Protocol. This trial protocol has been provided by the authors to give readers additional information about their work.

Transcription

1 Protocol This trial protocol has been provided by the authors to give readers additional information about their work. Protocol for: O Dell JR, Mikuls TR, Taylor TH, et al. Therapies for active rheumatoid arthritis after methotrexate failure. N Engl J Med 2013;369: DOI: /NEJMoa

2 This supplement contains the following items for A randomized double-blind comparative effectiveness trial in rheumatoid arthritis patients with active disease despite methotrexate : 1. Original protocol, final protocol, summary of changes. 2. Original statistical analysis plan, final statistical analysis plan, summary of changes

3 COOPERATIVE STUDIES PROGRAM #551 RHEUMATOID ARTHRITIS: COMPARISON OF ACTIVE THERAPIES IN PATIENTS WITH ACTIVE DISEASE DESPITE METHOTREXATE THERAPY (RACAT) James R. O Dell, M.D. Study Co-Chairman Omaha VA Healthcare System (USA) Edward C. Keystone MD, FRCP(C) Study Co-Chairman University of Toronto (Canada) Mary T. Brophy, M.D., M.P.H. Study Director MAVERIC Cooperative Studies Coordinating Center, Boston (USA) Robert A. Lew, Ph.D. Study Biostatistician MAVERIC Cooperative Studies Coordinating Center, Boston (USA) Stuart Warren, J.D., Pharm.D. Research Pharmacist Cooperative Studies Clinical Research Pharmacy Coordinating Center, Albuquerque (USA) Ciaran Phibbs, Ph.D. Health Economist Cooperative Studies Health Economics Research Center, Palo Alto (USA) Aslam Anis, M.A., Ph.D. Health Economist University of British Columbia, Vancouver (Canada)

4

5 UPDATES TO THE PROTOCOL On occasion, changes will be made to the study protocol. It is vital that these changes are reflected in your copy of the protocol. The following serves as a permanent record of all changes and additions that have occurred since initial printing. A new version of this page will be updated and mailed to each center every time one or more pages in the manual are modified. Please be sure to replace the old version of this page with the new version at the same time you replace old unmodified pages with the new modified pages. Also, be sure to update all copies of the Protocol located at your center. Each time a new update is received, please verify that all previous modified pages listed on this form are actually present in this protocol. Revision Date Old New Subject pages s February 26, sent to the sites February The following details changes that have been made since 1.2. February 26, For exclusion criterion #25, subcutaneously was replaced with intraarticular. February 26, Because there will no longer be an x-ray at the end of the observational follow-up period, the following sentence has been deleted: In addition, at termination of study all patients will have x-ray studies. February 26, Checkmark for X-ray at termination was deleted. February 26, Because there will no longer be an x-ray at the end of the observational follow-up period, the following sentence has been deleted: At your last long-term study visit, an additional x-ray will be taken to tell if there has been any damage to your joints.

6 TABLE OF CONTENTS I. TABLE OF CONTENTS.. i 1. ACRONYMS USED IN THIS DOCUMENT. v 2. EXECUTIVE SUMMARY... vii II. PROTOCOL THE NEED FOR A TRIAL PROBLEM TO BE ADDRESSED PRINCIPAL RESEARCH QUESTIONS TO BE ADDRESSED WHY IS A TRIAL NEEDED NOW? RELEVANT SYSTEMATIC REVIEWS HOW WILL THE RESULTS OF THE TRIAL BE USED? RISKS TO THE SAFETY OF PARTICIPANTS IN THE TRIAL THE PROPOSED TRIAL PROPOSED TRIAL DESIGN TRIAL INTERVENTIONS APPROACH FOR ALLOCATING PARTICIPANTS TO THE TRIAL GROUPS METHODS FOR PROTECTING AGAINST OTHER SOURCES OF BIAS INCLUSION/EXCLUSION CRITERIA DURATION OF TREATMENT FREQUENCY AND DURATION OF FOLLOW-UP OUTCOME MEASURES DETERMINATION OF OUTCOME MEASURES HEALTH ECONOMICS ISSUES SAMPLE SIZE RECRUITMENT RATE PROBLEMS WITH COMPLIANCE ESTIMATES OF LOSS TO FOLLOW-UP NUMBER OF CENTERS TO BE INVOLVED DETAILS OF PLANNED ANALYSIS PROPOSED FREQUENCY OF ANALYSIS PLANNED SUBGROUP ANALYSES PILOT STUDIES TRIAL MANAGEMENT DAY-TO-DAY MANAGEMENT OF THE TRIAL ROLES OF THE APPLICANTS AND CO-APPLICANTS DESCRIPTION OF STUDY COMMITTEES REFERENCES STUDY SCHEMA i

7 III. APPENDICES A. OUTCOME MEASURES PRIMARY: DISEASE ACTIVITY SCORE (DAS28) 33 a. RATIONALE b. FORM SECONDARY: a. RADIOGRAPHIC (SHARP SCORE) RATIONALE METHODS FORM 47 b. FUNCTIONAL (HAQ) FORM 51 c. ECONOMIC AND QUALITY OF LIFE EURO-QOL (EQ-5D) FORM HEALTH UTILITIES INDEX (HUI-3) HEALTHCARE UTILIZATION FORMS a. HEALTH SERVICES FORM.. 63 b. WORK PRODUCTIVITY FORM. 65 c. OUTPATIENT VISIT FORM.. 66 d. HOSPITALIZATION FORM e. LAB AND DIAGNOSTIC TEST FORM. 69 B. SERUM/PLASMA AND DNA BANKS SERUM AND PLASMA BANK SERUM AND PLASMA BANK PROCEDURES DNA BANK: PRINCIPALS, ORGANIZATION AND OPERATION.. 83 C. RELEVANT REVIEW ARTICLES THERAPEUTIC STRATEGIES FOR RHEUMATOID ARTHRITIS NEW DRUGS FOR RHEUMATOID ARTHRITIS 105 D. RATIONALE FOR CHOICE OF TNF-α BLOCKER E. HEALTH ECONOMICS AND QUALITY OF LIFE 123 A. SUMMARY 124 B. DIRECT HEALTH CARE COSTS 125 C. INDIRECT HEALTH CARE COSTS. 130 D. COST STANDARDIZATION. 131 E. UTILITIES AND QUALITY-ADJUSTED LIFE YEARS 131 F. ANALYSIS PLAN. 132 G. ADMINISTRATIVE POLICY. 134 H. PROJECTION OF LIKELY COST EFFECTIVENESS RESULT F. STATISTICS AND DATA MANAGEMENT STATISTICAL ANALYSIS AND POWER CALCULATIONS DATA MANAGEMENT 146 G. OPERATION AND ADMINISTRATION GOOD CLINICAL PRACTICES MONITORING SITE PERFORMANCE MONITORING ADVERSE EVENTS AND SERIOUS ADVERSE EVENTS PUBLICATIONS POLICY ii

8 H. HUMAN RIGHTS/INFORMED CONSENT A. INFORMED CONSENT PROCEDURES 159 B. INFORMED CONSENT CONTENT C. SURROGATE CONSENT D. RISKS AND BENEFITS. 160 E. TEMPLATE INFORMED CONSENT FORM F. TEMPLATE HIPAA WAIVER L. SITE INFORMATION M. ADDITIONAL FORMS COLLECTED FORM 00A PRE-SCREEN FORM 2. FORM 01A ELIGIBILITY A 3. FORM 01B ELIGILBILITY B 4. FORM 01C CONSENT CHECKLIST 5. FORM 02 DEMOGRAPHICS 6. FORM 03 HISTORY OF RHEUMATOID ARTHRITIS 7. FORM 04 PAST MEDICAL HISTORY 8. FORM 05 SOCIAL HISTORY AND ALLERGIES 9. FORM 06 MEDICATIONS 10. FORM 07 PHYSICAL ASSESSMENT 11. [FORM 08 JOINT COUNT/DAS28 INCLUDED IN APPENDIX A] 12. FORM 09 LABORATORY FORM 13. FORM 10 STUDY MEDICATION DISPENSING 14. FORM 11 STUDY MEDICATION COMPLIANCE 15. FORM 12 TEMPORARY DISCONTINUATION 16. FORM 13 PERMANENT DISCONTINUATION OF PROTOCOL-DRIVEN THERAPIES 17. [FORMS INCLUDED IN APPENDIX A] 18. FORM 19 JOINT PROCEDURE FORM 19. FORM 20 ENCOUNTER FORM 20. FORM 21 PHONE VISIT 21. FORM 22 ADVERSE EVENT FORM 22. FORM ID IDENTIFIABLE INFORMATION FORM 23. FORM SAE SERIOUS ADVERSE EVENT FORM 24. FORM SAEFU SERIOUS ADVERSE EVENT FOLLOW-UP FORM iii

9 iv

10 Commonly Used Acronyms ACR CBC CHF CIHR CRADO CRF CRRC CSPCC CSPCRPCC DAS DCF DMARDs DNA DSMB EDC EQ-5D ESR EULAR FDA GCP HAQ HCQ HIV HRC LFT MTX NSAIDs QALYs RA RAIN RF SAE SLE SMART SSZ TB TNF VA VAS WBC American College of Rheumatology Complete Blood Count Congestive Heart Failure Canadian Institute of Health Research Chief Research and Development Officer Case Report Form Canadian Rheumatology Research Consortium Cooperative Studies Coordinating Center Cooperative Studies Clinical Research Pharmacy Coordinating Center Disease Activity Score Data Correction Form Disease-modifying Antirheumatic Drugs Deoxyribonucleic Acid Data & Safety Monitoring Board Electronic Data Capture EuroQol-5 Dimensions Erythrocyte Sedimentation Rate European League Against Rheumatism Food and Drug Administration Good Clinical Practices Health Assessment Questionnaire Hydroxychloroquine Human Immunodeficiency Virus Human Rights Committee Liver Function Test Methotrexate Nonsteroidal Anti-inflammatory Drugs Quality Adjusted Life Years Rheumatoid Arthritis Rheumatoid Arthritis Investigational Network Rheumatoid Factor Serious Adverse Event Systemic Lupus Erythematosus Site Monitoring, Auditing and Resource Team Sulfasalazine Tuberculosis Tumor Necrosis Factor Veterans Affairs Visual Analog Scale White Blood Cell v

11 vi

12 Rheumatoid arthritis: Comparison of active therapies in patients with active disease despite methotrexate therapy Executive Summary Rheumatoid arthritis (RA) is a chronic inflammatory disease of the joints leading to joint destruction, with significant long-term morbidity and mortality. Early treatment of RA patients with disease-modifying antirheumatic drugs (DMARDs) significantly decreases these complications. Methotrexate (MTX) is an excellent, economical first-line DMARD used to treat a majority of RA patients. While most patients respond well to MTX, many continue to have active disease. Therefore, understanding how to best treat RA patients with active disease despite MTX therapy is critically important. Although a number of therapies with significantly different economic implications have been shown to be effective when added to MTX, no trial has directly compared active therapies. This study will compare therapeutic strategies using two regimens with proven efficacy when added to MTX therapy; a) hydroxychloroquine and sulfasalazine (cost ~ $1000 per year); b) the tumor necrosis factor inhibitor, etanercept (cost ~ $12,000 per year). We propose a bi-national multi-center randomized, double-blind equivalency trial comparing (A) the strategy of initially adding hydroxychloroquine and sulfasalazine to MTX in patients with active disease despite MTX, with a switch at 24 weeks to etanercept in nonresponders to (B) a strategy of adding etanercept to MTX, with a switch to hydroxychloroquine and sulfasalazine in nonresponders at 24 weeks. If we find that the strategy of first adding hydroxychloroquine and sulfasalazine to MTX identifies a subset of responsive patients and that there is no harm to nonresponders because of early rescue with etanercept, then this less expensive option should become the standard treatment for MTX resistant patients. Six hundred RA patients with active disease despite treatment with MTX as indicated by a Disease Activity Score with 28 joints (DAS28) of 4.4 units will be randomized. A DAS improvement of 1.2 (validated as clinically significant) at 24 weeks will be used to identify early nonresponder who will switch therapy. Subjects with a DAS28 improvement of > 1.2 at 24 weeks will remain on their initial therapy. The primary endpoint is achievement of a DAS28 of 3.2 (validated as representing low disease activity) at 48 weeks. The secondary endpoint is comparison of radiographic progression of disease at 48 weeks, as measured by the change in Sharp score. Economic and functional outcomes will be assessed and a serum and DNA bank will be established to evaluate potential biomarkers predictive of treatment response/toxicity and disease progression. This 30 month trial will recruit 600 subjects over 18 months. At the end of the 48 week blinded active therapy portion of the trial, the blind will be broken and data will be collected in an open fashion until all 600 patients have completed the 48 week portion of the trial. vii

