €733.2 ± 788.9 in year 1 to €615.0 ± 783.4 in year 2 and €481.0 ± 676.2    76 year 3. Hospitalization costs were incurred by a quarter to a fifth of patients, representing under 10% of costs each year. The mean hospitalization cost per patient was €498.4 ± 1318.9 in year 1, €526.2 ± 1325.7 in year 2 and €423.3 ± 1100.8 in year 3. Medical workups were consistently the smallest component of health resource utilisation, never exceeding 5% of yearly costs.               77 Table 2.2 Health Resource Use and Productivity Loss Costs by Year and Anti-TNF Use   Year 1 Year 2 Year 3 Patient Group Resource Component % patients with costs mean sd med % patients with costs mean sd med % patients with costs mean sd med All patients (n=708) Health Practitioners* 91 733 789 486 84 615 783 350 77 481 676 254  Medical Workups** 79 249 306 144 71 195 257 85 66 163 234 65  Anti-TNF agents 24 2515 5322 0 27 3205 6202 0 26 3287 6460 0  All Drugs 100 2680 5340 155 100 3339 6224 127 100 3396 6477 99  HRU Costs 100 4161 5901 1297 100 4675 6780 1107 100 4464 6859 856  Productivity Loss 30 843 2899 0 23 769 3368 0 24 497 1951 0  All Costs Excluding Anti-TNF agents 100 2489 3629 1228 100 2239 4159 872 100 1673 2718 706  All Costs 100 5004 6870 1547 100 5444 7937 1287 100 4961 7457 1063 Ever Received an anti-TNF agent (n=225) Health Practitioners* 99 997 886 722 94 909 922 577 86 702 886 399  Medical Workups** 96 427 374 324 91 328 303 245 83 265 296 168  Hospitalizations 31 666 1660 0 30 742 1601 0 24 445 1031 0  Anti-TNF agents 76 7913 6819 8510 86 10084 7191 12047 81 10344 7641 12694      All Drugs 100 8123 6802 8548 100 10265 7183 12186 100 10477 7650 12716 HRU Costs 100 10214 7073 11175 100 12244 7422 13437 100 11889 7895 13577 Productivity Loss 41 1523 3723 0 33 1391 4827 0 34 945 2900 0 All Costs Excluding Anti-TNF agents 100 3823 4341 2019 100 3552 5614 1677 100 2490 3635 1217 All Costs 100 11736 7883 12477 100 13635 8915 13880 100 12834 8667 13765                  78   Year 1 Year 2 Year 3 Patient Group Resource Component % patients with costs mean sd med % patients with costs mean sd med % patients with costs mean sd med Never Received an anti-TNF agent (n=483) Health Practitioners* 88 610 707 352 80 478 668 211 72 378 522 177  Medical Workups** 70 166 225 81 62 132 205 41 59 116 180 41  Hospitalizations 22 420 1119 0 20 426 1164 0 19 413 1133 0  Anti-TNF agents 0 0 0 0 0 0 0 0 0 0 0 0  All Drugs 100 145 223 99 100 113 143 75 100 98 145 53  HRU Costs 100 1341 1662 750 100 1149 1619 510 100 1004 1518 429  Productivity Loss 25 527 2360 0 19 479 2355 0 20 288 1241 0  All Costs Excluding Anti-TNF agents 100 1868 3057 850 100 1628 3090 566 100 1293 2060 521  All Costs 100 1868 3057 850 100 1628 3090 566 100 1293 2060 521 *Health practitioners includes physicians and other allied health professionals (nurse, physiotherapist, osteopath, etc.) **Medical workups include blood and urine tests, X-ray, ultrasound, scanner, MRI, bone densitometry, scintigraphy, respirometry, colonoscopy, mammography, and fibroscopy.    79 Table 2.3 Cost Component as Proportion of Total Costs by Year and Anti-TNF Use Patient group Cost Component Year 1 Year 2 Year 3 All patients (n=708) Health Practitioners 15% 11% 10%  Medical Acts 5% 4% 3%  Hospitalizations 10% 10% 9%  Anti-TNF agents 50% 59% 66%  All Drugs 54% 61% 68%  Productivity Loss 17% 14% 10%  All Costs Excluding Anti-TNF agents 50% 41% 34% Ever received an anti-TNF agent (n=225) Health Practitioners 8% 7% 5%  Medical Acts 4% 2% 2%  Hospitalizations 6% 5% 3%  Anti-TNF agents 67% 74% 81%  All Drugs 69% 75% 82%  Productivity Loss 13% 10% 7%  All Costs Excluding Anti-TNF agents 33% 26% 19% Never received an anti-TNF agent (n=483) Health Practitioners 33% 29% 29%  Medical Acts 9% 8% 9%  Hospitalizations 22% 26% 32%  Anti-TNF agents 0% 0% 0%  All Drugs 8% 7% 8%  Productivity Loss 28% 29% 22%  All Costs Excluding Anti-TNF agents 100% 100% 100%   2.3.2.2 Productivity Losses  Approximately a third of patients (30.1%) incurred productivity losses in year 1; this decreased to a quarter of patients in year 2 (23.4%) and year 3 (24.4%). The mean cost of productivity losses was €843.4 ± 2899.1 in year 1, €768.9 ± 3368.0 and €497.0 ± 1951.0 in year 3. In year 1, productivity losses represented 16.8% of total costs, compared to 14.1% of total costs in year 2 and 10.0% of total costs in year 3.     80 2.3.2.3 Total Costs  The mean annual total cost per patient was €5004.1 ± 6870.2 in year 1, €5444.1 ± 7936.9 in year 2, and €4960.6 ± 7457.4 in year 3. Mean annual total costs rose with anti-TNF use: patients who did not receive an anti-TNF agent had mean total costs of €1867.9 ± 3056.6 in year 1, €1628.2 ± 3090.3 in year 2 and €1292.9 ± 2059.6 in year 3, compared to €11,736.4 ± 7882.6 in year 1, €13,635.4 ± 8915.1 in year 2, and €12,833.9 ± 8667.4 in year 3 among patients who received an anti-TNF agent. In all years, anti-TNF agents represented over 50% of total costs, up to 66.3% in year 3. Overall, the mean 3-year total cost among patients was €15,408.76 ± 19,793.52. Patients who never received an anti-TNF agent had mean 3-year total costs of €4789.04 ± 6021.80 compared to €38,205.74 ± 19,829.33 among those who received an anti-TNF agent. The estimated total costs in the cohort over 3 years amounted to €10,909,399.9 ± 19574.0. Of this, patients who received an anti-TNF agent accounted for an estimated €8,596,291.6 ± 4860.99 or 78.8% of all costs in the cohort. Figures 2.1 and 2.2 show the trends in costs over the study period overall and for patients who received and did not receive an anti-TNF agent.  2.3.3 Factors Associated With Costs  2.3.3.1 Total Costs  In the final model (Table 2.4), each month of anti-TNF use was associated with a 23% increase (RR: 1.23; 95% CI 1.21, 1.24) in total costs. The presence of peripheral arthritis at baseline was associated with a 19% increase (RR: 1.19; 95% CI: 1.04, 1.37) in total costs, while every 10- point increase in average BASFI score was associated with a 18% increase in total costs (RR: 1.18; 95% CI: 1.15, 1.22). Although absolute costs were higher in year 2 than year 1 (Table 2.2), when adjusting for time on anti-TNF agents, costs in years 2 and 3 were significantly lower than in year 1 (Table 2.4). Baseline imaging status was not significantly associated with total costs and was excluded from the final model.    81 Figure 2.1 Yearly Costs Among Non-Biologics (Left) and Biologics Users (Right)            0 200 400 600 800 1000 1200 1400 1600 1800 2000  Biologics  Productivity Loss All Costs Excluding Biologics  All Costs Euros	Cost Components: Patients Never Using Biologics 0 2000 4000 6000 8000 10000 12000 14000 16000  Biologics  Productivity Loss All Costs Excluding Biologics  All Costs Cost Components: Patients Ever Using Biologics Year 1 Year 2 Year 3    82 Table 2.4 Models of Total Cost and Non-Biologic Costs Outcome Variable Rate Ratio (95% CI) P value Estimate* *Rate Ratio = [exp (estimate)] SE Estimate lower bound Estimate upper bound Min. estimate Max. estimate Min. Rate Ratio Max. Rate Ratio Total Costs Age at baseline (every 5-yr increase) 1.06 (1.02, 1.11) 0.006 0.060 0.022 0.017 0.102 0.055 0.065 1.06 1.07 Male vs female 0.75 (0.65, 0.87) 0.000 -0.287 0.074 -0.432 -0.142 -0.304 -0.265 0.74 0.77 Time on anti-TNF agents (in months) 1.23 (1.21, 1.24) <.0001 0.205 0.005 0.195 0.216 0.205 0.205 1.23 1.23 Year 2 vs Year 1 0.89 (0.78, 1.01) 0.063 -0.119 0.064 -0.244 0.007 -0.125 -0.114 0.88 0.89 Year 3 vs Year 1 0.71 (0.63, 0.81) <.0001 -0.339 0.062 -0.462 -0.217 -0.367 -0.321 0.69 0.73 Average BASFI score (every 10-point increase) 1.18 (1.15, 1.22) <.0001 0.168 0.015 0.137 0.198 0.164 0.172 1.18 1.19 Presence of peripheral arthritis at baseline 1.19 (1.04, 1.37) 0.014 0.177 0.072 0.036 0.318 0.166 0.198 1.18 1.22 Non-Biologic Costs    Age at baseline (every 5-yr increase) 1.10 (1.05, 1.16) 0.000 0.097 0.025 0.047 0.147 0.092 0.104 1.10 1.11 Male vs female 0.69 (0.58, 0.81) <.0001 -0.377 0.087 -0.548 -0.206 -0.398 -0.349 0.67 0.71 Time on anti-TNF agents (in months) 1.04 (1.02, 1.05) <.0001 0.037 0.008 0.022 0.052 0.037 0.038 1.04 1.04 Year 2 vs Year 1 0.87 (0.75, 1.00) 0.050 -0.143 0.073 -0.286 0.000 -0.151 -0.139 0.86 0.87 Year 3 vs Year 1 0.72 (0.63, 0.83) <.0001 -0.327 0.072 -0.470 -0.184 -0.357 -0.301 0.70 0.74 Average BASDAI score (every 10-point increase) 1.21 (1.16, 1.25) <.0001 0.187 0.019 0.151 0.224 0.185 0.192 1.20 1.21 Presence of peripheral arthritis at baseline 1.20 (1.02, 1.40) 0.024 0.181 0.080 0.024 0.338 0.170 0.198 1.18 1.22   83 2.3.3.2 Cost Excluding Anti-TNF Agents  In the final model (Table 2.4), each month of anti-TNF use was associated with a 4% increase (RR: 1.04; 95% CI 1.02, 1.05) in non-biologic costs. The presence of peripheral arthritis at baseline was associated with a 20% increase (RR: 1.20; 95% CI: 1.02, 1.40) in non-anti-TNF agents costs; every 10-point increase in average BASDAI score was associated with a 21% increase in non-biologic costs (RR: 1.21; 95% CI: 1.16, 1.25). Every 5-year increase in age was associated with a 10% increase (RR: 1.10; 95% CI 1.05, 1.16) in non-biologic costs, while males had 31% less costs relative to females (RR: 0.69; 95% 0.58, 0.81). Baseline imaging status was not significantly associated with costs excluding anti-TNF agents and was excluded from the final model.  2.4 Discussion  This study described costs related to health resource utilisation and work productivity losses among early SpA patients over three years. Our findings reflect the economic impact of anti-TNF agents, which represented by far the largest cost component among DESIR patients. This finding is consistent with evidence that anti-TNF agents are replacing work productivity loss as the largest component of total societal costs among SpA patients. In a 2012 review, Palla et al. 172 identified 15 COI studies conducted among AS patients since 2000, of which just three studies184-186 included the costs of anti-TNF agents, drawing the conclusion that work productivity costs were the largest component of total costs among AS patients.172 At least nine additional COI studies have since examined cost among SpA patients in the biologic era.187-194 Five of these studies found that anti-TNF agents were the greatest cost driver among SpA patients, with the proportion of total costs spent on anti-TNF agents varying from 53%190 to 93%.195  In our study, an analysis of early SpA patients in the biologic era, productivity loss may be expected to represent a lower proportion of costs given patients' better functional status and higher drug costs; that said, productivity loss costs among DESIR patients still represented between 10% and 17% of annual costs over the study period. In terms of absolute costs, work   84 productivity loss costs among DESIR patients (which ranged from €843 in year one to €497 in year 3) are consistent with those in a Dutch cohort of early SpA patients (approximately €422 over one year).232 We note that our estimates of productivity loss costs are conservative: we did not calculate the costs of presenteeism, i.e., the cost of reduced performance at work, which is another potentially important source of costs.233   Currently, there is increased interest in non-radiographic axSpA, defined as SpA in the absence of definite sacroiliac changes on X-Ray, which includes both patients who satisfy only the clinical criteria for SpA and those with active inflammation in the sacroiliac joints visible only on MRI.28 To our knowledge, this is the first longitudinal study of SpA patients that has explored the impact of radiographic damage on cost, distinguishing between damage on X-Ray with or without MRI inflammation. Interestingly, while preventing progression to radiographic SpA remains a high priority from the clinical perspective, we found no significant independent effect of baseline imaging status on total costs or costs excluding anti-TNF agents.  In this study, the most significant cost component was anti-TNF agents, and patients who ever received one accounted for over 75% of total costs in the cohort. While this was not surprising, the finding that approximately a quarter of DESIR patients had received an anti-TNF agent by the 6-month visit (anti-TNF use use at baseline was not permitted) was higher than anticipated, suggesting that anti-TNF use was already relatively common even among patients with early SpA in 2010. Our study also found peripheral arthritis, age, and functional ability to be related to total costs, indicating subgroups where anti-TNF use may be prioritized. In this study, we found that, in the first three years, treatment with anti-TNF agents leads to increased costs in other health resource use domains, not a reduction as is sometimes expected; follow-up is needed to determine whether this trend holds over a longer period.  This study has some limitations. The DESIR cohort includes only early SpA patients in France between 2007 and 2014; the results cannot necessarily be generalized to SpA patients with more advanced disease or in other settings at other times. Indeed, it is unknown whether similarly high rates of anti-TNF use exist among early SpA patients in other countries. Another recently-established early SpA cohort (SPACE) is ongoing in the Netherlands,208 but to our knowledge at   85 the time of writing the prevalence of anti-TNF use in this cohort has not yet been reported. A recent Canadian study of SpA patients seen clinically between 2003-2014 found that 48% had been treated with an anti-TNF agent, yet the mean disease duration (8 years) was much longer in this sample than in DESIR.234 In a cohort of Belgian patients, anti-TNF therapy was initiated in 44% of patients with definite AS and 16% of those with probable AS, for a total of 38%.235 Generalizable to early SpA patients in France, our findings will allow for future comparisons with other cohorts of early SpA patients, as well as the observation of trends within DESIR over time.  In terms of costing methods employed, we did not include non-SpA-related drugs, transportation, or presenteeism, or other patient out-of-pocket costs. Attendance declined over time and data imputation was required. Resource utilisation was self-reported and the potential exists for either underreported or overreported utilisation. In order to exclude the excess resource utilisation associated with study participation, we excluded the baseline visit from cost estimates. However, study participation could have also resulted in underestimation of resource utilisation following the baseline visit, if study visits replaced normal resource utilisation but were not reported as such. Finally, work productivity losses were estimated using average daily wages by professional category, estimated from the entire population of French workers. We did not further adjust wages for age and sex, as the age and sex distribution in DESIR was similar to that of the French working population (48% female and 77% aged 25-54 among French workers vs. 53% female and 54% female and 81% aged 25-54 in DESIR236). Although our valuation method provides a good estimate of productivity losses, we acknowledge that a greater level of accuracy could be achieved with direct self-report data on wage.  Despite these limitations, our findings are consistent with another recent description of early SpA patients237 and reflect patterns of health resource utilisation and work productivity loss within one of the largest early SpA cohorts to date. Given the size of the cohort and the detailed reporting of resource utilisation and productivity losses, this study may serve as a benchmark in terms of cost-of-illness among early SpA patients in the biologic era.     86 Chapter 3i   Measurable Definitions of Ankylosing Spondylitis Management Recommendations Are Needed for Use in Observational Studies  3.1 Introduction  Clinical practice guidelines for AS have been published by the ASAS and are applicable to all patients with axSpA.1 According to the guidelines, the goal of recommended treatment is to control symptoms and inflammation, prevent structural damage, and preserve function and social participation among SpA patients.1 By achieving these goals, it may be expected that recommended treatment should both reduce SpA-related costs and improve patients’ quality of life, as disease activity and functional ability are the most important predictors of these outcomes.178, 181, 184, 186, 238-240 However, the relationship between recommended treatment and patient outcomes requires careful evaluation. For ethical reasons, this evaluation is limited to observational studies, as patients may not be randomized to receive non-recommended care. In an observational study, a process must therefore be established to identify patients whose actual care is adherent and non-adherent, respectively, with treatment recommendations.  Currently, such a process among SpA patients is hindered by the structure of the ASAS practice guidelines, which are written as general principles and neglect to define specific actions that represent adherenceii or non-adherence with the recommendations. The problem of lack of specificity and absence of behaviour-based, clinical ‘action’ terms is not unique to the ASAS recommendations, but rather has been widely observed in clinical practice guidelines. In the literature, this problem has been chiefly discussed as it relates to the effective implementation of best practices: clinicians have been shown to be less likely to adhere to guidelines when                                                 i A version of Chapter 3 has been published: Harvard S, Gossec L, Pham T, Richette P, Dougados M, Anis A, Fautrel B. Measurable definitions of ankylosing spondylitis management recommendations are needed for use in observational studies. Joint Bone Spine. 2016 Jan;83(1):101-3. ii For the purpose of consistency, terms 'adherence' and 'adherent' are employed throughout this thesis to refer to the practice of following treatment recommendations. The original survey that was part of the study reported in this chapter used the synonyms 'compliance' and 'compliant', which were later found to be terms not preferred by rheumatologists, consistent with other reports. 329   87 behaviours and actions are not precisely specified.241 It has been noted that to be effectively implemented, guidelines require “actionable and precise” definitions of best practices, indicating who should do what, when, how, and in what patient group.242 When they are used systematically to measure practice implementation, such actionable definitions are known as ‘quality indicators’.242-245 In the context of research, data derived from quality indicators collected in practice are a valuable resource, and there is an outstanding need to develop quality indicators to help measure the implementation of clinical practice guidelines and evaluate their impact in SpA.246 However, the development of quality indicators is an intensive process, involving numerous stakeholders including patients, physicians and other caregivers.247 In the case of SpA, data from quality of care indicators collected in clinical practice may be years away from becoming a standard part of observational databases. At the same time, the amount of observational data available on SpA patients is widely expanding, given the establishment of several cohort studies,207, 212, 248, 249 including DESIR, a longitudinal study of 708 patients with early SpA symptoms in France.250 Although quality of care indicators are not yet available in SpA, there are nonetheless opportunities and imperatives to study the impact of recommended treatment on SpA patients. To do so requires a method of measuring recommended treatment using the data typically collected in longitudinal cohort studies and distinguishing patients on the basis of these data.  In this study, we aimed to establish, by expert consensus, measurable definitions of items in the ASAS practice guidelines that could be used in conjunction with observational data to identify SpA patients whose care is adherent or non-adherent, respectively, with recommendations.   88 3.2 Materials and Methods   3.2.1 Delphi Survey Development  A two-step process was implemented. First, a working group of 4 rheumatologists in France met in December 2012, to review the ASAS recommendations. Where possible, the working group proposed measurable definitions of items in the recommendations in terms of one or more specific clinical actions, while also identifying items that could not be transformed into actions. This was done with feasibility in mind, anticipating the application of the definitions using data from DESIR, an early SpA cohort study in which patient follow-up is every six months for the first year and every year thereafter.250 Proposals for measurable definitions of each item made by the working group were subsequently developed into a Delphi questionnaire for participant feedback. The draft questionnaire was reviewed for clarity and face validity by an external, English-speaking rheumatologist in Vancouver, Canada, and a small number of changes to wording were made following the review. The final questionnaire was made available online using Survey Monkey (a free-of-charge software in use in France) and optional enhanced security was selected. No personal information was solicited on the questionnaire. The study was reviewed and approved by the University of British Columbia Research Ethics Board.  3.2.2 Participant Recruitment   Invitations to participate in the Delphi study were sent by email to all 15 rheumatologists involved in the DESIR cohort study.250 All rheumatologists were based in Europe, with the majority located in France. Email invitations to participate included a description of the study, its objective, and methods. It specified that participants would view only aggregated responses from their colleagues, without linkage to individual identities. The email contained a link to the online survey, as well as an informed consent statement indicating that proceeding to the survey implied consent to participate, though participation was voluntary and could be stopped at any time.     89 3.2.3 Delphi Process  Following established methods,251-254 the Delphi questionnaire was completed in iterations (‘rounds’). In the questionnaire, participants were asked to agree to ‘drop’ items that the working group felt could not be transformed or to propose another option. For the remaining items, measurable definitions of recommended items were comprised of one or more actions. Participants were asked to agree or disagree with the definition as a whole, but were invited to give feedback on individual actions. Following Round 1, the percent agreement with each definition was presented, along with alternative actions or amendments to wording proposed by participants. Definitions that drew 75% or greater consensus from participants were accepted, while definitions that drew 25% consensus or less were eliminated; neither accepted nor eliminated definitions appeared in subsequent rounds. In Round 3 only, amendments to wording proposed by participants (which appeared only in Round Two) were eliminated if consensus was under 50% and accepted if consensus was over 75% (i.e., only amendments with consensus in between 50-74% in Round 2 re-appeared in Round 3). To help encourage consensus, the wording of response items in Round 3 was modified to a yes/no format (‘Yes, I can accept the definition’ vs. ‘No, I cannot accept the definition). The questionnaire was terminated at Round 3.   3.3 Results  3.3.1 Working Group Consensus and Definitions  Following the review of the ASAS practice guidelines, the working group reached the conclusion that directly identifying ‘adherence’ with recommendations would not be possible using observational data, but rather could only be achieved with quality indicators collected in practice specifically for that purpose. However, the working group agreed that identifying ‘non-adherence’ would be nonetheless be possible, as observational data could be more readily used to flag clinical actions that were clearly contradictory to the recommendations. Thus, measurable definitions were developed specifically in terms of actions that represented non-adherence with the recommended items.   90 Among 11 items in the ASAS recommendations, members of the working group agreed that 4 items could not be transformed into actions and therefore no measurable definitions were proposed for them. These items were: ‘General Treatment’, ‘Disease Monitoring’, ‘Analgesics’, and ‘Changes in Disease Course’. Questions on these items in the Delphi study asked if participants agreed that no action could be defined to represent non-adherence, or to propose another option.  3.3.2 Delphi Participant Feedback  Items that the working group found could not be transformed into actions drew high levels of agreement in Round 1: ‘General Treatment’ (100%), ‘Disease Monitoring’ (91.7%), ‘Analgesics’ (100%), ‘Changes in Disease Course’ (100%). These items were dropped, as was the item ‘Surgery’ (91.7%), for which one definition was proposed by the working group but which participants agreed could not be transformed into action (Table 3.2). These non-measurable items were the only ones to achieve high enough consensus to be terminated in Round 1. All other items drew either a proposed amendment or alternative definition from participants. Based on Round 1 feedback, Round 2 presented a number of possible definitions for and/or amendments to ‘Non-Pharmacological Therapy’, ‘Extra-Articular Manifestations and Co-Morbidities', 'Non-Steroidal Anti-Inflammatory Drugs', 'Disease-Modifying Anti-Rheumatic drugs', and ‘Anti-TNF Agents’ (Table 3.3).  Following Round 1 feedback, a proposal was made for a more sensitive definition for ‘Non-Pharmacological Therapy’, i.e., one more likely to detect non-adherence. This definition, comprised of a single action (‘if patient has 0 visits to physiotherapist by end of year 1 follow-up then treatment is non-adherent’) drew a higher level of consensus in Round 2 than the definition originally proposed by the working group (Table 3.3). In Round 3, this version was retained, but with a moderately high percentage of respondents (66%). Among the respondents who disagreed with the proposed definition, five provided feedback on their reasons. One participant felt that physiotherapy was not indicated for patients without severe disease: “…without modifications of mobility and no activity of the disease visits to the physiotherapist are not mandatory.” The remaining participants emphasized that independent or home exercises were equally effective:   91 “A significant number of patients do not require the help of a physiotherapist, even for practicing appropriate exercises. They can learn by themselves how to practice efficient physiotherapy. Sometimes, they don't even need specific exercises. Therefore non-adherent is too strong wording and cannot be generally applied.”   “A patient who regularly practices a sport activity but does not visit a physiotherapist keeps on being adherent for me.”  “Home exercises are effective. Regular exercise can be done without physiotherapist.”  “If the patient is performing home exercises this is also adherent.”  Regarding the item ‘Extra-Articular Manifestations and Co-Morbidities', a number of amendments were proposed in Round 1. These included the removal of two actions related to dermatologist care of pustulosis and psoriasis, respectively, as well as the addition of one action requiring cardiologist care of cardiovascular disease and one mandating the use of bone densitometry. These amendments appeared in Round 2, where the items regarding cardiologist care and bone densitometry were accepted (80%) and rejected (6.7%), respectively, with a high level of consensus. Although a significant number of participants agreed to drop the actions regarding dermatologist care of pustulosis (40%) and psoriasis (46.7%), these did not reach 50% agreement and did not appear in Round 3.  Regarding DMARD use, three amendments were proposed in Round 1 in terms of dropping actions from the working group’s proposed definition. The actions that some participants felt should be dropped were: ‘if patient has synovitis ≥3 at two consecutive visits and is not prescribed a DMARD at either of these visits’; ‘if at a given study visit, a patient is receiving SSZ and has no history of peripheral arthritis, IBD, or uveitis’; and ‘if patient has no history of peripheral arthritis, IBD, or uveitis and is receiving a DMARD other than MTX’. Amendments proposing to drop these actions appeared in Round 2, however, only one of these amendments received 50% or more agreement. The action in question was dropped (‘If patient has no history   92 of peripheral arthritis, IBD, or uveitis and is receiving a DMARD other than MTX’) while the others were retained.  One comment made during Round 2 argued strongly that physician opinion was a major point of the ASAS recommendation, and a requirement to start anti-TNF therapy. A new, modified definition was therefore proposed in Round 3, which included a PhGA criterion. Final definitions relating to the 6 items are presented in Table 3.4. The definitions for all items were accepted by 75% or more of participants, with the exception of the definition for ‘Non-Pharmacological Therapy’ (66% acceptance).   3.4 Discussion  In this study, experienced rheumatologists developed measurable definitions of the recommendations in the ASAS AS clinical practice guidelines, grouped under 6 of 11 domains of SpA care. These definitions are each comprised of one or more clinical actions, whereby the detection of one or more action in a single domain of SpA care is defined as non-adherence in that domain. The definitions were developed for applied purposes and actions that define non-adherence should be readily identifiable in clinical data typically collected in observational studies. In that the measures developed were derived from evidence-based management recommendations and reviewed by experts, we consider their clinical relevance and validity to be high. It should be noted that the failure to identify non-adherent actions in an observational database does not guarantee that a patient’s management is adherent. However, the detection of non-adherent actions will serve as an important marker that the patient’s care is inconsistent with the ASAS recommendations.  The decision to identify actions representing non-adherence rather than adherence was made by a working group of four experienced rheumatologists, following extensive discussion on how to interpret the ASAS recommendations. The working group consistently found that interpreting the recommendations in terms of adherence was complicated by the lack of statements regarding frequency and/or dose of interventions. At the same time, the group felt the recommendations were nonetheless clear about therapies that are not indicated in SpA management, whether   93 always or under certain circumstances. In these cases, the recommended frequency and dose were easily interpreted as ‘zero’. The fact that contra-indicated therapies were made clear, while the details on how to implement recommended practices were lacking, led to the decision to focus on actions which the recommendations advised against, i.e., definitions of ‘non-adherence’ comprised of one or more clinical actions.   In this study, two non-adherent actions were identified relative to anti-TNF agents, of which the first reflects the population of patients for whom an anti-TNF agent is indicated. Specifically, ASAS recommends the use of anti-TNF agents only among patients who satisfy the Modified New York AS criteria or ASAS criteria for axial Spa, who have active disease for 4 weeks or longer with a BASDAI of 4 or greater, who have had an adequate trial of 2 or more NSAIDs and a positive expert opinion.124 These criteria are clearly incorporated into the definition for the anti-TNF item, in which PhGA is used as a proxy for positive expert opinion. Importantly, this definitional approach may be applied in an observational study to identify the subset of SpA patients who satisfy the recommendations for an anti-TNF agent. This represents an important methodological contribution, as in observational studies examining the effectiveness of anti-TNF agents comparisons between patients using and not using anti-TNF agents must be done carefully in order to control for confounding by indication. In practice, this means comparisons should only be made within the subset of patients for whom anti-TNF agents are recommended, thereby comparing patients of comparable disease severity who are taking and not taking anti-TNF agents. To facilitate these comparisons, a method is needed to identify the subset of patients who satisfy the recommendations for anti-TNF use. The definition developed here may be used for this purpose, while noting its limitations.  