13 CSP 551 RHEUMATOID ARTHRITIS: COMPARISON OF ACTIVE THERAPIES IN PATIENTS WITH ACTIVE DISEASE DESPITE METHOTREXATE THERAPY August 2004 Planning Committee Members James R. O Dell, M.D.; Omaha, NE, USA Ed Keystone, M.D.; Toronto, ON, Canada Boulos Haraoui, M.D.; Montreal, QB, Canada Daniel O. Clegg, M.D.; Salt Lake City, UT, USA David Daikh, M.D.; San Francisco, CA, USA Salahuddin Kazi, M.D.; Dallas, TX, USA Gail S. Kerr, M.D.; Washington, DC, USA Ted Mikuls, M.D.; Omaha, NE, USA John T. Sharp, M.D.; Bainbridge Island, WA, USA Louis Fiore, M.D., M.P.H.; MAVERIC CSPCC, Boston, MA, USA Mary T. Brophy, M.D., M.P.H.; MAVERIC CSPCC, Boston, MA, USA Robert A. Lew, Ph.D.; MAVERIC CSPCC, Boston, MA, USA Theodore Colton, Sc.D.; MAVERIC CSPCC, Boston, MA, USA Daniel Deykin, M.D.; MAVERIC CSPCC, Boston, MA, USA Matthew H. Liang, M.D., M.P.H.; MAVERIC CSPCC, Boston, MA, USA James S. Kaufman, M.D.; MAVERIC CSPCC, Boston, MA, USA Stuart Warren, J.D., Pharm.D.; CSPCRPCC, Albuquerque, NM, USA Ciaran Phibbs, Ph.D.; HERC, Palo Alto, CA, USA Forest Baker, Ph.D.; HERC, Palo Alto, CA, USA Andreas Maetzel, M.D., M.Sc., Ph.D.; Toronto, ON, Canada Linda Li, Ph.D.; Ottawa, ON, Canada PRIVILEGED AND CONFIDENTIAL NOT TO BE DISSEMINATED BEYOND THE INTENDED FUNCTION AND USE 1 of 180

14 2 of 180

15 CSP 551 Protocol V1.3 Rheumatoid arthritis: Comparison of active therapies in patients with active disease despite methotrexate therapy 1. THE NEED FOR A TRIAL 1.1 Problem to be addressed The primary objective of this bi-national study is to determine whether two treatment strategies with proven efficacy but with significantly different economic implications are clinically equivalent in patients with rheumatoid arthritis who have active disease despite methotrexate therapy. This trial asks a critical, previously unaddressed, clinical question of major economic importance in rheumatoid arthritis patients who continue to have active joint disease despite methotrexate therapy. Methotrexate is an excellent, economical first-line therapy for a significant percentage of rheumatoid arthritis patients. (1-3) While patients respond well to methotrexate, many continue to have significant manifestations of disease which interfere with their functional status and results in structural damage to their joints. (1) Although a number of therapies have been shown to be effective when added to methotrexate in partial responders, no trial has directly compared active therapies. Which therapy to use in patients partially responsive to methotrexate is one of the most important questions that clinical rheumatologists face as they try to provide optimal care for their patients with rheumatoid arthritis and at the same time balance the significantly different economic implications of therapies in this class. (1) This study addresses this question by evaluating two therapeutic strategies each with proven efficacy: A. Combination methotrexate, hydroxychloroquine and sulfasalazine, which costs approximately $1000 per year; and B. Combination methotrexate and etanercept (a tumor necrosis factor inhibitor) which costs approximately $12,000 per year. Upfront costs of therapy with tumor necrosis factor (TNF) inhibitors are considerable for patients with rheumatoid arthritis and their health care providers. Whether these initial costs will be offset by future gains in quality of life and increased productivity is unknown. These therapies cost a minimum of $1000 per month. Currently, the Department of Veterans Affairs, Veterans Health Administration (VA) supplies three TNF inhibitors (etanercept, infliximab and adalimumab), and between March 1, 2003 and February 29, 2004 spent over $30 million dollars on these agents. If this trial shows that the strategy of first adding a combination of sulfasalazine and hydroxychloroquine to methotrexate is effective for a significant subset of patients who otherwise would have been given TNF inhibitors, it has the potential of saving the VA and other health care providers millions of dollars in patient care costs, while at the same time providing state-of-the-art care for patients. Additionally, the results of this trial will set the world-wide standards for how patients with active disease despite methotrexate therapy should be treated. To determine which of these two common strategies is best for the treatment of patients with rheumatoid arthritis who have active disease despite treatment with methotrexate, a study of 600 patients will be needed. Although patients like these are common (in the VA, 30% of rheumatoid 1 of 25 2/26/ of 180

16 CSP 551 Protocol V1.3 arthritis patients are currently receiving methotrexate therapy), no single clinic site has a large enough patient base to complete this study. Additionally, a number of sites will be necessary to assure that enrollment occurs within an 18-month period. We will involve approximately 15 VA and 10 Canadian Rheumatology Research Consortium (CRRC) sites with the enrollment of approximately 20 patients per site. Additionally, we have solicited the aid of the Rheumatoid Arthritis Investigational Network (RAIN). These networks are experienced in the development, and execution of investigator-initiated trials. In addition to providing the needed enrollment, the CRRC and RAIN will enroll a broader cross-section of patients that will enhance the generalizablity of the results. We believe that the VA and Canadian healthcare environments are ideal for conducting this critical study. Although a trial to address this question has long been needed, so far this study has not been done because trials in routine clinical practice are quite expensive and manufacturers have little incentive to compare active biological products (etanercept, infliximab or adalimumab) to less expensive, conventional, active therapies. Moreover, the FDA has not required these comparisons. Since the VA and the Canadian healthcare systems already provide these expensive biological therapies (at a cost of > $30 million per year in the VA alone), it would appear that they are ideally suited to address this question and, in the process, help clinicians around the world make better judgments for their patients with rheumatoid arthritis who have active disease despite methotrexate therapy. 1.2 Principal research questions to be addressed a. Primary Objective To determine whether two successful treatment strategies with significantly different economic implications are clinically equivalent as measured by the percentage of patients who reach a low disease activity state (Disease Activity Score for 28 joints (DAS28) 3.2; see Appendix A). In patients with rheumatoid arthritis who have active disease despite methotrexate therapy, the treatment strategies to be tested are: (1) Strategy A: the addition of hydroxychloroquine and sulfasalazine to methotrexate, with a switch over at 24 weeks to the addition of etanercept to methotrexate in nonresponders, compared to (2) Strategy B: initial therapy with etanercept and methotrexate and switch to the addition of hydroxychloroquine and sulfasalazine to methotrexate in non-responders at 24 weeks. b. Secondary Objective (1) To determine whether radiographic progression at 48 weeks will be equivalent in patients randomized to Strategy A compared to Strategy B. (2) To determine whether functional outcomes are equivalent in patients randomized to Strategy A compared to Strategy B. This is particularly important because the DAS28 does not incorporate functional outcomes. (3) To determine the cost-effectiveness of Strategy A compared to Strategy B. 2 of 25 2/26/ of 180

17 CSP 551 Protocol V1.3 c. Other Objectives To identify genetic factors and biomarkers that predict disease progression and success or toxicity of the different strategies, serum, plasma and DNA will be collected and banked for future use. Serum and plasma will be collected as part of the main protocol. Utilization of these samples will require an approved, funded protocol submitted by interested investigators. According to established VA DNA Bank procedures, a sub-protocol with a separate consent process is under development. This sub-protocol will be reviewed by the Scientific Advisory and Ethical Oversight Committees of the DNA Bank. Further information about the Serum/Plasma and DNA banks are provided in Appendix B. Many other analyses are of interest and are planned but are not objectives of this study. Therefore, these analyses will, for the most part, be hypothesis-generating. Analyses of this type include, but are not limited to: DAS28 responses at 24 weeks; DAS28 responses 3.2 at 48 weeks with switchers counted as failures; comparison based on the change in the DAS28 score; proportion of patients achieving an American College of Rheumatology Criteria (ACR20, ACR50, ACR70) response at 24 and 48 weeks; percentages of responders to each therapy after failure to respond at 24 weeks; European League Against Rheumatism (EULAR) response (those that improve by at least 1.2 and have a DAS28 of 5.1); correlation of DAS28 changes with radiographic outcomes; correlations between radiographic and quality of life changes; percent of subjects not progressing radiographically; percent achieving major clinical response; and proportion receiving steroid injections. Other analyses will be confirmatory in nature such as treatment comparisons of the continuous components of the DAS28 and ACR responses. 1.3 Why is a trial needed now? The treatment of patients with rheumatoid arthritis is dramatically different than it was a decade ago. (1,2) In the last ten years, a number of paradigm shifts (Table 1) have occurred that have moved therapy forward. (1,2) Two of these paradigm shifts, the use of combinations of diseasemodifying antirheumatic drugs (DMARDs) and the introduction of biological therapies for the treatment of rheumatoid arthritis, have direct relevance to this protocol. Table 1 Recent Paradigm Shifts in RA Therapy - Early Treatment with DMARDs - The Use of Combinations of DMARDs - The Use of Biological Agents - The Importance of Co-Morbidities Prior to the mid-1990s, combinations of disease-modifying drugs were rarely used to treat patients with rheumatoid arthritis. (3) Because of the studies published in the last decade, rheumatologists are currently using combination DMARD therapy to treat a substantial percentage of their patients, particularly patients partially responding to methotrexate. (4-20) These combinations have resulted in dramatically better outcomes for patients with rheumatoid arthritis. The first study to show the superiority of combination DMARD therapy over methotrexate monotherapy head-to-head was published in 1996 by O Dell, et al. (4) In this trial, triple therapy with sulfasalazine, hydroxychloroquine, and methotrexate was shown to be substantially more effective than methotrexate alone (77% excellent response rate versus 33% excellent response rate, p < 0.003). The efficacy of this triple combination has been confirmed 3 of 25 2/26/ of 180

18 CSP 551 Protocol V1.3 in studies from Finland where this combination was shown to be much more effective than sulfasalazine alone in inducing remissions in patients with early rheumatoid arthritis (7,21), in slowing progression of anatomical damage as measured radiographically and in preventing C1- C2 subluxation at five years. (22) Further, this combination was shown to be more effective than monotherapy (methotrexate, sulfasalazine or hydroxychloroquine) or treatment with double combinations (methotrexate-sulfasalazine or methotrexate-hydroxychloroquine) in patients with early rheumatoid arthritis. (8) In this study, the triple combination was more effective in producing a 50% improvement, remissions, and slowing radiographic progression of disease. Finally, in patients with established rheumatoid arthritis, the triple combination is superior to the combinations of methotrexate-sulfasalazine or methotrexate-hydroxychloroquine. (9) Biological therapies, particularly those directed against tumor necrosis factor alpha (TNF-α) have been strikingly successful in the treatment of patients with rheumatoid arthritis. (11-14, 23-25) Three agents, two monoclonal antibodies against TNF-α (infliximab and adalimumab) and one soluble receptor for TNF (etanercept), are now commercially available. Studies on the efficacy of these compounds have centered on patients who have had suboptimal responses to methotrexate. The standard design of these studies has been to study patients who have active disease despite methotrexate, randomizing them to the addition of either a TNF inhibitor or placebo. In such studies, all three agents have been shown to be significantly more effective than placebo. (12-14) ACR-20 responses for these three agents have ranged from 50% to 71%. In addition to the improvement in clinical response in this patient population, studies have shown that the patients benefit radiographically from these compounds with substantially less progression with infliximab, adalimumab, or etanercept compared to placebo. (13-17) Unfortunately, despite studying at least seven different therapies that are effective when added to methotrexate (5,9,10,12-18) (Figure 1), no trials have compared these seven active therapies to one another. Therefore, clinicians faced with the need to choose among the myriad of options have little basis for a rational approach. Importantly, data from all of the above-mentioned trials have shown that these strategies have been well-tolerated by patients with treatment-refractory disease. Therefore, this trial will compare treatments that are already known to have acceptable toxicities in this patient population. Finally, the cost of these therapies varies widely from $1,431/month for infliximab (minimum dose) to $65/month for the combination of sulfasalazine and hydroxychloroquine. (26) A study involving patients with rheumatoid arthritis who had not been treated with methotrexate has compared the cost effectiveness of TNF inhibitors versus triple therapy and found triple therapy to be much more cost effective by the parameters used in this trial. (27) No studies have directly addressed this important question in the type of patients to be studied in this trial. 4 of 25 2/26/ of 180

19 CSP 551 Protocol V1.3 Figure 1: ACR 20 responses in RA subjects with active disease despite methotrexate.* 80 Placebo Responders (%) Etanercept 20 Infliximab Anakinra Weinblatt(12) Lipsky (13) Cohen (15) Weinblatt(14) Tugwell (5) O Dell (9) Kremer (10) Adalimumab 48 CSA 36 SS Z SSZ HCQ SSZ & HCQ Leflunomide *All Patients Continue to Receive Baseline Methotrexate 1.4 Relevant systematic reviews (Appendix C) There have been two very recent (May and June of 2004) reviews of therapy of RA published in the New England Journal of Medicine. (1,28) The first review by Olsen and Stein concentrated on new therapies for rheumatoid arthritis, especially biological therapies including etanercept. (28) The second review by O Dell was a comprehensive review on therapy for RA and included up-to-date recommendations for management as well as outlining deficiencies in our current knowledge base. (1) The lack of any studies comparing active therapies in RA patients with active disease despite methotrexate therapy was one of the two most critical issues highlighted in this review. This trial was designed to answer this need directly. 1.5 How will the results of the trial be used? In the United States patients are frequently given TNF-inhibitors as the next step when they do not respond sufficiently to methotrexate. If we find that the strategy of first adding sulfasalazine and hydroxychloroquine to methotrexate (triple therapy) works as well for a subset of patients and that early rescue with a switch to TNF inhibition in non-responders results in no harm, clinicians will have a less expensive option to try first. 1.6 Risks to the safety of participants in the trial All active drugs in this trial are FDA approved and are used according to recommended guidelines. Details about the pharmacology and toxicology for each drug can be found in the Drug Information Report in the Operations Manual. 5 of 25 2/26/ of 180

20 CSP 551 Protocol V1.3 All patients will receive a proven effective therapy for their disease. All patients will have the opportunity to switch therapies if they have not shown clinically significant improvement at 24 weeks. It is expected that patients who receive etanercept will respond more rapidly than those who receive sulfasalazine and hydroxychloroquine. However, there is no data that suggests this small temporal delay in response will result in long-term, adverse clinical consequences. Other effective therapies are available for patients with active disease despite methotrexate treatment, including adding leflunomide, anakinra, abatacept, rituximab or another TNF inhibitor (infliximab or adalimumab) to methotrexate. At any time during the study, participants may withdraw from the study and return to their rheumatology clinic for standard care. 2. THE PROPOSED TRIAL 2.1 Proposed trial design This trial will be a double-blind, randomized, 48-week, two-arm trial. The trial will test the hypothesis that the two strategies are equivalent with respect to the percentage of patients in each arm achieving a DAS Patients with rheumatoid arthritis who have active disease despite 15 to 25 mg of methotrexate weekly will be offered an opportunity to participate. They will be randomized to receive either triple therapy (sulfasalazine and hydroxychloroquine added to their baseline methotrexate), or etanercept 50 mg subcutaneously weekly in addition to baseline methotrexate. Although three TNF inhibitors are approved for the treatment of RA, we have chosen etanercept for this trial (please see rationale in Appendix D). Clinical response will be measured using a DAS28 (Appendix A) which is a well validated measure of composite clinical disease activity. (29) If, at the end of 24 weeks, patients have not achieved a DAS28 improvement of 1.2 units they will, in a blinded fashion, be switched to the other therapy. Improvement of 1.2 or greater is recognized as clinically meaningful. (30) Thus, patients randomized to receive triple therapy at the onset who fail to achieve a DAS28 improvement of at least 1.2 will receive methotrexate plus etanercept for the second 24 weeks and similarly patients randomized to methotrexate plus etanercept who fail to achieve this level of improvement will receive triple therapy for the second 24 weeks. The timing of the switch (24 weeks) is important since most rheumatologists are unwilling to wait longer than this to obtain clinically meaningful improvement for their patients. Since both groups will be switched if the 24 week DAS28 goal (improvement of 1.2) is not achieved, the blind will be protected. Importantly, for the purpose of the intention to treat analysis of the primary clinical endpoint, outcomes of patients who switch will be attributed to initial treatment assignment. Furthermore, because patients may have a clinical response but still demonstrate radiologic progression, an important component of this trial is to determine whether the two strategies have similar effects on radiologic outcomes at 48 weeks. Therefore, we are testing the strategy of triple therapy first with rescue etanercept, versus etanercept first with rescue triple therapy. At the end of 48 weeks, the subject will exit the blinded portion of the trial and the code will be broken so that the clinicians will know what the subject has been receiving. If patients have achieved a DAS28 of 3.2, indicating good clinical response, clinicians will be encouraged to continue the current therapy. Descriptive data will continue to be collected on all these patients in an open fashion until all 600 patients have completed the 48 week portion of the trial. Since RA is a life-long disease this observational period will allow us to collect long term toxicity data as well as important effectiveness (clinical and radiographic), tolerability and economic data. 6 of 25 2/26/ of 180

21 CSP 551 Protocol V1.3 Importantly, the radiographic data on this population will continue to be read in a blinded manner. Figure 2: Study design Study Design Randomization DAS28 Improved 1.2? Primary Outcome: **DAS28 3.2? *SSZ + HCQ *Etanercept **No Yes Yes SSZ + HCQ Etanercept Yes Yes Open Continuation Open Continuation Start 24 Weeks 48 Weeks *All patients continue to receive methotrexate **Patients who switch at 24 weeks remain with their randomization group for analysis of primary outcome In summary, this unique design compares two treatment strategies instead of the traditional trial design where one treatment is compared to another in a static, inflexible way. Our design is dynamic and mirrors the way patients are treated in practice, allowing therapy to be changed at 24 weeks if patients have not had clinically significant improvement. This innovative design is critical for the question we are addressing. Patients may respond more rapidly to the addition of etanercept to methotrexate compared to sulfasalazine, hydroxychloroquine, and methotrexate. However, a substantial number of patients will respond to initial therapy with the three drug combination. If this trial shows that the two strategies are equivalent at 48 weeks, a significant percentage of patients will have been spared expensive therapies and health agencies can ethically recommend the strategy of using sulfasalazine, hydroxychloroquine, and methotrexate initially, which is much less expensive. The trial design also includes an economic evaluation of the alternate treatments performed concurrently as part of the clinical trial. The primary economic analysis will be a costeffectiveness analysis based on a lifetime model of cost and outcomes from both a societal and third party payer perspective that will estimate the incremental Cost per QALY (Quality Adjusted Life Year) gained when choosing one strategy over the other. Additional economic analyses, known as Budget Impact Analysis, will also be performed to assess the budgetary impact on the payer (VA or Canadian formulary authorities) of adopting these treatment strategies. Because resource costs and practice patterns are different between countries, the results will be presented in a number of different ways, including a US dollar analysis, a Canadian dollar analysis and a pooled analysis using alternate currency. Appropriate methodology for pooling international resource use data, extrapolating short term trial data to 7 of 25 2/26/ of 180

22 CSP 551 Protocol V1.3 lifetime models of cost and outcomes and undertaking second order sensitivity analysis to account for both modeling and extrapolation errors will be used. Details of the economic analyses are presented in Section 2.10 and Appendix E. 2.2 Trial interventions Eligible subjects must be on a stable dose of 15 to 25 mg of methotrexate per week (oral or subcutaneous) for at least four weeks prior to randomization, and on any dose of methotrexate for at least four months. If, in the judgment of their physicians, subjects have demonstrated intolerance to this dose of methotrexate then those on a stable minimum dose of 10 mg weekly will also be eligible for entry into the trial. All subjects will continue to receive methotrexate at their entry dose throughout the trial. Eligible, consenting subjects will be randomized to either: Strategy A Initial treatment for 24 weeks with: - sulfasalazine 1 gram daily orally for 6 weeks, then increased to 2 grams daily - hydroxychloroquine 400 mg daily orally - placebo subcutaneous injection weekly Assessment of response at 24 weeks: - if DAS28 improvement of 1.2, continue therapy as above - if DAS28 improvement is < 1.2, switch to: - etanercept 50mg subcutaneous injection weekly - placebo sulfasalazine tablets orally daily - placebo hydroxychloroquine tablets orally daily. Strategy B Initial treatment for 24 weeks with: - etanercept 50mg subcutaneous injection weekly - placebo sulfasalazine tablets orally daily - placebo hydroxychloroquine tablets orally daily. Assessment of response at 24 weeks: - if DAS28 improvement of 1.2, continue therapy as above - if DAS28 improvement is < 1.2, switch to: - sulfasalazine 1 gram daily orally for 6 weeks, then increased to 2 grams daily - hydroxychloroquine 400 mg daily orally - placebo subcutaneous injection weekly Study physicians may place subjects on a reduced dose of study hydroxychloroquine (or placebo) if necessary to keep the overall dose to below 6.5 mg/kg/day. Suggested dose schedules include taking 2 pills six of seven days (i.e. skipping Sundays) or taking 2 pills four days a week and 1 pill three days a week. Likewise, the dose of sulfasalazine (or placebo) may be reduced to one tablet bid (500 mg bid if active sulfasalazine) for patients who cannot tolerate two tablets bid (1 gram bid if active sulfasalazine). Study etanercept (or placebo) must follow the dosing schedule above. 8 of 25 2/26/ of 180

23 CSP 551 Protocol V1.3 All patients will continue to receive their baseline methotrexate at the dose and the route of administration that they started the trial on. Patients must receive folic acid, up to 3 mg/day (5 mg per week is an acceptable dose) at the discretion of their physician. Corticosteroids at 10 mg/day of prednisone (or equivalent) will be allowed and this dose will remain constant throughout the study. For disease flares, increases in oral steroid dose of 10 mg/day prednisone for 2 weeks may be given up to 2 times in each 6-month period, but not within one month of an evaluation point, or one intraarticular steroid injection (maximum dose 80 mg prednisone equivalent) will be permitted in each 6 month period, but not within 6 weeks of an evaluation point. NSAIDs will be allowed and physician will be encouraged, but not required, to keep these constant throughout the study. Analgesics should be stopped at least 12 hours prior to a joint evaluation. Use of live vaccines during the study period is also proscribed. Changes to medications or therapies for rheumatoid arthritis (i.e. the addition of a DMARD) will be considered protocol violations, except as described in this paragraph. Study staff will obtain information from the subject on all prior (taken within 90 days of the screening visit) and concomitant medications. Subjects should immediately inform study staff and/or investigator of any new concomitant medications or therapies as well as any changes in concomitant medications or therapies. 2.3 Approach for allocating participants to the trial groups To control for potential imbalance in the randomization, both stratification and blocking will be employed. Stratification by site is necessary because of possible regional differences in clinical practices. Randomization lists will be generated based upon permuted random blocks of variable size to assure approximate balance over time. Allocation will be approximately equal in both randomized treatment groups. The Boston VA Cooperative Studies Coordinating Center (CSPCC) will develop the randomization scheme and, with the assistance of a web-based randomization system, will allocate subjects to the two treatment arms. If the patient meets the inclusion/exclusion criteria, and consents to participate in the study, the baseline evaluation will be completed. The participating center will access the web-based randomization system to verify eligibility for randomization. After confirmation of eligibility, the patients will be randomized by the system and will begin treatment immediately. The Boston CSPCC will serve as the primary statistical and data management coordinating center for the trial. The Albuquerque CSPCRPCC will manage drug supply, accountability, and assignments. Eligibility criteria will be gathered, confirmed, and stored in a central electronic file. Drug assignments based on randomization will be entered in the same file. 2.4 Methods for protecting against other sources of bias Several steps were taken to reduce potential bias. First, subjects will be randomly assigned to treatments, distributing potential confounders evenly between treatment arms. Second, subjects and clinicians will be blinded to treatment assignment. Third, clinicians administering the DAS28 will be trained in how to perform the measure in a consistent way. Fourth, our secondary hypothesis comparing radiographic progression of disease uses a hard endpoint with two readers to address measurement error. Fifth, entry criteria will reduce the possibility of selection bias. Additionally, consent and data forms will also be made available in French to 9 of 25 2/26/ of 180