The definition of non-adherence developed here for the anti-TNF item does not take into account contraindications to anti-TNF use. Rather, it identifies the general group of patients for whom an anti-TNF agent is recommended on the basis of disease severity and treatment history, from which the number of patients with contraindications could be further subtracted. The rheumatologists who participated in this study did not propose a method for subtracting patients with contraindications; this may be a reflection of the small proportion of patients seen in practice who have contraindications to anti-TNF agents, or simply of the applied research   94 purpose for which the definitions were developed. Again, the definitions developed here aim to identify markers of non-adherent care using observational data, not to confirm individual patients’ quality of care. Given this limitation, the development of quality of care indicators for use and collection in clinical practice remains an important priority in SpA research. However, for use in observational studies, the definitions of non-adherence developed here, including that for anti-TNF agents, should be valuable. In identifying the subset of patients for whom anti-TNF use is recommended, there is a question as to where to place patients who satisfy the relevant disease severity and treatment history criteria but who have a contraindication to anti-TNF agents. These patients may fit more appropriately in the subset of patients for whom anti-TNF agents are recommended, as they are more clinically similar to other patients in this subset in terms of disease severity than to patients outside it.  As in all Delphi exercises, arbitrary cut-off points were used in this study to define consensus. Although for most items consensus was reached as defined by the pre-set threshold, the study results reflect an important level of disagreement between rheumatologists regarding clinical processes in SpA care. In more than one instance, the amendments to definitions of non-adherence proposed by participants reflected disputes around the validity of specific clinical actions. For example, a large proportion of participants (44%) voted to drop the action ‘If at a given study visit, a patient is receiving SSZ and has no history of peripheral arthritis, IBD, or uveitis’, which appeared in the definition of non-adherence for the DMARDs item. As the percent agreement to drop this element was under 50% in Round 2, the final definition of non-adherence for the DMARDs item contains this action. Nonetheless, a substantial proportion of participants expressed the view that this clinical practice is not necessarily contrary to SpA management recommendations. Furthermore, as discussed, the definition for ‘Non-Pharmacological Therapy’ was not accepted by the set threshold of 75% or more participants, reflecting an important level of disagreement surrounding the need for physiotherapy. In the case of ‘Non-Pharmacological Therapy’ no statement is made in the ASAS recommendations regarding the optimal frequency of physiotherapy and participants in this study proposed their own opinions on this matter. Perhaps not surprisingly, it was this item that was most contended in the Delphi study, with some participants voicing that physiotherapy is not in fact always indicated for SpA patients. In light of this disagreement among experts, adaptations to the   95 recommendations should be made to better define the specific clinical practices advised and discouraged in SpA care, as well as the frequency of recommended practice. Given that guidelines are created using available evidence, this implies more generally that further research is needed to inform the recommended intensity of interventions in AS, particularly physiotherapy.  One limitation of the definitions developed here is that they may need to be modified for observational studies with different follow-up intervals.  The original proposals made by the working group were developed taking into consideration the nature of follow-up in the DESIR study, an ongoing study of SpA in which follow-up is every six months in the first year and every year thereafter.250 Despite this limitation, the data collected in DESIR represent standard clinical data that are likely to be collected in any observational study of SpA. The definitions of non-adherence developed here may therefore be readily adapted for use in other studies. The impact, validity, and reliability of the definitions of non-adherence may be further validated in the course of application in other studies.  This study represents an important step in developing a method of measuring recommended treatment in observational studies of SpA patients. The development of such measures is vital in the context of the wide expansion in observational data on SpA patients, as well as the pressing need to evaluate outcomes on the basis of treatment, including the recommended use of anti-TNF agents. Particularly given that cost-effectiveness data are in high demand regarding anti-TNF use in SpA patients, the observational data available now should be taken advantage of to examine this question. The results of the current study will facilitate an analysis of the costs and benefits of following ASAS recommendations, including those regarding anti-TNF agents. The definitions developed here should also help estimate the uptake of recommended SpA care, as well as inform updates to SpA management recommendations. The process of developing measurable definitions of SpA recommendations management uncovered important gaps in consensus around frequency around recommended interventions, which should be addressed in future studies.    96 Table 3.1 Summary of ASAS Recommendations 1 Item Bullet Point General Treatment “Treatment should be tailored to:   1.1 The current manifestations of the disease (axial, peripheral, entheseal, extra-articular symptoms and signs).   1.2 The level of current symptoms, clinical findings, and prognostic criteria.   1.3 The general clinical status (age, gender, comorbidity, concomitant medications, psychosocial factors).” Disease Monitoring “Disease monitoring should include:   2.1 Patient history (e.g., questionnaires)   2.2 Clinical parameters   2.3 Laboratory tests   2.4 Imaging (according to the clinical presentation as well as the ASAS core set) “ Non-Pharmacological Treatment  3.1 The cornerstone of non-pharmacological treatment is patient education and regular exercise.   3.2 Home exercises are effective. Physical therapy with supervised exercises, land or water based, individually or in a group, should be preferred as these are more effective than home exercises.   3.3 Patient associations and self-help groups may be useful.    Extra-Articular Manifestations and  Co-Morbidities  4.1 The frequently observed extra-articular manifestations, for example, psoriasis, uveitis and IBD, should be managed in collaboration with the respective specialists.   4.2. Rheumatologists should be aware of the increased risk of cardiovascular disease and osteoporosis  Non-Steroidal Anti-Inflammatory Drugs  5.1 NSAID, including Coxibs, are recommended as first-line drug treatment for AS patients with pain and stiffness.   5.2 Continuous treatment with NSAID is preferred for patients with persistently active, symptomatic disease.   5.3 Cardiovascular, gastrointestinal and renal risks should be taken into account when prescribing NSAID.    Analgesics  6.1 Analgesics, such as paracetamol and opioid (like) drugs, might be considered for residual pain after previously recommended treatments have failed, are contraindicated, and/or poorly tolerated.     Glucocorticoids  7.1 Corticosteroid injections directed to the local site of musculoskeletal inflammation may be considered.   7.2 The use of systemic glucocorticoids for axial disease is not supported by evidence.    Disease Modifying Anti-Rheumatics Drugs  8.1 There is no evidence for the efficacy of DMARD, including sulfasalazine and methotrexate, for the treatment of axial disease.   8.2 Sulfasalazine may be considered in patients with peripheral arthritis.    Anti-TNF Therapy  Anti-TNF therapy should be given to patients with persistently high disease activity despite conventional treatments according to the ASAS recommendations.   There is no evidence to support the obligatory use of DMARD before or concomitant with anti-TNF therapy in patients with axial disease.  There is no evidence to support a difference in efficacy of the various TNF inhibitors on the axial and articular/entheseal disease manifestations; but in the presence of IBD a difference in gastrointestinal efficacy needs to be taken into account.   Switching to a second TNF blocker might be beneficial especially in patients with loss of response.  There is no evidence to support the use of biological agents other than TNF inhibitors in AS.    Surgery  10.1 Total hip arthroplasty should be considered in patients with refractory pain or disability and radiographic evidence of structural damage, independent of age.   10.2 Spinal corrective osteotomy may be considered in patients with severe disabling deformity.   10.3 In patients with AS and an acute vertebral fracture a spinal surgeon should be consulted.   Changes in Disease Course If a significant change in the course of the disease occurs, other causes than inflammation, such as a spinal fracture, should be considered and appropriate evaluation, including imaging, should be performed.     97 Table 3.2 Proposals Considered in Round 1 and Percent Agreement ASAS Item Option 1 from Working Group  Option 2 from Working Group   Proposal:   Agreement (%) Proposal: Agreement (%) ASAS Item Proposal: Option 1 from Working Group Agree Option 1 (%) Proposal: Option 2 from Working Group Agree Option 2 (%) General Treatment Drop  100 N/A N/A Disease Monitoring Drop  91.7 N/A 8.3 Non-Pharmacological Treatment Drop   33.3 Define 'non-adherence' by satisfaction of 1 criterion (a):   a. If at a given study visit, a patient’s chest expansion is below normal or occiput wall is greater than 0, AND at the next study visit patient has 0 visits to a physiotherapist  58.3 Extra-Articular Manifestations and Co-Morbidities Drop  16.7  Define 'non-adherence' by satisfaction of ≥1 of 4 criteria (ANY of a, b, c, d):     a. If at a given study visit, a patient has a new diagnosis of uveitis AND has not had an opthamologist consult by the next study visit   b. If at a given study visit, a patient has a new diagnosis of psoriasis AND has not had a dermatologist consult by the next study visit   c. If at a given study visit, a patient has a new diagnosis of pustulosis AND has not had a dermatologist consult by the next study visit  d. If at a given study visit, a patient has a new diagnosis of IBD AND has not had a gastroenterologist consult by the next study visit  d. If at a given study visit, a patient has a new diagnosis of IBD AND has not had a gastroenterologist consult by the next study visit 58.3 Non-Steroidal Anti-Inflammatory Drugs  Non-Steroidal Anti-Inflammatory Drugs – cont.  Define 'non-adherence' by satisfaction of ≥1 of 3 criteria:  a. If patient received their first DMARD before their first NSAID   b. If patient has diagnosis of renal insufficiency (i.e., creatinine clearance < 30 ml/ min) AND NSAID use is not interrupted within 15 days of that diagnosis (as assessed at next study visit)   c. If patient has history of GI event AND receives an NSAID without a concomitant PPI   41.7 Define 'non-adherence' as by satisfaction of ≥1 of the 3 criteria from Option 1 above (a, b, c, d) or 3 additional criteria below:   d. If patient has BASDAI Q2 =≥4 or BASDAI (Q5 + Q6)/2=≥4 at two consecutive visits and is not prescribed an NSAID at the second of those visits (*except if renal insufficiency as above is applicable)   e. If patient has a history of myocardial infarction and thereafter receives an NSAID other than Naproxen    f.  If patient has a history of unstable coronary heart disease and thereafter receives an NSAID other than Naproxen  33.3 Analgesics Drop 100 N/A N/A Glucocorticoids Drop 25.0 Define 'non-adherence' as follows, by satisfaction of 1 criterion (a):  a. If at a given study visit, a patient is receiving oral prednisone or equivalent AND has no history of uveitis, peripheral arthritis or inflammatory bowel disease 75.0        98 ASAS Item Option 1 from Working Group  Option 2 from Working Group  Disease Modifying Anti-Rheumatics Drugs Define 'non-adherence' by satisfaction of ≥1 of 4 criteria:   a. If patient has synovitis ≥3 at two consecutive visits and is not prescribed a DMARD at either of these visits b. If at a given study visit, a patient is receiving MTX and has no history of peripheral arthritis or psoriasis   c. If at a given study visit, a patient is receiving SSZ and has no history of peripheral arthritis, IBD, or uveitis  d. If patient has no history of peripheral arthritis, IBD, or uveitis and is receiving a DMARD other than MTX 63.6 N/A N/A Anti-TNF Therapy  Define 'non-adherence' by satisfaction of ≥1 of 2 criteria (a, b):  a. If at two consecutive study visits, patient has had at least 2 adequate therapeutic trials of NSAIDs and BASDAI is ≥4, AND an anti-TNF agent has not been prescribed at the 3rd visit    b. If patient is receiving a biological agent other than anti-TNF  50.0 Define 'non-adherence' by satisfaction of ≥1 of (a, b) from Option 1 or of 1 additional criterion:   c. If patient has IBD or uveitis and is receiving Etanercept 25.0 Surgery Drop 91.7 Define 'non-adherence' as by satisfaction of 1 criterion:   a. If patient has an acute vertebral fracture and a spinal surgeon is not consulted within X weeks  8.3 Changes in Disease Course Drop 100 N/A N/A   99 Table 3.3 Proposals Considered in Round 2 and Percent Agreement ASAS Item Proposal (NB: items in italics appeared first in Round 1; bold items appeared first in Round 2) Agree Option 1 (%) Non-Pharmacological Treatment Define 'non-adherence' by satisfaction of 1 criterion (a):   a. If at a given study visit, a patient’s chest expansion is below normal or occiput wall is greater than 0, AND at the next study visit patient has 0 visits to a physiotherapist  Define 'non-adherence' by satisfaction of 1 criterion (a) (proposed in Round 1):   a. If by the 12-month follow-up visit, patient has had 0 visits to the physiotherapist.  Drop 40.0    53.3   6.7 Extra-Articular Manifestations and  Co-Morbidities Accept Round 1 majority definition with amendment: Drop criterion b  b. If at a given study visit, a patient has a new diagnosis of psoriasis AND has not had a dermatologist consult by the next study visit    Accept Round 1 majority definition with amendment: Drop criterion c  c. If at a given study visit, a patient has a new diagnosis of pustulosis AND has not had a dermatologist consult by the next study visit    Accept Round 1 majority definition with amendment: Add new criterion e:  e. If by the 12-month follow-up visit, a bone densitometry is not available   Accept Round 1 majority definition with amendment: Add new criterion f: f.  If at a given study visit, a patient has a new cardiovascular event AND has not had a cardiologist consult by the next study visit 40.0    46.7    6.7   80.0 Non-Steroidal Anti-Inflammatory Drugs Define 'non-adherence' by satisfaction of ≥1 of 3 criteria:  a. If patient received their first DMARD before their first NSAID   b. If patient has diagnosis of renal insufficiency (i.e., creatinine clearance < 30 ml/ min) AND NSAID use is not interrupted within 15 days of that diagnosis (as assessed at next study visit)   c. If patient has history of GI event AND receives an NSAID without a concomitant PPI    Define 'non-adherence' by satisfaction of ≥1 of the 3 criteria from Option 1 or 3 additional criteria:   d. If patient has BASDAI Q2 =≥4 or BASDAI (Q5 + Q6)/2=≥4 at two consecutive visits and is not prescribed an NSAID at the second of those visits (*except if renal insufficiency as above is applicable)  e. If patient has a history of myocardial infarction and thereafter receives an NSAID other than Naproxen   f.  If patient has a history of unstable coronary heart disease and thereafter receives an NSAID other than Naproxen  Define 'non-adherence' by satisfaction of a single criterion:   a. If patient received their first DMARD before their first NSAID  Accept Round 1 majority definition with amendment: Modify criterion b ‘Indicator b should indicate 8 days (i.e., "If patient has diagnosis of renal insufficiency...AND NSAID use is not interrupted within 8 days of that diagnosis...")’     Accept Round 1 majority definition with amendment: Modify criterion c ‘Indicator c should define GI event to exclude dyspepsia’ 66.7      20.0        20.0   46.7   60.0 Disease Modifying Anti-Rheumatics Drugs Accept Round 1 majority definition with amendment: Drop criterion a a. If patient has synovitis ≥3 at two consecutive visits and is not prescribed a DMARD at either of these visits  Accept Round 1 majority definition with amendment: Drop criterion c c. If at a given study visit, a patient is receiving SSZ and has no history of peripheral arthritis, IBD, or uveitis  Accept Round 1 majority definition with amendment: Drop criterion d d. If patient has no history of peripheral arthritis, IBD, or uveitis and is receiving a DMARD other than MTX 33.3    44.4    77.8 Anti-TNF Therapy  Accept Round 1 majority definition with amendment: Modify criterion by adding an exception b. If patient is receiving a biological agent other than anti-TNF b.if patient is receiving a biological agent other than anti-TNF (*EXCEPTION: psoriatic patients may receive a biologic other than anti-TNF, but then cannot receive a concomitant anti-TNF)  Accept Round 1 majority definition with amendment: Add new criterion d d. If patient is prescribed anti-TNF with a concomitant DMARD    Accept Round 1 majority definition with amendment: Add new criterion e e. If patient is prescribed anti-TNF without pre-treatment with a DMARD while patient has synovitis >=3  81.8     9.1   27.3   100 Table 3.4 Final Quantitative Definitions of 6 of 11 ASAS Recommendations ASAS Item Quantitative Definition of Recommendation Deciding Round % Agreement/ Acceptance* Non-Pharmacological Therapy Define 'non-adherence' by satisfaction of 1 criterion (a):  a. If by the 12-month follow-up visit, patient has had 0 visits to the physiotherapist 3 66.7 Extra-Articular Manifestations and Co-Morbidities Define 'non-adherence' as by satisfaction of ≥1 of 5 criteria:  a. If at a given study visit, a patient has a new diagnosis of uveitis AND has not had an opthamologist consult by the next study visit  b. If at a given study visit, a patient has a new diagnosis of psoriasis AND has not had a dermatologist consult by the next study visit  c. If at a given study visit, a patient has a new diagnosis of pustulosis AND has not had a dermatologist consult by the next study visit  d. If at a given study visit, a patient has a new diagnosis of IBD AND has not had a gastroenterologist consult by the next study visit  e. If at a given study visit, a patient has a new cardiovascular event and has not had a cardiologist consult by the next study visit 2 80.0 Non-Steroidal Anti-Inflammatory Drugs Define 'non-adherence' by satisfaction of ≥1 of 3 criteria:  a. If patient received their first DMARD before their first NSAID  b. If patient has diagnosis of renal insufficiency (i.e., creatinine clearance < 30 ml/ min) and NSAID use is not interrupted within 15 days of that diagnosis (as assessed at next study visit)  c. If patient has history of GI event other than dyspepsia and receives an NSAID or Cox inhibitor without a concomitant PPI 3 93.3 Glucocorticoids Define 'non-adherence' by satisfaction of 1 criterion (a):  a. If at a given study visit, a patient is receiving oral prednisone or equivalent and has no history of uveitis, peripheral arthritis or inflammatory bowel disease 1 75.0 Disease-Modifying Anti-Rheumatic Drugs Define 'non-adherence' by satisfaction of ≥1 of 3 criteria:  a. If patient has synovitis ≥3 at two consecutive visits and is not prescribed a DMARD at either of these visits  b. If at a given study visit, a patient is receiving MTX and has no history of peripheral arthritis or psoriasis  c. If at a given study visit, a patient is receiving SSZ and has no history of peripheral arthritis, IBD, or uveitis 2 77.8 Anti-TNF Agents Define 'non-adherence' by satisfaction of ≥1 of 2 criteria:  a. If at two consecutive study visits, patient has had at least 2 adequate therapeutic trials of NSAIDs (i.e., minimum two NSAIDs over a 4-week period in total since symptom onset), BASDAI is ≥4, Physician's Global Assessment of Disease is ≥4 AND an anti-TNF agent has not been prescribed at the 3rd visit  b. If patient is receiving a biological agent other than anti-TNF (*EXCEPTION: psoriatic patients may receive a biologic other than anti-TNF, but then cannot receive a concomitant anti-TNF) 3 93.3 *Rounds 1 and 2 inquired as to ‘agreement’ with proposed definitions (“I agree with option” vs. “I propose another option”) while Round 3 inquired as to ‘acceptance’ (“I can accept the definition” vs. “I cannot accept the definition”).      101 Chapter 4i   Adherence to Anti-TNF Use Recommendations in Spondyloarthritis: Measurement and Impact in the DESIR Cohort  4.1 Introduction  The ASAS AS management recommendations apply to all patients with axSpA.1 Recommendations outline the use of medication, including NSAIDs, analgesics, DMARDs, glucocorticoids, and anti-TNF agents, as well as non-pharmacological therapy and specialist management of extra-articular symptoms. In general, recommended management aims to reduce symptoms and preserve patients' function and social participation 1. These outcomes are also associated with costs and quality of life among axSpA patients. 178, 181, 184, 186, 238-240   To our knowledge, no studies have examined to what extent axSpA care in clinical practice follows the ASAS recommendations, or how recommended care affects patient outcomes. One important barrier is the lack of validated methods to define or measure adherence to recommended axSpA care. Recently, we asked rheumatologists involved in DESIR, a longitudinal study of early axSpA patients 218, how adherence or non-adherence with the ASAS recommendations might be measured using observational data 255. In a Delphi process, rheumatologists developed a classification system based on markers of non-adherence, defined as clinical actions clearly discordant with the recommendations. Adherence was then defined in terms of the absence of markers of non-adherence. This system to define adherence, like any classification system, reflects the need to balance sensitivity and specificity according to the perceived consequences of both false negatives and false positives 256. Furthermore, like in the case of diagnostic or screening tests, there is bound to be a range of values that do not clearly indicate how to best classify the patient 257. Using observational data alone, perfect discrimination between adherence and non-adherence to axSpA management recommendations is unlikely. However, the system developed 255 provides a means to explore differences between axSpA patients with and without clear markers of non-adherent management.                                                 i A version of Chapter 4 has been accepted for publication: Harvard S, Guh D, Bansback N, Richette P, Saraux A, Fautrel B, Anis A. Adherence to Anti-TNF Use Recommendations in Spondyloarthritis: Measurement and Impact in the DESIR Cohort.   102 In this study, we aimed to evaluate the use of this classification system among DESIR patients, measuring costs and health status across groups defined by adherence to ASAS anti-TNF recommendations while controlling for adherence to other recommendations.  4.2 Methods  4.2.1 Data Source and Study Population  The DESIR cohort 218 includes 708 patients aged 18-50 with inflammatory back pain 6, 219 suggestive of SpA lasting >3 months and < 3 years. Patients with definite diagnosis of non-SpA back pain, history of anti-TNF use, or conditions affecting informed consent were excluded. Follow-up occurred every six months in the first two years and every year thereafter. Our analyses included data from the first three years, i.e., baseline plus follow-up visits at months 6-36. DESIR data include clinical history, quality of life (i.e., SF-36), and total health resource use and work productivity loss costs 258. Briefly, total costs were estimated in 2013 Euros using public cost data linked to self-reported, all-cause health resource utilisation (i.e., health practitioner visits, hospitalizations, medical workups, medications) and work productivity losses, calculated as number of work days lost multiplied by daily estimated wage. Patient out-of-pocket costs were not included. Missing data were imputed using Monte Carlo Markov Chain multiple imputation, last observation carried forward, probabilistic imputation, or negative values based on clinical expertise 258. The current analysis included patients who satisfied the ASAS criteria for axSpA 259.   4.2.2 Classification of Adherence   We used DESIR data to evaluate a classification system designed for use with observational data to define adherence to ASAS anti-TNF use and other care recommendations (4.1) 255. The definition of anti-TNF adherence considers timing of anti-TNF initiation relative to disease activity on BASDAI and PhGA ('Physician's Global Assessment', a proxy for positive expert opinion, cited by ASAS as a requirement for anti-TNF use). All patients who receive an adequate trial of NSAIDs who experience BASDAI and PhGA ≥4 at 2 consecutive visits 6 months apart   103 must receive an anti-TNF at the subsequent visit to be defined as adherent to recommendations; all anti-TNF use initiated before 2 consecutive visits with BASDAI and PhGA ≥4 is classed as adherent. Only psoriatic patients may receive a biologic other than anti-TNF and then cannot receive a concomitant anti-TNF. The system also defines adherence relative to recommended physiotherapy, specialist care for extra-articular manifestations and co-morbidities, and NSAID, glucocorticoid, and DMARD use (Table 4.1).   In preliminary analyses, many patients were missing data on NSAID use and few experienced ≥2 consecutive visits with BASDAI and PhGA ≥4. To have an adequate number of patients for analysis, the original definition of anti-TNF adherence 255 was adapted as follows: all anti-TNF users were assumed to have had an adequate trial of NSAIDs and patients with BASDAI and PhGA ≥4 at 2 consecutive visits 6 months apart had to receive an anti-TNF at the 2nd of those visits (rather than the subsequent visit) to be defined as adherent. All anti-TNF use initiated before 2 consecutive visits with BASDAI and PhGA ≥4 was considered adherent. Reasons for anti-TNF non-use were not evaluated (data unavailable). No other adherence definitions were adapted.  In classifying patients, we aimed to group patients of similar disease severity over equal observation periods, limiting potential confounding by indication as much as possible. To do so, each patient was assigned an index date. For anti-TNF users, the index date was date of anti-TNF initiation. For anti-TNF non-users, the index date was the 2nd consecutive visit with BASDAI + PhGA ≥4 or, where not applicable, the 2nd visit within the 2 consecutive visits with highest mean BASDAI prior to month 24; in the case of ≥1 pairs of consecutive visits with equal average BASDAI, the earliest pair was chosen. Classification of adherence to recommendations other than anti-TNF use was then done considering the period up to and including the index date only.  To explore the validity of adherence groupings, an intermediate analysis was undertaken in which patients were stratified by anti-TNF use (yes/no) and number of visits with 'high disease activity' pre-index, defined as both BASDAI and PhGA ≥4  at 0 visits, ≥1 non-consecutive visits, 2 consecutive visits, or >2 consecutive visits pre-index. Anti-TNF users and non-users in each pre-index disease activity group were compared for significant differences on baseline disease   104 severity markers, including baseline CRP, BASDAI, BASFI, sacroiilitis or spinal inflammation on X-Ray, CT or MRI, peripheral arthritis, and CRP one visit pre-index, using Chi-square, t-tests, and ANOVA as appropriate. Anti-TNF users across pre-index disease activity groups were compared on positive anti-TNF response following the ASAS definition (i.e., 50% relative BASDAI change or absolute change of 2 (on 0–10 scale) 124) using Chi-square. The analytic strategies for assigning patients to adherence groups in the main analysis and a sensitivity analysis were selected following these comparisons (Table 4.2).  In the main and sensitivity analyses, respectively, patients were classed using two alternate definitions of adherence to anti-TNF recommendations (Table 4.3). In the main analysis, patients with high disease activity at 2 consecutive visits who received an anti-TNF agent on the 2nd of those visits were classed as 'adherent' users, i.e., timely anti-TNF use. In the sensitivity analysis, these patients were classed as non-adherent users, i.e., late anti-TNF use (Table 4.3). Descriptive statistics were produced to compare the characteristics and outcomes of patients by adherence group, as well as subsets where appropriate.   4.2.3 Impact of Adherence Classifications  Regression models were developed to estimate total costs, costs excluding anti-TNF ('non-biologic costs'), and quality adjusted life years (QALYs) across groups defined by adherence to anti-TNF recommendations, while controlling for adherence to other recommendations. All dependent variables were calculated over the 1 year following the patient's index date. To estimate QALYs, SF-36 questionnaire data were converted into SF-6D health states and QALYs were calculated using corresponding utility scores following the area under the curve method 260. The primary independent variable in all models was adherence to anti-TNF recommendations, alternately defined in the main and sensitivity analyses. Adherence to other ASAS recommendations (i.e., physiotherapy, non-biologic drugs, specialist care, defined in Table 4.1) in the period up to and including the patient's index date were also tested as independent variables. Adherence to NSAID, glucocorticoid, and DMARD recommendations was tested as a single variable 'non-biologic drug recommendations'. Adherence to recommendations for specialist care for pustulosis and cardiac events was not examined due to few patients affected.   105 Given the risk of confounding by indication, sociodemographic and clinical variables were tested, including baseline age, sex, education, profession, smoking ('yes' vs. 'no/don't know'), baseline CRP and CRP one visit pre-index, baseline sacroiilitis or spinal inflammation on X-Ray, CT or MRI, peripheral arthritis, marital status, and number of months on anti-TNF. BASDAI and BASFI scores were not included as independent variables, as preliminary analyses suggested these were captured by adherence groupings.  Generalized linear regression models assuming gamma distribution with log link were used for costs outcomes, while linear models were used for the QALY outcome. In model development, independent variables were first tested in univariate models of each outcome and those significantly associated with outcomes at p<0.20 were included in multivariate models. Multivariate model selection was then done in a backward stepwise manner, beginning with all independent variables selected and removing those not associated with the outcome at p<0.05 and goodness-of-fit was assessed based on the Akaike Information Criterion (AIC). In all models, the reference group was adherent anti-TNF users. All analyses were performed using SAS 9.4.   4.3 Results  4.3.1 Classification of Adherence  A total of 469 patients met the ASAS criteria and were included in the analysis. Table 4.2 shows patients' clinical characteristics by anti-TNF use and timing of initiation relative to disease activity pre-index. Among patients who had 0 visits with high disease activity pre-index, anti-TNF users had significantly more baseline peripheral arthritis (63.0% vs. 38.2%; p=0.015), higher baseline BASDAI (33.9 ±  13.7 vs. 24.8 ±  13.6; p=0.003), and BASFI (33.2 ±  18.6 vs. 14.2 ±  14.0; p<.0001, and higher baseline CRP (17.0 ± 16.2 vs. 5.4 ± 7.3; -p=0.002) and CRP one visit pre-index. Among patients who had ≥1 more non-consecutive visits with high disease activity pre-index, anti-TNF users had significantly higher baseline BASDAI, BASFI, and CRP compared to non-users (Table 4.2) and significantly more baseline sacroiilitis/spinal inflammation visible on X-Ray, CT or MRI (77.8% vs. 58.5%; p=0.008). In both the main and   106 sensitivity analyses, patients with 0 or  ≥1 more non-consecutive visits with high disease activity pre-index were classed as adherent users (timely use).  