24 CSP 551 Protocol V1.3 allow as many subjects as possible to participate. Lastly, eligibility criteria will be applied similarly in both countries. 2.5 Inclusion/exclusion criteria A. Inclusion Criteria 1. All patients must fulfill ACR classification criteria for rheumatoid arthritis. 2. All patients must have been 16 years of age or older at time of diagnosis of rheumatoid arthritis. 3. All patients must be 18 years of age or older at the time of entry into the study. 4. All patients will have been receiving oral or subcutaneous methotrexate 15 to 25 mg/week (unless intolerant and on a minimum 10 mg/week) at a constant dose for at least one month, and on any methotrexate for no less than four months. 5. All patients will have active disease as defined by a DAS28 of If patients are receiving corticosteroids, they must have been on stable dose ( 10 mg prednisone or equivalent) for at least two weeks prior to screening. 7. If patients are using non-steroidal anti-inflammatory drugs (NSAIDs), they must be on stable doses for at least one week prior to enrollment. 8. If patients have taken leflunomide, cyclosporine, gold, Anakinra, azathioprine, or penicillamine in combination with methotrexate, they must have stopped this therapy at least 2 months prior to randomization. 9. Laboratory tests must meet the following criteria within 2 weeks of randomization: a. Serum creatinine 1.8 mg/dl b. Hemoglobin 9 g/dl c. WBC 3000 mc/l d. Neutrophils 1000 mc/l e. Platelets 100,000 mc/l f. Serum transaminase levels (ALT or AST) not exceeding 1.2 times upper limits of normal g. Albumin above lower limit of normal 10. All patients must be capable of giving informed consent and able to adhere to study visit schedule. 11. Subject or designee must have the ability to self-inject investigational product or have a caregiver who can inject subcutaneous injections 12. Negative PPD, or PPD of <5 mm with no chest x-ray findings consistent with tuberculosis, within 3 months of randomization. B. Exclusion Criteria: 1. Previous intolerance to methotrexate (unless able to tolerate at least 10 mg/week) 2. Sensitivity to study medications 3. Previous treatment with methotrexate, sulfasalazine or hydroxychloroquine in combination with each other. 4. No bed or wheelchair-bound patients 5. Active tuberculosis or evidence of latent tuberculosis (positive PPD skin test or a chest x-ray with evidence of tuberculosis) without previous adequate therapy. 6. Previous treatment with a TNF-α inhibitor (etanercept, infliximab or adalimumab) 10 of 25 2/26/ of 180

25 CSP 551 Protocol V Evidence of important acute or chronic infections (no IV antibiotics within 1 month, and no PO antibiotics within 2 weeks) 8. Pregnant or nursing women 9. Women of childbearing potential or their partners who are not practicing an acceptable form of birth control as defined by investigator 10. Active substance abuse or psychiatric illness likely to interfere with protocol completion 11. History of multiple sclerosis, transverse myelitis, or optic neuritis 12. History of macular degeneration 13. New York Heart Association Class III or IV congestive heart failure 14. Active malignancy (other than in situ cervical cancer or non-melanoma skin cancer), or history of lymphoma 15. History of HIV 16. History of any opportunistic infection to include but not limited to Pneumocystis carinii, aspergillosis, histoplasmosis, or atypical mycobacterium 17. History of porphyria 18. Diagnosis of SLE or seronegative spondyloarthropathy or any other form of concomitant arthritis (osteoarthritis is permitted) 19. Diagnosis of psoriasis unless rheumatoid factor positive 20. Any significant unstable medical condition considered a contraindication by investigator 21. Any participation in another investigational drug study during the 90 days preceding randomization. 22. Receipt of a live vaccine within 90 days of study entry. 23. History of oral or IV cyclophosphamide use 24. Life expectancy less than 2 years 25. Receipt of steroid injection, intravenous, intramuscular, or intraarticular, within 30 days of randomization. 2.6 Duration of treatment Subjects will be in the blinded treatment portion of the trial for a total of 48 weeks following randomization. After completion of this portion of the trial, subjects will be followed in the open observational portion of the trial for a maximum of 82 weeks, depending on when they entered the trial. 2.7 Frequency and duration of follow-up The study schema is presented in Exhibit 1. During the blinded therapy portion of the trial, patients will be seen in the clinic every six weeks for laboratory (CBC, creatinine, LFTs) and adverse event monitoring. Clinical evaluation with a joint count, vital signs, completion of the Health Assessment Questionnaire (HAQ) and pain scores will occur every 12 weeks. Full evaluations including the DAS28, physician global assessment, quality of life and functional surveys, and x-rays will occur every 24 weeks. In the observational portion of the trial, subjects will be followed by phone call or visit every 12 weeks for adverse event monitoring and in a study visit every 24 weeks with a DAS28, physician global assessment, quality of life and functional surveys, pain scores and adverse event monitoring. 11 of 25 2/26/ of 180

26 CSP 551 Protocol V Outcome measures a. Primary Outcome Measure The primary outcome measure is the proportion of patients achieving a DAS28 of 3.2 or less at 48 weeks. This target value is a validated measure of low disease activity. The DAS28 is an absolute measure of disease activity, which can be compared both within and across subjects. This measure stands in contrast to other criteria that assess disease activity relative to a starting point, and thus are only comparable within subjects. A detailed description of the DAS28 is in Appendix A. b. Secondary Outcome Measures (1) Radiographic progression will be assessed by obtaining standard PA x-rays of the hand and wrist and AP x-rays of the feet, which will be graded according to the Van der Heijdemodified Sharp score. (32) A description of the X-ray procedures and the method for grading Sharp score are included in Appendix A. (2) Functional outcomes will be measured using the Health Assessment Questionnaire (HAQ, Appendix A) (33), which is the standard, most widely used and best validated functional outcome assessment measure in rheumatoid arthritis patients. (3) Patient self reported quality of life will be measured at baseline and at follow-up using the Health Utilities Index (HUI) and the EQ-5D (Euro QOL). Both the HUI and the EQ-5D have been validated as valid and responsive measures in patients with RA and correlate well with other disease specific and generic measures of health related quality of life in this patient population. Patients will be monitored closely at each visit for adverse medical events (e.g. adverse experiences, hospitalization, and other complications) whether or not related to study participation. An adverse medical event is defined as any adverse change from the patient s baseline conditions, including any significant clinical or laboratory test value abnormality that occurs during the course of the study, regardless of its association with the study treatment. An assessment of treatment relatedness is required for each event reported. (See Appendix G.) 2.9 Determination of outcome measures The DAS28 is a well-validated composite outcome measure that incorporates a tender and swollen joint count of 28 joints, a measure of systemic inflammation (ESR) and a patient general assessment of health on a visual analog scale all into one measure. Study coordinators and physicians responsible for joint counts will be trained to ensure standardization. ESR determinations will be done at the local laboratory and patients will mark the visual analog scale. The results of the three components will be sent to CSPCC where the DAS28 will be calculated. Determination of modified Sharp score will be graded by two readers who will read each subject s x-rays in a blinded sequence. Change over 48 weeks in the mean Sharp score of the two readers will be used to determine radiographic progression. 12 of 25 2/26/ of 180

27 CSP 551 Protocol V1.3 For the functional and economic outcomes standard questionnaires will be completed by the patient and scored according to validated methods. Criteria for Withdrawing Subjects Subjects may be temporarily or permanently withdrawn from the study medications at any time subject to their request or the request of their treating rheumatologist. Patients will be withdrawn for adverse events or toxicities at any time (see Appendix G). An effort will be made to follow as many of the patients as possible in the blinded portion of the trial as well as in the follow up portion. All patients who exit the trial before completion of the 48 weeks will have complete evaluations including x-rays done when they exit. Other reasons for removal from investigational medication or observation might include: Administrative decision by investigator Withdrawal of consent The patient changes or discontinues their form of hormonal contraception (if using a hormonal form of contraception) Pregnancy Ineligibility Significant protocol deviation Patient non-compliance (e.g. missed visits or doses) Serious infection 2.10 Health economics issues The health economics analysis for this trial will evaluate the cost effectiveness of the following strategies for the treatment of rheumatoid arthritis (RA) in patients with active disease refractory to methotrexate therapy. Triple therapy: methotrexate, sulfasalazine, hydroxychloroquine Anti-TNF therapy: etanercept (Enbrel) plus methotrexate As is noted elsewhere in this proposal, this trial is a comparison of two treatment strategies, not a direct comparison of the two different pharmaceutical therapies. Economic analysis will be central to this evaluation because anti-tnf therapy is more than 10 times as expensive as triple therapy; about $12,000 compared to about $1,000 per year for the VA, but costs for anti-tnf therapy may be considerably higher in other healthcare environments. Given that the existing trial data indicate marginal gains in effectiveness for anti-tnf therapy, compared to triple therapy, the added value of anti-tnf therapy is uncertain. These strategies will be evaluated using standard cost-effectiveness analysis methods. The two strategies will be evaluated relative to each other to identify which therapy is dominated by another (e.g., strategies that are more costly and less effective than available alternatives). In practice, however, most strategies are not dominant or dominated. Typically, more effective treatments are more expensive. In these situations, the cost-effectiveness analysis helps answer the question: is the incremental benefit worth the added cost? Costs will include direct treatment costs and costs associated with lost productivity (indirect costs). For a complete picture of direct treatment costs, it is necessary to collect utilization data from several sources. The analysis is complicated by the fact that that this is a bi-national study 13 of 25 2/26/ of 180

28 CSP 551 Protocol V1.3 with each country having differing costs and pratice patterns. To deal with issue we have developed the appropriate data collection and analysis strategy for each country respectively. In the VA, we have developed a method that first uses administrative data, and then uses Medicare data, provider bills, and self-report data to fill in any gaps. The indirect costs will be estimated from respondent s self-report. Alternate methods, appropriate for the context, will be used to capture costs and utilization for patients in the Canadian and non-va arms of the study (details provided in Appendix E). Outcomes will be expressed in Quality-Adjusted Life Years (QALYs). In the primary analysis, QALYs will be periodically measured using the Health Utilities Index (HUI) version 3. The HUI- 3 is standardized multi-attribute utility instrument. (34-36) Because many of the European studies have used the EuroQol (EQ-5D) to measure QALYs, we will also use the EQ-5D to facilitate comparisons with these other studies. (37, 38) We will also collect self-reported general health on a 10-cm visual analog scale (VAS). A VAS is usually included as part of the EQ-5D. Data from the VAS will provide an alternative measure of patient utility and we will use this in a sensitivity analysis. The VAS has also been used in prior studies of RA, and thus including this measure will enable us to compare our results to prior studies. (The HUI and EQ-5D are provided in Appendix A.) The cost-effectiveness analysis will follow widely accepted and recommended methods. (39, 40) Costs and outcomes will be collected during the study. A societal perspective will be adopted so that all health care costs, including patient-incurred costs (e.g. caregiver costs) and the direct costs of health care, will be considered. The cost-effectiveness analysis will use the patients life span as the time horizon. Trial data will then be used as inputs into decision analytic models to project the trial results to obtain lifetime health care cost and QALYs for each treatment group. Stochastic analysis will be undertaken using cost-effectiveness acceptability curves to express uncertainty around cost-effectiveness ratios based on mean costs and outcomes. (41-43) Cost-effectiveness ratios will be calculated to compare trial participants lifetime cost and outcomes. Existing literature and VA clinical data will be used to identify starting points for the economic model, to determine necessary assumptions, and to identify parameters to be varied in the sensitivity analysis; when these data are ambiguous or incomplete, consensus from the clinical investigators will be sought. A sensitivity analysis will be done to reflect uncertainty of the input parameters and to test the robustness of the model. (40) While cost-effectiveness analysis is the standard for evaluating medical technologies, VA managers and Canadian formulary authorities are usually more concerned with their own narrower perspectives and how a change in therapy will affect their budgets. Thus, we will conduct additional budget impact analyses. First, we will estimate the lifetime model, only considering VA costs. Second, we will estimate the short-term (one and two years) effect on VA expenditures. The later is especially important with respect to VA policy, as VA managers are most concerned about the immediate budget effects. Both of these analyses do not require any additional information; they just involve restricting the model to VA costs and limiting the time horizon. For both of these additional analyses, Canadian and RAIN costs will be converted to VA costs. Additional analyses will also be conducted from the Canadian perspective, converting U.S. costs to Canadian costs. 14 of 25 2/26/ of 180