Among patients who had 2 consecutive visits with high disease activity pre-index, anti-TNF users had significantly higher baseline BASDAI (64.5 ±  10.0 vs. 54.8 ±  15.7; p=.002) and BASFI (52.7 ±  21.6 vs. 39.6 ±  24.4; p=0.029) and more baseline peripheral arthritis (85.0% vs. 54.7%; p=0.015) compared to anti-TNF non-users (Table 4.2). Also, the 20 anti-TNF users 2 consecutive visits with high disease activity had significantly lower positive response to anti-TNF therapy compared to anti-TNF users with 0 visits (44.4% vs. 5.0%; p=0.003), ≥1 visits (58.0% vs. 5.0% p=0.0001) or ≥2 visits of high disease activity pre-index (50.0% vs. 5%; p=0.002) (4.2). The 20 anti-TNF users with 2 consecutive visits of high disease activity pre-index were classed as adherent users (timely anti-TNF use) in the main analysis, and as non-adherent users (late anti-TNF use) in the sensitivity analysis.  Patients classed as adherent non-users are described in Table 4.7. Table 4.4 shows clinical characteristics, cost outcomes, and health status of patients classed as adherent users (timely anti-TNF use), non-adherent users (late anti-TNF use), and non-adherent non-users (unmet anti-TNF need) in the main analysis and sensitivity analysis, respectively. A subset analysis of the 20 anti-TNF users classed as adherent users in the main analysis and as non-adherent users in the sensitivity analysis indicated these patients had the lowest health status post-index (0.49 ± 0.13), as well as the highest total costs (19,586 ± 8,263), non-biologic costs (7,987 ± 6,939), non-biologic health resource use costs (3,500 ± 3,127), and productivity loss costs (4,487 ± 7,196).    Across all groups defined in the main analysis, a total of 208 patients (44.3%) had treatment non-adherent to physiotherapy recommendations, i.e., did not receive ≥1 physiotherapy visit in the first year.  A total of 39 patients (8.3%) had treatment non-adherent to one or more non-biologic drug recommendations. Non-adherence to specialist care for uveitis, psoriasis, or inflammatory bowel disease was infrequently observed (Table 4.4 and Table 4.7).     107 4.3.2 Impact of Adherence Classifications  4.3.2.1 Cost Outcomes  Table 4.5 shows the multivariate models of cost outcomes produced in the main and sensitivity analyses. In the main analysis, non-adherent users (late anti-TNF use) and adherent users (timely anti-TNF use) showed no significant differences in total costs (RR: 0.86; 95% CI: 0.54, 1.36; p=0.516) or non-biologic costs (RR: 0.72; 95% CI: 0.44, 1.18; p=0.187). Relative to adherent users, non-adherent non-users (unmet anti-TNF need), had significantly lower total costs (RR: 0.11; 95% CI: 0.08, 0.15; p<0.001) and significantly lower non-biologic costs (RR: 0.56; 95% CI: 0.39, 0.79; p<0.001). In the main analysis, age and female sex were associated with increased total and non-biologic costs; being unmarried was associated with decreased non-biologic costs (Table 4.5). Other independent variables tested in univariate models, including non-adherence to other recommendations, were not significant in multivariate models in the main or sensitivity analyses.  In the sensitivity analysis, non-adherent users (late anti-TNF use) and adherent users (timely anti-TNF use) showed no significant differences in total costs (RR: 0.94; 95% CI: 0.65, 1.37; p=0.753). However, non-adherent, i.e., 'late', anti-TNF users had significantly increased non-biologic costs (RR: 1.58; 95% CI: 1.06, 2.36; p=0.026) relative to adherent users. Relative to adherent users (timely anti-TNF use), non-adherent non-users (unmet anti-TNF need) had significantly lower total costs (RR: 0.11; 95% CI: 0.08, 0.15; p<0.001) and significantly lower non-biologic costs (RR: 0.68; 95% CI: 0.48, 0.98; p=0.036).   108 Table 4.1 Rheumatologist-Proposed Definitions of Adherence to ASAS Recommendations  ASAS Item Adherence Definition Physiotherapy**  Define 'non-adherence' by satisfaction of 1 criterion (a): a. If by the 12-month follow-up visit, patient has had 0 visits to the physiotherapist Extra-Articular Manifestations and Co-Morbidities Define 'non-adherence' as by satisfaction of ≥1 of 5 criteria: a. If at a given study visit, a patient has a new diagnosis of uveitis AND has not had an ophthalmologist consult by the next study visit b. If at a given study visit, a patient has a new diagnosis of psoriasis AND has not had a dermatologist consult by the next study visit c. If at a given study visit, a patient has a new diagnosis of pustulosis AND has not had a dermatologist consult by the next study visit d. If at a given study visit, a patient has a new diagnosis of IBD AND has not had a gastroenterologist consult by the next study visit e. If at a given study visit, a patient has a new cardiovascular event and has not had a cardiologist consult by the next study visit Non-Steroidal Anti-Inflammatory Drugs Define 'non-adherence' by satisfaction of ≥1 of 3 criteria: a. If patient received their first DMARD before their first NSAID b. If patient has diagnosis of renal insufficiency (i.e., creatinine clearance < 30 ml/ min) and NSAID use is not interrupted within 15 days of that diagnosis (as assessed at next study visit) c. If patient has history of GI event other than dyspepsia and receives an NSAID or Cox inhibitor without a concomitant PPI Glucocorticoids Define 'non-adherence' by satisfaction of 1 criterion (a): a. If at a given study visit, a patient is receiving oral prednisone or equivalent and has no history of uveitis, peripheral arthritis or inflammatory bowel disease Disease-Modifying Anti-Rheumatic Drugs Define 'non-adherence' by satisfaction of ≥1 of 3 criteria: a. If patient has synovitis ≥3 at two consecutive visits and is not prescribed a DMARD at either of these visits b. If at a given study visit, a patient is receiving MTX and has no history of peripheral arthritis or psoriasis c. If at a given study visit, a patient is receiving SSZ and has no history of peripheral arthritis, IBD, or uveitis Anti-TNF Agents Define 'non-adherence' by satisfaction of ≥1 of 2 criteria: a. If at two consecutive study visits, patient has had at least 2 adequate therapeutic trials of NSAIDs (i.e., minimum two NSAIDs over a 4-week period in total since symptom onset), BASDAI is ≥4, PhGA Assessment of Disease is ≥4 AND an anti-TNF agent has not been prescribed at the 3rd visit b. If patient is receiving a biological agent other than anti-TNF† (EXCEPTION: psoriatic patients may receive a biologic other than anti-TNF, but then cannot receive a concomitant anti-TNF) *Originally defined as 'compliance'; **ASAS item is 'Non-Pharmacological Therapy', but was defined in exclusively on the basis of physiotherapy; †No DESIR patients were receiving biological agents other than anti-TNF   109 Table 4.2 Patient Clinical Characteristics by Anti-TNF Use and Disease Activity Pre-Index   0 visits BASDAI & PhGA ≥ 4 pre-index  ≥ 1 non-consecutive visits BASDAI & PhGA ≥ 4 pre-index 2 consecutive visits BASDAI & PhGA ≥ 4 pre-index >2 consecutive visits BASDAI & PhGA ≥ 4 pre-index Clinical marker Anti-TNF Use (n=30) No Anti-TNF Use (n=173) p value Anti-TNF Use (n=85) No Anti-TNF Use (n=82) p value Anti-TNF Use (n=20) No Anti-TNF Use (n=67) p value Anti-TNF Use (n=29) No Anti-TNF use (n=0) Baseline BASDAI 33.9 ±  13.7 24.8 ±  13.6 0.003 57.5 ±  13.1 51.1 ±  15.6 0.005 64.5 ±  10.0 54.8 ±  15.7 0.002 57.2 ±  14.2 NA Baseline BASFI 33.2 ±  18.6 14.2 ±  14.0 <.0001 39.7 ±  19.7 32.7 ±  21.0 0.030 52.7 ±  21.6 39.6 ±  24.4 0.029 46.0 ±  20.6 NA Peripheral Arthritis 17 ( 63.0%) 66 ( 38.2%) 0.015 56 ( 69.1%) 52 ( 63.4%) 0.440 17 (85.0%) 35 (54.7%) 0.015 17 ( 70.8%) NA Baseline sacroiilitis or spinal inflammation on X-Ray, CT or MRI 21 ( 77.8%) 105 ( 60.7%) 0.087 63 ( 77.8%) 48 ( 58.5%) 0.008 10 ( 50.0%) 32 ( 50.0%) 1.000 14 ( 58.3%) NA Baseline CRP 17.0 ± 16.2 5.4 ± 7.3 0.002 20.0 ± 24.1 9.1 ± 12.7 0.001 9.3 ± 10.5 7.0 ± 10.7 0.463 10.2 ± 12.0 NA CRP one visit prior to index date 14.8 ± 16.0 4.9 ± 6.8 0.004 18.9 ± 23.8 7.5 ± 10.7 0.001 9.4 ± 10.9 6.3 ± 7.6 0.327 5.6 ± 6.1 NA Positive Anti-TNF Response 12 ( 44.4%) NA   47 ( 58.0%) NA   1 (5.0%)§  NA   12 ( 50.0%) NA Analtyic strategy for main analysis Class anti-TNF use as adherent ('timely use') Class anti-TNF non-use as adherent ('no anti-TNF need')   Class anti-TNF use as adherent ('timely use') Class anti-TNF non-use as adherent ('no anti-TNF need')   Class anti-TNF use as adherent ('timely use') Class anti-TNF non-use as non-adherent ('unmet anti-TNF need')   Class anti-TNF use as non-adherent ('late use') NA   110 Analytic strategy for sensitivity analysis Class anti-TNF use as adherent ('timely use') Class anti-TNF non-use as adherent ('no anti-TNF need')   Class anti-TNF use as adherent ('timely use') Class anti-TNF non-use as adherent ('no anti-TNF need')   Class anti-TNF use as non-adherent ('late use') Class anti-TNF non-use as non-adherent ('unmet anti-TNF need')   Class anti-TNF use as non-adherent ('late use') NA §Positive anti-TNF response in this group significantly different from all other anti-TNF user groups, all pairwise comparisons at p<0.003   111 4.3.2.2 Health Outcomes  Table 4.6 shows the multivariate model of QALY outcomes in the main and sensitivity analyses. In the main analysis, there were no significant differences in health status between adherent users (timely anti-TNF use) and non-adherent users (late anti-TNF use), nor between adherent users and non-adherent non-users (unmet anti-TNF need). Baseline post-secondary education was associated with significantly higher health status, while smoking and female sex was associated with significantly lower health status (Table 4.6). Other independent variables tested in the main analysis, including non-adherence to other recommendations, were not significant in multivariate models. In the sensitivity analysis, non-adherent anti-TNF users (late anti-TNF use) had significantly lower health status relative to adherent users (timely use) (-0.06; 95% CI: -0.09, 0.-0.03; p=0.001). Non-adherent non-users (unmet anti-TNF need) also had significantly lower health status than adherent 'timely' users (-0.04; 95% CI: -0.07, 0.-0.01; p=0.016).  Table 4.3 Alternate Definitions Used to Classify Patients Based on Adherence to Anti-TNF Recommendations MAIN ANALYSIS  No Anti-TNF Use Anti-TNF Use Disease activity pre-index 0 or ≥1 non-consecutive visits  or BASDAI & PhGA ≥ 4 for anti-TNF non-users  Maximum 2 consecutive visits BASDAI & PhGA  ≥ 4 for anti-TNF users   Adherent Non-User  (No Anti-TNF Need)  Adherent User  (Timely Anti-TNF Use) >2 consecutive visits BASDAI & PhGA ≥ 4  Non-Adherent Non-User  (Unmet Anti-TNF Need)  Non-Adherent User  (Late Anti-TNF Use) SENSITIVITY ANALYSIS  No Anti-TNF Use Anti-TNF Use Disease activity pre-index 0 or ≥1 non-consecutive visits BASDAI & PhGA ≥ 4  Adherent Non-User  (No Anti-TNF Need)  Adherent User  (Timely Anti-TNF Use) ≥2 consecutive visits BASDAI & PhGA ≥ 4  Non-Adherent Non-User  (Unmet Anti-TNF Need)  Non-Adherent User  (Late Anti-TNF Use)   112 Table 4.4 Characteristics and Outcomes of Patients Defined as Adherent Users, Non-Adherent Users, and Non-Adherent Non-Users   ADHERENCE CLASSIFICATION 1: MAIN ANALYSIS  ADHERENCE CLASSIFICATION 2: SENSITIVITY ANALYSIS Patient characteristics ADHERENT USER: TIMELY ANTI-TNF USE (n=135) NON-ADHERENT USER: LATE ANTI-TNF USE (n=29) NON-ADHERENT NON-USER: UNMET ANTI-TNF NEED (n=67) ADHERENT USER: TIMELY ANTI-TNF USE (n=115) NON-ADHERENT USER: LATE ANTI-TNF USE (n=49) NON-ADHERENT NON-USER: UNMET ANTI-TNF NEED (n=67) Baseline age 33.8 ± 9.7 33.8 ± 7.5 33.7 ± 7.9 33.5 ± 10.0 34.7 ± 7.8 33.7 ± 7.9 Male  64 ( 50.0%) 9 ( 37.5%) 24 ( 37.5%) 57 ( 52.8%) 16 ( 36.4%) 24 ( 37.5%) Post secondary education 65 ( 50.8%) 11 ( 45.8%) 34 ( 53.1%) 58 ( 53.7%) 18 ( 40.9%) 34 ( 53.1%) Married 74 ( 57.8%) 18 ( 75.0%) 43 ( 67.2%) 59 ( 54.6%) 33 ( 75.0%) 43 ( 67.2%) Academic or executive-level occupation 18 ( 14.1%) 4 ( 16.7%) 11 ( 17.2%) 16 ( 14.8%) 6 ( 13.6%) 11 ( 17.2%) Peripheral arthritis 90 ( 70.3%) 17 ( 70.8%) 35 ( 54.7%) 73 ( 67.6%) 34 ( 77.3%) 35 ( 54.7%) Baseline disease duration 1.6 ± 0.9 1.6 ± 1.0 1.5 ± 0.9 1.6 ± 0.8 1.6 ± 0.9 1.5 ± 0.9 Baseline sacroiilitis or spinal inflammation on X-Ray, CT or MRI 94 ( 73.4%) 14 ( 58.3%) 32 ( 50.0%) 84 ( 77.8%) 24 ( 54.5%) 32 ( 50.0%) Baseline CRP 17.8 ± 21.2 10.2 ± 12.0 7.0 ± 10.7 19.2 ± 22.2 9.8 ± 11.2 7.0 ± 10.7 Baseline HLAB27 Positive 97 ( 75.8%) 20 ( 83.3%) 47 ( 73.4%) 81 ( 75.0%) 36 ( 81.8%) 47 ( 73.4%) No comorbidities at baseline 99 ( 77.3%) 19 ( 79.2%) 46 ( 71.9%) 81 ( 75.0%) 37 ( 84.1%) 46 ( 71.9%) Mean BASDAI Pre-Index 53.6 ± 17.4 64.5 ± 10.5 58.2 ± 11.0 51.3 ± 17.7 64.5 ± 10.3 58.2 ± 11.0 Mean BASFI Pre-Index 39.5 ± 20.8 48.6 ± 21.0 40.9 ± 21.4 37.0 ± 19.8 50.0 ± 21.0 40.9 ± 21.4 Baseline Physician Assessment of Disease Activity  3.4 ± 2.7 3.9 ± 2.4 5.5 ± 1.3 2.8 ± 2.5 5.0 ± 2.3 5.5 ± 1.3 Non-adherent on physiotherapy recommendations 48 ( 37.5%) 12 ( 50.0%) 21 ( 32.8%) 42 ( 38.9%) 18 ( 40.9%) 21 ( 32.8%) Non-adherent on non-biologic drug recommendations 15 ( 11.7%) 5 ( 20.8%) 3 (  4.7%) 13 ( 12.0%) 7 ( 15.9%) 3 (  4.7%) Non-adherent on specialist care for uveitis 0 (0%) 1 (  4.2%) 0 (0%) 0 (0%) 1 (  2.3%) 0 (0%)   113   ADHERENCE CLASSIFICATION 1: MAIN ANALYSIS  ADHERENCE CLASSIFICATION 2: SENSITIVITY ANALYSIS Patient characteristics ADHERENT USER: TIMELY ANTI-TNF USE (n=135) NON-ADHERENT USER: LATE ANTI-TNF USE (n=29) NON-ADHERENT NON-USER: UNMET ANTI-TNF NEED (n=67) ADHERENT USER: TIMELY ANTI-TNF USE (n=115) NON-ADHERENT USER: LATE ANTI-TNF USE (n=49) NON-ADHERENT NON-USER: UNMET ANTI-TNF NEED (n=67) Non-adherent on specialist care for psoriasis 7 (  5.5%) 0 (0%) 2 (  3.1%) 7 (  6.5%) 0 (0%) 2 (  3.1%) Non-adherent on specialist care for IBD 1 (  0.8%) 0 (0%) 0 (0%) 1 (  0.9%) 0 (0%) 0 (0%) Positive response to anti-TNF therapy 60 ( 46.9%) 12 ( 50.0%) na 59 ( 54.6%) 13 ( 29.5%) na Time on anti-TNF (in months) 1 year from index 10.3 ± 3.3 10.0 ± 3.9 na 10.5 ± 3.2 9.7 ± 3.7 na QALY 0.586 ± 0.155 0.595 ± 0.122 0.572 ± 0.135 0.604 ± 0.153 0.548 ± 0.136 0.572 ± 0.135 Total costs 16061 ± 7686 14281 ± 8529 2092 ± 2880 15408 ± 7433 16692 ± 8730 2092 ± 2880 Non-biologic Costs 3341 ± 4835 2350 ± 2607 2092 ± 2880 2481 ± 3796 4913 ± 5742 2092 ± 2880 Non-biologic HR Costs 2014 ± 2384 1717 ± 1689 1265 ± 1392 1739 ± 2127 2527 ± 2579 1265 ± 1392 Work Productivity Costs 1327 ± 4215 633 ± 1169 827 ± 2234 742 ± 3112 2385 ± 5232 827 ± 2234   114 Table 4.5 Models of Impact of Adherence Classifications on Cost Outcomes ADHERENCE CLASSIFICATION 1: MAIN ANALYSIS Outcome Variable EST (SE) 95% CI p value RR* Total Cost Age at baseline (year increase) 0.02 (0.01) 0.02 (0.01, 0.04) < .0001 1.03 (1.01, 1.04) Female (vs. Male) 0.52 (0.10) 0.52 (0.32, 0.72) < .0001 1.69 (1.38, 2.06) Baseline smoking status (Yes vs. No) -0.05 (0.10) -0.05 (-0.25, 0.16) 0.6605 0.96 (0.78, 1.17) Adherent NonUser (vs. Adherent User) -2.75 (0.12) -2.75 (-2.98, -2.52) < .0001 0.06 (0.05, 0.08) Non-Adherent Non-User (vs. Adherent User) -2.24 (0.17) -2.24 (-2.57, -1.92) < .0001 0.11 (0.08, 0.15) Non-Adherent User (vs. Adherent User) -0.15 (0.24) -0.15 (-0.61, 0.31) 0.5156 0.86 (0.54, 1.36) Non-Biologic Costs Age at baseline (year increase) 0.03 (0.01) 0.03 (0.02, 0.05) < .0001 1.03 (1.02, 1.05) Female (vs. Male) 0.64 (0.11) 0.64 (0.43, 0.85) < .0001 1.90 (1.53, 2.35) Unmarried (vs. Married) -0.30 (0.12) -0.30 (-0.54, -0.06) 0.0131 0.74 (0.58, 0.94) Baseline smoking status (Yes vs. No) 0.09 (0.11) 0.09 (-0.13, 0.31) 0.4256 1.09 (0.88, 1.36) Adherent Non-User (vs. Adherent User) -1.05 (0.13) -1.05 (-1.30, -0.80) <  .0001 0.35 (0.27, 0.45) Non-Adherent Non-User (vs. Adherent User) -0.59 (0.18) -0.59 (-0.94, -0.24) 0.001 0.56 (0.39, 0.79) Non-Adherent User (vs. Adherent User) -0.33 (0.25) -0.33 (-0.83, 0.16) 0.1873 0.72 (0.44, 1.18) ADHERENCE CLASSIFICATION 2: SENSITIVITY ANALYSIS Total Cost Age at baseline (year increase) 0.02  (0.01) 0.02 (0.01, 0.04) < .0001 1.03 (1.01, 1.04) Female (vs. Male) 0.52 (0.10) 0.52 (0.32, 0.72) < .0001 1.69 (1.38, 2.06) Baseline smoking status (Yes vs. No) -0.04 (0.10) -0.04 (-0.25, 0.16) 0.6704 0.96 (0.78, 1.17) Adherent Non-User (vs. Adherent User) -2.74 (0.12) -2.74 (-2.99  -2.50) < .0001 0.06 (0.05, 0.08) Non-Adherent Non-User (vs. Adherent User) -2.24 (0.17) -2.24 (-2.57, -1.90) < .0001 0.11 (0.08, 0.15) Non-Adherent User (vs. Adherent User) -0.06 (0.19) -0.06 (-0.43, 0.31) 0.753 0.94 (0.65, 1.37) Non-Biologic Costs Age at baseline (year increase) 0.03 (0.01) 0.03 (0.02, 0.05) < .0001 1.03 (1.02, 1.05) Female (vs. Male) 0.61 (0.11) 0.61 (0.39, 0.82) < .0001 1.84 (1.48, 2.28) Unmarried (vs. Married) -0.29 (0.12) -0.29 (-0.52, -0.05) 0.0174 0.75 (0.59, 0.95) Baseline smoking status (Yes vs. No) 0.07 (0.11) 0.07 (-0.15, 0.29) 0.5588 1.07 (0.86, 1.33) Adherent Non-User (vs. Adherent-User) -0.85 (0.13) -0.85 (-1.11, -0.59) < .0001 0.43 (0.33, 0.55) Non-Adherent Non-User (vs. Adherent User) -0.38 (0.18) -0.38 (-0.74, -0.03) 0.0359 0.68 (0.48, 0.98) Non-Adherent User (vs. Adherent User) 0.46 (0.21) 0.46 (0.05, 0.86) 0.0263 1.58 (1.06, 2.36) Adherent Non-User=No Anti-TNF Need; Non-Adherent Non-User=Unmet Anti-TNF Need; Adherent User=Timely Anti-TNF Use; Non- Adherent User=Late Anti-TNF Use; *Rate Ratio obtained by [exp (EST)];   115 Table 4.6 Models of Impact of Adherence Classification on QALY Outcome ADHERENCE CLASSIFICATION 1: MAIN ANALYSIS Outcome Variable EST (SE) 95% CI p value QALY Post-secondary education (vs. none) 0.04 (0.01) 0.04 (0.02, 0.06) < .0001 Female (vs. Male) -0.03 (0.01) -0.03 (-0.05, -0.01) 0.001 Baseline smoking status (Yes vs. No) -0.04 (0.01) -0.04 (-0.06, -0.02) < .0001 Adherent Non-User (vs. Adherent User) 0.07 (0.01) 0.07 (0.04, 0.09) < .0001 Non-Adherent Non-User (vs. Adherent User) -0.02 (0.02) -0.02 (-0.05, 0.01) 0.1444 Non-Adherent User (vs. Adherent User) -0.01 (0.02) -0.01 (-0.06, 0.03) 0.4962 ADHERENCE CLASSIFICATION 2: SENSITIVITY ANALYSIS QALY Post-secondary education (vs. none) 0.04 (0.01) 0.04 (0.02, 0.05) 0.0002 Female (vs. Male) -0.03 (0.01) -0.03 (-0.04, -0.01) 0.0025 Baseline smoking status (Yes vs. No) -0.04 (0.01) -0.04 (-0.05, -0.02) 0.0002 Adherent Non-User (vs. Adherent User) 0.05 (0.01) 0.05 (0.03, 0.07) < .0001 Non-Adherent Non-User (vs. Adherent User) -0.04 (0.02) -0.04 (-0.07, -0.01) 0.0164 Non-Adherent User (vs. Adherent User) -0.06 (0.02) -0.06 (-0.09, -0.03) 0.0005 Adherent Non-User=No Anti-TNF Need; Non-Adherent Non-User=Unmet Anti-TNF Need; Adherent User=Timely Anti-TNF Use; Non- Adherent User=Late Anti-TNF Use    116 Table 4.7 Characteristics of Adherent Non-Users (n=255) Patient characteristics Mean  ± SD or n (%) Baseline age 32.3 ± 8.3 Male  142 ( 55.7%) Post secondary education 180 ( 70.6%) Married 162 ( 63.5%) Academic or executive-level occupation 76 ( 29.8%) Peripheral arthritis 118 ( 46.3%) Baseline disease duration 1.5 ± 0.9 Baseline sacroiilitis or spinal inflammation on X-Ray, CT or MRI 153 ( 60.0%) Baseline CRP 6.7 ± 9.6 Baseline HLAB27 Positive 230 ( 90.2%) No comorbidities at baseline 203 ( 79.6%) Mean BASDAI Pre-Index 31.9 ± 17.8 Mean BASFI Pre-Index 18.1 ± 17.1 Baseline Physician Assessment of Disease Activity  2.9 ± 1.9 Non-adherent on physiotherapy recommendations 127 ( 49.8%) Non-adherent on non-biologic drug recommendations 16 ( 6.3%) Non-adherent on specialist care for uveitis 0 (0%) Non-adherent on specialist care for psoriasis 9 (3.5%) Non-adherent on specialist care for IBD 2 (0.8%) Positive response to anti-TNF therapy na Time on anti-TNF (in months) 1 year from index na QALY 0.674 ± 0.156 Total costs 1108 ± 1848 Non-biologic Costs 1108 ± 1848 Non-biologic HR Costs 898 ± 1391 Work Productivity Costs 214 ± 1070     117 4.4 Discussion   The ASAS recommendations advise that anti-TNF therapy should be prescribed to patients with 4 or more weeks of high disease activity 1. To measure adherence to these recommendations using observational data over 6-month intervals, a definition of adherence must specify the number of consecutive visits with high disease activity that should be interpreted as evidence of sustained activity over 4 weeks. A recent classification system proposed one such definition255, which we explored using DESIR data to compare patient characteristics and outcomes across groups defined by anti-TNF use and high disease activity (BASDAI + PhGA ≥4) over 6-month intervals.   As rheumatologists proposed that 'early' anti-TNF users are best defined as adherent 255, one goal of this study was to evaluate the validity of classing as adherent all anti-TNF users who received an anti-TNF before experiencing high disease activity at 2 consecutive visits. Comparing anti-TNF users and non-users within strata of patients with 0 or ≥1 non-consecutive visits of high disease activity pre-index, we found that anti-TNF users had significantly higher disease activity than anti-TNF non-users. We believe this supports rheumatologists' proposal to class early anti-TNF users as adherent. However, its failure to distinguish premature anti-TNF use, and its excess costs, is a crucial flaw of the proposed classification system.  We also evaluated the validity of classing as adherent patients who experienced high disease activity at 2 consecutive visits, but who received an anti-TNF on the 2nd of those visits. Importantly, the 2 consecutive visit cut-off permits patients to experience up to 6 months of high disease activity before anti-TNF initiation, which could result in classifying as adherent some patients whose anti-TNF initiation might otherwise be considered late. In our intermediate analysis, we found that the 20 anti-TNF users who experienced high disease activity at exactly 2 consecutive visits pre-index had the lowest rate of positive anti-TNF response of all anti-TNF users. We chose to classify these 20 patients as adherent users in the main analysis, but as non-adherent 'late' users in the sensitivity analysis. A subset analysis suggested these 20 patients had poorer outcomes compared to the overall groups defined as non-adherent users in the main and sensitivity analyses. These findings suggest that the difference in results between the main and   118 sensitivity analyses was driven by these 20 patients, raising the question whether anti-TNF initiation among these patients was indeed later than optimal.    The discrepancy between the main and sensitivity analyses here suggests that the impact of adherence to anti-TNF recommendations is highly sensitive to the definition of adherence employed. In the main analysis, with patients permitted up to 2 consecutive visits with high disease activity, no benefit of adherence was apparent. However, in a sensitivity analysis allowing only 0 or ≥1 non-consecutive visits with high disease activity, specific benefits of adherence were demonstrated: adherent 'timely' anti-TNF users had significantly lower non-biologic costs compared to non-adherent 'late' anti-TNF users, and they had significantly better health status than both non-adherent 'late' anti-TNF users and non-adherent non-users. We interpret the findings of the sensitivity analysis as preliminary evidence that adherence to anti-TNF use recommendations may reduce non-biologic costs and increase quality of life among patients who warrant anti-TNF therapy. The findings of our main analysis suggest that the previously proposed definition of adherence to anti-TNF recommendations 255 has the potential to misclassify as 'adherent' some anti-TNF users whose therapy initiation may have been later than optimal.   This study has limitations. For one, we required patients to receive an anti-TNF sooner than proposed by rheumatologists255. While, a priori, this raised the concern that some patients would be prematurely classed as 'non-adherent', the results of the sensitivity analysis suggested the opposite concern (i.e., misclassification of late users as adherent) was more pertinent. Also, we could not explain why patients did not receive an anti-TNF agent, though possible reasons include patient refusal, contraindication to treatment, or lack of 'positive expert opinion'. Importantly, although ASAS cites positive expert opinion as a requirement for anti-TNF use, the criteria that should inform the expert's opinion are not defined quantitatively: the recommendations state only "expert opinion should consider clinical features (history and examination) as well as either serum acute phase reactant levels or imaging results, such as radiographs demonstrating rapid progression or MRI scans indicating inflammation."124 This is problematic in developing a method to detect positive expert opinion using observational data. We used PhGA as a proxy for positive expert opinion, as this reflects the physician's opinion on   119 disease severity. Given the inclusion of this proxy, results pertaining to patients classed as non-adherent non-users should be interpretable as the consequences of not receiving an anti-TNF agent, for any reason, despite having high disease activity as assessed by the rheumatologist. However, the PhGA variable may not capture all reasons for a lack of positive expert opinion; consequently, no anti-TNF non-users can be classed with certainty as being 'non-adherent' to recommendations using the system explored here. For the purpose of further research, the nature of 'positive expert opinion' should be elaborated by ASAS, as this undefined criterion acts as a strong barrier to measuring adherence.   As in all studies using observational data to compare patients on the basis of treatment, the results of this study are limited by potential confounding by indication. We note that disease severity markers, including BASDAI, BASFI, CRP, and baseline sacroiliitis/spinal inflammation, were insignificant in multivariate models, meaning disease severity was effectively captured by adherence groupings. Nonetheless, possible residual confounding by indication should be considered when interpreting our findings. As well, radiographic progression in SpA occurs slowly261, and this study's limited one-year observation period means that full impact of anti-TNF therapy on long-term outcomes has not been captured. As placebo-controlled studies of anti-TNF use among axSpA patients have generally lasted only 12-16 weeks 122, 198, this study provides comparatively long-term data on the effect of anti-TNF agents. However, longer-term assessment of anti-TNF users will be needed to understand the impact of adherence over time.  This study has examined the measurement of adherence to the ASAS anti-TNF use recommendations. The results show that the impact of adherence is highly sensitive to the definition of adherence employed. A classification system proposed for defining adherence 255 has substantial limitations, including failure to define premature anti-TNF use and to distinguish anti-TNF non-users who are adherent to recommendations despite high disease activity. While benefits of adherence to anti-TNF use recommendations were not demonstrated when using one definition of adherence, benefits were observed when using an alternate definition. This discrepancy highlights the need to refine and validate methods to measure adherence to axSpA anti-TNF recommendations and its corresponding impact.    120 Chapter 5i  Access Criteria for Anti-TNF Agents in Spondyloarthritis:  Influence on Comparative 1-Year Cost-Effectiveness Estimates  5.1 Introduction  Anti-TNF agents, including infliximab, etanercept, adalimumab, golimumab, and certolizumab pegol, significantly reduce disease activity and improve functional ability among patients with SpA, including AS and nr-axSpA patients.121, 122, 262 However, because of their similar high cost and potential side-effects, most health systems worldwide restrict access to all anti-TNF agents to SpA patients meeting specific clinical criteria. Van den Berg et al. described 23 different criteria sets from various international settings that designate which SpA patients are eligible to receive anti-TNF therapy.2 Some of these criteria sets represent clinical recommendations, while others are reimbursement criteria. All of the criteria sets differ in terms of the diagnosis, disease activity level, and history of treatment failure required to begin anti-TNF therapy. Some criteria sets limit anti-TNF agents to patients with AS, a severe form of SpA in which bone damage is visible on X-Ray; others approve anti-TNF use among patients with nr-axSpA, the term for SpA prior to the development of this radiographic damage. Many criteria sets that allow anti-TNF use by nr-axSpA patients require them to have sacroiilitis or spinal inflammation visible on MRI and/or elevated acute-phase reactants, such as CRP or ESR, while others do not incorporate these additional markers.   The variation in these criteria sets means that patient access to anti-TNF therapy is more difficult in some settings than others. For example, only an estimated 50% or less of all SpA patients have AS,263 meaning far fewer SpA patients will be treated with anti-TNF therapy in settings that require radiographic damage. The prevalence of other clinical criteria commonly cited in anti-TNF access criteria also varies: elevated acute phase reactants such as CRP or ESR are present in only approximately 40–50 % of patients with AS,47 while sacroiilitis visible on MRI appears to be present in less than half of patients with nr-axSpA.264, 265 Currently, there is a lack of evidence                                                 i A version of Chapter 5 has been submitted for publication: Harvard S, Guh D, Bansback N, Richette P, Saraux A, Fautrel B, Anis A. Access Criteria for Anti-TNF Agents in Spondyloarthritis:  Influence on Comparative 1-Year Cost-Effectiveness Estimates.   121 to indicate how many SpA patients possess the unique combinations of clinical characteristics demanded by different sets of anti-TNF criteria across various settings. However, it is clear that by requiring anti-TNF users to meet clinical criteria present in only a portion of SpA patients, fewer individuals will be treated with anti-TNF therapy than if it were available to all. Importantly, the burden of SpA in terms of disease activity and impairment to be comparable among AS and nr-axSpA patients,266-268 indicating the need to treat both populations.  By limiting the number of patients treated, anti-TNF access criteria may be seen as a means of reducing the total budget impact269 of anti-TNF agents. However, the cost-effectiveness of limiting anti-TNF therapy to patients meeting any particular set of clinical criteria has not been demonstrated. To date, some attention has been paid to the relative cost-effectiveness of anti-TNF agents in AS patients versus nr-axSpA patients,121 although the results are considered inconclusive. This is due in part to heterogeneity in the probability and magnitude of anti-TNF response observed across the small number of anti-TNF trials in nr-axSpA,270-274 which, notably, have included patients with different clinical characteristics. Although a meta-analysis indicates a slightly lower effect of anti-TNF therapy in the nr-axSpA population compared to AS,121 evidence from certain trials that have included both populations suggests the effect may be the same if patients are similar in terms of CRP levels, HLA-B27 positivity, and presence of MRI inflammation.274, 275 Unique combinations of clinical characteristics, such as those cited in anti-TNF access criteria, have not been studied in terms of their influence on the estimated cost-effectiveness of anti-TNF therapy.  The DESIR cohort is a longitudinal study of early SpA that provides clinical and cost data on a clinically heterogeneous population of both AS and nr-axSpA patients in France. Our objective was to explore how many DESIR patients would possess the unique clinical characteristics required to receive anti-TNF therapy in select settings. We then aimed to estimate anti-TNF cost-effectiveness over one year among each of these subsets of patients, with the goal of determining whether the current French restrictions on anti-TNF access 276 are the most cost-effective in that setting relative to other potential restrictions.      122 5.2 Methods  5.2.1 Study Setting and Data Source  The current study was an analysis of data from the DESIR cohort, a 10-year longitudinal study of 708 early SpA patients recruited from 25 centres across France between October 2007 and April 2010.250 The DESIR cohort is a clinically heterogeneous SpA population whose characteristics have been extensively described.59, 277, 278 At study entry, all patients were aged 18-50 and had symptoms of inflammatory back pain6, 219 that had lasted > 3 months and <3 years and was suggestive of SpA according to a rheumatologist's assessment. Follow-up visits occurred every six months in the first two years and every year thereafter. Data from the first three years of DESIR follow-up, i.e., baseline visit (n=708) plus follow-up visits at months 6 (n=704), 12 (n=698), 18 (n=691), 24 (n=692), 36 (n=631) were available for this analysis.  The DESIR database contains clinical, quality of life, and cost data. The clinical data include many of the parameters commonly cited in access criteria for anti-TNF agents,2 including diagnosis; disease activity according to the BASDAI and PhGA; X-Ray, MRI, and CT findings; acute phase reactants (e.g., CRP); and treatment history. The quality of life data collected in DESIR is derived from the SF-36.  The DESIR cost data were derived from a recent cost-of-illness study, for which detailed costing methods, unit costs, and data sources have been described.258 In summary, costing was conducted from a limited societal perspective, including all-cause direct medical costs (i.e., health resource use) and indirect costs (i.e., productivity losses), but excluding direct non-medical costs (e.g., transportation, devices, caregiver expenses), and expressed in 2013 Euros. Direct medical costs were grouped into health practitioner visits, hospitalizations including emergency room visits and surgeries, medical workups, and medications. Total direct non-medical costs were calculated as the reported number of consumed units of each cost component, multiplied by the corresponding unit cost, and summed across all categories and patients. Indirect costs were valued by multiplying the reported number of work days lost by a daily   123 estimated wage per patient in 2013 Euros, which was based on reported professional category and average population wage data.228, 228 The age and sex distribution of DESIR the cohort was compared to that of the population of French workers from which average population wages were obtained and wages were not further adjusted for age and sex. Missing cost and clinical data were imputed using Monte Carlo Markov Chain multiple imputation, last observation carried forward, probabilistic imputation, or with negative values based on clinical expertise, as appropriate.258  5.2.2 Selection of Anti-TNF Access Criteria   Most patient characteristics cited in anti-TNF access criteria2 are routinely collected in clinical practice for multiple purposes. Using DESIR clinical data, it is possible to assess patient satisfaction of the anti-TNF access criteria in place in France276 and numerous other settings. For the purpose of the analysis, we sought to select a practical number of sets of access criteria with clinically meaningful differences between them and the French criteria, i.e., sets citing different markers of disease severity whose prevalence would vary within the DESIR cohort. The selection process was undertaken by the research team, which included a rheumatologist (BF), epidemiologist (SH), and biostatistician (DG) with knowledge of the DESIR cohort and database. By consensus, four sets of access criteria were selected, including those from Canada,279 Germany,280 Hong Kong,281 and the UK.282 Based on their respective criteria, these sets were anticipated to result in multiple, distinct (though potentially overlapping) subsets of DESIR patients defined as eligible for anti-TNF therapy.  5.2.3 Creation of 'Study Population' Datasets   We created five separate datasets containing the DESIR patients who satisfied the diagnosis and disease severity criteria for anti-TNF access in France,276 Canada,279 Germany,280 Hong Kong,281 and UK,282 respectively. These datasets were created to represent five separate 'study populations' of patients, each comprised of anti-TNF users and non-users who satisfied the same set of anti-TNF access criteria. As patients could satisfy multiple criteria sets, unique patients   124 could appear in more than one study population dataset. However, as the five study population datasets were separate, only anti-TNF users and non-users who satisfied the same criteria could be compared to each other. This was done to help limit confounding by indication, as patients satisfying the same anti-TNF access criteria have comparable disease severity on a number of specific measures. Satisfaction of the treatment failure criterion, i.e., insufficient response to NSAIDs, was assumed for all patients.  In creating the five study population datasets, specific rules were applied in a basecase analysis and subsequently varied in sensitivity analyses. In all analyses, patients were required to satisfy the relevant criteria set no later than month 24. In the basecase analysis, anti-TNF use (yes/no) was defined based on the patient's experience in the one year following the date of criteria satisfaction, which was taken as the index date for all patients. In the sensitivity analyses, anti-TNF use (yes/no) was defined over the entire study period, with date of criteria satisfaction taken as the index date for anti-TNF non-users and date of anti-TNF initiation taken as the index date for anti-TNF users. In all analyses, outcomes were observed in the one year following the index date. Because the start point for the one year observation period was defined differently in the basecase and sensitivity analyses, patients who were classed as anti-TNF non-users in the basecase analysis could be classed as anti-TNF users in the sensitivity analyses.   To be included in the basecase analysis, anti-TNF users were required not to have initiated anti-TNF therapy prior to criteria satisfaction (rule 1). Anti-TNF users were further required to have initiated the anti-TNF <6 months after criteria satisfaction (rule 2). Consequently, anti-TNF users who initiated anti-TNF before criteria satisfaction or > 6 months after criteria satisfaction were excluded from the basecase analysis. These rules were applied in order to include only anti-TNF users with a similar length of anti-TNF exposure in the basecase analysis, in which outcomes were observed following the date of criteria satisfaction rather than therapy initiation.  In the first sensitivity analysis, anti-TNF users were permitted to receive the anti-TNF prior to criteria satisfaction (rule 1 lifted). In the second sensitivity analysis, anti-TNF users were permitted to receive the anti-TNF > 6 months after criteria satisfaction (rule 2 lifted). In the third sensitivity analysis, anti-TNF users were permitted to receive the anti-TNF at any time point   125 (rules 1 and 2 lifted). A separate sensitivity analysis was conducted to explore the impact of simulating a 24-week stopping rule for anti-TNF non-responders, defined as patients who did not achieve a 50% relative change or absolute change of 2 on the BASDAI scale124 one visit post-therapy initiation. In this analysis, anti-TNF costs accumulated by non-responders after 24 weeks of therapy were excluded. Additional sensitivity analyses were conducted to explore the impact of excluding indirect costs in all scenarios.  5.2.4 Descriptive Statistics  Sociodemographic and clinical characteristics at baseline and at time of criteria satisfaction among patients in each of the five basecase study population datasets were described in terms of mean (SD) and frequency (%) as appropriate. Descriptive statistics were also produced to describe the number of anti-TNF users in the DESIR cohort who did not satisfy any of the selected criteria sets (and were therefore excluded from analysis) as well as the number of anti-TNF non-responders in the basecase study population datasets and their total time on anti-TNF therapy.  5.2.5 Adjustment of Costs and QALYs  To control for differences between anti-TNF users and non-users, we used linear regression models to estimate adjusted total costs (i.e., direct medical plus indirect costs) in the one year post-index. Independent variables considered to be potential confounders of the relationship between anti-TNF use and costs were first tested in univariate models of costs, specifically age, sex, education, marital status, disease duration, smoking (yes vs. no/do not know), HLA-B27 status and presence of peripheral arthritis at baseline, and the Bath Ankylosing Spondylitis Functional Index (BASFI), BASDAI, PhGA, CRP, and SF-36 at the patient's index date. The same variables were then each tested in preliminary multivariate models of costs that included BASFI (the strongest predictor of costs in univariate analyses) and anti-TNF use. Independent variables that changed the coefficient for anti-TNF use by more than 10% in the preliminary multivariate models were included in the final multivariate costs model.    126 Total QALYs in the one year post-index were calculated using SF6D utility weights derived from SF-36 health states, following the area under the curve method.260, 283 Again to control for differences between anti-TNF users and non-users, we used linear regression models to derive adjusted mean QALYs. Independent variables as above were first tested in univariate models then in preliminary multivariate models that included SF-36 at time of criteria satisfaction (the strongest predictor of QALY in univariate analyses) and anti-TNF use. Independent variables that changed the coefficient for anti-TNF use by more than 10% in preliminary multivariate models were included in the final multivariate QALY model.  5.2.6 Cost-effectiveness Analysis Using Adjusted Costs and QALYs  For each of the five study population datasets, we calculated the ICER comparing the costs and QALYs of anti-TNF users versus non-users, i.e., the incremental cost per additional QALY gained by treating with an anti-TNF, using the standard formula: [(Cost anti-TNF- Cost no anti-TNF)/(QALYs anti-TNF- QALYs no anti-TNF)]. To explore the range of uncertainty around mean costs and QALYs, we used non-parametric bootstrapping,284 repeating the same procedures for each study population datasets (i.e., group of patients satisfying a given criteria set). Specifically, 10,000 bootstrap samples were generated (i.e., by sampling with replacement), stratified by anti-TNF users and non-users. For each bootstrapped sample, linear regression models were fitted for costs and QALYs; although the models were fitted separately, the data used were from the same samples, meaning the interdependence of costs and QALYs was accounted for. Adjusted mean costs and QALYs and hence the incremental costs and QALYs were then estimated from the models. The 2.5th and 97.5th percentiles of the bootstrapped distribution were used to estimate 95% CI for the incremental costs and QALYs.  5.3 Results  Table 5.1 shows the individual components of the selected criteria sets and their satisfaction at baseline by the 708 patients in DESIR, as well as the proportion of patients who ever satisfied the full criteria set over the study period. Criteria sets from the UK and Hong Kong both required a diagnosis of AS, while those from Canada, France, and Germany were inclusive of nr-axSpA   127 patients. Anti-TNF access criteria from France were satisfied by the largest number of DESIR patients (197/708; 27.8%), followed by Germany (175/708; 25.1%), Canada (169/708; 23.8%), the UK (86/708; 12.1%) and Hong Kong (61/708; 8.6%).  Table 5.1 Selected Criteria Sets: Satisfaction Among All DESIR Patients (n=708) Criteria Set Origin Patients ever satisfying criteria set (n) and percent of total (N=708)   Diagnosis and Disease Severity Criteria France n=197 (27.8%) Ÿ BASDAI ≥4 Ÿ New York criteria for AS OR involvement of SJ OR spine by X-ray, CT or MRI Ÿ Physician's Global Assessment ≥4 Germany n=175 (24.7%) Ÿ BASDAI ≥4  Ÿ ASAS criteria Ÿ Positive MRI* Ÿ Elevated CRP *§         *Minimum 1 of 2 criteria Canada n=169 (23.8%) Ÿ BASDAI ≥4 Ÿ Sacroiilitis or spinal inflammation on X-Ray, CT or MRI Ÿ Elevated CRP or ESR* Ÿ Inflammatory lesions in the sacroiliac joints and/or spine on MRI* Ÿ Expert opinion*†        *Minimum 2 of 3 criteria United Kingdom n=86  (12.1%) Ÿ BASDAI ≥4 Ÿ Modified New York criteria for AS Ÿ Spinal pain in last week 4/10 VAS  Hong Kong n=61  (8.6%) Ÿ BASDAI ≥4 Ÿ Modified New York criteria for AS Ÿ Morning stiffness ≥45 mins Ÿ Inflammatory back pain ≥40 /100 VAS Ÿ Patient Global Assessment ≥40/100 VAS  †Positive expert opinion defined in the analysis as Physician's Global Assessment ≥4; §NA at baseline, values reported for month 6; CT=computerized tomography; MRI=magnetic resonance imaging; BASDAI=Bath Ankylosing Spondylitis Disease Activity Index; AS=ankylosing spondylitis; CRP=C-reactive protein; ESR=erythrocyte sedimentation rate; SJ=sacroiliac joint; VAS=visual analogue scale   Table 5.2 shows the characteristics of anti-TNF users and non-users in each of the basecase study population datasets. The proportion of anti-TNF users was highest among patients who met the Hong Kong criteria (32/61; 52.5%), followed by the UK (40/86; 46.5%), Canada (71/169; 42.0%), France (80/197; 40.6%), and Germany (67/175; 38.3%). Among a total 225 anti-TNF users in the DESIR cohort, 107 (47.6%) never satisfied the French anti-TNF access criteria,   128 while 94 (41.8%) never satisfied any of the selected criteria sets and were thus excluded from the analysis. The characteristics of excluded anti-TNF users are shown in Table 5.3.   129 Table 5.2 Characteristics of Anti-TNF Users and Non-Users by Criteria Set   Canada n=169 France n=197 UK n=86 Germany n=175 Hong Kong n=61   Anti-TNF Non-User (n=98) Anti-TNF User (n=71) Anti-TNF Non-User (n=117) Anti-TNF User (n=80) Anti-TNF Non-User (n=46) Anti-TNF User (n=40) Anti-TNF Non-User (n=108) Anti-TNF User (n=67) Anti-TNF Non-User (n=29) Anti-TNF User (n=32) At baseline                      Age 32.8 ± 8.5 33.5 ± 9.3 33.2 ± 8.7 33 ± 9.1 33 ± 7.6 32.6 ± 9.6 32.7 ± 7.9 33.1 ± 9.4 35 ± 8 31.8 ± 8.5 Male  53 (54.1%) 40 (56.3%) 58 (49.6%) 59 (53.8%) 27 (58.7%) 24 (60%) 59 (54.6%) 37 (55.2%) 17 (58.6%) 19 (59.4%) Post-secondary education 55 (56.1%) 42 (59.2%) 65 (55.6%) 48 (60%) 22 (47.8%) 20 (50%) 60 (55.6%) 35 (59.7%) 11 (37.9%) 14 (43.8%) Married 58 (59.2%) 40 (59.2%) 71 (60.7%) 48 (60%) 24 (52.2%) 23 (57.5%) 69 (63.9%) 40 (59.7%) 17 (58.6%) 19 (59.4%) Smoking  41 (41.8%) 36 (50.7%) 50 (42.7%) 38 (47.5%) 23 (50.0%) 22 (55.0%) 44 (40.7%) 35 (52.2%) 12 (41.4%) 19 (59.4%) Disease Duration 1.5 ± 0.9 1.6 ± 0.8 1.6 ± 0.9 1.6 ± 0.8 1.7 ± 0.9 1.6 ± 0.8 1.5 ± 0.9 1.6 ± 0.8 1.7 ± 0.9 1.6 ± 0.8 Peripheral Arthritis 54 (55.1%) 49 (69%) 65 (55.6%) 55 (68.8%) 22 (47.8%) 25 (62.5%) 56 (51.9%) 46 (68.7%) 13 (44.8%) 21 (65.6%) HLA: Positive 59 (60.2%) 44 (62%) 66 (56.4%) 48 (60%) 32 (69.6%) 25 (62.5%) 83 (76.9%) 46 (68.7%) 18 (62.1%) 19 (59.4%) Radiographic sacroiliitis 42 (42.9%) 40 (56.3%) 49 (41.9%) 44 (55.0%) 28 (60.9%) 30 (75.0%) 49 (45.4%) 41 (61.2%) 15 (51.7%) 25 (78.1%) Sacroiilitis or spine inflammation on MRI 67 (68.4%) 49 (69.0%) 67 (57.3%) 49 (61.3%) 25 (54.3%) 26 (65.0%) 80 (74.1%) 52 (77.6%) 12 (41.4%) 21 (65.6%) BASDAI ≥4 74 (75.5%) 67 (94.4%) 91 (77.8%) 76 (95.0%) 34 (73.9%) 39 (97.5%) 78 (72.2%) 62 (92.5%) 21 (72.4%) 31 (96.9%) CRP >10 mg/L 24 (24.5%) 31 (43.7%) 25 (21.4%) 33 (41.3%) 8 (17.4%) 21 (52.5%) 26 (24.1%) 32 (47.8%) 5 (17.2%) 14 (43.8%) Physician's Global Assessment ≥4 76 (77.6%) 66 (93.0%) 92 (78.6%) 75 (93.8%) 33 (71.7%) 38 (95.0%) 73 (67.6%) 61 (91.0%) 20 (69.0%) 31 (96.9%) Inflammatory back pain ≥4 79 (80.6%) 68 (95.8%) 96 (82.1%) 76 (95.0%) 40 (87.0%) 40 ( 100%) 88 (81.5%) 64 (95.5%) 25 (86.2%) 32 ( 100%) Morning stiffness ≥45 mins 53 (54.1%) 47 (66.2%) 62 (53.0%) 56 (70.0%) 20 (43.5%) 31 (77.5%) 57 (52.8%) 47 (70.1%) 18 (62.1%) 30 (93.8%) Patient Global Assessment ≥4 73 (74.5%) 65 (91.5%) 88 (75.2%) 74 (92.5%) 33 (71.7%) 38 (95.0%) 79 (73.1%) 60 (89.6%) 23 (79.3%) 31 (96.9%) At criteria satisfaction                     BASDAI  56.2 ± 12.7 60.6 ± 11.4 55.4 ± 12.2 60.9 ±10.9 52.8 ± 10.5 58.9 ± 11.9 55.5 ± 12.7 58.7 ± 10.8 56.1 ± 12 59.3 ±11.7 BASFI  35.8 ± 21.8 43.9 ± 20.2 35.9 ± 21.5 45 ± 19.8 30.8 ± 20 47.4 ± 20.8 36.5 ± 23.5 42.1 ± 19.7 39 ± 21.6 46.5 ± 22 SF-6D 0.6 ± 0.1 0.5 ± 0.1 0.6 ± 0.1 0.5 ± 0.1 0.6 ± 0.1 0.5 ± 0.1 0.6 ± 0.1 0.5 ± 0.1 0.6 ± 0.1 0.5 ± 0.1 Physician's Global Assessment  5.3 ± 1.2 6.7 ± 1.6 5.3 ± 1.2 6.8 ± 1.5 4.7 ± 1.7 6.8 ± 1.7 4.5 ± 1.8 6.4 ± 2 5 ± 1.6 6.7 ± 1.6 CRP 10.9 ± 18.6 20.1 ± 24.6 9.5 ± 17.2 18.8 ± 23.8 7.7 ± 9.3 22.3 ± 25.5 10.9 ± 12.5 19.7 ± 23.7 7.6 ± 10.2 15.8 ± 20.6 *Statistics presented are: Mean ± SD, or N (%)     130 Table 5.3 Characteristics of Anti-TNF Users not Satisfying Selected Criteria Sets   Not Satisfying Any (n=94) Not Satisfying Canada (n=118) Not Satisfying France (n=107) Not Satisfying UK (n=177) Not Satisfying Germany (n=120) Not Satisfying Hong Kong (n=188) At baseline              Age 34.14 ± 8.48 34.35 ± 8.76 34.41 ± 8.64 34.3 ± 8.78 34.62 ± 8.72 34.21 ± 8.89 Male  35 (37.2%) 41 (34.7%) 38 (35.5%) 69 (39%) 45 (37.5%) 77 (41%) Post-secondary education 51 (54.3%) 60 (50.8%) 55 (51.4%) 95 (53.7%) 64 (53.3%) 101 (53.7%) Married 59 (62.8%) 75 (63.6%) 69 (64.5%) 113 (63.8%) 75 (62.5%) 118 (62.8%) Smoking  33 (35.1%) 42 (35.6%) 39 (36.4%) 69 (39%) 41 (34.2%) 72 (38.3%) Disease Duration 1.47 ± 0.88 1.46 ± 0.88 1.45 ± 0.89 1.52 ± 0.88 1.47 ± 0.85 1.53 ± 0.87 Peripheral Arthritis 66 (70.2%) 84 (71.2%) 76 (71%) 127 (71.8%) 83 (69.2%) 133 (70.7%) HLA: Positive 45 (47.9%) 60 (50.8%) 55 (51.4%) 95 (53.7%) 51 (42.5%) 104 (55.3%) Radiographic sacroiliitis 12 (12.8%) 20 (16.9%) 16 (15%) 39 (22%) 20 (16.7%) 47 (25%) Sacroiilitis or spine inflammation on MRI 14 (14.9%) 17 (14.4%) 18 (16.8%) 59 (33.3%) 14 (11.7%) 65 (34.6%) BASDAI ≥4 69 (73.4%) 92 (78%) 80 (74.8%) 144 (81.4%) 94 (78.3%) 152 (80.9%) CRP >10 mg/L 22 (23.4%) 25 (21.2%) 25 (23.4%) 42 (23.7%) 26 (21.7%) 49 (26.1%) Physician's Global Assessment ≥4 72 (76.6%) 90 (76.3%) 79 (73.8%) 142 (80.2%) 94 (78.3%) 150 (79.8%) Inflammatory back pain ≥4 81 (86.2%) 102 (86.4%) 94 (87.9%) 157 (88.7%) 103 (85.8%) 168 (89.4%) Morning stiffness ≥45 mins 51 (54.3%) 70 (59.3%) 60 (56.1%) 99 (55.9%) 68 (56.7%) 100 (53.2%) Patient Global Assessment ≥4 81 (86.2%) 104 (88.1%) 93 (86.9%) 156 (88.1%) 106 (88.3%) 165 (87.8%) At time of treatment initiation             BASDAI  42.26 ± 22.35 44.44 ± 21.9 42.84 ± 22.3 41.87 ± 22.44 44.2 ± 21.92 40.92 ± 22.56 BASFI  32.16 ± 23.2 34.52 ± 23.24 33.48 ± 23.26 32.28 ± 22.69 34.19 ± 22.52 32.07 ± 22.58 SF-6D 0.62 ± 0.12 0.61 ± 0.12 0.62 ± 0.12 0.62 ± 0.12 0.61 ± 0.12 0.63 ± 0.12 Physician's Global Assessment  3.85 ± 2.57 3.98 ± 2.59 3.81 ± 2.61 3.76 ± 2.58 4.02 ± 2.52 3.72 ± 2.61 CRP 6.53 ± 17.51 6.03 ± 15.71 6.44 ± 16.49 5.63 ± 14.01 5.96 ± 15.78 6 ± 14.33 *Statistics presented are: Mean ± SD, or N (%)     131 Table 5.4 Incremental Costs, QALYs, and ICERS by Criteria Set  Canada n=169 France n=197 UK n=86 Germany n=175 Hong Kong n=61  Anti-TNF Non-User (n=98) Anti-TNF User (n=71) Anti-TNF Non-User (n=117) Anti-TNF User (n=80) Anti-TNF Non-User (n=46) Anti-TNF User (n=40) Anti-TNF Non-User (n=108) Anti-TNF User (n=67) Anti-TNF Non-User (n=29) Anti-TNF User (n=32) Unadjusted                   Total Costs (Mean ± SD) 1981 ± 2812 15448 ± 6155 2117 ± 2976 15311 ± 5801 1426 ± 1852 16475 ± 7506 2044 ± 3351 15339 ± 6453 1697 ± 2447 15132 ± 6548 Health Practitioner 655 ± 745 896 ± 772 640 ± 720 1000 ± 939 440 ± 573 971 ± 1034 605 ± 750 953 ± 1013 472 ± 722 835 ± 682 Hospital 532 ± 1289 710 ± 2010 615 ± 1505 678 ± 1913 530 ± 1326 646 ± 1394 454 ± 1090 731 ± 2036 695 ± 1541 591 ± 1184 Medical Act 178 ± 238 399 ± 362 185 ± 247 396 ± 352 155 ± 239 449 ± 417 154 ± 228 386 ± 313 162 ± 330 433 ± 388 Medication 149 ± 135 12511 ± 5252 143 ± 132 12364 ± 5114 126 ± 100 12742 ± 5844 145 ± 125 12409 ± 5497 138 ± 120 11981 ± 4977 Productivity 467 ± 1893 932 ± 2796 534 ± 1987 872 ± 2411 176 ± 526 1667 ± 4059 687 ± 2661 859 ± 2776 231 ± 672 1292 ± 3271 Total Costs: Med (IQR) 913 (362, 2343) 14172 (12741,17495) 920 (364,2343) 14293 (12806,17537) 839 (326,1621) 14963 (13110,18742) 865 (323,2092) 14111 (12825,16356) 809 (298,1681) 14404 (12579,17222) SF6D (Mean ±SD)                   Index 0.579 ± 0.101 0.532 ± 0.078 0.579 ± 0.098 0.533 ± 0.075 0.602 ± 0.088 0.528 ± 0.074 0.594 ± 0.105 0.536 ± 0.079 0.594 ± 0.078 0.531 ± 0.073 6M after Index Visit 0.645 ± 0.108 0.644 ± 0.130 0.638 ± 0.105 0.635 ± 0.128 0.644 ± 0.103 0.630 ± 0.130 0.650 ± 0.111 0.652 ± 0.132 0.609 ± 0.094 0.624 ± 0.128 12M after Index Visit 0.647 ± 0.118 0.620 ± 0.123 0.639 ± 0.115 0.612 ± 0.122 0.639 ± 0.108 0.609 ± 0.112 0.646 ± 0.124 0.632 ± 0.123 0.633 ± 0.099 0.595 ± 0.107 Total QALY (Mean ± SD) 0.63 ± 0.09 0.61 ± 0.10 0.62 ± 0.09 0.60 ± 0.10 0.63 ± 0.08 0.60 ± 0.10 0.63 ± 0.09 0.62 ± 0.10 0.61 ± 0.07 0.59 ± 0.09 Adjusted (Mean ± SE)                   Total Costs 1703 ± 366 15741 ± 802 1763 ± 347 15773 ± 687 756 ± 644 16952 ± 1160 1874 ± 408 15617 ± 852 722 ± 946 15764 ± 1229 Total QALY 0.61 ± 0.01 0.63 ± 0.01 0.61 ± 0.01 0.62 ± 0.01 0.61 ± 0.01 0.63 ± 0.02 0.62 ± 0.01 0.64 ± 0.01 0.59 ± 0.02 0.62 ± 0.02   132 Table 5.4 shows the unadjusted and adjusted costs of patients in the five basecase study population datasets. In final multivariate models, costs were adjusted for smoking and HLA-B27 status at baseline, and BASFI, PhGA, and CRP at date of criteria satisfaction; QALYs were adjusted for age, sex, education, smoking, HLA-B27 status and peripheral arthritis at baseline, and SF-36, PhGA, and CRP at date of criteria satisfaction.  Table 5.5 Comparative Estimates of Costs, QALYs, and ICERs: Basecase Analysis   Anti-TNF User Anti-TNF Non-User Increment (User vs Non-User) 95% CI* Canada     Costs 15,741 1,703 14,038 (12179, 15991) QALYs 0.626 0.609 0.017 (-0.008, 0.042) ICER   818,186 (330082, Dominated) France     Costs 15,773 1,763 14,010 (12,423 , 15,694) QALYs 0.620 0.607 0.013 (-0.011,  0.036) ICER   1,105,859 (375227, Dominated) UK     Costs 16,952 756 16,195 (13327, 19181) QALYs 0.627 0.606 0.021 (-0.015,  0.059) ICER   766,217 (264164, Dominated) Germany     Costs 15,617 1,874 13,743 (11848, 15759) QALYs 0.642 0.617 0.025 ( 0.001,  0.050) ICER   545,808 (272286,  18727278) Hong Kong     Costs 15,764 722 15,042 (11825, 18898) QALYs 0.619 0.586 0.033 (-0.011,  0.076) ICER   456,850 (189636, Dominated) *Lower bound=2.5th, upper bound=97.5th percentile of bootstrapped distribution   Table 5.5 shows the incremental costs and QALYs and ICERs over one year for each of the five study populations in the basecase analysis. The most favourable cost-effectiveness point estimate was derived from the study population satisfying the Hong Kong criteria (ICER €456,850), followed by Germany (€545,808), the UK (€766,217), and Canada (€818,186). The highest   133 ICER was derived from the study population satisfying the French criteria (€1,105,859) However, as shown in Figure 5.1, the confidence intervals surrounding the point estimates for the incremental costs and QALYs derived from each of the five study populations were overlapping, indicating uncertainty in the results of the analysis.  Figure 5.1 Confidence Intervals Around ICERs in Each of the Five Study Populations      A positive anti-TNF response one visit post-therapy initiation was achieved by approximately half of anti-TNF users who satisfied the criteria from Canada (n=39; 54.9%), France (n=42; 52.5%), and Germany (35; 51.5%), respectively, and by approximately forty percent of anti-TNF users who satisfied criteria from the UK (n=17; 42.5%) and Hong Kong (n=13; 40.6%). In each of the five study populations, 90% or more of non-responders continued anti-TNF therapy for   134 one or more years (Table 5.6). In the sensitivity analysis that examined the effect of excluding costs accumulated past 24 weeks by anti-TNF non-responders, the incremental cost per QALY was reduced by approximately 25% (France: €857,992 vs. €1,105,859; Canada: € 626,459 vs. €818,186; Germany: € 422,568 vs. €545,808); UK €578,899 vs. €766,217; Hong Kong €335,418 vs. €456,850) (Table 5.7). Consistent with this finding, utility gain was observed to be lower among anti-TNF non-responders compared to responders (Table 5.8).  Table 5.6 Total Time on Anti-TNF Therapy Among Non-Responders (From Initiation Until End of Follow-Up) Interval Canada (N=32) France (N=38) UK (N=23) Germany (N=32) Hong Kong (N=19) < 12weeks   1 ( 2.6%)       12-24 weeks 0 ( 0.0%) 0 ( 0.0%) 0 ( 0.0%) 2 ( 6.3%) 0 ( 0.0%) 24 weeks to 1 year 1 ( 3.1%) 1 ( 2.6%) 1 ( 4.3%) 1 ( 3.1%) 1 ( 5.3%) 1-2 years 11 (34.4%) 13 (34.2%) 7 (30.4%) 10 (31.3%) 7 (36.8%) 2-3 years 13 (40.6%) 15 (39.5%) 9 (39.1%) 13 (40.6%) 5 (26.3%) 3-4 years 7 (21.9%) 8 (21.1%) 6 (26.1%) 6 (18.8%) 6 (31.6%)  Table 5.7 Comparative Estimates of Costs, QALYs, and ICERs: Sensitivity Analysis Excluding Non-Responder Anti-TNF Costs Past 24 Weeks    Anti-TNF User Anti-TNF Non-User Increment (User vs Non-User) 95% CI* Canada     Costs 12,405 1,656 10,749 ( 8868, 12789) QALYs 0.626 0.609 0.017 (-0.008, 0.042) ICER   626,459 (247149, Dominated) France     Costs 12,566 1,696 10,870 ( 9144, 12751) QALYs 0.620 0.607 0.013 (-0.011, 0.036) ICER   857,992 (284242, Dominated) UK     Costs 12,973 737 12,236 ( 8,981 , 15,769) QALYs 0.627 0.606 0.021 (-0.015, 0.059) ICER   578,899 (187442, Dominated) Germany     Costs 12,321 1,682 10,640 -877,112,728 QALYs 0.642 0.617 0.025 ( 0.001, 0.050) ICER   422,568 ( 206749, 14587057)   135   Anti-TNF User Anti-TNF Non-User Increment (User vs Non-User) 95% CI* Hong Kong     Costs 11,935 891 11,044 ( 7243, 15470) QALYs 0.619 0.586 0.033 (-0.011 ,  0.076) ICER   335,418 (124073, Dominated) *Lower bound=2.5th, upper bound=97.5th percentile of bootstrapped distribution   Table 5.8 Utility Gain Six and Twelve Months Post-Therapy Initiation in Anti-TNF Responders and Non-Responders   Canada France UK Germany HK Unadjusted  (Mean ± SD)           All users           Index 0.53 ±  0.08 0.53 ±  0.08 0.53 ±  0.07 0.54 ±  0.08 0.53 ±  0.07 6M after Index Visit 0.64 ±  0.13 0.63 ±  0.13 0.63 ±  0.13 0.65 ±  0.13 0.62 ±  0.13 12M after Index Visit 0.62 ±  0.12 0.61 ±  0.12 0.61 ±  0.11 0.63 ±  0.12 0.60 ±  0.11 Responders           Index 0.55 ±  0.09 0.55 ±  0.09 0.53 ±  0.08 0.55 ±  0.09 0.55 ±  0.08 6M after Index Visit 0.68 ±  0.12 0.67 ±  0.12 0.65 ±  0.12 0.69 ±  0.12 0.67 ±  0.11 12M after Index Visit 0.65 ±  0.12 0.64 ±  0.12 0.62 ±  0.10 0.66 ±  0.11 0.60 ±  0.10 Non-Responders           Index 0.52 ±  0.06 0.52 ±  0.06 0.52 ±  0.07 0.52 ±  0.06 0.52 ±  0.07 6M after Index Visit 0.60 ±  0.13 0.59 ±  0.12 0.62 ±  0.14 0.61 ±  0.13 0.59 ±  0.13 12M after Index Visit 0.59 ±  0.12 0.58 ±  0.12 0.61 ±  0.12 0.60 ±  0.13 0.59 ±  0.11   In the sensitivity analysis using the basecase study population, but excluding indirect costs, all ICERs became more favourable (Table 5.9). In all additional sensitivity analyses, i.e., including anti-TNF users who initiated therapy prior to and/or 6-12 months after criteria satisfaction, anti-TNF agents were dominated in all study populations (Table 5.9); this finding did not change upon the exclusion of indirect costs (data not shown).        136 Table 5.9 Comparative Estimates of Costs, QALYs, and ICERs in Sensitivity Analyses Basecase population, excluding indirect costs     Anti-TNF User Anti-TNF Non-User  Incremental Cost Incremental QALY ICER Canada Cost 14,704 1,284 13,420 0.017 782,141 Canada Qaly 0.626 0.609    France Cost 14,777 1,286 13,490 0.013 1,064,838 France Qaly 0.620 0.607    UK Cost 14,920 979 13,941 0.021 659,561 UK Qaly 0.627 0.606    Germany Cost 14,652 1,378 13,273 0.025 527,169 Germany Qaly 0.642 0.617    HK Cost 14,177 1,013 13,164 0.033 399,813 HK Qaly 0.619 0.586       Including anti-TNF users who received therapy prior to criteria satisfaction (rule 1 lifted) Canada Cost 16,319 1,837 14,482 -0.023 Dominated Canada Qaly 0.609 0.631    France Cost 15,952 2,073 13,879 -0.023 Dominated France Qaly 0.604 0.627    UK Cost 16,783 1,556 15,227 -0.014 Dominated UK Qaly 0.614 0.628    Germany Cost 16,476 1,850 14,626 -0.016 Dominated Germany Qaly 0.625 0.641    HK Cost 16,200 1,498 14,702 -0.016 Dominated HK Qaly 0.597 0.614    Including anti-TNF users who received therapy 6-12 months after criteria satisfaction (rule 2 lifted) Canada Cost 16,254 2,071 14,182 -0.027 Dominated Canada Qaly 0.608 0.635    France Cost 16,091 2,256 13,835 -0.028 Dominated France Qaly 0.603 0.631    UK Cost 16,166 1,784 14,382 -0.021 Dominated UK Qaly 0.609 0.630    Germany Cost 15,726 2,060 13,666 -0.023 Dominated Germany Qaly 0.617 0.639    HK Cost 15,939 1,708 14,231 -0.020 Dominated HK Qaly 0.596 0.616    Including anti-TNF users who received therapy prior to or 6-12 months after criteria satisfaction (rule 1 and 2 lifted) Canada Cost 16,216 2,065 14,152 -0.030 Dominated Canada Qaly 0.605 0.635    France Cost 15,940 2,227 13,713 -0.028 Dominated France Qaly 0.602 0.630    UK Cost 16,155 1,796 14,360 -0.021 Dominated UK Qaly 0.608 0.629    Germany Cost 15,933 1,994 13,938 -0.024 Dominated Germany Qaly 0.616 0.640    Hong Kong Cost 15,767 1,747 14,020 -0.020 Dominated Hong Kong Qaly 0.592 0.612       137 5.4 Discussion  To our knowledge, this is the first study to explore what proportion of SpA patients in a single cohort possesses the unique combination of clinical characteristics demanded by select sets of anti-TNF access criteria. We found that the proportion of DESIR patients eligible to receive anti-TNF therapy ranged from 9% to 28%, depending on the criteria set considered. For illustrative purposes, we note that assuming a SpA prevalence of 0.43% in France,285 this may translate to as few as 39 or as many as 120 people per 100,000 population per year being recommended anti-TNF therapy. At an estimated cost of €13,000 for a full year of anti-TNF therapy,258 the additional 81 people treated under the less restrictive access conditions would have an annual budget impact of €1.05 million. One of the contributions of the present study is in highlighting the potential role of anti-TNF access criteria, as at the current cost of anti-TNF therapy even a small number of additional patients treated will correspond to a large increase in health budgets, which may or may not represent good value for the public.  This study focused on the comparative cost-effectiveness of selected criteria sets in the French setting, and the absolute ICER values generated here should be interpreted with caution. The ICERs were produced using data over a single year using real-world data, and it should be stressed that these cannot be compared to ICERs from models that employ a lifetime horizon, estimate treatment effectiveness using RCT data, or assume that non-responders will be withdrawn from treatment. Lifetime cost-effectiveness models have the important capability of acknowledging that not all benefits of anti-TNF therapy will be realized within a short time frame; in general, anti-TNF agents appear more cost-effective in models with longer time horizons.196 Recently, the UK's NICE121 reported upperbound ICERs of £66,529 per QALY for AS patients and £34,232 per QALY for nr-axSpA patients based on lifetime cost-effectiveness models. These are vastly more favourable than the ICERs estimated here, reflecting, in part, the impact of including the latent benefits of anti-TNF therapy. However, including these predicted benefits required extrapolating outcomes beyond periods for which observed data are available. In contrast, the present analysis has provided important observed data on the costs and benefits associated with anti-TNF use in a real-world setting. Importantly, the NICE models assume that all non-responders will discontinue therapy at 12-weeks,121 yet we found that the vast majority of   138 non-responders continued therapy for a year or more. The continuation of therapy among non-responders appears to be one reason that ICERs estimated in the basecase analysis here are less favourable than those estimated by the NICE: in a sensitivity analysis, we found that ICERs were reduced by approximately 25% by simulating a 24-week stopping rule. As well, it may be noted that we found only modest QALY gains associated with anti-TNF use over one year, though utility gains were up to 0.03 units higher among anti-TNF responders compared to non-responders. Few studies have directly reported utility gain associated with anti-TNF use, and the NICE cost-effectiveness models predicted utility from BASDAI and BASFI scores using an algorithm that has not been externally evaluated.121 It is difficult to determine whether the utility gain associated with anti-TNF use among DESIR patients is similar to what was predicted by the NICE, and there is an outstanding need for studies to report observed utility gain among patients using anti-TNF therapy.  