29 CSP 551 Protocol V1.3 For a further description of the Economic Analysis, please see Appendix E Sample size A one-sided equivalence comparison test will be used for the primary hypothesis test. This test will compare the proportion of subjects in each arm of the study who reach the primary endpoint of DAS28 of 3.2. This target value is a validated measure and represents a very low disease activity (Appendix A). The null hypothesis is that the upper limit on the 95% one-sided confidence interval on the difference between two proportions of patients achieving a DAS28 of 3.2 exceeds 12.5%. Because it is expected that etanercept response will be higher than the triple therapy response the proposed test is one-sided equivalence. The boundary or limit on equivalence is 12.5%. With a type 1 error of 5% and a sample size of N=300 per study arm, the one-sided comparison of etanercept with triple therapy has 92% power to reject the null hypothesis. The study has 90% power allowing for up to an 8% dropout rate. In recent placebocontrolled trials, differences of 20% or more were shown, indicating 12.5% is an appropriate limit. (12-14) See Appendix F for further detail on sample size calculations and statistical analysis. The difference of 12.5% is small compared to differences between treatments observed in recent randomized clinical trials. In the 2003 ARMADA trial the group of patients receiving 20-mg adalimumab and the groups receiving 40 mg and 80 mg adalimumab differed by 18% and 19% in terms of the proportion of ACR20 responders after 24 weeks. (12) Similar results were observed for the ACR50 endpoint. (44) In the 2004 trial comparing the costimulation blocker CTLA4Ig to placebo the groups differed by 25% in terms of the proportion of ACR20 responders after 24 weeks. (20) Assuming these proportions have a common standard deviation of 0.5, the corresponding effect sizes are double these differences. In these terms the proposed trial has a limit on effect size of 0.25 compared to trials reported in the literature with effects sizes near Note that the endpoint of the proposed trial is the proportion of patients achieving a specific DAS28 score, rather than an ACR20. Hence, our assertions assume that these distinct endpoints have very similar effect sizes. Effect size is a unit-free measurement. We designed this equivalence trial by choosing a difference between the two treatments that is small enough that clinicians believe it is not clinically meaningful. Typically such a choice depends on the effect sizes observed in relevant clinical trials. Our equivalence point of 12.5% was based on discussions with the seven rheumatologists on our Planning Committee, selected for their expertise in rheumatoid arthritis and clinical trials. The sample size for the secondary radiographic outcome was based on an on-protocol analysis. With a type I error of 5% and a sample size of N = 240 subjects per arm, the one-sided comparison of patients on Strategy A with patients on Strategy B, the test has 90% power to reject the null hypothesis. The standard deviation of the 12-month change in Sharp scores is assumed to be 5.0 units, the value for active treatments as reported by Keystone. (51) The calculation allows for a 20% overall dropout rate because in this case patients who drop out for any reason will not have radiographic evaluation at 12 months. 15 of 25 2/26/ of 180

30 CSP 551 Protocol V1.3 Power calculations for the economic and functional outcomes are presented in Appendices E and F respectively Recruitment rate Recruitment will occur over the first 18 months of the study. Recruitment of 1 to 2 patients per month at each of the participating centers is expected. VA and Canadian sites will be selected from major regional rheumatology referral centers with large rheumatology clinics. The Rheumatoid Arthritis Investigational Network (RAIN) was established in 1989 and is recognized internationally for its innovative investigator-initiated research. RAIN has a long track record of successful recruitment and trial completion and has performed several of the trials on which this study is based. (4,9,45) Additionally, RAIN was the number one enroller (among 16 centers) in the recently completed NIH-funded osteoarthritis knee trial. Please also see Appendix L Problems with compliance Compliance, while always a concern, is not expected to be a major problem in this study. This evaluation is based on the two published trials that most closely mirror this study. (9,12) Our trial will have several key advantages over previously published trials: all patients will be receiving active therapies of proven efficacy and all patients will be switched to the other treatment if they have not achieved significant clinical benefit by 24 weeks. With both the patient and the treating physician aware that there is no placebo group (both treatments have proven efficacy in this patient population) and that all patients have the opportunity to switch to the other strategy, we expect excellent compliance. Other studies in RA have shown that if patients are doing well compliance and withdrawals are minimized. (4,6,7) Compliance will be assessed via pill counts and diary review Estimates of loss to follow-up This two-armed randomized clinical trial seeks to enroll a total of 600 subjects. With an intentto-treat analysis, the final sample sizes remain about 300 per arm because all subjects who stop study medications for reasons related to treatment are regarded as treatment failures (side effects and withdrawal of consent) even though they are no longer on blinded medications at week 48. About 5% of the subjects are expected to stop study medications due to adverse treatment effects before week 48. Among patients receiving the active treatment in the 24-week ARMADA trial, only 2.3% withdrew for adverse reactions attributed to the treatment. (14) This is a rate of 4.6% per year. For this proposed trial these subjects will be regarded as treatment failures. At most another 5% per year will drop out for reasons unrelated to treatment such as a severe episode of an unrelated condition, death or moving away. These subjects will be regarded as study dropouts. Hence, the overall dropout rate for any reason should be significantly less than 8% Number of centers to be involved This trial represents collaboration between four cooperative research groups: VA Cooperative Studies, Canada s Institutes of Health Research (CIHR), the Canadian Rheumatology Research Consortium (CRRC) and the Rheumatoid Arthritis Investigators Network (RAIN). VA and CIHR have a well-established collaboration, and VA CSP has worked with a number of independent non-va US groups like RAIN. Thirty-five centers will be involved in this study: 16 of 25 2/26/ of 180

31 CSP 551 Protocol V VA centers recruiting 20 subjects each; 10 Canadian centers, recruiting 20 subjects each; and 10 RAIN centers recruiting 10 subjects each Details of planned analysis a. Primary Hypothesis The primary hypothesis will be tested using the binary outcome at 48 weeks of a DAS as evidence of a response (successful outcome) and a score > 3.2 as a non-response (failure). The treatments will be considered equivalent if the upper limit on the one-sided 95% confidence limit on the difference between the two success rates is less than 12.5%. It is expected that initial treatment with etanercept will have the higher success rate. With a binary endpoint for the primary outcome variable, the proposed test is a one-sided equivalence test of the null hypothesis that the upper limit on the 95% confidence interval on the difference between two proportions exceeds 12.5%, the boundary on equivalence. Hence, if this upper confidence limit (on the difference between the initial treatment etanercept strategy response rate and the initial treatment triple therapy strategy) is no more than 12.5% then we will reject the null hypothesis of nonequivalence and accept the alternative hypothesis of equivalence. In addition to the formal test comparing the upper confidence limit on the difference between proportions to 12.5%, we will carry out a descriptive analysis. We will report the proportions, their difference, and the odds ratio. We will compare the two treatment strategies in the primary hypothesis using a one-sided Fisher s exact test. This analysis will be followed by further, confirmatory analyses using multiple logistic regression modeling to take account of the effects of baseline covariates on the primary hypotheses. Our analytic strategy is to use PROC GENMOD in the SAS statistical package and with the following baseline covariates: age, sex, prior treatment with sulfasalazine or hydroxychloroquine, duration of treatment with methotrexate, disease duration, baseline ESR, and concurrent prednisone use. In addition, further logistic regression modeling will be undertaken to include all treatment by baseline covariate interactions to assess whether there may be relevant treatment-covariate interactions that merit consideration of subgroup analyses. Equivalence tests comparing two treatment strategies wherein one strategy is postulated to be not worse than the standard are referred to as tests of non-inferiority and therefore typically set up as one-sided tests. (50) All the proposed tests for the primary and secondary hypotheses are tests of non-inferiority. b. Secondary Hypotheses (1) The radiologic hypothesis will be tested with a one-sided equivalence test of the null hypothesis that the difference in the two mean Sharp scores for the two treatment strategies is not zero. We will reject the null hypothesis of non-equivalence if the upper limit of the one-sided 95% confidence interval on the difference is 1.4 Sharp score units or less. A consensus of our Planning Committee rheumatologists was that differences in Sharp score of 2.0 units or less are clinically unremarkable. We shall also use the independent samples t-test. Further analysis will account for baseline Sharp score as a covariate. (2) We will assess the level of disability at baseline and after 12 months using the Disability Index of the Health Assessment Questionnaire (HAQ). (See Appendix A) The primary 17 of 25 2/26/ of 180

32 CSP 551 Protocol V1.3 hypothesis tests the equivalence of the two treatment strategies. In particular, the proposed test is a one-sided equivalence test of the null hypothesis that the difference between the two means is not zero. The two means are the mean 12 month change in HAQ score for patients initially on triple therapy and the mean 12 month change in HAQ score for patients initially on Etanercept. The boundary or limit on equivalence is a difference between the means of 0.2 HAQ score units where the HAQ score ranges from 0 to 3. We will reject the null hypothesis of non-equivalence and accept the alternative hypothesis of equivalence if the upper limit of the one-sided 95% confidence interval on the difference between the means is 0.2 units or less. To test equivalence, the upper limit of the 95% confidence interval for the difference between the means will be compared to 0.2 HAQ score units. To confirm that confounding has not produced a misleading result, the hypotheses will be retested using the mixed effects model multiple regression done with PROC MIXED in the SAS statistical package. Fixed covariates will include age, sex, and baseline severity of disease. Study site and subject will be random effects. (3) For economic analyses, please see section 2.10 and Appendix E Proposed frequency of analysis Interim monitoring will focus on patient intake, overall and within hospital; hospital adherence to protocol; and baseline comparability of treatment groups. A single analysis of the effect of treatment strategies will be conducted at the end of the trial. The committee has considered and rejected the inclusion of interim analysis for several reasons. (1) There are no data on the effects of switching treatment from etanercept to triple therapy or from triple therapy to etanercept. (2) Subjects have little chance of receiving ineffective treatment beyond 24 weeks because the intention-to-treat design allows for early rescue and the switching rule allows a treatment change. (3) Data gathered after the 48-week trial ends will also strengthen the cost analysis for this equivalence trial Planned subgroup analyses In other trials of rheumatoid arthritis, no subgroup analyses have been performed, and none will be performed for this trial Pilot studies No pilot study has been carried out using this design. 3. TRIAL MANAGEMENT 3.1 Day-to-day management of the trial The VA Cooperative Studies Coordinating Center (CSPCC Boston) will manage the study in conjunction with the Study Chairman s Offices. Dr. O Dell s office will manage the trial for VA and RAIN sites, while Dr. Keystone s office will manage the trial for Canadian sites. The CSPCC will collect and hold all trial-related data in one central location and will be responsible for data quality issues for all sites. The Health Economics Resource Center (HERC) will monitor health economics data and will extract secondary data to coordinate the economic analysis for the VA and RAIN sites. The Pharmacoeconomics at St. Paul s Hospital in Vancouver will coordinate the Canadian Health Economic data collection. Appropriate protocols will be developed to coordinate and pool the health economic data for both countries 18 of 25 2/26/ of 180

33 CSP 551 Protocol V1.3 and undertake all analysis. In addition, all participating investigators and coordinators will be trained at the same meetings. All site personnel involved with the conduct of this study will be certified in Good Clinical Practices (GCP) and Human Subjects Protection training. GCP training will be conducted at the organizational (kick-off) meeting and subsequent annual meetings. The Canadian Study Chairman s office will oversee study monitoring for its sites and will handle day-to-day questions, timeliness of data submission, and clinical and logistic issues internally. The US Study Chairman s office will handle similar tasks for the non-va and VA sites. General leadership and cross-country coordination will be shared between the two countries. After the study is approved, the Study Chairman and the Boston CSPCC will prepare an Operations Manual which will be provided to the study sites as a guide to the operation and management of the study as well as a technical reference manual. A training session will occur at the study kick-off meeting for all study participants in order to assure uniformity in patient management and data collection procedures, and to train the participants in study procedures and criteria. Study procedures will be reinforced by the use of regular conference calls, particularly in the first few months of the study and by the periodic distribution of a study newsletter. All study personnel will attend group meetings during the enrollment period when study procedures again will be discussed in detail. The Study Chairman s Office will be available to clarify study procedures by telephone, fax and . If the Executive Committee (see Section 3.3) determines that a procedure must be changed, the participating sites will be informed by conference call and/or newsletter and an updated section of the Operations Manual pertinent to the changed procedure will be provided to all sites. The trial will be conducted in compliance with Good Clinical Practices (see Appendix G). As outlined in section 2.3 subjects will be randomized using the Albuquerque CSPCRPCC webbased randomization system. The Boston CSPCC will collect and manage clinical trial data for this trial using a web-based Electronic Data Capture (EDC) approach. EDC is a process by which clinical subjects and study data are managed through web based electronic data capture forms. The system is similar to a paper based approach with case report forms (CRF) and data querying through a data correction (DCF) process, but instead with web based management tools. The system allows site coordinators to enter patient data, track patients through a clinical study, respond to electronic data queries, and correct patient data all online. Thus the process of data entry, clarification, and resolution is made easier and more efficient allowing for clinical data management to occur within a shorter time frame. The system, and data management in general, are further described in Appendix G. 3.2 Roles of the applicants and co-applicants a. Role of the named applicants Dr. James O Dell has established an international reputation for his work in directing clinical trials in rheumatoid arthritis. He created and still directs a trial network (Appendix L) that has been responsible for producing many important investigator initiated clinical trials in rheumatoid arthritis. These trials have included pioneering work in triple therapy (methotrexate, sulfasalazine and hydroxychloroquine [4,9,42]), extensive work in minocycline in early rheumatoid arthritis (46-48) and studies on predicting responses to therapy with genetic markers (45). Recently, 19 of 25 2/26/ of 180