In this study, we were unable to confirm whether France's restrictions are the most cost-effective in that setting relative to other potential restrictions over the short term; the uncertainty around the results in the basecase analysis indicates all of the criteria sets compared here may be equally cost-effective. However, the study makes a number of observations that highlight the potential for anti-TNF access regulations to shape the therapy's cost-effectiveness, in part by defining the target population for initiation, which both influences the likelihood of anti-TNF response and determines the appropriate population of non-users for comparison. In this study, we found that between 41% and 55% of anti-TNF users across the five study population datasets achieved a BASDAI 50 response, and the mean SF6D utility gain one year following anti-TNF initiation was higher among responders compared to non-responders. At the same time, cost-effectiveness estimates did not vary directly in accordance with the proportion of BASDAI 50 responders: the most favourable ICER was derived from the Hong Kong criteria dataset, though it had the lowest proportion of responders. This discrepancy appears to result from the lower utility among the anti-TNF non-users in the Hong Kong criteria dataset, which translated to a larger incremental difference in QALYs compared to other study population datasets. These findings underscore that, to maximize cost-effectiveness, anti-TNF therapy must be directed to patients mostly likely to experience substantial improvement in quality of life when compared to conventional care, and there is a strong need to inform anti-TNF access criteria with evidence to characterize this   139 patient population. To date, a good deal of research has demonstrated predictors of anti-TNF response, both among AS patients286-289 and nr-axSpA patients.266, 270-272, 290 However, there are shortcomings in this evidence base, with more data derived from RCT populations270-272, 275, 291 than observational cohorts266, 286, 287, 290 and more evidence on certain markers (e.g., CRP272, 287, 291) than others (e.g., HLA-B27290). Furthermore, few studies have modelled anti-TNF response based on combinations of clinical characteristics, which should be more useful for decision-making than single predictors.289 Perhaps most importantly, most studies have defined anti-TNF response in binary terms using clinical measures such as the BASDAI,124 ASAS40,123 or ASDAS292 and the effect on quality of life of achieving a response as defined by these measures has not been established.293  The results of this study suggest that one crucial strategy to improve anti-TNF cost-effectiveness is to ensure treatment discontinuation by anti-TNF users not experiencing clinical benefit. To implement this strategy, it would be useful to confirm minimally important differences on common quality of life measures,294, 295 to encourage clinicians to measure the benefits of anti-TNF therapy in terms of quality of life, and to help patients and providers engage in a shared decision-making process around discontinuation. We acknowledge that enforcement of regulations surrounding anti-TNF therapy is challenging, as reflected by the fact that 40% of anti-TNF users in DESIR did not satisfy the French anti-TNF access criteria. However, the potential for anti-TNF access criteria to shape the cost-effective use of these agents should not be ignored, and the rationale for initiating therapy only among patients likely to benefit- and for discontinuing therapy when appropriate- should be known by patients and providers.  Certain limitations of this study should be noted. For one, results differed depending on which anti-TNF users were included in the analysis: when anti-TNF users who received therapy prior to criteria satisfaction or six to twelve months after criteria satisfaction were included, anti-TNF therapy was dominated in all scenarios. This finding points to a possible role for timing of anti-TNF initiation in determining the therapy's cost-effectiveness; however, the results could also be explained by unmeasured, time-variant confounders. In general, the data analyzed here were derived from a non-randomized study, meaning all results are subject to confounding by indication. To help control for this, we made comparisons only between anti-TNF users and non-  140 users who satisfied the same set of access criteria and we further adjusted costs and QALYs for known confounders, though residual confounding cannot be ruled out. In terms of other study limitations, we assumed that all patients met treatment failure criteria, which were defined differently in the selected criteria sets. It should be noted that up to a third of SpA patients may achieve clinical remission with NSAIDs alone296 and anti-TNF therapy will necessarily be more cost-effective if used only by patients who have failed this less costly treatment. The present study did not evaluate the number of NSAIDs that should be tried before anti-TNF therapy in order to maximize cost-effectiveness, which is a limitation.  Despite its limitations, this study is unique in having used data from a heterogeneous, real-world population of SpA patients to demonstrate the influence of patient characteristics on anti-TNF cost-effectiveness estimates. In line with the initiative to incorporate economic evidence into clinical guidelines,297, 298 future research should focus on confirming what combination of patient characteristics best predicts quality of life improvement following anti-TNF therapy and informing anti-TNF access criteria with this evidence. As a substantial number of anti-TNF users in the DESIR cohort did not satisfy any of the selected criteria sets, and as discontinuation of anti-TNF therapy following non-response was infrequently observed, this study further calls for a discussion as to the practical application of regulations surrounding anti-TNF therapy.      141 Chapter 6  Discussion: Informing Economic Questions and Treatment Recommendations with Observational Data  6.1. Introduction  The overarching goal of this thesis was to use observational data from the DESIR cohort to address economic questions in the context of SpA treatment recommendations. The purpose of this chapter is to review its contributions to current knowledge according to each of its stated objectives, as well as the challenges encountered in the research process and the limitations affecting the results. The final section aims to place the thesis in the context of evidence-based paradigms in the health sciences, and to comment on its significance and implications, including future directions for research and policy.  6.2 Contribution of This Thesis to Current Knowledge, Challenges Encountered, and Limitations  6.2.1 Objective 1: The Quantification of the Cost of Illness From the Societal Perspective Among DESIR Patients and Identification of Factors Associated With Costs in the First Three Years of Follow-Up  The DESIR COI study was the first COI study to be conducted among a population of chiefly nr-axSpA patients. A primary finding of the DESIR COI study is that costs were largely driven by the quarter of patients who were using anti-TNF agents, which represented over 50% of total costs each year and which accounted for nearly 70% of costs in year three. The only SpA COI study to examine the effect of imaging status on costs, the DESIR COI study found that patterns of X-Ray and MRI findings at baseline were not associated with total costs or non-biologic costs. A foundational study, the main contribution of the DESIR COI study was in estimating the cost outcomes required to address economic research questions using data from the DESIR cohort.  Challenges encountered in the DESIR COI study arose in part from the scope of DESIR data. For one, the DESIR study does not collect information on patient out-of-pocket costs, meaning   142 the DESIR COI study excludes specific, identifiable cost components. Data on work productivity loss in DESIR are also limited to self-reported workdays lost, and do not include information on presenteeism or unpaid work productivity loss. Similarly, no direct data on income are collected, meaning wages in the DESIR COI study were estimated by linking self-reported data on professional category to public data on wages. The use of average wages rather than self-reported income data will have had an unknown effect on final estimates, and the estimated value of work productivity loss in the DESIR cohort will not be comparable to estimates that include presenteeism and unpaid work productivity loss.   To date, a standardized list of costing domains, such as that proposed by Merkesdal,3 has not been widely adopted in economic evaluations in rheumatology, and none was used in the DESIR COI study. As a matter of methodological choice, the DESIR COI study was an all-costs analysis, rather than costs specific to SpA. From a conceptual perspective, there is a strong argument in favour of including all costs, as manifestations of chronic health conditions and co-morbidities intersect and it may be difficult to designate which costs are SpA-related and which are not. By including costs stemming from non-SpA-related health resource use and work productivity loss, estimates from the DESIR COI study will be higher than those from studies that limited their scope to SpA-related costs. There remains an inherent difficulty in comparing the results of COI studies conducted in different settings, and as COI studies are conducted among other European early SpA cohorts208, 299, 300 it may be challenging, though interesting, to compare their results to those obtained in DESIR.   6.3.2 Objective 2: The Development of a Method to Measure Adherence to Specific Elements of the ASAS Treatment Recommendations  The second objective of this thesis was to collaborate with rheumatologists to develop measurable definitions of adherence to the 2010 ASAS SpA treatment recommendations. In the first phase of this research, a working group found that the lack information on frequency and/or dose of interventions made it impossible for them to directly define adherence to the ASAS recommendations. In the second phase of the research, a larger group of rheumatologists agreed that several items in the ASAS recommendations could not be translated into clinical actions.   143 This was an important observation given that clinicians are less likely to adhere to guidelines when actions are not precisely specified.241 The second phase of the research also uncovered disagreement between rheumatologists regarding recommended therapies. For example, rheumatologists differed in their interpretation of recommendations surrounding physiotherapy, with many voicing that this treatment is effectively optional if the patient's mobility is unaffected, or if they are performing home exercises instead. In general, feedback from rheumatologists pointed to a need to strengthen the ASAS recommendations with more information on the recommended frequency and dose of treatments and on patient characteristics that may influence treatment effects.   Some particular challenges arose in the Delphi study in the context of defining adherence to anti-TNF use recommendations. In DESIR, follow-up visits occur every six months, while the ASAS recommendations advise that anti-TNF therapy may be prescribed to patients with ≥4 weeks of high disease activity.1 Rheumatologists were therefore required to decide what number of consecutive 6-month follow-up visits with high disease activity should be interpreted as evidence of sustained disease activity over 4 weeks. On this matter, rheumatologists expressed that the label 'non-adherent' to recommendations should not be applied prematurely on the basis of observational data. In their proposed classification system, no anti-TNF users are defined as non-adherent on the basis of receiving therapy earlier than recommended, and patients who experience one full year of high disease activity before initiating anti-TNF therapy are still defined as adherent. As the ASAS recommendations do not describe when an anti-TNF 'must' be prescribed, it is difficult to know whether the latency period allowed by the classification system is appropriate or meaningful.  Another challenge confronted by rheumatologists was in choosing a proxy variable for 'expert opinion', cited as a requirement for anti-TNF use in the ASAS recommendations. Importantly, the ASAS recommendations do not define in quantitative terms the criteria that the expert should consider in deciding whether to prescribe anti-TNF therapy. The recommendations mention clinical features, including history and examination, as well as acute phase reactant levels and imaging results (X-Ray or MRI),124 but do not speak to specific findings that should prompt anti-TNF initiation. This leaves experts to interpret for themselves the available literature on patient   144 characteristics that predict a positive anti-TNF response and complicated the selection of a proxy variable (or variables) for expert opinion. Ultimately, rheumatologists chose to use the PhGA variable as a proxy variable for expert opinion; yet since expert opinion is poorly defined, it is unknown what PhGA is meant to capture. The inclusion of the PhGA variable is a limitation to the final classification system, as there is marked disagreement between rheumatologists’ opinions regarding which patients should start anti-TNF therapy.301   Overall, the biggest challenge to defining adherence to the ASAS recommendations was the non-specific language employed in them, particularly with respect to anti-TNF therapy. This characteristic of the ASAS recommendations is not necessarily a fault of their developers, but reflects insufficient knowledge on which SpA patients benefit most from anti-TNF therapy. Understandably, rheumatologists were reluctant to describe care as non-adherent without very strong evidence to justify the label, and the ASAS recommendations did not contain clear enough statements to constitute this evidence. In the end, rheumatologists' definition of adherence to anti-TNF recommendations incorporates few factors: a singular importance is attached to the number of consecutive visits that a patient has with both BASDAI and PhGA ≥4 but no anti-TNF therapy. Furthermore, the classification system fails to define any unnecessarily early anti-TNF use, which is an important shortcoming since this would come at great monetary cost. In hindsight, considering that it was established early on that the ASAS recommendations were not clear or quantifiable, it might be concluded that developing any classification system to define adherence to the ASAS recommendations was premature.  6.3.3 Objective 3: The Evaluation of the Impact of Adherence to Measurable Elements of Recommended SpA Care Among DESIR Patients in Terms of Costs and Health Benefits  This study appears to be the first to describe patterns of adherence to the ASAS SpA treatment recommendations, using definitions proposed by rheumatologists. The study found that nearly half of DESIR patients did not see a physiotherapist in the first year, making the lack of physiotherapy the most common form of non-adherent management in the DESIR cohort. Overall, the descriptive findings of this study suggest that SpA care in France is generally concordant with the ASAS recommendations regarding anti-TNF use recommendations, non-  145 biologic drug, and specialist care for IBD, uveitis, or psoriasis care. However, the definitions of adherence used in this study may have been insensitive; the descriptive findings of this study are limited by this potential.  As only small numbers of patients were defined as non-adherent to the different ASAS recommendations, measuring the effect of specific types of adherence was a challenge. Ultimately, the analytic portion of this study focused on the effect of adherence to anti-TNF recommendations, while controlling for the effect of adherence to physiotherapy and non-biologic drug recommendations. Very small numbers of patients were defined as non-adherent to non-biologic drug recommendations, which may account for the finding that this form of adherence was not associated with outcomes. A greater number of patients was defined as non-adherent to physiotherapy recommendations, yet this form of adherence was also found to be unrelated to outcomes among DESIR patients. The source of these insignificant findings could be the physiotherapy adherence definition: unless it were a proxy for engagement with physiotherapy relative to non-engagement, a minimum of one physiotherapy visit in a single year would not likely be associated with outcomes. At the same time, the apparent lack of effect is consistent with another observational study using DESIR data by Escalas et al.,302 which also failed to show benefits of physiotherapy. These findings from DESIR counter data from RCTs suggesting that physiotherapy improves pain and function in axSpA patients.303-307 The discrepancy raises the question of whether the benefits of physiotherapy are difficult to measure in observational studies, or whether physiotherapy is indeed ineffective at improving outcomes among axSpA patients outside the trial setting. If the latter is true, it raises the question as to whether a lack of standardization or quality of real-world physiotherapy programs explains the difference.302   Exploring adherence to anti-TNF use recommendations, only in a sensitivity analysis were benefits demonstrated: adherent, 'timely' anti-TNF use was associated with lower non-biologic costs compared to non-adherent 'late' use, and with significantly better health status compared to both non-adherent late use and non-adherent non-use. This finding seems to point to the potential for earlier initiation of anti-TNF therapy to result in greater benefit, yet a practical interpretation of the finding is challenging. Rather simplistically, the results suggest that there may be cost and   146 health benefits to initiating anti-TNF within six months when patients and physicians agree that disease activity is high, estimated at a 58% reduction in non-biologic costs and a 0.06 unit QALY increase. While pointing to the potential importance of timing of anti-TNF initiation, this does not provide further information to help determine which patients warrant anti-TNF therapy. Furthermore, the apparent benefits of adherence seem to have been driven by the 20 patients who received an anti-TNF specifically upon their second consecutive 6-month visit with high disease activity. These patients had worse outcomes (e.g., a positive anti-TNF response rate of only 5%) than all other patients, including patients who received an anti-TNF upon their third visit with high disease activity. This pattern is not easily interpretable, and the possibility that the apparent benefits of adherence are a result of chance should not be overlooked.   The main contribution of this study is in showing that the impact of adherence to recommendations is inherently dependent on the definition of adherence. The study's analytic strategy was informed by a classification system developed by rheumatologists, who attempted to interpret the ASAS recommendations in a quantitative way in order to define adherence. Yet the 2010 ASAS recommendations, by all accounts, contain little quantitative information. Ultimately, the challenges encountered in this quantitative analysis recall those of the qualitative study that preceded it. It is almost impossible to tell what constitutes adherence to the 2010 ASAS recommendations, and therefore it is difficult to observe or interpret the benefits of adherence.   6.3.4 Objective 4: The Assessment of Current French Anti-TNF Access Restrictions in Terms of the Cost-Effectiveness in That Setting Relative to Other Potential Restrictions  This study made three important observations. First, variations in anti-TNF access criteria translated to meaningful differences in the number of DESIR patients that would be eligible for anti-TNF therapy in selected settings, highlighting the importance of anti-TNF access criteria in shaping the budget impact of these agents. Second, the proportion of anti-TNF responders varied across subsets defined by access criteria, and average utility gain one year following anti-TNF initiation was higher among responders than non-responders. At the same time, the proportion of anti-TNF users classed as responders was not the only factor that influenced cost-effectiveness:   147 the utility level among the anti-TNF non-users included in the comparison was also a crucial determinant of cost-effectiveness. These findings served to illustrate how the estimated cost-effectiveness of anti-TNF therapy is influenced by heterogeneity in anti-TNF treatment effects relative to conventional care. Third, although previous cost-effectiveness models have assumed prompt discontinuation of anti-TNF therapy by non-responders, this was rare among real anti-TNF non-responders in the DESIR cohort. By simulating a 24-week stopping rule, this analysis demonstrated the large extent to which the continuation of anti-TNF therapy among non-responders increases costs.  One challenge encountered in this study was the power to detect differences between subsets- only a small proportion of DESIR patients satisfied the selected sets of anti-TNF access criteria, particularly those from the UK (13%) and Hong Kong (9%). The results of the analysis were ultimately inconclusive, as bootstrap confidence intervals around the five cost-effectiveness estimates overlapped. Cost-effectiveness estimates were also shown to depend on whether outcomes in anti-TNF users were observed following the date of criteria satisfaction or following the time of treatment initiation, a finding which is difficult to interpret. As differences between patient subsets created without a strong biological rationale should be interpreted with caution,308 the findings of this study may be best taken as a demonstration of the potential for anti-TNF access criteria to influence cost-effectiveness. At this level, the study points to the need to inform access criteria with better evidence of patient characteristics that predict meaningful quality of life gains from anti-TNF therapy, but also to observe health outcomes among subsets of anti-TNF non-users. If certain clinical characteristics predict good health outcomes with conventional care alone, treating patients with those characteristics with anti-TNF agents will reduce the cost-effectiveness of anti-TNF therapy.  6.4 Significance of Findings and Implications  There is a longstanding initiative to inform health care decisions with research evidence as a means to optimize efficiency and outcomes. This initiative is reflected in the overlapping paradigms of EBM, HTA, and, more recently, CER, each of which has been defined differently by health care organizations in Europe, Canada, and the US.138 According to the definition of   148 CER given by US IOM, this thesis shares a number of characteristics with this new paradigm in particular (Table 6.1). In CER it is emphasized that average treatment effects from RCTs may not translate to routine practice settings,309 that not all patients respond to treatments in the same way, and that there is a need to establish what will provide meaningful perceived benefit for the individual.310, 311 This focus means that CER makes certain types of research a priority, specifically those that allow researchers to observe the effects of interventions among large, heterogeneous populations representative of clinical practice. Thus, in a departure from paradigms that uphold RCTs as the gold standard in research design, CER stresses the value and importance of observational studies. Furthermore, in measuring benefits, CER prioritizes the use of outcome measures that are important to patients, replacing the clinical outcomes commonly used in RCTs by PROs, including ones that account for patient preferences.312 This thesis, by virtue of using observational data from a real-world population of SpA patients, by analyzing outcomes within patient subgroups, and by accounting for patient preferences by linking HRQoL to utilities, aligns with CER priorities.  Table 6.1 Shared Characteristics of CER and Current Work CER Current work Directly informs a specific clinical decision from the patient perspective or a health policy decision from the population perspective Informs decision-making relevant to development of treatment recommendations  Compares at least two alternative interventions, each with the potential to be “best practice” Compares treatment adherent to recommendations versus non-adherent to recommendations  Describes results at the population and subgroup levels Describes outcomes among subsets of patients adherent versus non-adherent to treatment recommendations Measures outcomes— both benefits and harms— that are important to patients Measures costs and QALYs Employs methods and data sources appropriate for the decision of interest Employs clinical and cost data on patients to whom treatment recommendations apply; employs restriction and statistical adjustment to account for confounding by indication Conducted in settings that are similar to those in which the intervention will be used in practice Conducted evaluation using observational data derived from routine clinical practice in France     149 In their framework of evidence-based paradigms (Figure 6.1), Luce et al.138 position CER as an accepted source of input into clinical guidelines. This has important implications for the future, as clinical guidelines informed by CER will move away from making recommendations solely on the basis of average effects observed in RCTs, and start to incorporate evidence from observational studies, including evidence of treatment effect heterogeneity.310 However, there is disagreement as to what role economic evaluation has to play in informing clinical guidelines,137 whether as an input to CER or via HTA, the paradigm traditionally seen as informing reimbursement decisions.138  Ultimately, this debate raises two questions relevant to the development of clinical guidelines. First, in the absolute sense, should clinical guidelines incorporate economic considerations? In the absolute sense, some authors have argued 'no', on the grounds that this equals 'rationing'-137, 313 others 'yes', on the basis that only by considering the costs and benefits of an intervention together can patients and providers understand its value.314  This thesis embraces the latter position, taken by the American College of Physicians,314 by the UK's NICE,297 and by the Australian government,136 for example. Second, if clinical guidelines are to incorporate economic considerations, what sources of economic evidence should be included? Those who promote an official role for economic evaluation within the CER paradigm have argued in favour of greater use of economic evidence based on real-world effectiveness data and PROs specifically.315 This thesis supports this view, and has demonstrated some of the unique input that observational studies can add to economic evaluations in the SpA field.    150 Figure 6.1 Redefined Relationships of Evidence Processes. Reproduced from Luce BR, Drummond M, Jönsson B, Neumann PJ, Schwartz JS, Siebert U, Sullivan SD. EBM, HTA, and CER: clearing the confusion. Milbank Q. 2010 Jun;88(2):256-76.    151 In January 2017, around the time this thesis was completed, ASAS published an update to their SpA treatment recommendations, in which "for the first time, cost considerations received a prominent place".316 While the 2010 ASAS recommendations make no mention of the word 'cost',1 implying the belief that economic considerations are outside their purview, the 2016 update contains numerous statements relative to the cost of anti-TNF agents specifically. To some extent, this reflects an evolution in SpA field in the perceived role of clinical guidelines, indicating new support for at least acknowledging economic considerations. However, although the recommendations assert that "costs have been taken into account at all times during the development of these recommendations"316 the authors voice a conscious choice to exclude formal cost-effectiveness evidence. In a series of statements, ASAS makes their position on the matter of costs explicit, in clear opposition to principles of economic evaluation:  "Treatment of axSpA should aim at the best care and must be based on a shared decision between the patient and the rheumatologist....'Best care’ is an important concept and closely relates to overarching principle 2: ‘to maximise health-related quality of life’. But ‘best care’ here refers to the ‘best possible care’ for individual patients, and still prevails when costs of treatment are taken into account."  "axSpA incurs high individual, medical and societal costs, all of which should be considered in its management by the treating rheumatologist...only if the outcome for the patient is expected to be similar under either treatment, healthcare costs can drive the choice."   "Needless to say that—for the quality of life of patients with axSpA—principles of ‘best care’ and ‘shared decision-making’ should outweigh cost considerations, but the latter remain significant."  Thus, although the 2016 ASAS recommendations move to acknowledge cost considerations, they ultimately suggest that interventions that afford improvement to health outcomes are to be pursued at any cost. With the introduction of biosimilars, (biological anti-TNF agents highly similar to the original products- i.e., 'bio-originators'- examined in this thesis), it appears that ASAS is implicitly supporting the use of these lower-cost agents over bio-originators with their statement "only if the outcome for the patient is expected to be similar under either treatment, healthcare costs can drive the choice".316 Yet the cost-effectiveness of biosimilars has not been established any more than that of bio-originators. In excluding formal cost-effectiveness evidence from consideration, the ASAS position effectively ignores, or disputes, the fact that   152 anti-TNF therapy will inevitably defer resources from other programs- even programs that may be highly valued by SpA patients themselves.   The position taken here is that SpA patients and the public would experience greater net benefit if the ASAS recommendations were to incorporate formal evidence from economic evaluations, acknowledging the opportunity costs of anti-TNF therapy. This position is not simply an ideological one, but is supported by empirical work that demonstrates the net loss in health that results from overinvestment in therapies that are cost-effective only at high WTP thresholds.317 As noted by Claxton,317 parties who stand to benefit from an intervention, including manufacturers, patients, and clinicians, exert great pressure on reimbursement authorities to encourage funding regardless of the intervention's cost-effectiveness. In contrast, the parties who stand to bear the health consequences of the diverted resources are unidentified, leaving them without an advocate. However, as anti-TNF therapy benefits only a portion of SpA patients, this unidentified group necessarily comprises members of the very population for which ASAS is meant to advocate. In order to better advocate for all SpA patients, ASAS would be well advised to incorporate formal cost-effectiveness evidence into their anti-TNF use recommendations, working with reimbursement authorities, not against them, to ensure these anti-TNF agents provide good value for all.136, 297  Beyond supporting the incorporation of economic evidence into the ASAS recommendations, this thesis supports establishing a formal role for economic evaluation within the CER paradigm and aligning the ASAS recommendations with CER priorities. This would translate to the greater use of economic evidence based on observational studies of SpA patients that 1) estimate the effects of anti-TNF agents in SpA populations representative of routine clinical practice; 2) account for anti-TNF response heterogeneity; and 3) incorporate outcomes that are important to SpA patients. In fact, this thesis demonstrates how the goals of CER are closely linked to one another, and to the goals of economic evaluation, in the context of SpA. Overall, the thesis points to the need to account for the heterogeneity in quality of life improvement that anti-TNF agents afford to SpA patients in real-world clinical practice, to quantify the economic implications of this variation by performing CEA within SpA patient subsets,318 and to inform SpA treatment recommendation with this evidence. Thus informed, SpA treatment recommendations could   153 describe in detail which patients should initiate anti-TNF therapy in order to maximize their value for patients and the public. Measurable definitions of adherence to recommendations could then be developed in collaboration with stakeholders, and the impact of adherence to recommendations on costs and health outcomes could be more easily observed and used to inform decision-making.   6.4.1 Research and Policy Directions  It is clear that additional research will be required to support the goals identified by this thesis. At the same time, the findings here suggest some policy measures could be taken immediately to help improve the cost-effectiveness of anti-TNF therapy in the interim. The following section outlines research and policy directions supported by this thesis.  