34 CSP 551 Protocol V1.3 because of his reputation in this field, he was selected to write the review article on RA therapy in the New England Journal of Medicine (1). Dr. Ed Keystone is also a well recognized and respected for his clinical trial work in rheumatoid arthritis on an international level. Dr. Keystone directs the Canadian Rheumatology Research Consortium (CRRC) and serves on most major pharmaceutical advisory boards. He has been particularly active in investigations involving cutting-edge biological therapies for the treatment of rheumatoid arthritis (10,14,24,25). Dr. Keystone has published extensively in this area and is a frequent invited speaker for rheumatoid arthritis programs around the world. b. Responsibilities of the staff employed on the grant Staff employed on the grant will vary between countries, but will include research nurses, data managers and support persons. Staff will be responsible for coordination of the study within the country (Chairman s Office) and recruitment of study participants (Research Nurse). Other staff will be responsible for the pharmacy aspects of the study (Albuquerque PCC) and data management (Boston CSPCC). c. Role of the named collaborators The roles of the Health Economists and Radiologists are described in the attached protocols (Appendices E and A). d. Role of the Study Director and Trial Statistician Mary T. Brophy, M.D., M.P.H., will be the Study Director for the Boston CSPCC Statistical and Data Management teams. Dr. Brophy is a co-director of the coordinating center and is involved in a number of CSP trials. Robert A. Lew, Ph.D., will be the trial statistician. Dr. Lew has several years of clinical trials experience in both industry and academic settings. 3.3 Description of Study Committees The groups charged with monitoring the various aspects of the study will be the Executive Committee, Data and Safety Monitoring Board (DSMB), Human Rights Committee, and the Study Group. These committees will meet at regular intervals according to the Cooperative Studies guidelines: prior to the beginning of patient intake, nine months later, and every twelve months thereafter. In addition, at the midpoint of the study, the Cooperative Studies Evaluation Committee (CSEC) will monitor and review the progress of the study. Executive Committee The Executive Committee is the management and decision-making body for the operational aspects of the study and will monitor the performance of participating medical centers and the quality of data collected. The Executive Committee will formulate publication plans and will oversee the publication and presentation of all data from the study. (See Appendix G) Data and Safety Monitoring Board The Data and Safety Monitoring Board (DSMB) will review the progress of the study and will monitor patient intake, outcomes, adverse events, and other issues related to patient safety. The DSMB makes recommendations to the Chief Research and Development Officer (CRADO) about whether the study should continue or be stopped. The DSMB will consist of experts in the fields 20 of 25 2/26/ of 180

35 CSP 551 Protocol V1.3 of rheumatology, clinical trials, biostatistics, and ethics. These experts will not be participants in the trial and will not have participated in the planning of the protocol. The DSMB will consider safety or other circumstances as grounds for early termination, including either compelling internal or external evidence of treatment differences or the unfeasibility of addressing the study hypothesis (e.g., poor patient intake, poor adherence to the protocol). At each of its meetings during the study period, the DSMB will review the randomization rates and assess the difference between the actual and the projected rates, as well as the impact of these assessments on overall trial size. If the study enrollment is inadequate, the reasons for exclusion may be scrutinized and actions may be suggested. An assessment of whether the trial should be continued will be made followed by recommendations, as appropriate. All serious adverse events will be reported on a regular basis to the DSMB for their review. Unexpected serious adverse events will be reported to the DSMB as soon as they become known based upon the consensus of the Study Chairman, the Study Director, the Biostatistician, Boston CSPCC, and the Study Pharmacist. The Study Biostatistician will provide the appropriate data to the DSMB at specified intervals for this purpose. Conditional power estimates will be provided to the DSMB to assist them in making their decisions and recommendations. Human Rights Committee The Human Rights Committee (HRC) at the Coordinating Center will review the study annually to ensure proper protection of patients rights and safety. The Study Biostatisticians will present the HRC with a report about the progress of the study and ethical issues relevant to the HRC. In the interim, the HRC may be asked to convene if there is any serious adverse event requiring its attention. An HRC member will regularly attend each meeting of the DSMB and an HRC member will site-visit one study site annually to determine that the patients rights and safety are being properly protected. Study Group The Study Group, which consists of all participating investigators and study coordinators, will meet annually to discuss the progress of the study and any problems encountered during the conduct of the trial. 21 of 25 2/26/ of 180

36 CSP 551 Protocol V1.3 REFERENCES: 1. O Dell, JR. Therapeutic Strategies for Rheumatoid Arthritis. N Engl J Med 2004;350: O Dell, JR. Treating rheumatoid arthritis early: a window of opportunity: Arthritis Rheum 2002;46: Mikuls T and O'Dell J. The treatment of rheumatoid arthritis: current trends in therapy. Arthritis Rheum 1999;42:S O Dell JR, Haire CE, Erikson N, et al. Treatment of rheumatoid arthritis with methotrexate alone, sulfasalazine, and hydroxychloroquine, or a combination of all three medications. N Engl J Med 1996;334: Tugwell P, Pincus T. Yocum D, et al. Combination therapy with cyclosporine and methotrexate in severe rheumatoid arthritis. N Engl J Med 1995;333: Boers M, Verhoeven AC, Markusse HM, et al. Randomized comparison of combined step-down prednisolone, methotrexate and sulphasalazine with sulphasalazine alone in early rheumatoid arthritis. Lancet 1997;350: Möttönen T, Hannonen P, Leirisalo-Repo M, et al. Comparison of combination therapy with single-drug therapy in early rheumatoid arthritis: a randomised trial. Lancet 1999; 353: Calgüneri M. Pay S, Caliskaner Z, et al. Combination therapy versus monotherapy for the treatment of patients with rheumatoid arthritis. Clin Exp Rheumatol 1999;17: O Dell JR, Leff R, Paulsen G, et al. Treatment of rheumatoid arthritis with methotrexate and hydroxychloroquine, methotrexate and sulfasalazine, or a combination of three medications. Arthritis Rheum 2002;46: Kremer JM. Concomitant leflunomide therapy in patients with active rheumatoid arthritis despite stable doses of methotrexate. A randomized, double-blind, placebocontrolled trial. Ann Intern Med 2002;137: Maini RN, Breedveld FC, Kalden JR, et al. Therapeutic efficacy of multiple intravenous infusions of anti-tumor necrosis factor alpha monoclonal antibody combined with lowdose weekly methotrexate in rheumatoid arthritis. Arthritis Rheum 1998;41: Weinblatt ME, Kremer JM, Bankhurst AD, et al. A trial of etanercept, a recombinant tumor necrosis factor receptor:fc fusion protein, in patients with rheumatoid arthritis receiving methotrexate. N Engl J Med 1999;340: Lipsky PE, Van Der Heijde DMFM, St. Clair EW, et al. Infliximab and methotrexate in the treatment of rheumatoid arthritis. N Engl J Med 2000;343: Weinblatt ME, Keystone EC, Furst DE, et al. Adalimumab, a fully human anti-tumor necrosis factor monoclonal antibody, for the treatment of rheumatoid arthritis in patients taking concomitant methotrexate. Arthritis Rheum 2003;48(1): Klareskog L, van der Heijde D, de Jager JP, et al. Therapeutic effect of the combination of etanercept and methotrexate compared with each treatment alone in patients with rheumatoid arthritis: double-blind randomised controlled trial. Lancet 2004; 363(9410): Breedveld FC, Weisman MH, Kavanaugh AF, Cohen SB, Pavelka K, van Vollenhoven R, Sharp J, Perez JL, Spencer-Green GT for the PREMIER Investigators. A multicenter, randomized, double-blind clinical trial of combination therapy with adalimumab plus 22 of of 180

37 CSP 551 Protocol V1.3 methotrexate versus methotrexate alone or adalimumab alone in patients with early, aggressive rheumatoid arthritis who had not had previous methotrexate treatment. Arthritis Rheum 2006; 54: St. Clair EW, van der Heijde DM, Smolen JS, Maini RN, Bathon JM, Emery P, et al. Combination of infliximab and methotrexate therapy for early rheumatoid arthritis: double-blind randomized controlled trial. Arthritis Rheum 2004;50: Cohen S, Hurd E, Cush J, et al. Treatment of rheumatoid arthritis with anakinra, a recombinant human interleukin-1 receptor antagonist, in combination with methotrexate. Arthritis Rheum 2002;46: Edwards JCW, Szczepanski L, Szechinski J, et al. Efficacy and safety of rituximab, a B- cell targeted chimeric monoclonal antibody: a randomized, placebo-controlled trial in patients with rheumatoid arthritis. Arthritis Rheum 2002;46:S Kremer J, Westhovens R, Leon M, et al. A phase 2B multi-center, randomized, doubleblind, placebo controlled study to evaluate the safety and efficacy of two different doses of CTLA4Ig administered intravenously to subjects with active rheumatoid arthritis while receiving methotrexate. Arthritis Rheum 2002;46:S Mottonen T, Hannonen P, Korpela M, et al. Delay to institution of therapy and induction of remission using single-drug or combination-disease-modifying antirheumatic drug therapy in early rheumatoid arthritis. Arthritis Rheum 2002;46(4): Neva MH. FIN-RACo Trail Group. Combination drug therapy retards the development of rheumatoid atlantoaxial subluxations. Arthritis Rheum 2000; 43(11): Elliott MJ, Maini RN, Feldmann M, et al. Randomized, double-blind comparison of chimeric monoclonal antibody to tumor necrosis factor alpha (ca2) versus placebo in rheumatoid arthritis. Lancet 1994; 344: Moreland LW, Baumgartner SW, Schiff MH, et al. Treatment of rheumatoid arthritis with a recombinant human tumor necrosis factor receptor (p75)-fc fusion protein. N Engl J Med 1997;337: Bathon JM, Martin RW, Fleischmann RM, et al. A comparison of etanercept and methotrexate in patients with early rheumatoid arthritis. N Engl J Med 2000;343: Drugs for Rheumatoid Arthritis Medical Letter 2003; 1: Choi HK, Seeger JD, Kuntz KM. A cost effectiveness analysis of treatment options for methotrexate-naive rheumatoid arthritis. J Rheumatol 2002;29: Olsen NJ, Stein CM. New drugs for rheumatoid arthritis. N Engl J Med 2004;350: Prevoo MLL, van t Hof MA, Kuper HH, et al. Modified disease activity scores that include twenty-eight-joint counts: development and validation in a prospective longitudinal study of patients with rheumatoid arthritis. Arthritis Rheum 1995;38: Van Gestel AM, Prevoo MLL, van t Hof MA, et al. Development and validation of the European League Against Rheumatism response criteria for rheumatoid arthritis. Arthritis Rheum 1996;39: Van Gestel AM, Haagsma CJ, van Riel PLCM. Validation of rheumatoid arthritis improvement criteria that include simplified joint counts. Arthritis Rheum 1998;41: Van der Heijde D. How to read radiographs according to the Sharp/van der Heijde method. J Rheumatol 1999;25: of of 180

38 CSP 551 Protocol V Fries J, Spitz P, Kraines R, Holman H. Measurements of patient outcome in arthritis. Arthritis Rheum 1980;23: Torrance GW, Furlong W, Feeny D, Boyle M. Multi-attribute preference functions. Health Utilities Index. Pharmacoeconomics. 1995;7(6): Feeny D, Furlong W, Boyle M, Torrance GW. Multi-attribute health status classification systems. Health Utilities Index. Pharmacoeconomics. 1995;7(6): Bosch JL, Hunink MG. Comparison of the Health Utilities Index Mark 3 (HUI3) and the EuroQol EQ-5D in patients treated for intermittent claudication. Qual Life Res. 2000;9(6): EuroQolGroup. EuroQol--a new facility for the measurement of health-related quality of life. Health Policy. 1990;16: Kind P. The EuroQol instrument: an index of health-related quality of life. In: Spilker B, ed. Quality of Life and Pharmacoeconomics in Clinical Trials; 1996: Drummond MF, O'Brien B, Stoddart GL, Torrance GW. Methods for the economic evaluation of health care programmes. Second Edition ed. Oxford: Oxford University Press; Gold MR, Siegel JE, Russell LB, Weinstein MC, eds. Cost-Effectiveness in Health and Medicine. Oxford: Oxford University Press; Briggs AH. A Bayesian approach to stochastic cost-effectiveness analysis. Health Econ. 1999;8(3): Stinnett AA, Mullahy J. Net health benefits: a new framework for the analysis of uncertainty in cost-effectiveness analysis. Med Decis Making. 1998;18(2 Suppl):S van Hout BA, Al MJ, Gordon GS, Rutten FF. Costs, effects and C/E-ratios alongside a clinical trial. Health Econ. 1994;3(5): American College of Rheumatology Subcommittee on Rheumatoid Arthritis Guidelines Update: Guidelines for the management of rheumatoid arthritis. Arthritis Rheum 2002;46: O Dell JR, Nepom BS, Haire C, et al. HLA-DRB1 typing in rheumatoid arthritis: predicting response to specific treatments. Ann Rheum Dis 1998;57: O'Dell JR, Haire C, Erikson N, Drymalski W, Palmer W, Maloley P, Klassen LW, Wees S, Moore GF. Efficacy of triple DMARD therapy in patients with RA with suboptimal response to methotrexate. J Rheum 23(suppl 44):72-74, O'Dell JR, Haire CE, Palmer W, Drymalski W, Wees S, Blakely K, Churchill M, Eckhoff PJ, Weaver A, Doud D, Erikson N, Dietz F, Olson R, Maloley P, Klassen LW, Moore GF. Treatment of early rheumatoid arthritis with minocycline or placebo. Arth Rheum 40: , O'Dell JR, Paulsen G, Haire CE, Blakely K, Palmer W, Wees S, Eckhoff PJ, Klassen LW, Churchill M, Doud D, Weaver A, Moore GF. Treatment of early seropositive rheumatoid arthritis with minocycline: Four-year followup of a double-blind, placebo-controlled trial. Arthritis Rheum 42: , O Dell JR, Blakely KW, Mallek JA, Eckhoff PJ, Leff RD, Wees SJ, Sems KM, Fernandez AM, Palmer WR, Klassen LW, Paulsen GA, Haire CE, Moore GF. Treatment of early seropositive rheumatoid arthritis. A two-year, double-blind comparison of minocycline and hydroxychloroquine. Arthritis Rheum 44; , Blackwelder WC. Equivalence trials. s in Biostatistics in Clinical Trials, Redmond C and Colton T, Eds. John Wiley & Sons Ltd, Chichester of of 180