6.4.1.1 Continue Work to Identify Which SpA Patients Are Likely to Respond to Anti-TNF Agents, Defining 'Response' in Terms of Meaningful Utility Gain  The 2010 ASAS recommendations "are based primarily on inadequate response to conventional therapies and less on the expectation that an anti-TNF agent will be effective in a particular patient".289 In part, this reflects the limitations in the evidence base available to inform the 2010 recommendations: even today, although there is research demonstrating individual predictors of a 'positive' anti-TNF response using clinical measures, the work has not yet translated to a perfect description of the SpA patients that will respond to anti-TNF agents.319 In the 2016 ASAS recommendations, some additional statements have been inserted on this topic, with the authors asserting that elevated CRP is the strongest predictor of a good response to anti-TNF therapy, followed by inflammation on MRI. Yet again the recommendations defer to rheumatologists, stating: "The task force hopes that rheumatologists will take CRP and (when available) MRI into consideration when deciding about the appropriateness of starting a bDMARD, irrespective of whether radiographic sacroiliitis is present or not".316 Thus, despite pointing to the rationale for considering these predictors, the 2016 ASAS recommendations still advise the use of expert opinion in making the final determination regarding anti-TNF initiation. As the work here suggests, this is likely to promote a pattern of anti-TNF use whereby one   154 patient will respond for every two that are treated. From an economic perspective, this 50% response pattern stands to have important consequences, particularly since the 2016 ASAS recommendations advise assessing treatment response after 'at least 12 weeks'. This suggests that each trial of anti-TNF agents among non-responders will translate to a minimum €3,000 investment without appreciable benefit in terms of BASDAI improvement. Thus, even before considering formal cost-effectiveness evidence, there is a strong economic rationale for research to describe as accurately as possible the subsets of SpA patients that will experience a positive anti-TNF response. Performing formal CEA within these subsets318 would then provide optimal evidence to inform clinical guidelines.  Guided by the principles of CER as well as economic evaluation, future research to identify patients likely to respond to anti-TNF therapy would best support decision-making if response were defined in terms of a meaningful gain in utility. This points to two simple, yet important priorities for future research. For one, numerous trials of anti-TNF agents to date have collected quality of life data using the SF36 or EQ5D, yet these outcomes have not been consistently reported or linked to utilities.121 Additional analysis of existing trial data is warranted in order to directly compare utility gain in anti-TNF users and non-users, controlling for patient characteristics. This approach would better characterize the net benefit of anti-TNF therapy in different subsets of SpA patients, taking into consideration all of the factors that contribute to treatment effect heterogeneity. As outlined by Kravitz,320 these factors include risk without treatment (e.g., what are the expected health outcomes without anti-TNF therapy), responsiveness (e.g., effect of anti-TNF therapy on reducing SpA symptoms), vulnerability (e.g., probability of adverse effects of anti-TNF therapy), and patient preferences (e.g., to what extent the overall effects of anti-TNF use or non-use correspond to health states preferred by patients). In short, for anti-TNF therapy to be cost-effective, it must be directed to patients who are likely to do poorly under conventional care, to experience a meaningful reduction in SpA symptoms, to avoid substantial side effects, and, ultimately, to prefer the health state associated with anti-TNF therapy. A first priority for research, therefore, is to measure the effects of anti-TNF therapy in terms of utility- "the ultimate arbiter of treatment success".320 A second, related priority is to define a threshold for meaningful utility gain among SpA patients. Although research has been conducted to establish the MID in utility, there is wide variation in MID values corresponding to   155 specific instruments and populations, as well as uncertainty as to which instrument is most suitable for capturing MID in utility.295 Currently, an OMERACT initiative is underway to identify shortcomings in PROs in use in economic evaluations,321 which should help identify the optimal tool for measuring utility in SpA patients and ultimately inform the definition of meaningful benefit from anti-TNF therapy.  6.4.1.2 Use Actionable Wording in Treatment Recommendations and Develop 'Quality Indicators' to Measure Adherence to Recommendations   To be effectively implemented, clinical guidelines must contain “actionable and precise” definitions of best practices, which then become the basis for formal quality indicators.242 Speaking on the topic of AS in 2007, Zochling and Braun expressed uncertainty as to whether the ASAS recommendations contained information specific enough to form the basis for quality indicators, voicing the need for a collaborative approach to refine outstanding issues before quality indicators can be developed.246 Ten years later, the need for quality indicators in SpA is still unfulfilled; however, the evidence base to inform SpA treatment recommendations continues to be strengthened. There is every reason to champion the development of quality indicators alongside the next revision of SpA treatment recommendations, as this will provide the means to evaluate the implementation and effect of treatment recommendations and to target areas for improvement. The development of quality indicators should be pursued in collaboration with SpA patients, rheumatologists and other stakeholders, who can determine together what are appropriate and realistic standards for care.322 Importantly, some level of non-adherence to recommendations should be both expected and accepted: some reasons for intentional non-adherence have been described and assessed as valid, including contraindications and patient decision.323 Thus, target endpoints, i.e., optimal levels of adherence to recommendations and achievement of quality indicators, should be determined by stakeholders taking these factors into consideration. Finally, considering the rationale for including economic evidence in SpA treatment recommendations, the collaborative process to develop quality indicators may include stakeholders with expertise in economic evaluation. Predecent for the involvement of health economists in guideline development has been set in the UK, where dedicated 'guideline development groups' (GDG) follow a process to inform national treatment recommendations   156 with cost-effectiveness evidence.324 In proposing recommendations, the GDG states the specific treatment and population to which the recommendation applies, and uses action-based wording to help translate the recommendation into clinical practice.325  6.4.1.3 Consider Innovative Policies Linking Anti-TNF Coverage to Outcome and/or Evidence Development; Emphasize Shared Goals of Clinical and Reimbursement Guidelines by Supporting Appropriate Anti-TNF Discontinuation  Research to identify the SpA patients most likely to experience meaningful benefit from anti-TNF therapy will not be accomplished overnight, and, for an indefinite period, anti-TNF therapy will almost certainly be initiated in more SpA patients than will experience meaningful benefit. This suggests two priorities for policy-making. First, reimbursement authorities may be advised to consider innovative policies that would link the coverage of anti-TNF agents to outcomes and/or evidence development. Several such policies have been outlined by Walker et al.,326 who assert that the choice between them will depend in part on the power possessed by the reimbursement authority. In settings where the reimbursement authority has the power to determine the effective price for an intervention, reimbursement may be linked to clinical outcomes at the individual patient level. This may be accomplished through numerous potential mechanisms, including money-back guarantees (i.e., the reimbursement authority is refunded by the manufacturer if the patient does not achieve a specified outcome), 'conditional treatment continuation' (i.e., the reimbursement authority pays for treatment continuation only in patients achieving a specified outcome), and price linked to outcome schemes (i.e., the price paid for the intervention corresponds directly to the level of improvement it affords).326 In settings where the reimbursement authority has the power to ensure that research is conducted, there is an additional option to institute a policy whereby an intervention is funded only for patients who are participating in research (an 'Only In Research' policy) or funded only upon the condition that additional research will be conducted ('Only With Research' policy). Importantly, these policies are only attractive when the cost of additional research is seen to be a good investment, and are not advised when existing evidence already strongly suggests that an intervention is not cost-effective. In the French setting, a number of different policies linking reimbursement to outcomes and/or evidence development could be considered, since, as noted by Walker et al.326,   157 the reimbursement authority in France has both the power to set the effective price of interventions and to ensure that research is conducted.   In addition to innovative pricing policies for anti-TNF agents, there is a strong need to foster collaboration between clinical guideline developers and reimbursement authorities to reinforce their shared goals, including the appropriate discontinuation of anti-TNF therapy among non-responders. While reimbursement authorities have a clear economic interest in supporting anti-TNF discontinuation, ASAS and other guideline developers should have a clinical interest in the same goal. This follows from the fact that the long-term effects of anti-TNF agents have not been conclusively established, including their effects on radiographic progression327 but also on the risk of malignancy or other latent side effects.328 Thus, from a clinical perspective, there is reason to avoid exposing patients to anti-TNF therapy unless the short-term benefits are sure to outweigh the potential long-term harms.  6.4.1.4 Invest in Observational Research Including Evaluation of Adherence to Recommendations   Some of the most significant findings of this thesis were the simplest: over 40% of anti-TNF users in the DESIR cohort did not satisfy any of the selected anti-TNF access criteria sets, and the majority of anti-TNF non-responders continued therapy for a year or more. These findings are important not because they are sophisticated, but because they describe the real-world use of anti-TNF agents, including patterns with economic implications. This underscores the value of observational research in informing health policy. Yet, as noted, certain challenges encountered in the course of this research arose from shortcomings in the observational data available. This included limitations in the cost components captured, the six-month intervals in which disease activity data were available, and in the number of patients available for subset analysis. This experience points to the potential advantage of making further investment in observational research, optimizing evidence for decision-making. To start, a more exhaustive approach could be taken to collecting clinical and cost data in SpA cohort studies, capturing more parameters and at more frequent intervals. Dependent on their future development, the collection of quality indicators within SpA cohorts would allow for the evaluation of the implementation of treatment   158 recommendations and their impact on health and economic outcomes. Furthermore, an initiative could be taken to amalgamate data across existing cohorts,204-210 which would enhance the opportunity for subset analyses specifically. Indeed, it would be interesting to re-apply the methods used here but using a larger database, in which greater numbers of patients would fall into unique subsets, facilitating a better assessment of the heterogeneity in anti-TNF cost-effectiveness. This would be an exciting opportunity to realize the full potential of the methods applied in this thesis, which incorporated some of the latest means to overcome the challenges inherent in observational research.310  6.5 Conclusion  The field of SpA is unique in the extent to which it highlights the intersection between clinical and economic questions in health research. The increased attention to SpA in the last decade may be viewed, at least in part, as having emerged as a result of the advent of anti-TNF therapy, which from the perspective of its developers there is an economic incentive to deliver to as many SpA patients as possible. Whether there is an incentive to do so from the societal perspective has yet to be conclusively determined. In parallel to this question, one remains as to whether SpA treatment recommendations should include or exclude evidence from economic evaluations. Based on the lessons learned in the current work, the conclusion here is that including economic evidence in treatment recommendations would help achieve the same patient-oriented goals that are being increasingly embraced by patients and providers. From a natural place within the CER paradigm, the economic evidence base to support SpA treatment recommendations should continue to be built systematically, with studies that acknowledge the heterogeneity of this unique patient group in the real world.      159 References  1. Braun J, van den Berg R, Baraliakos X, et al. 2010 update of the ASAS/EULAR recommendations for the management of ankylosing spondylitis. Ann Rheum Dis. 2011;70:896-904. 2. van den Berg R, Stanislawska-Biernat E, van der Heijde DM. Comparison of recommendations for the use of anti-tumour necrosis factor therapy in ankylosing spondylitis in 23 countries worldwide. Rheumatology. 2011;50:2270-2277. 3. Merkesdal S, Ruof J, Huelsemann JL, et al. Development of a matrix of cost domains in economic evaluation of rheumatoid arthritis. J Rheumatol. 2001;28:657-661. 4. Kellgren JH. Diagnostic criteria for population studies. Bull Rheum Dis. 1962;13:291-292. 5. Moll JM, Wright V. New York clinical criteria for ankylosing spondylitis. A statistical evaluation. Ann Rheum Dis. 1973;32:354-363. 6. Calin A, Porta J, Fries JF, Schurman DJ. Clinical history as a screening test for ankylosing spondylitis. JAMA. 1977;237:2613-2614. 7. Rudwaleit M, Metter A, Listing J, Sieper J, Braun J. Inflammatory back pain in ankylosing spondylitis: a reassessment of the clinical history for application as classification and diagnostic criteria. Arthritis Rheum. 2006;54:569-578. 8. Sieper J, van der Heijde D, Landewe R, et al. New criteria for inflammatory back pain in patients with chronic back pain: a real patient exercise by experts from the Assessment of SpondyloArthritis international Society (ASAS). Ann Rheum Dis. 2009;68:784-788. 9. van der Linden S, Valkenburg HA, Cats A. Evaluation of diagnostic criteria for ankylosing spondylitis. A proposal for modification of the New York criteria. Arthritis Rheum. 1984;27:361-368. 10. Moll JM, Haslock I, Macrae IF, Wright V. Associations between ankylosing spondylitis, psoriatic arthritis, Reiter's disease, the intestinal arthropathies, and Behcet's syndrome. Medicine (Baltimore). 1974;53:343-364. 11. Prakash S, Bansal R, Rajagopalan P, Malaviya AN. Immunological studies in seronegative spondyloarthropathies. Br J Rheumatol. 1983;22:146-150. 12. Calin A. Spondyloarthropathy in Caucasians and non-Caucasians. J Rheumatol Suppl. 1983;10:16-19. 13. Prakash S, Mehra NK, Bhargava S, Malaviya AN. HLA B27 related 'unclassifiable' seronegative spondyloarthropathies. Ann Rheum Dis. 1983;42:640-643. 14. Khan MA, van der Linden SM. A wider spectrum of spondyloarthropathies. Semin Arthritis Rheum. 1990;20:107-113. 15. Amor B, Dougados M, Mijiyawa M. Criteria of the classification of spondylarthropathies. Rev Rhum Mal Osteoartic. 1990;57:85-89. 16. Brewerton DA. Causes of arthritis. Lancet. 1988;2:1063-1066. 17. Dougados M, van der Linden S, Juhlin R, et al. The European Spondylarthropathy Study Group preliminary criteria for the classification of spondylarthropathy. Arthritis Rheum. 1991;34:1218-1227. 18. Braun J, Bollow M, Eggens U, Konig H, Distler A, Sieper J. Use of dynamic magnetic resonance imaging with fast imaging in the detection of early and advanced sacroiliitis in spondylarthropathy patients. Arthritis Rheum. 1994;37:1039-1045. 19. Ahlstrom H, Feltelius N, Nyman R, Hallgren R. Magnetic resonance imaging of sacroiliac joint inflammation. Arthritis Rheum. 1990;33:1763-1769.   160 20. Oostveen J, Prevo R, den Boer J, van de Laar M. Early detection of sacroiliitis on magnetic resonance imaging and subsequent development of sacroiliitis on plain radiography. A prospective, longitudinal study. J Rheumatol. 1999;26:1953-1958. 21. Mau W, Zeidler H, Mau R, et al. Clinical features and prognosis of patients with possible ankylosing spondylitis. Results of a 10-year followup. J Rheumatol. 1988;15:1109-1114. 22. Rudwaleit M, Khan MA, Sieper J. The challenge of diagnosis and classification in early ankylosing spondylitis: do we need new criteria? Arthritis Rheum. 2005;52:1000-1008. 23. Sieper J, Rudwaleit M. Early referral recommendations for ankylosing spondylitis (including pre-radiographic and radiographic forms) in primary care. Ann Rheum Dis. 2005;64:659-663. 24. van der Heijde D, Bellamy N, Calin A, Dougados M, Khan MA, van der Linden S. Preliminary core sets for endpoints in ankylosing spondylitis. Assessments in Ankylosing Spondylitis Working Group. J Rheumatol. 1997;24:2225-2229. 25. van der Heijde D, van der Linden S, Bellamy N, Calin A, Dougados M, Khan MA. Which domains should be included in a core set for endpoints in ankylosing spondylitis? Introduction to the ankylosing spondylitis module of OMERACT IV. J Rheumatol. 1999;26:945-947. 26. van der Heijde D, Calin A, Dougados M, Khan MA, van der Linden S, Bellamy N. Selection of instruments in the core set for DC-ART, SMARD, physical therapy, and clinical record keeping in ankylosing spondylitis. Progress report of the ASAS Working Group. Assessments in Ankylosing Spondylitis. J Rheumatol. 1999;26:951-954. 27. Rudwaleit M, Landewe R, van der Heijde D, et al. The development of Assessment of SpondyloArthritis international Society classification criteria for axial spondyloarthritis (part I): classification of paper patients by expert opinion including uncertainty appraisal. Ann Rheum Dis. 2009;68:770-776. 28. Rudwaleit M, van der Heijde D, Landewe R, et al. The development of Assessment of SpondyloArthritis international Society classification criteria for axial spondyloarthritis (part II): validation and final selection. Ann Rheum Dis. 2009;68:777-783. 29. Rudwaleit M, van der Heijde D, Landewe R, et al. The Assessment of SpondyloArthritis International Society classification criteria for peripheral spondyloarthritis and for spondyloarthritis in general. Ann Rheum Dis. 2011;70:25-31. 30. Zeidler H, Amor B. The Assessment in Spondyloarthritis International Society (ASAS) classification criteria for peripheral spondyloarthritis and for spondyloarthritis in general: the spondyloarthritis concept in progress. Ann Rheum Dis. 2011;70:1-3. 31. Ruof J, Stucki G. Validity aspects of erythrocyte sedimentation rate and C-reactive protein in ankylosing spondylitis: a literature review. J Rheumatol. 1999;26:966-970. 32. Spoorenberg A, van der Heijde D, de Klerk E, et al. Relative value of erythrocyte sedimentation rate and C-reactive protein in assessment of disease activity in ankylosing spondylitis. J Rheumatol. 1999;26:980-984. 33. Tse SM, Laxer RM. New advances in juvenile spondyloarthritis. Nat Rev Rheumatol. 2012;8:269-279. 34. Rudwaleit M, Khan MA, Sieper J. The challenge of diagnosis and classification in early ankylosing spondylitis: do we need new criteria? Arthritis Rheum. 2005;52:1000-1008. 35. Dean LE, Jones GT, MacDonald AG, Downham C, Sturrock RD, Macfarlane GJ. Global prevalence of ankylosing spondylitis. Rheumatology. 2014;53:650-657. 36. van Tubergen A. The changing clinical picture and epidemiology of spondyloarthritis. Nat Rev Rheumatol. 2015;11:110-118.   161 37. Reveille JD, Witter JP, Weisman MH. Prevalence of axial spondylarthritis in the United States: estimates from a cross-sectional survey. Arthritis Care Res. 2012;64:905-910. 38. Zochling J, Braun J. Assessments in ankylosing spondylitis. Best Pract Res Clin Rheumatol. 2007;21:699-712. 39. Sieper J, Rudwaleit M, Baraliakos X, et al. The Assessment of SpondyloArthritis international Society (ASAS) handbook: a guide to assess spondyloarthritis. Ann Rheum Dis. 2009;68 Suppl 2:ii1-44. 40. U.S. Department of Health and Human Services FDA Center for Drug Evaluation and Research, U.S. Department of Health and Human Services FDA Center for Biologics Evaluation and Research, U.S. Department of Health and Human Services FDA Center for Devices and Radiological Health. Guidance for industry: patient-reported outcome measures: use in medical product development to support labeling claims: draft guidance. Health Qual Life Outcomes. 2006;4:79. 41. Wahl ER, Yazdany J. Challenges and Opportunities in Using Patient-reported Outcomes in Quality Measurement in Rheumatology. Rheum Dis Clin North Am. 2016;42:363-375. 42. Carron P, Van Praet L, Van den Bosch F. Peripheral manifestations in spondyloarthritis: relevance for diagnosis, classification and follow-up. Curr Opin Rheumatol. 2012;24:370-374. 43. Ramiro S, Stolwijk C, van Tubergen A, et al. Evolution of radiographic damage in ankylosing spondylitis: a 12 year prospective follow-up of the OASIS study. Ann Rheum Dis. 2015;74:52-59. 44. Rudwaleit M, Khan MA, Sieper J. The challenge of diagnosis and classification in early ankylosing spondylitis: do we need new criteria? Arthritis Rheum. 2005;52:1000-1008. 45. Bennett AN, McGonagle D, O'Connor P, et al. Severity of baseline magnetic resonance imaging-evident sacroiliitis and HLA-B27 status in early inflammatory back pain predict radiographically evident ankylosing spondylitis at eight years. Arthritis Rheum. 2008;58:3413-3418. 46. Poddubnyy D, Sieper J. Similarities and differences between nonradiographic and radiographic axial spondyloarthritis: a clinical, epidemiological and therapeutic assessment. Curr Opin Rheumatol. 2014;26:377-383. 47. Reveille JD. Biomarkers for diagnosis, monitoring of progression, and treatment responses in ankylosing spondylitis and axial spondyloarthritis. Clin Rheumatol. 2015;34:1009-1018. 48. Bredella MA, Steinbach LS, Morgan S, Ward M, Davis JC. MRI of the sacroiliac joints in patients with moderate to severe ankylosing spondylitis. AJR Am J Roentgenol. 2006;187:1420-1426. 49. Rudwaleit M, Haibel H, Baraliakos X, et al. The early disease stage in axial spondylarthritis: results from the German Spondyloarthritis Inception Cohort. Arthritis Rheum. 2009;60:717-727. 50. Arends S, Brouwer E, van der Veer E, et al. Baseline predictors of response and discontinuation of tumor necrosis factor-alpha blocking therapy in ankylosing spondylitis: a prospective longitudinal observational cohort study. Arthritis Res Ther. 2011;13:R94. 51. Wallis D, Haroon N, Ayearst R, Carty A, Inman RD. Ankylosing spondylitis and nonradiographic axial spondyloarthritis: part of a common spectrum or distinct diseases? J Rheumatol. 2013;40:2038-2041. 52. Jones SD, Steiner A, Garrett SL, Calin A. The Bath Ankylosing Spondylitis Patient Global Score (BAS-G). Br J Rheumatol. 1996;35:66-71. 53. Nhan DT, Caplan L. Patient-Reported Outcomes in Axial Spondyloarthritis. Rheum Dis Clin North Am. 2016;42:285-299.   162 54. Jones SD, Steiner A, Garrett SL, Calin A. The Bath Ankylosing Spondylitis Patient Global Score (BAS-G). Br J Rheumatol. 1996;35:66-71. 55. Kviatkovsky MJ, Ramiro S, Landewe R, et al. The Minimum Clinically Important Improvement and Patient-acceptable Symptom State in the BASDAI and BASFI for Patients with Ankylosing Spondylitis. J Rheumatol. 2016;43:1680-1686. 56. Calin A, Garrett S, Whitelock H, et al. A new approach to defining functional ability in ankylosing spondylitis: the development of the Bath Ankylosing Spondylitis Functional Index. J Rheumatol. 1994;21:2281-2285. 57. Fernandez-Espartero C, de Miguel E, Loza E, et al. Validity of the ankylosing spondylitis disease activity score (ASDAS) in patients with early spondyloarthritis from the Esperanza programme. Ann Rheum Dis. 2014;73:1350-1355. 58. Spoorenberg A, van Tubergen A, Landewe R, et al. Measuring disease activity in ankylosing spondylitis: patient and physician have different perspectives. Rheumatology. 2005;44:789-795. 59. Desthieux C, Molto A, Granger B, Saraux A, Fautrel B, Gossec L. Patient-physician discordance in global assessment in early spondyloarthritis and its change over time: the DESIR cohort. Ann Rheum Dis. 2016;75:1661-1666. 60. Guyatt GH, Veldhuyzen Van Zanten SJ, Feeny DH, Patrick DL. Measuring quality of life in clinical trials: a taxonomy and review. CMAJ. 1989;140:1441-1448. 61. Haywood KL, Garratt AM, Dziedzic K, Dawes PT. Generic measures of health-related quality of life in ankylosing spondylitis: reliability, validity and responsiveness. Rheumatology. 2002;41:1380-1387. 62. Guyatt GH, Feeny DH, Patrick DL. Measuring health-related quality of life. Ann Intern Med. 1993;118:622-629. 63. Fitzpatrick R, Davey C, Buxton MJ, Jones DR. Evaluating patient-based outcome measures for use in clinical trials. Health Technol Assess. 1998;2:i-iv, 1-74. 64. Haywood KL, Garratt AM, Dziedzic K, Dawes PT. Patient centered assessment of ankylosing spondylitis-specific health related quality of life: evaluation of the Patient Generated Index. J Rheumatol. 2003;30:764-773. 65. Haywood KL, Garratt AM, Jordan KP, Healey EL, Packham JC. Evaluation of ankylosing spondylitis quality of life (EASi-QoL): reliability and validity of a new patient-reported outcome measure. J Rheumatol. 2010;37:2100-2109. 66. Kotsis K, Voulgari PV, Drosos AA, Carvalho AF, Hyphantis T. Health-related quality of life in patients with ankylosing spondylitis: a comprehensive review. Expert Rev Pharmacoecon Outcomes Res. 2014;14:857-872. 67. Doward LC, Spoorenberg A, Cook SA, et al. Development of the ASQoL: a quality of life instrument specific to ankylosing spondylitis. Ann Rheum Dis. 2003;62:20-26. 68. Guillemin F, Challier B, Urlacher F, Vancon G, Pourel J. Quality of life in ankylosing spondylitis: validation of the ankylosing spondylitis Arthritis Impact Measurement Scales 2, a modified Arthritis Impact Measurement Scales Questionnaire. Arthritis Care Res. 1999;12:157-162. 69. Guillemin F, Challier B, Urlacher F, Vancon G, Pourel J. Quality of life in ankylosing spondylitis: validation of the ankylosing spondylitis Arthritis Impact Measurement Scales 2, a modified Arthritis Impact Measurement Scales Questionnaire. Arthritis Care Res. 1999;12:157-162. 70. Zochling J. Measures of symptoms and disease status in ankylosing spondylitis: Ankylosing Spondylitis Disease Activity Score (ASDAS), Ankylosing Spondylitis Quality of Life Scale   163 (ASQoL), Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), Bath Ankylosing Spondylitis Functional Index (BASFI), Bath Ankylosing Spondylitis Global Score (BAS-G), Bath Ankylosing Spondylitis Metrology Index (BASMI), Dougados Functional Index (DFI), and Health Assessment Questionnaire for the Spondylarthropathies (HAQ-S). Arthritis Care Res. 2011;63 Suppl 11:S47-58. 71. Wiebe S, Guyatt G, Weaver B, Matijevic S, Sidwell C. Comparative responsiveness of generic and specific quality-of-life instruments. J Clin Epidemiol. 2003;56:52-60. 72. Hurst NP, Kind P, Ruta D, Hunter M, Stubbings A. Measuring health-related quality of life in rheumatoid arthritis: validity, responsiveness and reliability of EuroQol (EQ-5D). Br J Rheumatol. 1997;36:551-559. 73. Jaeschke R, Singer J, Guyatt GH. Measurement of health status. Ascertaining the minimal clinically important difference. Control Clin Trials. 1989;10:407-415. 74. Ware JE,Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care. 1992;30:473-483. 75. Walters SJ, Brazier JE. What is the relationship between the minimally important difference and health state utility values? The case of the SF-6D. Health Qual Life Outcomes. 2003;1:4. 76. Yang X, Fan D, Xia Q, et al. The health-related quality of life of ankylosing spondylitis patients assessed by SF-36: a systematic review and meta-analysis. Qual Life Res. 2016. 77. Beresniak A, Russell AS, Haraoui B, Bessette L, Bombardier C, Duru G. Advantages and limitations of utility assessment methods in rheumatoid arthritis. J Rheumatol. 2007;34:2193-2200. 78. Kobelt G, Andlin-Sobocki P, Brophy S, Jonsson L, Calin A, Braun J. The burden of ankylosing spondylitis and the cost-effectiveness of treatment with infliximab (Remicade). Rheumatology. 2004;43:1158-1166. 79. Kobelt G, Andlin-Sobocki P, Maksymowych WP. Costs and quality of life of patients with ankylosing spondylitis in Canada. J Rheumatol. 2006;33:289-295. 80. Ariza-Ariza R, Hernandez-Cruz B, Lopez-Antequera G, Navarro-Sarabia F. Variables related to utility in patients with ankylosing spondylitis. Clin Rheumatol. 2009;28:207-211. 81. Kawalec P, Malinowski K. Disease activity, quality of life and indirect costs of reduced productivity at work, generated by Polish patients with ankylosing spondylitis. Reumatologia. 2015;53:301-308. 82. Chen HH, Chen DY, Chen YM, Lai KL. Health-related quality of life and utility: comparison of ankylosing spondylitis, rheumatoid arthritis, and systemic lupus erythematosus patients in Taiwan. Clin Rheumatol. 2016. 83. Boonen A, Patel V, Traina S, Chiou CF, Maetzel A, Tsuji W. Rapid and sustained improvement in health-related quality of life and utility for 72 weeks in patients with ankylosing spondylitis receiving etanercept. J Rheumatol. 2008;35:662-667. 84. Verstappen SM, Jacobs JW, van der Heijde DM, et al. Utility and direct costs: ankylosing spondylitis compared with rheumatoid arthritis. Ann Rheum Dis. 2007;66:727-731. 85. Braun J, McHugh N, Singh A, Wajdula JS, Sato R. Improvement in patient-reported outcomes for patients with ankylosing spondylitis treated with etanercept 50 mg once-weekly and 25 mg twice-weekly. Rheumatology. 2007;46:999-1004. 86. Dougados M, Tsai WC, Saaibi DL, et al. Evaluation of Health Outcomes with Etanercept Treatment in Patients with Early Nonradiographic Axial Spondyloarthritis. J Rheumatol. 2015;42:1835-1841.   164 87. Haywood KL, Garratt AM, Dziedzic K, Dawes PT. Generic measures of health-related quality of life in ankylosing spondylitis: reliability, validity and responsiveness. Rheumatology. 2002;41:1380-1387. 88. Van Tubergen A, Boonen A, Landewe R, et al. Cost effectiveness of combined spa-exercise therapy in ankylosing spondylitis: a randomized controlled trial. Arthritis Rheum. 2002;47:459-467. 89. Bansback N, Harrison M, Brazier J, et al. Health state utility values: a description of their development and application for rheumatic diseases. Arthritis Rheum. 2008;59:1018-1026. 90. Torrance GW. Measurement of health state utilities for economic appraisal. J Health Econ. 1986;5:1-30. 91. Zhang W, Bansback N, Anis AH. Measuring and valuing productivity loss due to poor health: A critical review. Soc Sci Med. 2011;72:185-192. 92. Zhang W, Chiu JA, Bansback N, Anis AH. An update on the measurement of productivity losses due to rheumatoid diseases. Best Pract Res Clin Rheumatol. 2012;26:585-597. 93. van der Weijden MA, Boonen A, van der Horst-Bruinsma IE. Problems in work participation and resource use should not be underestimated in patients with early spondyloarthritis. J Rheumatol. 2014;41:2413-2420. 94. van der Burg LR, Ter Wee MM, Boonen A. Effect of biological therapy on work participation in patients with ankylosing spondylitis: a systematic review. Ann Rheum Dis. 2012;71:1924-1933. 95. Boonen A, van der Heijde D, Landewe R, et al. Work status and productivity costs due to ankylosing spondylitis: comparison of three European countries. Ann Rheum Dis. 2002;61:429-437. 96. Boonen A, Brinkhuizen T, Landewe R, van der Heijde D, Severens JL. Impact of ankylosing spondylitis on sick leave, presenteeism and unpaid productivity, and estimation of the societal cost. Ann Rheum Dis. 2010;69:1123-1128. 97. Beaton D, Bombardier C, Escorpizo R, et al. Measuring worker productivity: frameworks and measures. J Rheumatol. 2009;36:2100-2109. 98. Stolwijk C, Castillo-Ortiz JD, Gignac M, Luime J, Boonen A, OMERACT Worker Productivity Group. Importance of Contextual Factors When Measuring Work Outcome in Ankylosing Spondylitis: A Systematic Review by the OMERACT Worker Productivity Group. Arthritis Care Res. 2015;67:1316-1327. 99. Ward MM, Kuzis S. Risk factors for work disability in patients with ankylosing spondylitis. J Rheumatol. 2001;28:315-321. 100. Forejtova S, Mann H, Stolfa J, et al. Factors influencing health status and disability of patients with ankylosing spondylitis in the Czech Republic. Clin Rheumatol. 2008;27:1005-1013. 101. Boonen A, van der Heijde D, Landewe R, et al. Work status and productivity costs due to ankylosing spondylitis: comparison of three European countries. Ann Rheum Dis. 2002;61:429-437. 102. Ramos-Remus C, Hernandez-Rios G, Duran-Barragan S, et al. Fifteen-year trends of long-term disability and sick leaves in ankylosing spondylitis. Clin Rheumatol. 2011;30:361-367. 103. Gordeev VS, Maksymowych WP, Schachna L, Boonen A. Understanding presenteeism in patients with ankylosing spondylitis: contributing factors and association with sick leave. Arthritis Care Res. 2014;66:916-924. 104. Healey EL, Haywood KL, Jordan KP, Garratt A, Packham JC. Impact of ankylosing spondylitis on work in patients across the UK. Scand J Rheumatol. 2011;40:34-40.   165 105. Maksymowych WP, Gooch KL, Wong RL, Kupper H, van der Heijde D. Impact of age, sex, physical function, health-related quality of life, and treatment with adalimumab on work status and work productivity of patients with ankylosing spondylitis. J Rheumatol. 2010;37:385-392. 106. Ward MM, Deodhar A, Akl EA, et al. American College of Rheumatology/Spondylitis Association of America/Spondyloarthritis Research and Treatment Network 2015 Recommendations for the Treatment of Ankylosing Spondylitis and Nonradiographic Axial Spondyloarthritis. Arthritis Care Res. 2016;68:151-166. 107. Rohekar S, Chan J, Tse SM, et al. 2014 Update of the Canadian Rheumatology Association/spondyloarthritis research consortium of Canada treatment recommendations for the management of spondyloarthritis. Part I: principles of the management of spondyloarthritis in Canada. J Rheumatol. 2015;42:654-664. 108. Field MJ, Lohr KN, eds. Guidelines for Clinical Practice: From Development to use . Washington, DC: National Academy Press; 1992. 109. Evidence-Based Medicine Working Group. Evidence-based medicine. A new approach to teaching the practice of medicine. JAMA. 1992;268:2420-2425. 110. Luce B, Cohen RS. Health technology assessment in the United States. Int J Technol Assess Health Care. 2009;25 Suppl 1:33-41. 111. Institute of Medicine. Initial National Priorities for Comparative Effectiveness Research. National Academies Press; 2009. 112. Schwartz JA, Pearson SD. Cost consideration in the clinical guidance documents of physician specialty societies in the United States. JAMA Intern Med. 2013;173:1091-1097. 113. Zochling J, van der Heijde D, Dougados M, Braun J. Current evidence for the management of ankylosing spondylitis: a systematic literature review for the ASAS/EULAR management recommendations in ankylosing spondylitis. Ann Rheum Dis. 2006;65:423-432. 114. Zochling J, van der Heijde D, Burgos-Vargas R, et al. ASAS/EULAR recommendations for the management of ankylosing spondylitis. Ann Rheum Dis. 2006;65:442-452. 115. Rohekar S, Chan J, Tse SM, et al. 2014 Update of the Canadian Rheumatology Association/Spondyloarthritis Research Consortium of Canada Treatment Recommendations for the Management of Spondyloarthritis. Part II: Specific Management Recommendations. J Rheumatol. 2015;42:665-681. 116. Taurog JD, Chhabra A, Colbert RA. Ankylosing Spondylitis and Axial Spondyloarthritis. N Engl J Med. 2016;374:2563-2574. 117. Dagfinrud H, Kvien TK, Hagen KB. The Cochrane review of physiotherapy interventions for ankylosing spondylitis. J Rheumatol. 2005;32:1899-1906. 118. Sieper J, Poddubnyy D. New evidence on the management of spondyloarthritis. Nat Rev Rheumatol. 2016;12:282-295. 119. Escalas C, Trijau S, Dougados M. Evaluation of the treatment effect of NSAIDs/TNF blockers according to different domains in ankylosing spondylitis: results of a meta-analysis. Rheumatology. 2010;49:1317-1325. 120. Keyser FD, Mielants H, Veys EM. Current use of biologicals for the treatment of spondyloarthropathies. Expert Opin Pharmacother. 2001;2:85-93. 121. Corbett M, Soares M, Jhuti G, et al. Tumour necrosis factor-alpha inhibitors for ankylosing spondylitis and non-radiographic axial spondyloarthritis: a systematic review and economic evaluation. Health Technol Assess. 2016;20:1-334, v-vi.   166 122. Maxwell LJ, Zochling J, Boonen A, et al. TNF-alpha inhibitors for ankylosing spondylitis. Cochrane Database Syst Rev. 2015;(4):CD005468. doi:CD005468. 123. Brandt J, Listing J, Sieper J, Rudwaleit M, van der Heijde D, Braun J. Development and preselection of criteria for short term improvement after anti-TNF alpha treatment in ankylosing spondylitis. Ann Rheum Dis. 2004;63:1438-1444. 124. van der Heijde D, Sieper J, Maksymowych WP, et al. 2010 Update of the international ASAS recommendations for the use of anti-TNF agents in patients with axial spondyloarthritis. Ann Rheum Dis. 2011;70:905-908. 125. Baraliakos X, van den Berg R, Braun J, van der Heijde D. Update of the literature review on treatment with biologics as a basis for the first update of the ASAS/EULAR management recommendations of ankylosing spondylitis. Rheumatology. 2012;51:1378-1387. 126. McAlister FA, van Diepen S, Padwal RS, Johnson JA, Majumdar SR. How evidence-based are the recommendations in evidence-based guidelines? PLoS Med. 2007;4:e250. 127. Krumholz HM. Guideline recommendations and results: the importance of the linkage. Ann Intern Med. 2007;147:342-343. 128. Woolf S, Schunemann HJ, Eccles MP, Grimshaw JM, Shekelle P. Developing clinical practice guidelines: types of evidence and outcomes; values and economics, synthesis, grading, and presentation and deriving recommendations. Implement Sci. 2012;7:61-5908-7-61. 129. Worrall G, Chaulk P, Freake D. The effects of clinical practice guidelines on patient outcomes in primary care: a systematic review. CMAJ. 1997;156:1705-1712. 130. Lugtenberg M, Burgers JS, Westert GP. Effects of evidence-based clinical practice guidelines on quality of care: a systematic review. Qual Saf Health Care. 2009;18:385-392. 131. Grimshaw JM, Russell IT. Effect of clinical guidelines on medical practice: a systematic review of rigorous evaluations. Lancet. 1993;342:1317-1322. 132. Hepner KA, Rowe M, Rost K, et al. The effect of adherence to practice guidelines on depression outcomes. Ann Intern Med. 2007;147:320-329. 133. Basedow M, Esterman A. Assessing appropriateness of osteoarthritis care using quality indicators: a systematic review. J Eval Clin Pract. 2015;21:782-789. 134. Mahmood S, Lesuis N, van Tuyl LH, van Riel P, Landewe R. Quality in rheumatoid arthritis care. Best Pract Res Clin Rheumatol. 2015;29:664-679. 135. Wallace JF, Weingarten SR, Chiou CF, et al. The limited incorporation of economic analyses in clinical practice guidelines. J Gen Intern Med. 2002;17:210-220. 136. Antioch KM, Drummond MF, Niessen LW, Vondeling H. International lessons in new methods for grading and integrating cost effectiveness evidence into clinical practice guidelines. Cost Eff Resour Alloc. 2017;15:1-017-0063-x. eCollection 2017. 137. Garber AM, Sox HC. The role of costs in comparative effectiveness research. Health Aff (Millwood). 2010;29:1805-1811. 138. Luce BR, Drummond M, Jonsson B, et al. EBM, HTA, and CER: clearing the confusion. Milbank Q. 2010;88:256-276. 139. Edwards RT, Charles JM, Lloyd-Williams H. Public health economics: a systematic review of guidance for the economic evaluation of public health interventions and discussion of key methodological issues. BMC Public Health. 2013;13:1001-2458-13-1001. 140. Drummond M. Cost-of-illness studies: a major headache? Pharmacoeconomics. 1992;2:1-4. 141. Larg A, Moss JR. Cost-of-illness studies: a guide to critical evaluation. Pharmacoeconomics. 2011;29:653-671.   167 142. Felli JC, Hazen GB. Sensitivity analysis and the expected value of perfect information. Med Decis Making. 1998;18:95-109. 143. Berger ML, Murray JF, Xu J, Pauly M. Alternative valuations of work loss and productivity. J Occup Environ Med. 2001;43:18-24. 144. Koopmanschap MA, Rutten FF, van Ineveld BM, van Roijen L. The friction cost method for measuring indirect costs of disease. J Health Econ. 1995;14:171-189. 145. Krol M, Brouwer W, Rutten F. Productivity costs in economic evaluations: past, present, future. Pharmacoeconomics. 2013;31:537-549. 146. Koopmanschap MA. Cost-of-illness studies. Useful for health policy? Pharmacoeconomics. 1998;14:143-148. 147. World Health Organization.  WHO Guide to Identifying the Economic Consequences of Disease and Injury. 2009. 148. Tarricone R. Cost-of-illness analysis. What room in health economics? Health Policy. 2006;77:51-63. 149. Sliwczynski A, Raciborski F, Klak A, et al. Prevalence of ankylosing spondylitis in Poland and costs generated by AS patients in the public healthcare system. Rheumatol Int. 2015;35:1361-1367. 150. Onukwugha E, McRae J, Kravetz A, Varga S, Khairnar R, Mullins CD. Cost-of-Illness Studies: An Updated Review of Current Methods. Pharmacoeconomics. 2016;34:43-58. 151. Byford S, Torgerson DJ, Raftery J. Economic note: cost of illness studies. BMJ. 2000;320:1335. 152. Goossens ME, Rutten-van Molken MP, Vlaeyen JW, van der Linden SM. The cost diary: a method to measure direct and indirect costs in cost-effectiveness research. J Clin Epidemiol. 2000;53:688-695. 153. Chan KA, Hernandez-Diaz S. Pharmacoepidemiology and rheumatic disorders. Rheum Dis Clin North Am. 2004;30:835-50, vii. 154. Sculpher MJ, Claxton K, Drummond M, McCabe C. Whither trial-based economic evaluation for health care decision making? Health Econ. 2006;15:677-687. 155. Cohen DJ, Reynolds MR. Interpreting the results of cost-effectiveness studies. J Am Coll Cardiol. 2008;52:2119-2126. 156. Anglemyer A, Horvath HT, Bero L. Healthcare outcomes assessed with observational study designs compared with those assessed in randomized trials. Cochrane Database Syst Rev. 2014;(4):MR000034. doi:MR000034. 157. Walter SD. Effects of interaction, confounding and observational error on attributable risk estimation. Am J Epidemiol. 1983;117:598-604. 158. Frick KD. Comparative Effectiveness Research Would Contribute More to Ethical Policy Making if Cost Were Considered. AMA J Ethics. 2015;17:651-655. 159. Barton P, Bryan S, Robinson S. Modeling in the economic evaluation of health care: selecting the appropriate approach. J Health Serv Res Policy. 2004;9:110-118. 160. Brennan A, Chick SE, Davies R. A taxonomy of model structures for economic evaluation of health technologies. Health Econ. 2006;15:1295-1310. 161. Sheldon TA. Problems of using modeling in the economic evaluation of health care. Health Econ. 1996;5:1-11. 162. Eze-Nliam CM, Zhang Z, Weiss SA, Weintraub WS. Cost-effectiveness Assessment of Cardiac Interventions: Determining a Socially Acceptable Cost Threshold. Interv Cardiol. 2014;6:45-55.   168 163. Briggs A, Fenn P. Confidence intervals or surfaces? Uncertainty on the cost-effectiveness plane. Health Econ. 1998;7:723-740. 164. O'Brien BJ, Drummond MF, Labelle RJ, Willan A. In search of power and significance: issues in the design and analysis of stochastic cost-effectiveness studies in health care. Med Care. 1994;32:150-163. 165. Briggs AH, Wonderling DE, Mooney CZ. Pulling cost-effectiveness analysis up by its bootstraps: a non-parametric approach to confidence interval estimation. Health Econ. 1997;6:327-340. 166. Nixon RM, Wonderling D, Grieve RD. Non-parametric methods for cost-effectiveness analysis: the central limit theorem and the bootstrap compared. Health Econ. 2010;19:316-333. 167. McCabe C, Claxton K, Culyer AJ. The NICE cost-effectiveness threshold: what it is and what that means. Pharmacoeconomics. 2008;26:733-744. 168. Neumann PJ, Cohen JT, Weinstein MC. Updating cost-effectiveness--the curious resilience of the $50,000-per-QALY threshold. N Engl J Med. 2014;371:796-797. 169. MacNeil Vroomen J, Bosmans JE, Eekhout I, et al. The Cost-Effectiveness of Two Forms of Case Management Compared to a Control Group for Persons with Dementia and Their Informal Caregivers from a Societal Perspective. PLoS One. 2016;11:. doi:10.1371/journal.pone.0160908. 170. Noyes K, Singer EA, Messing EM. Cancer Care Economics of Bladder Cancer: Cost-Enhancing Factors and Possible Cost Reduction. Curr Opin Urol. 2008;18:533-539. 171. Boonen A, van der Heijde D. Review of the costs of illness of ankylosing spondylitis and methodologic notes. Expert Rev Pharmacoecon Outcomes Res. 2005;5:163-181. 172. Palla I, Trieste L, Tani C, et al. A systematic literature review of the economic impact of ankylosing spondylitis. . Clin Exp Rheum. 2012;October 16. 173. Ara RM, Packham JC, Haywood KL. The direct healthcare costs associated with ankylosing spondylitis patients attending a UK secondary care rheumatology unit. Rheumatology. 2008;47:68-71. 174. Boonen A, van der Heijde D, Landewe R, et al. Direct costs of ankylosing spondylitis and its determinants: an analysis among three European countries. Ann Rheum Dis. 2003;62:732-740. 175. Boonen A, van der Heijde D, Landewe R, et al. Costs of ankylosing spondylitis in three European countries: the patient's perspective. Ann Rheum Dis. 2003;62:741-747. 176. Boonen A, van den Heuvel R, van Tubergen A, et al. Large differences in cost of illness and wellbeing between patients with fibromyalgia, chronic low back pain, or ankylosing spondylitis. Ann Rheum Dis. 2005;64:396-402. 177. Boonen A, Brinkhuizen T, Landewe R, van der Heijde D, Severens JL. Impact of ankylosing spondylitis on sick leave, presenteeism and unpaid productivity, and estimation of the societal cost. Ann Rheum Dis. 2010;69:1123-1128. 178. Huscher D, Merkesdal S, Thiele K, et al. Cost of illness in rheumatoid arthritis, ankylosing spondylitis, psoriatic arthritis and systemic lupus erythematosus in Germany. Ann Rheum Dis. 2006;65:1175-1183. 179. Torres TM, Ferraz MB, Ciconelli RM. Resource utilisation and cost of ankylosing spondylitis in Brazil. Clin Exp Rheumatol. 2010;28:490-497. 180. Verstappen SM, Jacobs JW, van der Heijde DM, et al. Utility and direct costs: ankylosing spondylitis compared with rheumatoid arthritis. Ann Rheum Dis. 2007;66:727-731.   169 181. Ward MM. Functional disability predicts total costs in patients with ankylosing spondylitis. Arthritis Rheum. 2002;46:223-231. 182. Younes M, Jalled A, Aydi Z, et al. Socioeconomic impact of ankylosing spondylitis in Tunisia. Joint Bone Spine. 2010;77:41-46. 183. Zhu TY, Tam LS, Lee VW, et al. Costs and quality of life of patients with ankylosing spondylitis in Hong Kong. Rheumatology. 2008;47:1422-1425. 184. Kobelt G, Sobocki P, Mulero J, Gratacos J, Pocovi A, Collantes-Estevez E. The burden of ankylosing spondylitis in Spain. Value Health. 2008;11:408-415. 185. Strombeck B, Englund M, Bremander A, et al. Cost of illness from the public payers' perspective in patients with ankylosing spondylitis in rheumatological care. J Rheumatol. 2010;37:2348-2355. 186. Rafia R, Ara R, Packham J, Haywood KL, Healey E. Healthcare costs and productivity losses directly attributable to ankylosing spondylitis. Clin Exp Rheumatol. 2012;30:246-253. 187. Azevedo VF, Rossetto CN, Lorencetti PG, Tramontin MY, Fornazari B, Araujo DV. Indirect and direct costs of treating patients with ankylosing spondylitis in the Brazilian public health system. Rev Bras Reumatol Engl Ed. 2016;56:131-137. 188. Cooksey R, Husain MJ, Brophy S, et al. The Cost of Ankylosing Spondylitis in the UK Using Linked Routine and Patient-Reported Survey Data. PLoS One. 2015;10:e0126105. 189. Lee TJ, Park BH, Kim JW, Shin K, Lee EB, Song YW. Cost-of-illness and quality of life in patients with ankylosing spondylitis at a tertiary hospital in Korea. J Korean Med Sci. 2014;29:190-197. 190. Petrikova A, Dolezal T, Klimes J, Vocelka M, Sedova L, Kolar J. The economic burden of the ankylosing spondylitis in the Czech Republic: comparison between 2005 and 2008. Rheumatol Int. 2013;33:1813-1819. 191. Sliwczynski A, Raciborski F, Klak A, et al. Prevalence of ankylosing spondylitis in Poland and costs generated by AS patients in the public healthcare system. Rheumatol Int. 2015;35:1361-1367. 192. Tu L, Rai JC, Cao S, Lin Z, Hu Z, Gu J. Costs and work limitation of patients with ankylosing spondylitis in China. Clin Exp Rheumatol. 2014;32:661-666. 193. Kvamme MK, Lie E, Kvien TK, Kristiansen IS. Two-year direct and indirect costs for patients with inflammatory rheumatic joint diseases: data from real-life follow-up of patients in the NOR-DMARD registry. Rheumatology. 2012;51:1618-1627. 194. Tsifetaki N, Migkos MP, Papagoras C, Voulgari PV, Athanasakis K, Drosos AA. Counting Costs under Severe Financial Constraints: A Cost-of-Illness Analysis of Spondyloarthropathies in a Tertiary Hospital in Greece. J Rheumatol. 2015;42:963-967. 195. Olivieri I, Cortesi PA, de Portu S, et al. Long-term costs and outcomes in psoriatic arthritis patients not responding to conventional therapy treated with tumour necrosis factor inhibitors: the extension of the Psoriatic Arthritis Cost Evaluation (PACE) study. Clin Exp Rheumatol. 2016;34:68-75. 196. Gaujoux-Viala C, Fautrel B. Cost effectiveness of therapeutic interventions in ankylosing spondylitis: a critical and systematic review. Pharmacoeconomics. 2012;30:1145-1156. 197. Bansback N, Maetzel A, Drummond M, et al. Considerations and preliminary proposals for defining a reference case for economic evaluations in ankylosing spondylitis. J Rheumatol. 2007;34:1178-1183.   170 198. Corbett M, Soares M, Jhuti G, et al. Tumour necrosis factor-alpha inhibitors for ankylosing spondylitis and non-radiographic axial spondyloarthritis: a systematic review and economic evaluation. Health Technol Assess. 2016;20:1-334. 199. Sieper J, van der Heijde D, Dougados M, et al. Efficacy and safety of adalimumab in patients with non-radiographic axial spondyloarthritis: results of a randomised placebo-controlled trial (ABILITY-1). Ann Rheum Dis. 2013;72:815-822. 200. Landewe R, Braun J, Deodhar A, et al. Efficacy of certolizumab pegol on signs and symptoms of axial spondyloarthritis including ankylosing spondylitis: 24-week results of a double-blind randomised placebo-controlled Phase 3 study. Ann Rheum Dis. 2014;73:39-47. 201. Wailoo A, Hernandez M, Philips C, Brophy S, Siebert S. Modeling Health State Utility Values in Ankylosing Spondylitis: Comparisons of Direct and Indirect Methods. Value Health. 2015;18:425-431. 202. Landewe R, Dougados M, Mielants H, van der Tempel H, van der Heijde D. Physical function in ankylosing spondylitis is independently determined by both disease activity and radiographic damage of the spine. Ann Rheum Dis. 2009;68:863-867. 203. Ramiro S, van der Heijde D, van Tubergen A, et al. Higher disease activity leads to more structural damage in the spine in ankylosing spondylitis: 12-year longitudinal data from the OASIS cohort. Ann Rheum Dis. 2014;73:1455-1461. 204. Heuft-Dorenbosch L, Spoorenberg A, van Tubergen A, et al. Assessment of enthesitis in ankylosing spondylitis. Ann Rheum Dis. 2003;62:127-132. 205. Fernandez-Espartero C, de Miguel E, Loza E, et al. Validity of the ankylosing spondylitis disease activity score (ASDAS) in patients with early spondyloarthritis from the Esperanza programme. Ann Rheum Dis. 2014;73:1350-1355. 206. Rudwaleit M, Haibel H, Baraliakos X, et al. The early disease stage in axial spondylarthritis: results from the German Spondyloarthritis Inception Cohort. Arthritis Rheum. 2009;60:717-727. 207. Atkinson MD, Brophy S, Siebert S, et al. Protocol for a population-based ankylosing spondylitis (PAS) cohort in Wales. BMC Musculoskelet Disord. 2010;11:197-2474-11-197. 208. van den Berg R, de Hooge M, Rudwaleit M, et al. ASAS modification of the Berlin algorithm for diagnosing axial spondyloarthritis: results from the SPondyloArthritis Caught Early (SPACE)-cohort and from the Assessment of SpondyloArthritis international Society (ASAS)-cohort. Ann Rheum Dis. 2013;72:1646-1653. 209. Sepriano A, Landewe R, van der Heijde D, et al. Predictive validity of the ASAS classification criteria for axial and peripheral spondyloarthritis after follow-up in the ASAS cohort: a final analysis. Ann Rheum Dis. 2016;75:1034-1042. 210. Dougados M, d'Agostino MA, Benessiano J, et al. The DESIR cohort: a 10-year follow-up of early inflammatory back pain in France: study design and baseline characteristics of the 708 recruited patients. Joint Bone Spine. 2011;78:598-603. 211. Baraliakos X, Braun J. Spondyloarthritides. Best Pract Res Clin Rheumatol. 2011;25:825-842. 212. Rudwaleit M, Haibel H, Baraliakos X, et al. The early disease stage in axial spondylarthritis: results from the German Spondyloarthritis Inception Cohort. Arthritis Rheum. 2009;60:717-727. 213. Zink A, Thiele K, Huscher D, et al. Healthcare and burden of disease in psoriatic arthritis. A comparison with rheumatoid arthritis and ankylosing spondylitis. J Rheumatol. 2006;33:86-90.   171 214. Barlow JH, Wright CC, Williams B, Keat A. Work disability among people with ankylosing spondylitis. Arthritis Rheum. 2001;45:424-429. 215. Boonen A, Chorus A, Miedema H, et al. Withdrawal from labour force due to work disability in patients with ankylosing spondylitis. Ann Rheum Dis. 2001;60:1033-1039. 216. Palla I, Trieste L, Tani C, et al. A systematic literature review of the economic impact of ankylosing spondylitis. Clin Exp Rheumatol. 2012;30:S136-41. 217. Reveille JD, Ximenes A, Ward MM. Economic considerations of the treatment of ankylosing spondylitis. Am J Med Sci. 2012;343:371-374. 218. Dougados M, Etcheto A, Molto A, et al. Clinical presentation of patients suffering from recent onset chronic inflammatory back pain suggestive of spondyloarthritis: The DESIR cohort. Joint Bone Spine. 2015;82:345-351. 219. Rudwaleit M, Metter A, Listing J, Sieper J, Braun J. Inflammatory back pain in ankylosing spondylitis: a reassessment of the clinical history for application as classification and diagnostic criteria. Arthritis Rheum. 2006;54:569-578. 220. Caisse Nationale de l'Assurance Maladie (France). L'Assurance Maladie en Ligne. Available at: http://www.ameli.fr/assures/soins-et-remboursements/combien-serez-vous-rembourse/consultations/les-consultations-en-metropole/dans-le-cadre-du-parcours-de-soins-coordonnes.php. Accessed August 26, 2014. 221. Institut de recherche et de documentation en économie de la santé. Institut de recherche et de documentation en économie de la santé. Available at: http://www.irdes.fr. Accessed August 26, 2014. 222. Institut de recherche et de documentation en économie de la santé. Eco-Santé Base de Données en Licence Ouverte. Available at: http://www.ecosante.fr/index2.php?base=FRAN&langh=FRA&langs=FRA&sessionid=. Accessed August 26 2014, 2014. 223. Vidal Group. Dictionnaire Vidal: L'information de référence sur les produits de santé. Available at: http://vidalfrance.com. 224. Sécurité Sociale L'Assurance Maladie. Medic'AM. Available at: http://www.ameli.fr/l-assurance-maladie/statistiques-et-publications/donnees-statistiques/medicament/medic-am-2008-2012.php. Accessed March 3, 2015. 225. Agence Technique de l’Information sur l’Hospitalisation. Utilisation des codes diagnostics principaux ou actes classants dans les bases   . Available at: http://stats.atih.sante.fr/mco/catalogmco.php. Accessed March 3, 2015. 226. Agence Technique de l’Information sur l’Hospitalisation. Tarifs MCO et HAD   . Available at: http://www.atih.sante.fr/?id=0001000021FF morey. Accessed March 3, 2015. 227. Inflation.eu: Worldwide inflation data. Available at: http://www.inflation.eu/inflation-rates/france/historic-inflation/cpi-inflation-france.aspx. Accessed May 6, 2015. 228. National Institute of Statistics and Economic Studies. Publications et statistiques pour la France ou les régions   . Available at: http://www.insee.fr/fr/themes/theme.asp?theme=4. Accessed March 3, 2015. 229. Pan W. Akaike's information criterion in generalized estimating equations. Biometrics. 2001;57:120-125. 230. Shen C, Chen Y. Model selection of generalized estimating equations with multiply imputed longitudinal data. Biom J. 2013;55:899-911. 231. IMORI S. Model Selection Criterion Based on the Multivariate Quasi-Likelihood for Generalized Estimating Equations. Scandinavian Journal of Statistics. 2015;42:1214-1224.   172 232. van der Weijden MA, Boonen A, van der Horst-Bruinsma IE. Problems in work participation and resource use should not be underestimated in patients with early spondyloarthritis. J Rheumatol. 2014;41:2413-2420. 233. Boonen A, Brinkhuizen T, Landewe R, van der Heijde D, Severens JL. Impact of ankylosing spondylitis on sick leave, presenteeism and unpaid productivity, and estimation of the societal cost. Ann Rheum Dis. 2010;69:1123-1128. 234. Bedaiwi M, Sari I, Thavaneswaran A, Ayearst R, Haroon N, Inman RD. Fatigue in Ankylosing Spondylitis and Nonradiographic Axial Spondyloarthritis: Analysis from a Longitudinal Observation Cohort. J Rheumatol. 2015;42:2354-2360. 235. Vander Cruyssen B, Ribbens C, Boonen A, et al. The epidemiology of ankylosing spondylitis and the commencement of anti-TNF therapy in daily rheumatology practice. Ann Rheum Dis. 2007;66:1072-1077. 236. Institut National de la statistique et des études économiques. Emploi- Population Active. Available at: http://www.insee.fr/fr/mobile/etudes/document.asp?reg_id=0&ref_id=T13F041. 237. van der Weijden MA, Boonen A, van der Horst-Bruinsma IE. Problems in work participation and resource use should not be underestimated in patients with early spondyloarthritis. J Rheumatol. 2014;41:2413-2420. 238. Ward MM, Kuzis S. Risk factors for work disability in patients with ankylosing spondylitis. J Rheumatol. 2001;28:315-321. 239. Zhu TY, Tam LS, Lee VW, et al. Costs and quality of life of patients with ankylosing spondylitis in Hong Kong. Rheumatology. 2008;47:1422-1425. 240. Zhu TY, Tam LS, Leung YY, et al. Socioeconomic burden of psoriatic arthritis in Hong Kong: direct and indirect costs and the influence of disease pattern. J Rheumatol. 2010;37:1214-1220. 241. Michie S, Johnston M. Changing clinical behaviour by making guidelines specific. BMJ. 2004;328:343-345. 242. Bosch M, Tavender E, Bragge P, Gruen R, Green S. How to define 'best practice' for use in Knowledge Translation research: a practical, stepped and interactive process. J Eval Clin Pract. 2013;19(5):763-8.  243. Harr DS, Balas EA, Mitchell J. Developing quality indicators as educational tools to measure the implementation of clinical practice guidelines. Am J Med Qual. 1996;11:179-185. 244. Bombardier C, Mian S. Quality indicators in rheumatoid arthritis care: using measurement to promote quality improvement. Ann Rheum Dis. 2013;72 Suppl 2:ii128-31. 245. Desai SP, Yazdany J. Quality measurement and improvement in rheumatology: rheumatoid arthritis as a case study. Arthritis Rheum. 2011;63:3649-3660. 246. Zochling J, Braun J. Quality indicators, guidelines and outcome measures in ankylosing spondylitis. Clin Exp Rheumatol. 2007;25:147-152. 247. Pasman HR, Brandt HE, Deliens L, Francke AL. Quality indicators for palliative care: a systematic review. J Pain Symptom Manage. 2009;38:145-156. 248. Ramiro S, van der Heijde D, van Tubergen A, et al. Higher disease activity leads to more structural damage in the spine in ankylosing spondylitis: 12-year longitudinal data from the OASIS cohort. Ann Rheum Dis. 2014;73:1455-1461. 249. Hellgren K, Smedby KE, Backlin C, et al. Ankylosing spondylitis, psoriatic arthritis, and risk of malignant lymphoma: a cohort study based on nationwide prospectively recorded data from Sweden. Arthritis Rheumatol. 2014;66:1282-1290.   173 250. Dougados M, d'Agostino MA, Benessiano J, et al. The DESIR cohort: a 10-year follow-up of early inflammatory back pain in France: study design and baseline characteristics of the 708 recruited patients. Joint Bone Spine. 2011;78:598-603. 251. Jones J, Hunter D. Consensus methods for medical and health services research. BMJ. 1995;311:376-380. 252. Berthelot JM, De Bandt M, Morel J, et al. A tool to identify recent or present rheumatoid arthritis flare from both patient and physician perspectives: the 'FLARE' instrument. Ann Rheum Dis. 2012;71:1110-1116. 253. Gossec L, Fautrel B, Flipon E, et al. Safety of biologics: elaboration and validation of a questionnaire assessing patients' self-care safety skills: the BioSecure questionnaire. An initiative of the French Rheumatology Society Therapeutic Education section. Joint Bone Spine. 2013;80:471-476. 254. Pham T, van der Heijde D, Calin A, et al. Initiation of biological agents in patients with ankylosing spondylitis: results of a Delphi study by the ASAS Group. Ann Rheum Dis. 2003;62:812-816. 255. Harvard S, Gossec L, Pham T, et al. Measurable definitions of ankylosing spondylitis management recommendations are needed for use in observational studies. Joint Bone Spine. 2016;83:101-103. 256. Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clin Chem. 1993;39:561-577. 257. Coste J, Pouchot J. A grey zone for quantitative diagnostic and screening tests. Int J Epidemiol. 2003;32:304-313. 258. Harvard S, Guh D, Bansback N, et al. Costs of early spondyloarthritis: estimates from the first 3 years of the DESIR cohort. RMD Open. 2016;2. 259. Rudwaleit M, van der Heijde D, Landewe R, et al. The development of Assessment of SpondyloArthritis international Society classification criteria for axial spondyloarthritis (part II): validation and final selection. Ann Rheum Dis. 2009;68:777-783. 260. Brazier J, Roberts J, Deverill M. The estimation of a preference-based measure of health from the SF-36. J Health Econ. 2002;21:271-292. 261. Poddubnyy D, Sieper J. Radiographic progression in ankylosing spondylitis/axial spondyloarthritis: how fast and how clinically meaningful? Curr Opin Rheumatol. 2012;24:363-369. 262. Callhoff J, Sieper J, Weiss A, Zink A, Listing J. Efficacy of TNF-alpha blockers in patients with ankylosing spondylitis and non-radiographic axial spondyloarthritis: a meta-analysis. Ann Rheum Dis. 2015;74:1241-1248. 263. Reveille JD. Epidemiology of spondyloarthritis in North America. Am J Med Sci. 2011;341:284-286. 264. van den Berg R, de Hooge M, van Gaalen F, Reijnierse M, Huizinga T, van der Heijde D. Percentage of patients with spondyloarthritis in patients referred because of chronic back pain and performance of classification criteria: experience from the Spondyloarthritis Caught Early (SPACE) cohort. Rheumatology. 2013;52:1492-1499. 265. Joven BE, Navarro-Compan V, Rosas J, et al. Diagnostic Value and Validity of Early Spondyloarthritis Features: Results from the Esperanza Cohort. Arthritis Care Res. 2017. doi: 10.1002/acr.23017.   174 266. Ciurea A, Scherer A, Exer P, et al. Tumor necrosis factor alpha inhibition in radiographic and nonradiographic axial spondyloarthritis: results from a large observational cohort. Arthritis Rheum. 2013;65:3096-3106. 267. Kiltz U, Baraliakos X, Karakostas P, et al. Do patients with non-radiographic axial spondylarthritis differ from patients with ankylosing spondylitis? Arthritis Care Res. 2012;64:1415-1422. 268. Rudwaleit M, Haibel H, Baraliakos X, et al. The early disease stage in axial spondylarthritis: results from the German Spondyloarthritis Inception Cohort. Arthritis Rheum. 2009;60:717-727. 269. Sullivan SD, Mauskopf JA, Augustovski F, et al. Budget impact analysis-principles of good practice: report of the ISPOR 2012 Budget Impact Analysis Good Practice II Task Force. Value Health. 2014;17:5-14. 270. Dougados M, van der Heijde D, Sieper J, et al. Symptomatic efficacy of etanercept and its effects on objective signs of inflammation in early nonradiographic axial spondyloarthritis: a multicenter, randomized, double-blind, placebo-controlled trial. Arthritis Rheumatol. 2014;66:2091-2102. 271. Haibel H, Rudwaleit M, Listing J, et al. Efficacy of adalimumab in the treatment of axial spondylarthritis without radiographically defined sacroiliitis: results of a twelve-week randomized, double-blind, placebo-controlled trial followed by an open-label extension up to week fifty-two. Arthritis Rheum. 2008;58:1981-1991. 272. Sieper J, van der Heijde D, Dougados M, et al. Efficacy and safety of adalimumab in patients with non-radiographic axial spondyloarthritis: results of a randomised placebo-controlled trial (ABILITY-1). Ann Rheum Dis. 2013;72:815-822. 273. Barkham N, Keen HI, Coates LC, et al. Clinical and imaging efficacy of infliximab in HLA-B27-Positive patients with magnetic resonance imaging-determined early sacroiliitis. Arthritis Rheum. 2009;60:946-954. 274. Landewe R, Braun J, Deodhar A, et al. Efficacy of certolizumab pegol on signs and symptoms of axial spondyloarthritis including ankylosing spondylitis: 24-week results of a double-blind randomised placebo-controlled Phase 3 study. Ann Rheum Dis. 2014;73:39-47. 275. Song IH, Weiss A, Hermann KG, et al. Similar response rates in patients with ankylosing spondylitis and non-radiographic axial spondyloarthritis after 1 year of treatment with etanercept: results from the ESTHER trial. Ann Rheum Dis. 2013;72:823-825. 276. Pham T, Fautrel B, Dernis E, et al. Recommendations of the French Society for Rheumatology regarding TNFalpha antagonist therapy in patients with ankylosing spondylitis or psoriatic arthritis: 2007 update. Joint Bone Spine. 2007;74:638-646. 277. Costantino F, Aegerter P, Dougados M, Breban M, D'Agostino MA. Two Phenotypes Are Identified by Cluster Analysis in Early Inflammatory Back Pain Suggestive of Spondyloarthritis: Results From the DESIR Cohort. Arthritis Rheumatol. 2016;68:1660-1668. 278. Molto A, Tezenas du Montcel S, Wendling D, Dougados M, Vanier A, Gossec L. Disease activity trajectories in early axial spondyloarthritis: results from the DESIR cohort. Ann Rheum Dis. 2016. 279. Maksymowych WP, Gladman D, Rahman P, et al. The Canadian Rheumatology Association/ Spondyloarthritis Research Consortium of Canada treatment recommendations for the management of spondyloarthritis: a national multidisciplinary stakeholder project. J Rheumatol. 