39 CSP 551 Protocol V Keystone EC, Kavanaugh AF, Sharp JT, et al. Radiographic, clinical, and functional outcomes of treatment with adalimubab (a human anti-tumor necrosis factor monoclonal antibody) in patients with active rheumatoid arthritis receiving concomitant methotrexate therapy. Arthritis Rheum 2004;50(5): of of 180

40 28 of 180

41 -3 to 0 mos* Baseline* Week 6 Week 12 Week 18 Week 24 Week 30 Week 36 Week 42 Week 48 q 12 wks q 24 wks Termination Evaluate inclusion/exclusion criteria X X Review ICF X Sign ICF X Patient Report: HAQ X X X X X X X HUI X X X X X EQ5D X X X X X Current pain rating X X X X X AM Stiffness X X X X X Patient global score X X X X X Patient report of current meds X X X X X Physician Report: History and physical X Physician global assessment X X X X X Vitals X X X X X X X DAS28 X X X X X Joint count X X X X X X X X General health assessment (VAS) X X X X X X X Start folate Rx X X-ray X X X X Adverse event monitoring X X X X X X X X X X X Laboratory and Diagnostic Monitoring: Urine pregnancy test (pre-menopausal women only) X ESR X X X X CBC X X X X X X X X X Creatinine X X X X X X X X X LFTs X X X X X X X X X G6PD X PPD X Chest X-ray (?) X Bank DNA X Bank serum X X X X Opthalmic evaluation X X Events: Switch/count as failure of * Please note, "-3 to 0 mos" and "Baseline" visits may occur on the same day, provided the subject has had an opportunity to review the consent form prior to signing it. Randomization will occur after eligibility is confirmed, including required lab tests. 29 of 180

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91 Summary of Protocol Amendments Amendment 1 (approved and distributed to the sites on August 17, 2007) included the following changes: Allowed subjects treated with MTX combined with either sulfasalazine or hydroxychloroquine to be eligible for the trial provided that such treatment was limited to 4 weeks or less and had ended at least 4 weeks prior to enrollment Changed the requirement for a baseline eye exam to include only subjects who had received hydroxychloroquine for at least 48 weeks Clarified inclusion/exclusion criteria regarding tuberculosis and chest x-rays for tuberculosis screening to allow subjects to enroll if they meet one of the following criteria: negative PPD; positive PPD < 5mm, with a negative chest x-ray; or a positive PPD > 5mm treated for at least 28 days with INH Amendment 2 (approved and distributed to sites on April 11, 2008) included the following changes: Allowed subjects undergoing MTX monotherapy for a minimum of 3 months and subjects previously exposed to brief (<5 weeks)tnf inhibitor therapy to be eligible for the trial Added 90 day timeframe for PPD testing prior to randomization Added information concerning serious skin reactions from etanercept to the informed consent form Changed the HIPAA authorization to allow collection of social security numbers Removed G6PD testing from the baseline assessments Amendment 3 (approved and distributed to the sites on December 3, 2008) included the following changes: Allowed subjects with stable albumin up to 1.0 g/dl below the lower limit of normal to be eligible for the trial Allowed subjects who have no tuberculosis exposure or travel history to be enrolled up to 180 days after their last PPD Allowed subjects with macular degeneration into the trial, with ophthalmologist approval Amendment 4 (approved and distributed to the sites on February 23, 2010) included the following changes: Decreased sample size to 450 patients Modified testing of the primary endpoint to be a continuous outcome: change in DAS28 over 48 weeks Allowed patients with a DAS-CRP to qualify for the study when the DAS28 < 4.4, but ESR <10 and 10 tender and swollen joints

92 Amended the role of the Human Rights Committee to be an advisory committee Allowed one non-articular steroid injection (60mg methylprednisone or equivalent) during each 24-week portion of the trial, but not within 6 weeks of an evaluation point

93 1.0 1 of 59 Title: Statistical Analysis Plan (SAP) MAVERIC Study Specific Documents DATE OF ISSUE OCTOBER 2008 DOCUMENT DESCRIPTION This document describes the statistical analysis plan for the RACAT randomized clinical trial. AUTHOR: Robert A. Lew, PhD SUBMITTED BY: Robert A. Lew, PhD DOCUMENT/RECORD HISTORY: DATE VERSION RECORD HISTORY 09/09/ Initial AUTHORIZATION SIGNATURES NAME FUNCTION TITLE SIGNATURE DATE James O Dell MD, Principal Investigator Mary Brophy, MD, MPH Study Director Robert Lew, PhD Study Biostatistician Erika Holmberg, MPH Project Manager Ryan Ferguson, MPH Quality Officer 1

94 1.0 2 of 59 Title: Statistical Analysis Plan (SAP) 1. INTRODUCTION STUDY OBJECTIVES INVESTIGATIONAL PLAN OVERALL STUDY DESIGN AND PLAN EFFICACY ENDPOINTS SAFETY VARIABLES NOT DEFINED IN PROTOCOL STATISTICAL METHODS STUDY SUBJECTS DISPOSITION OF SUBJECTS ANALYSIS POPULATIONS: DEMOGRAPHICS AND BASELINE CHARACTERISTICS DEMOGRAPHICS MEDICAL HISTORY AND COMORBIDITIES GENERAL CONSIDERATIONS EFFICACY EVALUATION PRIMARY EFFICACY ANALYSIS SECONDARY EFFICACY ANALYSIS SAFETY EVALUATION MORTALITY ADVERSE EVENTS DETERMINATION OF SAMPLE SIZE MISSING VALUES MULTIPLICITY INTERIM ANALYSES DATA MONITORING COMMITTEE REPORTING OUTPUT LISTS OF TABLES, FIGURES AND LISTINGS SUBJECT DISPOSITION SHELLS BASELINE CHARACTERISTICS SHELLS EFFICACY SHELLS SAFETY SHELLS

95 1.0 3 of 59 Title: Statistical Analysis Plan (SAP) 1. Introduction This document outlines the statistical methods for the display, summary and analysis of data collected in the Department of Veterans Affairs Cooperative Studies s study #551 entitled Rheumatoid Arthritis: Comparison of Active Therapies in Patients with Active Disease Despite Methotrexate Therapy. The purpose of this document is to provide guidelines from which the analysis will proceed. Deviations from these guidelines will be documented. The following documents were used in preparation of this statistical analysis plan (SAP): Clinical Study Protocol CSP #551 Case Report Forms (CRF) from above entitled Protocol MAVERIC SOP: Drafting and Approving Statistical Analysis Plans 2. Study Objectives The primary objective of the study is to determine in subjects with rheumatoid arthritis who have active disease despite methotrexate whether two successful treatment strategies are clinically equivalent as measured by the percentage of patients who reach a low disease activity state (Disease Activity Score for 28 joints (DAS28) 3.2). Treatment Strategies: (1) Strategy A (Triple Therapy): the addition of hydroxychloroquine and sulfasalazine to methotrexate, with a switch over at 24 weeks to the addition of etanercept to methotrexate in non-responders, compared to (2) Strategy B (Biologic Therapy): initial therapy with etanercept and methotrexate and switch to the addition of hydroxychloroquine and sulfasalazine to methotrexate in non-responders at 24 weeks. Secondary Objectives: (1) To determine whether radiographic progression at 48 weeks will be equivalent in patients randomized to Strategy A compared to Strategy B. (2) To determine whether functional outcomes are equivalent in patients randomized to Strategy A compared to Strategy B. This is particularly important because the DAS28 does not incorporate functional outcomes. (3) To determine the cost-effectiveness of Strategy A compared to Strategy B. Other Objectives: To identify genetic factors and biomarkers that predict disease progression and success or toxicity of the different strategies, serum, plasma and DNA will be collected and banked for future use. To carry out other descriptive and exploratory analyses; analyses directly related to the primary hypothesis, analyses of other endpoints of interest, and cost analyses. 3

96 1.0 4 of 59 Title: Statistical Analysis Plan (SAP) 3. Investigational Plan 3.1. Overall Study Design and Plan This is a double-blind, randomized, 48-week, two-arm trial. The trial will test the hypothesis that the two strategies are equivalent with respect to the percentage of patients in each arm achieving a DAS Patients with rheumatoid arthritis who have active disease despite 15 to 25 mg of methotrexate weekly will be offered an opportunity to participate. They will be randomized to receive either triple therapy (sulfasalazine and hydroxychloroquine added to their baseline methotrexate), or etanercept 50 mg subcutaneously weekly in addition to baseline methotrexate. Clinical response will be measured using a DAS28, which is a well validated measure of composite clinical disease activity. If, at the end of 24 weeks, patients have not achieved a DAS28 improvement of 1.2 units they will, in a blinded fashion, be switched to the other therapy. Improvement of 1.2 or greater is recognized as clinically meaningful. Thus, patients randomized to receive triple therapy at the onset who fail to achieve a DAS28 improvement of at least 1.2 will receive methotrexate plus etanercept for the second 24 weeks and similarly patients randomized to methotrexate plus etanercept who fail to achieve this level of improvement will receive triple therapy for the second 24 weeks. The timing of the switch (24 weeks) is important since most rheumatologists are unwilling to wait longer than this to obtain clinically meaningful improvement for their patients. Since both groups will be switched if the 24 week DAS28 goal (improvement of 1.2) is not achieved, the blind will be protected. Importantly, for the purpose of the intention to treat analysis of the primary clinical endpoint, outcomes of patients who switch will be attributed to initial treatment assignment. Furthermore, because patients may have a clinical response but still demonstrate radiologic progression, an important component of this trial is to determine whether the two strategies have similar effects on radiologic outcomes at 48 weeks. Therefore, we are testing the strategy of triple therapy first with rescue etanercept, versus etanercept first with rescue triple therapy. At the end of 48 weeks, the subject will exit the blinded portion of the trial and the code will be broken so that the clinicians will know what the subject has been receiving. If patients have achieved a DAS28 of 3.2, indicating good clinical response, clinicians will be encouraged to continue the current therapy. Descriptive data will continue to be collected on all these patients in an open fashion until all 600 patients have completed the 48 week portion of the trial. Since RA is a life-long disease this observational period will allow us to collect long term toxicity data as well as important effectiveness (clinical and radiographic), tolerability and economic data. Importantly, the radiographic data on this population will continue to be read in a blinded manner. 4

97 1.0 5 of 59 Title: Statistical Analysis Plan (SAP) Study Design Randomization DAS28 Improved 1.2? Primary Outcome: **DAS28 3.2? *SSZ + HCQ *Etanercept **No Yes Yes SSZ + HCQ Etanercept Yes Yes Start 24 Weeks 48 Weeks Open Continuation Open Continuation *All patients continue to receive methotrexate **Patients who switch at 24 weeks remain with their randomization group for analysis of primary outcome In summary, this unique design compares two treatment strategies instead of the traditional trial design where one treatment is compared to another in a static, inflexible way. The design is dynamic and mirrors the way patients are treated in practice, allowing therapy to be changed at 24 weeks if patients have not had clinically significant improvement. This innovative design is critical for the question we are addressing. Patients may respond more rapidly to the addition of etanercept to methotrexate compared to sulfasalazine, hydroxychloroquine, and methotrexate. However, a substantial number of patients will respond to initial therapy with the three drug combination. If this trial shows that the two strategies are equivalent at 48 weeks, a significant percentage of patients will have been spared expensive therapies and health agencies can ethically recommend the strategy of using sulfasalazine, hydroxychloroquine, and methotrexate initially, which is much less expensive. 3.2 Efficacy and Safety Variables Efficacy endpoints The efficacy endpoints for this study are as follows Primary Efficacy endpoints proportion of patients achieving a DAS28 of 3.2 or less at 48 weeks Secondary Efficacy endpoints Radiographic progression will be assessed by obtaining standard PA x-rays of the hand and wrist and AP x-rays of the feet, which will be graded according to the Van der Heijde-modified Sharp score Functional outcomes will be measured using the Health Assessment Questionnaire (HAQ) Patient self reported quality of life will be measured at baseline and at follow-up using the Health Utilities Index (HUI) and the EQ-5D (Euro QOL). The cost-effectiveness of Strategy A compared to Strategy B. 5