2007;34:2273-2284.   175 280. Kiltz U, Sieper J, Kellner H, et al. German Society for Rheumatology S3 guidelines on axial spondyloarthritis including Bechterew's disease and early forms: 8.4 Pharmaceutical therapy, 8.5 Evaluation of therapy success of pharmaceutical measures. Z Rheumatol. 2014;73 Suppl 2:78-96. 281. Mok CC. Consensus Statements on the Indications and Monitoring of Anti-tumor Necrosis Factor (TNF) Therapy for Rheumatic Diseases in Hong Kong. Hong Kong Bulletin on Rheumatic Diseases. 2005;5. 282. Keat A, Barkham N, Bhalla A, et al. BSR guidelines for prescribing TNF-alpha blockers in adults with ankylosing spondylitis. Report of a working party of the British Society for Rheumatology. Rheumatology. 2005;44:939-947. 283. Whitehead SJ, Ali S. Health outcomes in economic evaluation: the QALY and utilities. Br Med Bull. 2010;96:5-21. 284. Dias S, Sutton AJ, Welton NJ, Ades AE. Evidence synthesis for decision making 6: embedding evidence synthesis in probabilistic cost-effectiveness analysis. Med Decis Making. 2013;33:671-678. 285. Costantino F, Talpin A, Said-Nahal R, et al. Prevalence of spondyloarthritis in reference to HLA-B27 in the French population: results of the GAZEL cohort. Ann Rheum Dis. 2015;74:689-693. 286. Arends S, Brouwer E, van der Veer E, et al. Baseline predictors of response and discontinuation of tumor necrosis factor-alpha blocking therapy in ankylosing spondylitis: a prospective longitudinal observational cohort study. Arthritis Res Ther. 2011;13:R94. 287. Lord PA, Farragher TM, Lunt M, et al. Predictors of response to anti-TNF therapy in ankylosing spondylitis: results from the British Society for Rheumatology Biologics Register. Rheumatology. 2010;49:563-570. 288. Rudwaleit M, Listing J, Brandt J, Braun J, Sieper J. Prediction of a major clinical response (BASDAI 50) to tumour necrosis factor alpha blockers in ankylosing spondylitis. Ann Rheum Dis. 2004;63:665-670. 289. Vastesaeger N, van der Heijde D, Inman RD, et al. Predicting the outcome of ankylosing spondylitis therapy. Ann Rheum Dis. 2011;70:973-981. 290. Glintborg B, Sorensen IJ, Ostergaard M, et al. Ankylosing Spondylitis versus Nonradiographic Axial Spondyloarthritis: Comparison of Tumor Necrosis Factor Inhibitor Effectiveness and Effect of HLA-B27 Status. An Observational Cohort Study from the Nationwide DANBIO Registry. J Rheumatol. 2017;44:59-69. 291. Sieper J, Landewe R, Rudwaleit M, et al. Effect of certolizumab pegol over ninety-six weeks in patients with axial spondyloarthritis: results from a phase III randomized trial. Arthritis Rheumatol. 2015;67:668-677. 292. Lukas C, Landewe R, Sieper J, et al. Development of an ASAS-endorsed disease activity score (ASDAS) in patients with ankylosing spondylitis. Ann Rheum Dis. 2009;68:18-24. 293. van der Heijde D, Joshi A, Pangan AL, et al. ASAS40 and ASDAS clinical responses in the ABILITY-1 clinical trial translate to meaningful improvements in physical function, health-related quality of life and work productivity in patients with non-radiographic axial spondyloarthritis. Rheumatology. 2016;55:80-88. 294. Walters SJ, Brazier JE. What is the relationship between the minimally important difference and health state utility values? The case of the SF-6D. Health Qual Life Outcomes. 2003;1:4. 295. Walters SJ, Brazier JE. Comparison of the minimally important difference for two health state utility measures: EQ-5D and SF-6D. Qual Life Res. 2005;14:1523-1532.   176 296. Sieper J, Lenaerts J, Wollenhaupt J, et al. Efficacy and safety of infliximab plus naproxen versus naproxen alone in patients with early, active axial spondyloarthritis: results from the double-blind, placebo-controlled INFAST study, Part 1. Ann Rheum Dis. 2014;73:101-107. 297. Drummond M. Clinical Guidelines: A NICE Way to Introduce Cost-Effectiveness Considerations? Value Health. 2016;19:525-530. 298. Antioch KM, Drummond MF, Niessen LW, Vondeling H. International lessons in new methods for grading and integrating cost effectiveness evidence into clinical practice guidelines. Cost Eff Resour Alloc. 2017;15:1-017-0063-x. eCollection 2017. 299. Sepriano A, Landewe R, van der Heijde D, et al. Predictive validity of the ASAS classification criteria for axial and peripheral spondyloarthritis after follow-up in the ASAS cohort: a final analysis. Ann Rheum Dis. 2016;75:1034-1042. 300. Tomero E, Mulero J, de Miguel E, et al. Performance of the Assessment of Spondyloarthritis International Society criteria for the classification of spondyloarthritis in early spondyloarthritis clinics participating in the ESPERANZA programme. Rheumatology. 2014;53:353-360. 301. Pham T, Landewe R, van der Linden S, et al. An international study on starting tumour necrosis factor-blocking agents in ankylosing spondylitis. Ann Rheum Dis. 2006;65:1620-1625. 302. Escalas C, Dalichampt M, Dougados M, Poiraudeau S. Evaluation of physiotherapy in a prospective cohort of early axial spondyloarthritis. Data from the DESIR cohort. Joint Bone Spine. 2016;83:185-190. 303. Masiero S, Poli P, Bonaldo L, et al. Supervised training and home-based rehabilitation in patients with stabilized ankylosing spondylitis on TNF inhibitor treatment: a controlled clinical trial with a 12-month follow-up. Clin Rehabil. 2013;28:562-572. 304. Masiero S, Bonaldo L, Pigatto M, Lo Nigro A, Ramonda R, Punzi L. Rehabilitation treatment in patients with ankylosing spondylitis stabilized with tumor necrosis factor inhibitor therapy: a randomized controlled trial. J Rheumatol. 2011;38:1335-1342. 305. Giannotti E, Trainito S, Arioli G, Rucco V, Masiero S. Effects of physical therapy for the management of patients with ankylosing spondylitis in the biological era. Clin Rheumatol. 2014;33:1217-1230. 306. Kjeken I, Bo I, Ronningen A, et al. A three-week multidisciplinary in-patient rehabilitation programme had positive long-term effects in patients with ankylosing spondylitis: randomized controlled trial. J Rehabil Med. 2013;45:260-267. 307. Ciprian L, Lo Nigro A, Rizzo M, et al. The effects of combined spa therapy and rehabilitation on patients with ankylosing spondylitis being treated with TNF inhibitors. Rheumatol Int. 2013;33:241-245. 308. Parker AB, Naylor CD. Subgroups, treatment effects, and baseline risks: some lessons from major cardiovascular trials. Am Heart J. 2000;139:952-961. 309. Greenfield S, Kravitz R, Duan N, Kaplan SH. Heterogeneity of treatment effects: implications for guidelines, payment, and quality assessment. Am J Med. 2007;120:S3-9. 310. Sox HC, Goodman SN. The methods of comparative effectiveness research. Annu Rev Public Health. 2012;33:425-445. 311. Brouwers MC, Thabane L, Moher D, Straus SE. Comparative effectiveness research paradigm: implications for systematic reviews and clinical practice guidelines. J Clin Oncol. 2012;30:4202-4207.   177 312. Ahmed S, Berzon RA, Revicki DA, et al. The use of patient-reported outcomes (PRO) within comparative effectiveness research: implications for clinical practice and health care policy. Med Care. 2012;50:1060-1070. 313. Neumann PJ, Sanders GD. Cost-Effectiveness Analysis 2.0. N Engl J Med. 2017;376:203-205. 314. Owens DK, Qaseem A, Chou R, Shekelle P, Clinical Guidelines Committee of the American College of Physicians. High-value, cost-conscious health care: concepts for clinicians to evaluate the benefits, harms, and costs of medical interventions. Ann Intern Med. 2011;154:174-180. 315. Callahan KP, Bridges JF. Using comparative effectiveness research to inform decision-making: is there a role of economic evaluation? J Comp Eff Res. 2012;1:299-301. 316. van der Heijde D, Ramiro S, Landewe R, et al. 2016 update of the ASAS-EULAR management recommendations for axial spondyloarthritis. Ann Rheum Dis. 2017. 317. Claxton K, Sculpher M, Palmer S, Culyer AJ. Causes for concern: is NICE failing to uphold its responsibilities to all NHS patients? Health Econ. 2015;24:1-7. 318. Sculpher M. Subgroups and heterogeneity in cost-effectiveness analysis. Pharmacoeconomics. 2008;26:799-806. 319. Maneiro JR, Souto A, Salgado E, Mera A, Gomez-Reino JJ. Predictors of response to TNF antagonists in patients with ankylosing spondylitis and psoriatic arthritis: systematic review and meta-analysis. RMD Open. 2015;1:e000017-2014-000017. eCollection 2015. 320. Kravitz RL, Duan N, Braslow J. Evidence-based medicine, heterogeneity of treatment effects, and the trouble with averages. Milbank Q. 2004;82:661-687. 321. Trenaman L, Boonen A, Guillemin F, et al. OMERACT Quality-adjusted Life-years (QALY) Working Group: Do Current QALY Measures Capture What Matters to Patients? J Rheumatol. 2017. 322. Campbell SM, Braspenning J, Hutchinson A, Marshall M. Research methods used in developing and applying quality indicators in primary care. Qual Saf Health Care. 2002;11:358-364. 323. Arts DL, Voncken AG, Medlock S, Abu-Hanna A, van Weert HC. Reasons for intentional guideline non-adherence: A systematic review. Int J Med Inform. 2016;89:55-62. 324. Wonderling D, Sawyer L, Fenu E, Lovibond K, Laramee P. National Clinical Guideline Centre cost-effectiveness assessment for the National Institute for Health and Clinical Excellence. Ann Intern Med. 2011;154:758-765. 325. Hill J, Bullock I, Alderson P. A summary of the methods that the National Clinical Guideline Centre uses to produce clinical guidelines for the National Institute for Health and Clinical Excellence. Ann Intern Med. 2011;154:752-757. 326. Walker S, Sculpher M, Claxton K, Palmer S. Coverage with evidence development, only in research, risk sharing, or patient access scheme? A framework for coverage decisions. Value Health. 2012;15:570-579. 327. Keat A, Bennett AN, Gaffney K, Marzo-Ortega H, Sengupta R, Everiss T. Should axial spondyloarthritis without radiographic changes be treated with anti-TNF agents? Rheumatol Int. 2017;37:327-336. 328. Pereira R, Lago P, Faria R, Torres T. Safety of Anti-TNF Therapies in Immune-Mediated Inflammatory Diseases: Focus on Infections and Malignancy. Drug Dev Res. 2015;76:419-427. 329. Vrijens B, De Geest S, Hughes DA, et al. A new taxonomy for describing and defining adherence to medications. Br J Clin Pharmacol. 2012;73:691-705.   178 Appendix   Handling Missing Clinical and Cost Data  Missing cost and clinical data were imputed using the MCMC MI procedure, the LOCF method, probabilistic imputation, or with negative values based on clinical expertise, as appropriate.  Specifically, for clinical variables in which fluctuations are normal and commonly observed clinically (i.e., Ankylosing Spondylitis Disease Activity Score (ASDAS)-CRP70, BASDAI, BASFI, and Health Assessment Questionnaire for the Spondylarthropathies (HAQ-S70)), missing data was handled with MCMC imputation, which was considered the best strategy for representing this variability. For clinical variables observed to have greater stability, either the LOCF method or probabilistic imputation was used in lieu of MCMC imputation, as these methods provide better control over the variability of values imputed; LOCF was used if all patients had at least baseline data, while probabilistic imputation was used if baseline data were missing.  Following these rules, number of comorbid conditions and marital status were imputed by LOCF. Data on the satisfaction of ASAS imaging and ASAS clinical criteria were imputed probabilistically based on the distribution of non-missing values of those variables (i.e., the likelihood in the sample overall of satisfying these criteria). Data on the presence/absence of coxitis, IBD, uveitis, psoriasis/pustulosis were imputed with negative values based on clinical expertise, i.e., the rationale that these are relatively uncommon conditions, which may appear and disappear (meaning other strategies for dealing with missing data could result in excess positive values).   For missing cost data, separate assumptions were made for patients who attended the follow-up visit (partially incomplete data) and for patients who missed a follow-up visit (fully incomplete data). For patients who attended the visit, imputation of cost data was done only in cases where patients indicated resource use took place but did not provide sufficient details to allow for costing (i.e., missing or unintelligible responses to open-ended questions). This imputation was   179 executed at the smallest resource component level and was done only for resource components for which either 1.5% or more of data were missing across all patients and visits, or for which less than 1.5% data of were missing but for which unit costs were high (e.g., surgery). For patients who attended the visit, no imputation was done for productivity loss costs (i.e., missing data was assumed to indicate no productivity loss). In cases where data were missing for professional occupation, values were imputed based on data from the next-closest visit (either LOCF or next observation carried back).  For patients who missed a visit and therefore had fully incomplete data, missing data were imputed at the following levels: anti-inflammatories; analgesics; corticoids; traditional DMARDs; biologics; medical acts; physician visits; all hospitalizations; productivity loss. The MCMC method was used to impute values for all resource components except biologics, for which the LOCF method was used, given the clinical observation that most patients remain on a stable dose of biologics following initiation.   For all cost data imputation, predictors entered into the MCMC model included health resource costs plus age, sex, marital status, ethnicity, disease duration, age at disease onset, ASDAS-CRP, BASDAI, BASFI, HAQ-S, ASAS imaging and clinical criteria, peripheral arthritis at baseline, enthesitis at baseline, IBD (Crohn's disease or colitis), uveitis, psoriasis or pustulosis, coxitis grade (right and left) and comorbidity score (one point per self-reported comorbidity on a list of 23 comorbidities covering cardiovascular disease, gastrointestinal events, endocrine disorders, viral infections, tuberculosis, and Gougerot-Sjogren syndrome). For patients who missed the visit, the MCMC procedure considered non-missing data collected at previous visits. For patients who attended the visit, the MCMC procedure took into consideration non-missing data collected at previous visits and the most current visit. Data from 'future' visits were not taken into account when imputing data at a given time point.        180 Supplementary Table A.1: Unit Costs by Resource Component (References in Supplementary Table A.1a) Resource Component Category Adjusted to 2013 Y/N Derivation of Unit Cost Raw Unit Cost (RUC) 1 RUC 2 RUC 3 2013 Adjusted Unit Cost (AUC) 1 2013 AUC 2 Total Unit Cost (2013) Source Unit Cost 1 Source Unit Cost 2 Source Unit Cost 3 Productivity Loss Farmer Yes a 103   108.52  108.52 SSP – Agreste 2010. (1) Insee, DADS 2010 (2) (as multiplier)  Productivity Loss Academic Yes a 118   124.32  124.32 Insee, DADS 2010. (2)   Productivity Loss Artisan or small business Yes a 64   67.43  67.43 Insee, DADS 2010. (2)   Productivity Loss Intermediate professions Yes a 67   70.59  70.59 Insee, DADS 2010. (2)   Productivity Loss Employees Yes a 46   48.47  48.47 Insee, DADS 2010. (2)   Productivity Loss Tradespeople Yes a 49   51.63  51.63 Insee, DADS 2010. (2)   Physician Visit Acupuncturist Yes- unit cost 2 only b 23 19.7  23.00 20.76 43.76 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Anatomic-cyto pathologist Yes- unit cost 2 only b 28 9.6  28.00 10.11 38.11 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Anesthetist Yes- unit cost 2 only b 28 64  28.00 67.43 95.43 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Cardiologist Yes- unit cost 2 only b 49 19.5  49.00 20.55 69.55 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Surgeon Yes- unit cost 2 only b 28 87  28.00 91.66 119.66 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Orthopedic Surgeon Yes- unit cost 2 only b 28 61.6  28.00 64.90 92.90 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Dental Surgeon Yes- unit cost 2 only b 28 105.8  28.00 111.47 139.47 Ameli.fr 2013. (3) Eco Sante 2010. (4)    181 Resource Component Category Adjusted to 2013 Y/N Derivation of Unit Cost Raw Unit Cost (RUC) 1 RUC 2 RUC 3 2013 Adjusted Unit Cost (AUC) 1 2013 AUC 2 Total Unit Cost (2013) Source Unit Cost 1 Source Unit Cost 2 Source Unit Cost 3 Physician Visit Endocrinologist Yes- unit cost 2 only b 28 23.3  28.00 24.55 52.55 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Gastroenterologist Yes- unit cost 2 only b 28 35  28.00 36.88 64.88 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Generalist Yes- unit cost 2 only b 23 8.2  23.00 8.64 31.64 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Gynecologist Yes- unit cost 2 only b 28 32.5  28.00 34.24 62.24 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Homeopath Yes- unit cost 2 only b 23 20  23.00 21.07 44.07 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Nurse Yes- unit cost 2 only c 11.95 1.3  12.37 1.37 13.74 Ameli.fr 2013. (5) Eco Sante 2010. (4)  Physician Visit Internist Yes- unit cost 2 only b 28 33.5  28.00 35.30 63.30 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Physiotherapist Yes- unit cost 2 only d 21.2 0.36  21.20 0.38 21.58 Ameli.fr 2013. (6) Eco Sante 2010. (4)  Physician Visit Nephrologist Yes- unit cost 2 only b 28 19.3  28.00 20.33 48.33 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Neurosurgeon Yes- unit cost 2 only b 43.7 102.3  43.70 107.78 151.48 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Neuropsychologist Yes- unit cost 2 only b 43.7 23.7  43.70 24.97 68.67 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Neurologist Yes- unit cost 2 only b 43.7 25.7  43.70 27.08 70.78 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Opthalmologist Yes- unit cost 2 only b 28 22.9  28.00 24.13 52.13 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Otolaryngologist Yes- unit cost 2 only b 28 23  28.00 24.23 52.23 Ameli.fr 2013. (3) Eco Sante 2010. (4)                 182 Resource Component Category Adjusted to 2013 Y/N Derivation of Unit Cost Raw Unit Cost (RUC) 1 RUC 2 RUC 3 2013 Adjusted Unit Cost (AUC) 1 2013 AUC 2 Total Unit Cost (2013) Source Unit Cost 1 Source Unit Cost 2 Source Unit Cost 3 Physician Visit Pediatrician Yes- unit cost 2 only b 28 20.5  28.00 21.60 49.60 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Pulmonologist Yes- unit cost 2 only b 28 19  28.00 20.02 48.02 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Psychiatrist Yes- unit cost 2 only b 43.7 27.8  43.70 29.29 72.99 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Physiatrist Yes- unit cost 2 only b 28 25  28.00 26.34 54.34 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Radiologist Yes- unit cost 2 only b 28 24.2  28.00 25.50 53.50 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Rhumatologist Yes- unit cost 2 only b 28 17.9  28.00 18.86 46.86 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Midwife Yes- unit cost 2 only b 23 1.6  23.00 1.69 24.69 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Stomatologist Yes- unit cost 2 only b 28 96.4  28.00 101.57 129.57 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Urologist Yes- unit cost 2 only b 28 43.4  28.00 45.73 73.73 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Chiropractor Yes- unit cost 2 only b 50   50.00 0.00 50.00 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Liberal' Physicians Yes- unit cost 2 only b 23 25.2  23.00 26.55 49.55 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit MEP Yes- unit cost 2 only b 23 17.5  23.00 18.44 41.44 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Physician Visit Specialist Yes- unit cost 2 only b 25 32.1  25.00 33.82 58.82 Ameli.fr 2013. (3) Eco Sante 2010. (4)  Medical Acts Blood analysis Yes e 39.11504803   40.50  40.50 Ameli.fr 2013. (5) Biolam 2011 (7)                 183 Resource Component Category Adjusted to 2013 Y/N Derivation of Unit Cost Raw Unit Cost (RUC) 1 RUC 2 RUC 3 2013 Adjusted Unit Cost (AUC) 1 2013 AUC 2 Total Unit Cost (2013) Source Unit Cost 1 Source Unit Cost 2 Source Unit Cost 3 Medical Acts Urine analysis Yes f 1.079825682   1.12  1.12 Biolam 2011  (7)   Medical Acts X-Ray No g See Suppl. Table 2      Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9)  Medical Acts Ultrasound No g See Suppl. Table 2      Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9)  Medical Acts Scanner No h See Suppl. Table 2  115.59 1.50   Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9) Ameli.fr 2013 archivage (9) Medical Acts MRI No i See Suppl. Table 2  207.18 1.50   Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9) Ameli.fr 2013 archivage (9) Medical Acts Fibroscopy No j See Suppl. Table 2      Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9)  Medical Acts Colonoscopy No j See Suppl. Table 2      Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9)  Medical Acts Mammography No j See Suppl. Table 2      Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9)  Medical Acts Respirometry No j See Suppl. Table 2      Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9)                 184 Resource Component Category Adjusted to 2013 Y/N Derivation of Unit Cost Raw Unit Cost (RUC) 1 RUC 2 RUC 3 2013 Adjusted Unit Cost (AUC) 1 2013 AUC 2 Total Unit Cost (2013) Source Unit Cost 1 Source Unit Cost 2 Source Unit Cost 3 Medical Acts Bone densitometry No j See Suppl. Table 2      Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9)  Medical Acts Scintigraphy No j See Suppl. Table 2      Class. Comm. des Actes Medicaux Version 31 2013 (8). Ameli.fr 2013. forfait technique (9)  All Hospitalizations Hospitalizations Yes k See Suppl. Table 3      ATIH 2012.(10)   All Hospitalizations Surgeries Yes k See Suppl. Table 3      ATIH 2012.(10)   All Hospitalizations Emergency room visits Yes k See Suppl. Table 3      ATIH 2012.(10)   Drugs Antiinflammatories Yes l See Suppl. Table 4      Ameli.fr 2013. (11)   Drugs Analgesics Yes l See Suppl. Table 4      Ameli.fr 2013. (11)   Drugs Biologics Yes l See Suppl. Table 4      Ameli.fr 2013. (11)   Drugs TrDMARDs Yes l See Suppl. Table 4      Ameli.fr 2013. (11)                               185 Resource Component Category Adjusted to 2013 Y/N Derivation of Unit Cost Raw Unit Cost (RUC) 1 RUC 2 RUC 3 2013 Adjusted Unit Cost (AUC) 1 2013 AUC 2 Total Unit Cost (2013) Source Unit Cost 1 Source Unit Cost 2 Source Unit Cost 3 Drugs Corticoids Yes l See Suppl. Table 4      Ameli.fr 2013. (11)   a Daily wage for category= annual income in all agriculture groups reported by Agreste divided mean work days over all categories reported by Insee b Service fee plus average value of extra billing c 30 mins of nurse care plus transport plus average value of extra billing d 80 mins of physiotherapy for rheumatology care e Blood collection fee plus standard analysis (ESR, CRP, hemology including platelets, thrombocytes alone, transaminases, creatinine, creatinine clearance) f Urine Protein g Assigned based on clinical details; service fee h Assigned based on clinical details; service fee plus technical fee plus archival fee i Standardized service fee plus technical fee plus archival fee j Standardized service fee k Assigned based on clinical details; mean cost of DRG l Total reimbursement amount/number of boxes=price per box. Price per box/total mg in box=price per mg.      186 Supplementary Table A.1a: References for Unit Costs Listed in Supplementary Table A.1  Reference Number (Suppl. Table 1) Reference 1 http://agreste.agriculture.gouv.fr/IMG/xls_donnees_primeur273.xls 2 http://www.google.ca/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CB4QFjAA&url=http%3A%2F%2Fwww.insee.fr%2Ffr%2Fffc%2Ffigure%2FNATSEF04143.xls&ei=FHbiU86yM8-AogS8_4HoDQ&usg=AFQjCNGL9gQ41LmRWQE-bvcv5QS0b65xDw&bvm=bv.72676100,d.cGU 3 http://www.ameli.fr/assures/soins-et-remboursements/combien-serez-vous-rembourse/consultations/les-consultations-en-metropole/dans-le-cadre-du-parcours-de-soins-coordonnes.php 4 http://www.ecosante.fr/index2.php?base=FRAN&langh=FRA&langs=FRA&sessionid= 5 http://www.ameli.fr/professionnels-de-sante/infirmiers/votre-convention/les-tarifs-conventionnels.php 6 http://www.ameli.fr/professionnels-de-sante/masseurs-kinesitherapeutes/votre-convention/les-tarifs-conventionnels.php 7 http://www.ameli.fr/l-assurance-maladie/statistiques-et-publications/donnees-statistiques/biologie/biolam-2010-2012.php 8 http://www.ameli.fr/accueil-de-la-ccam/index.php 9 http://www.ameli.fr/accueil-de-la-ccam/regles-de-facturation/facturer-en-ccam/dispositions-generales-et-dispositions-diverses.php 10 http://www.atih.sante.fr/?id=0001000021FF morey 11 http://www.ameli.fr/l-assurance-maladie/statistiques-et-publications/donnees-statistiques/medicament/retroced-am-2010-2013.php    187 Supplementary Table A.2: Medical Workups Fees From the Classification Commune des Actes Medicaux Version 31 2013 (8)  CCAM CODE  Service Fee  LAQK003 23.9 LAQK005 34.6 HBQK002 21.3 MFQK002 20.0 ZBQK002 21.3 LDQK001 31.9 LEQK001 31.9 LEQK002 74.5 LFQK002 51.9 NAQK015 20.0 NAQK016 33.3 NAQK017 46.6 NAQK071 23.5 NAQK049 31.3 NEQK010 20.0 LGQK001 20.0 LHQK004 101.1 YYYY163 159.6 MAQK003 22.6 MAQK001 33.8 MBQK001 20.0 MDQK001 20.0 MDQK002 20.0 MZQK003 29.3 NCQK001 20.0 NFQK001 20.0 NFQK002 25.9 NDQK001 20.0 NDQK002 23.9 NZQK005 29.3 ZCQK002 20.0 LJQK015 23.9 JKQH001 73.2 JZQH002 85.1 ZCQM005 75.6 JHQM001 37.1 JQQM010 48.4 JQQM018 81.9 JQQM016 74.0   188 CCAM CODE  Service Fee  EBQM001 73.7 EJQM004 75.6 EFQM001 73.7 JNQM001 35.7 JNQM001 35.7 BZQM003 73.7 ZCQM006 55.3 MLQM001 55.3 PCQM001 37.8 QZQM001 37.8 ZCQJ006 56.7 JHQM002 36.9 PBQM001 37.8 JAQM003 55.3 JAQM004 55.3 QEQM001 41.6 KCQM001 36.9 PBQM002 36.9 JDQJ001 55.3 JQQJ037 33.4 AEQM001 37.8 ZCQM008 56.7 EBQM002 110.6 JGQJ001 56.7 ZCQM001 75.6 HCQM001 36.9 JAQM001 73.7 ZBQH001 25.3 ZCQK005 25.3 ZCQK004 50.5 NZQK002 25.3 LHQK001 25.3 NZQK002 25.3 LHQK005 25.3 MZQK002 25.3 ACQK001 25.3 ACQH002 50.5 LAQK013 25.3 LAQK013 25.3 ZCQK003 25.3 ZCQK004 50.5 ZCQH001 50.5 ACQH002 50.5   189 CCAM CODE  Service Fee  ACQH004 50.5 HCQH002 50.5 HEQE002 96.0 HHQE005 203.5* QEQK004 66.4 GLQP002 76.8 PAQK007 40.0 ZZQL017 174.6 ZCQK002 20.0 JKHD001 9.6 EBQF004 64.1 EKQH001 25.3 HEQD003 79.8 MZQH001 79.8 CDQP010 23.3 CDQP010 23.3 CDQP010 23.3 DEQP003 13.5 QZQP001 54.1 JKHA002 16.5 QZQP001 54.1 HLHB001 43.8 KCHJ001 38.4 QZHA003 34.9 EQQP005 19.5 DEQP003 13.5 DEQP003 13.5 DEQP003 13.5 DEQP003 13.5 BENA001 87.8 AMQP009 106.4 QZJB002 23.6 JLQE002 37.0 JLQE002 37.0 JKGD003 112.6 JDQH001 57.2 JDQE003 36.0 QEHB001 20.5 QEHJ003 19.2 QZQP001 54.1 QZQP001 54.1 YYYY172 37.8 YYYY172 37.8   190 CCAM CODE  Service Fee  EFQM001 73.7 DEQP003 13.5 YYYY172 37.8 ECQM002 73.7 DEQP003 73.7 DEQP003 73.7 AAQP007 57.6 AAQP007 57.6 AHQB027 51.8 AAQP007 57.6 AAQP007 57.6 DEQP003 73.7 AHQB027 51.8 AHQB027 51.8 AHQB027 51.8 AHQB027 51.8 AAQP007 57.6 HEQE002 96.0 GEQE008 164.6 HEQE002 96.0 HGQD002 112.5 ZCQN002 69.0 ZCQK005 25.3 AFLB007 44.9 JZQD001 118.9 JZQD001 118.9 HBQK002 21.3 JKHD001 9.6 GEQE008 164.6 GEQE009 165.6 GEQE010 166.6 BGQP004 36.9 JKHD001 9.6 JKHD002 10.6 JKHD003 11.6 JKHD004 12.6 JKHD005 13.6 JKHD006 14.6 HEQE002 96.0 JKHD003 11.6 DEQP005 77.0 DEQP005 77.0 DEQP006 78.0   191 CCAM CODE  Service Fee  LHLH001 34.2 LHLH001 34.2 LHLH001 34.2 LHLH001 34.2 QZLP001 8.6 HEQD003 79.8 EQQF006 96.0 BHQP002 17.3 BHQP002 17.3 BHQP002 17.3 BHQP002 17.3 BHQP002 17.3 HBQK002 21.3 HBQK002 21.3 ZZQL016 89.5 ZZQL016 89.5 ZZQL016 89.5 HEQD002 86.4 LHHH001 9.6 LHHH001 9.6 BLQP001 53.8 JEHD001 8.6 HJQE002 21.7 KCQL003 109.7 KCQL003 109.7 HJQE001 57.6 AMQP009 106.4 JHFB001 133.4 JHFB001 133.4 HBQK002 21.3 DAQM003 165.3 DAQM003 165.3 FGRP007 26.0 DAQM003 165.3 DAQM003 165.3 CERP005 57.6 KCHB001 38.4 GLQP008 67.2 HGQD002 112.5 HGQD002 112.5 ECQH010 25.3 MEQH001 79.8 FEQL006 53.1   192 CCAM CODE  Service Fee  NFCC002 447.3 AHQB027 51.8 DEQP003 13.5 FEQL006 53.1 FEQL006 53.1 *includes cost of anesthesia      193 Supplementary Table A.3: DRGs Linked to ATIH Cost File by Hospitalization Component ATIH 2012 (10)  Hospital Component  6 digit basecase DRG 5 digit DRG for sensitivity analyses ER 01C13J 01C13 ER 02C051 02C05 ER 02C05J 02C05 ER 02C07J 02C07 ER 03C071 03C07 ER 03C091 03C09 ER 03C17J 03C17 ER 03K021 03K02 ER 04M15T 04M15 ER 05C17J 05C17 ER 05K14Z 05K14 ER 05K201 05K20 ER 06C091 06C09 ER 06C111 06C11 ER 06C121 06C12 ER 06C141 06C14 ER 06C191 06C19 ER 06C221 06C22 ER 06K02Z 06K02 ER 06K04J 06K04 ER 06M093 06M09 ER 07C091 07C09 ER 07C131 07C13 ER 08C021 08C02 ER 08C12J 08C12 ER 08C131 08C13 ER 08C211 08C21 ER 08C271 08C27 ER 08C321 08C32 ER 08C341 08C34 ER 08C351 08C35 ER 08C371 08C37 ER 08C37J 08C37 ER 08C381 08C38 ER 08C38J 08C38 ER 08C391 08C39 ER 08C421 08C42 ER 08C441 08C44 ER 08C44J 08C44   194 Hospital Component  6 digit basecase DRG 5 digit DRG for sensitivity analyses ER 08C45J 08C45 ER 08C461 08C46 ER 08C481 08C48 ER 08M141 08M14 ER 08M211 08M21 ER 09K02J 09K02 ER 10C121 10C12 ER 11C051 11C05 ER 12C08J 12C08 ER 12C111 12C11 ER 13C031 13C03 ER 13C071 13C07 ER 13C07J 13C07 ER 13C08J 13C08 ER 13C12J 13C12 ER 13C171 13C17 ER 13K02Z 13K02 ER 14C08A 14C08 ER 14Z13A 14Z13 ER 17C031 17C03 ER 23Z03Z  ER 28Z17Z 28Z17 Hospital 01C13J 01C13 Hospital 01M041 01M04 Hospital 01M091 01M09 Hospital 01M111 01M11 Hospital 01M12T 01M12 Hospital 01M17T 01M17 Hospital 01M221 01M22 Hospital 01M331 01M33 Hospital 01M34T 01M34 Hospital 01M35Z 01M35 Hospital 02C05J 02C05 Hospital 02M081 02M08 Hospital 02M082 02M08 Hospital 03M04T 03M04 Hospital 03M05T 03M05 Hospital 04K02J 04K02 Hospital 04M053 04M05 Hospital 04M111 04M11 Hospital 05M121 05M12 Hospital 05M151 05M15 Hospital 06C081 06C08   195 Hospital Component  6 digit basecase DRG 5 digit DRG for sensitivity analyses Hospital 06C141 06C14 Hospital 06K04J  Hospital 06K04J 06K04 Hospital 06M02T 06M02 Hospital 06M03T 06M03 Hospital 06M041 06M04 Hospital 06M071 06M07 Hospital 06M11T 06M11 Hospital 06M12T 06M12 Hospital 06M18T 06M18 Hospital 06M191 06M19 Hospital 07C131 07C13 Hospital 07M041 07M04 Hospital 07M101 07M10 Hospital 07M14Z 07M14 Hospital 08C14J 08C14 Hospital 08C391 08C39 Hospital 08C501 08C50 Hospital 08M101 08M10 Hospital 08M151 08M15 Hospital 08M15T 08M15 Hospital 08M281 08M28 Hospital 08M291 08M29 Hospital 08M29T 08M29 Hospital 08M301 08M30 Hospital 08M35Z 08M35 Hospital 08M36Z 08M36 Hospital 09K02J 09K02 Hospital 09M051 09M05 Hospital 09M08T 09M08 Hospital 10C121 10C12 Hospital 10M081 10M08 Hospital 10M08T 10M08 Hospital 10M13Z 10M13 Hospital 10M163 10M16 Hospital 10M182 10M18 Hospital 11M062 11M06 Hospital 11M161 11M16 Hospital 11M16T 11M16 Hospital 13C091 13C09 Hospital 13K04Z 13K04 Hospital 14C05Z 14C05 Hospital 14C08A 14C08   196 Hospital Component  6 digit basecase DRG 5 digit DRG for sensitivity analyses Hospital 14M03A 14M03 Hospital 14M03B 14M03 Hospital 14M03T 14M03 Hospital 14Z13A 14Z13 Hospital 14Z16T 14Z16 Hospital 14Z16Z 14Z16 Hospital 16M11T 16M11 Hospital 16M15Z 16M15 Hospital 18M021 18M02 Hospital 19M02T 19M02 Hospital 19M07T 19M07 Hospital 20Z041 20Z04 Hospital 20Z061 20Z06 Hospital 21M05T 21M05 Hospital 23M02T 23M02 Hospital 23M061 23M06 Hospital 23M19Z 23M19 Hospital 23M20Z 23M20 Hospital 28Z17Z 28Z17 Surgery 01C13J 01C13 Surgery 02C051 02C05 Surgery 02C05J 02C05 Surgery 02C07J 02C07 Surgery 03C071 03C07 Surgery 03C091 03C09 Surgery 03C17J 03C17 Surgery 03K021 03K02 Surgery 04M15T 04M15 Surgery 05C17J 05C17 Surgery 05K14Z 05K14 Surgery 05K201 05K20 Surgery 06C091 06C09 Surgery 06C111 06C11 Surgery 06C121 06C12 Surgery 06C141 06C14 Surgery 06C191 06C19 Surgery 06C221 06C22 Surgery 06K02Z 06K02 Surgery 06K04J 06K04 Surgery 06M093 06M09 Surgery 07C091 07C09 Surgery 07C131 07C13 Surgery 08C021 08C02   197 Hospital Component  6 digit basecase DRG 5 digit DRG for sensitivity analyses Surgery 08C12J 08C12 Surgery 08C131 08C13 Surgery 08C211 08C21 Surgery 08C271 08C27 Surgery 08C321 08C32 Surgery 08C341 08C34 Surgery 08C351 08C35 Surgery 08C371 08C37 Surgery 08C37J 08C37 Surgery 08C381 08C38 Surgery 08C38J 08C38 Surgery 08C391 08C39 Surgery 08C421 08C42 Surgery 08C441 08C44 Surgery 08C44J 08C44 Surgery 08C45J 08C45 Surgery 08C461 08C46 Surgery 08C481 08C48 Surgery 08M141 08M14 Surgery 08M211 08M21 Surgery 09K02J 09K02 Surgery 10C121 10C12 Surgery 11C051 11C05 Surgery 12C08J 12C08 Surgery 12C111 12C11 Surgery 13C031 13C03 Surgery 13C071 13C07 Surgery 13C07J 13C07 Surgery 13C08J 13C08 Surgery 13C12J 13C12 Surgery 13C171 13C17 Surgery 13K02Z 13K02 Surgery 14C08A 14C08 Surgery 14Z13A 14Z13 Surgery 17C031 17C03 Surgery 23Z03Z  Surgery 28Z17Z 28Z17        198 Supplementary Table A.4: Difference, Hospitalization Basecase and Sensitivity Estimates   Year 1 Year 2 Year 3 mean sd mean sd mean sd Basecase Minus S1 (Maximum DRG)  -1590.3 4394.2 -1691.5 4667.0 -1399.3 3723.2 Basecase Minus S2 (Mean DRG)  -609.3 1728.3 -638.8 1848.3 -511.2 1405.5 Basecase Minus S3 (Median DRG)  -516.1 1434.9 -526.4 1559.3 -403.8 1112.2 Basecase Minus S4 (DRG Minimum)  135.8 623.8 136.7 492.1 126.0 523.2 Basecase Minus S5 (Next Higher DRG)  -387.1 1121.3 -389.3 1097.4 -304.3 794.4 Basecase Minus S6 (Next Lower DRG)  116.2 489.1 127.4 434.4 111.0 441.5                

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0357139/manifest

Comment

Related Items