98 1.0 6 of 59 Title: Statistical Analysis Plan (SAP) Safety Variables While the safety variables for this study are not defined in the protocol, the following will be evaluated in keeping with the VA s mission to protect the safety of our patients. Overall mortality rate at 48 weeks The number of adverse events at 48 weeks and percent of subjects with adverse events Number of serious adverse events at 48 weeks and percent of subjects with serious adverse events. Any unresolved adverse events will be followed for up to 30 days after the end of the study. The study executive committee will determine whether continued follow-up after the end of study is required. 4. Statistical Methods 4.1 Study Subjects Disposition of Subjects The number of subjects screened, enrolled, and completing the study will be presented overall and by center subdivided by demographic categories (i.e., gender, race, etc.) Reasons for the early termination of subjects as noted on Form 23 will be summarized. Subjects that are lost to follow-up will be summarized as one category. The following listings will be provided: A by-subject listing of completion/termination details A by-subject listing of protocol violations and deviations Analysis Populations: Three datasets will be constructed: the ITT analysis set for the definitive analyses, validation analysis sets (primarily the per protocol analysis subset), and the safety analysis dataset. All enrolled subjects included in the subject listings (by study ID number) will be included in the Intention-To-Treat (ITT) analysis provided they were randomized to the study and received at least one dose of study treatment (either Triple or Biologic). The safety analysis population includes all subjects enrolled. A variety of datasets will be used for confirmatory analyses. Foremost among these is the Evaluable Per-Protocol (EPP) analysis. Patients in this analysis subset of the ITT analysis set will have had no major protocol violations and have completed follow up (48 week visit for end point assessment). We will repeat the primary analyses on the EPP subset to confirm or not confirm the results found in the analysis of the ITT patients. Other confirmatory analyses will be carried out on subsets of the ITT patients, stratified by patient characteristics such as age, sex, and disease history. 6

99 1.0 7 of 59 Title: Statistical Analysis Plan (SAP) 4.2 Demographics and Baseline Characteristics Demographics Demographic details (e.g., age, sex, ethnicity, race, marital status, and smoking status) will be summarized overall and by center Medical History and Comorbidities The pertinent medical histories (i.e., comorbidities), and history of RA will be summarized overall and by center. Frequencies and percentages will be reported. 4.3 General Considerations In general, the number of observations, mean, median, standard deviation, minimum, and maximum will be calculated for continuous variables. The number of decimals places will be as follows: minimum and maximum to the same number of decimal places as present in the raw data, mean and median to plus one decimal place, standard deviation to plus two decimal places. Frequencies and percentages will be calculated for categorical data. Baseline is defined as the randomization visit. Missing endpoint data will not be imputed. Missing covariate data will be imputed in selected circumstances as described in Little and Rubin (2002). 1 All planned analyses for the final clinical study report will be performed using a two-sided nominal significance level of Efficacy Evaluation Primary Efficacy Analysis The primary hypothesis will be tested using the binary outcome at 48 weeks of a DAS as evidence of a response (successful outcome) and a score > 3.2 as a non-response (failure). Equivalence tests comparing two treatment strategies wherein one strategy is postulated to be not worse than the standard are referred to as tests of non-inferiority and therefore typically set up as one-sided tests. All the proposed tests for the primary and secondary hypotheses are tests of noninferiority. Using P(A) and P(B) to denote the proportion of subjects achieving low disease activity state for Strategy A and Strategy B respectively, the hypotheses to be tested can be stated as H 0 : P(B) > P(A) H a : P(B) = P(A) The treatments will be considered equivalent if the upper limit on the one-sided 95% confidence limit on the difference between the two success rates is less than 12.5%. It is expected that initial treatment with etanercept will have the higher success rate. With a binary endpoint for the primary outcome variable, the proposed test is a one-sided equivalence test of the null hypothesis that the upper limit on the 95% confidence interval on the difference between two proportions exceeds 12.5%, the boundary on equivalence. 1 Little, Roderick JA, and Rubin, Donald B (2002), Statistical analysis with missing data. John Wiley & Sons (New York; Chichester). 7

100 1.0 8 of 59 Title: Statistical Analysis Plan (SAP) Hence, if this upper confidence limit (on the difference between the initial treatment etanercept strategy response rate and the initial treatment triple therapy strategy) is no more than 12.5% then we will reject the null hypothesis of non-equivalence and accept the alternative hypothesis of equivalence. In addition to the formal test comparing the upper confidence limit on the difference between proportions to 12.5%, we will carry out a descriptive analysis. We will report the proportions, their difference, and the odds ratio of biological treatment compared to triple therapy treatment. We will also report the proportion and number of subjects who switched treatment regimens at week 24. We will compare the two treatment strategies in the primary hypothesis using a one-sided Fisher s exact test. We will repeat the descriptive analysis, stratified by 4 groups accounting for treatment switch, with the groups being: (1) triple therapy throughout the entire study, (2) triple therapy switched to etanercept, (3) etanercept throughout the entire study, and (4) etanercept switched to triple therapy. In addition, further logistic regression modeling will be undertaken to include all treatment by baseline covariate interactions to assess whether there may be relevant treatment-covariate interactions that merit consideration of subgroup analyses. This analysis will be followed by further, confirmatory analyses using multivariable logistic regression modeling to take account of the effects of baseline covariates on the primary hypotheses. Our analytic strategy is to use PROC GENMOD in the SAS statistical package to model the Odds Ratio of biologic treatment compared to triple therapy. This method will estimate the odds of achieving DAS at 48 weeks among those who started on etanercept (biologic) therapy compared to those who started on triple regimen, after adjusting for following baseline covariates: age, sex, prior treatment with sulfasalazine or hydroxychloroquine, duration of treatment with methotrexate, disease duration, baseline ESR, and concurrent prednisone use. The multivariable logistic regression model can be expressed as follows: Log (OR) = α + β 1 X 1 + β 2 X 2 + γ X 1 *X 2 + [Z] Where X 1 is the indicator variable for Triple therapy (1=Triple, 0=Biologic), X 2 is the indicator variable for a switch in therapy (1= switched, 0=not switched), and [Z] is the vector of baseline covariates adjusted in the model. The parameter β 1 will estimate the independent effect of Triple therapy compared to Biologic therapy, the parameter β 2 will estimate the independent effect of switching therapies, and the parameter γ will estimate the additional effect of switching from Triple therapy to Biologic therapy. Sample Size and Statistical Power Considerations A one-sided equivalence comparison test will be used for the primary hypothesis test. This test will compare the proportion of subjects in each arm of the study who reach the primary endpoint of DAS28 of 3.2. This target value is a validated measure and represents a very low disease activity. The null hypothesis is that the upper limit on the 95% one-sided confidence interval on the difference between two proportions of patients achieving a DAS28 of 3.2 exceeds 12.5%. Because it is expected that etanercept response will be higher than the triple therapy response the proposed test is one-sided equivalence. The boundary or limit on equivalence is 12.5%. With a 8

101 1.0 9 of 59 Title: Statistical Analysis Plan (SAP) type 1 error of 5% and a sample size of N=300 per study arm, the one-sided comparison of etanercept with triple therapy has 92% power to reject the null hypothesis for the test of the difference between proportions. The study has 90% power allowing for up to an 8% dropout rate. In recent placebo-controlled trials, differences of 20% or more were shown, indicating 12.5% is an appropriate limit. The difference of 12.5% is small compared to differences between treatments observed in randomized clinical trials. In the 2003 ARMADA trial the group of patients receiving 20-mg adalimumab and the groups receiving 40 mg and 80 mg adalimumab differed by 18% and 19% in terms of the proportion of ACR20 responders after 24 weeks. Similar results were observed for the ACR50 endpoint. In the 2004 trial comparing the costimulation blocker CTLA4Ig to placebo the groups differed by 25% in terms of the proportion of ACR20 responders after 24 weeks. Assuming these proportions have a common standard deviation of 0.5, the corresponding effect sizes are double these differences. In these terms the proposed trial has a limit on effect size of 0.25 compared to trials reported in the literature with effects sizes near Note that the endpoint of the proposed trial is the proportion of patients achieving a specific DAS28 score, rather than an ACR20. Hence, our assertions assume that these distinct endpoints have very similar effect sizes. Effect size is a unit-free measurement. We designed this equivalence trial by choosing a difference between the two treatments that is small enough that clinicians believe it is not clinically meaningful. Typically such a choice depends on the effect sizes observed in relevant clinical trials. Our equivalence point of 12.5% was based on discussions with the seven rheumatologists on our Planning Committee, selected for their expertise in rheumatoid arthritis and clinical trials. Losses to Follow-up This two-armed randomized clinical trial seeks to enroll a total of 600 subjects. With an intent-totreat analysis (see section 4), the final sample sizes remain about 300 per arm because all subjects who stop study medications for reasons related to treatment are regarded as treatment failures (side effects and withdrawal of consent) even though they are no longer on blinded medications at week 48. About 5% of the subjects are expected to stop study medications due to adverse treatment effects before week 48. Among patients receiving the active treatment in the 24-week ARMADA trial, only 2.3% withdrew for adverse reactions attributed to the treatment. This is a rate of 4.6% per year. For this trial these subjects will be regarded as treatment failures. At most another 5% per year will drop out for reasons unrelated to treatment such as a severe episode of an unrelated condition, death or moving away. These subjects will be regarded as study dropouts. Hence, the overall dropout rate for any reason should be significantly less than 8% Secondary Efficacy Analysis (1) The radiologic hypothesis will be tested with a one-sided equivalence test of the null hypothesis that the difference in the two mean Sharp scores for the two treatment strategies is not zero. We will reject the null hypothesis of non-equivalence if the upper limit of the one-sided 95% confidence interval on the difference is 1.4 Sharp score units or less. A consensus of our Planning Committee rheumatologists was that differences in Sharp score of 2.0 units or less are clinically 9

102 of 59 Title: Statistical Analysis Plan (SAP) unremarkable. We shall also use the independent samples t-test. Further analysis will account for baseline Sharp score as a covariate. (2) We will assess the level of disability at baseline and after 12 months using the Disability Index of the Health Assessment Questionnaire (HAQ). The primary hypothesis tests the equivalence of the two treatment strategies. In particular, the proposed test is a one-sided equivalence test of the null hypothesis that the difference between the two means is not zero. The two means are the mean 12 month change in HAQ score for patients initially on triple therapy and the mean 12 month change in HAQ score for patients initially on Etanercept. The boundary or limit on equivalence is a difference between the means of 0.2 HAQ score units where the HAQ score ranges from 0 to 3. We will reject the null hypothesis of non-equivalence and accept the alternative hypothesis of equivalence if the upper limit of the one-sided 95% confidence interval on the difference between the means is 0.2 units or less. To test equivalence, the upper limit of the 95% confidence interval for the difference between the means will be compared to 0.2 HAQ score units. To confirm that confounding has not produced a misleading result, the hypotheses will be retested using the mixed effects model multiple regression done with PROC MIXED in the SAS statistical package. Fixed covariates will include age, sex, and baseline severity of disease. Study site and subject will be random effects. We propose to carry out other descriptive and exploratory analyses; analyses directly related to the primary hypothesis, analyses of other endpoints of interest, and cost analyses. These analyses will, for the most part, be hypothesis-generating. Analyses of this type include, but are not limited to: DAS28 responses at 24 weeks; DAS28 responses 3.2 at 48 weeks with switchers counted as failures; comparison based on the change in the DAS28 score; proportion of patients achieving an American College of Rheumatology Criteria (ACR20, ACR50, ACR70) response at 24 and 48 weeks; percentages of responders to each therapy after failure to respond at 24 weeks; European League Against Rheumatism (EULAR) response (those that improve by at least 1.2 and have a DAS28 of 5.1); correlation of DAS28 changes with radiographic outcomes; correlations between radiographic and quality of life changes; percent of subjects not progressing radiographically; percent achieving major clinical response; and proportion receiving steroid injections. Other analyses will be confirmatory in nature such as treatment comparisons of the continuous components of the DAS28 and ACR responses Cost analyses The health economics analysis for this trial will evaluate the cost effectiveness of the following strategies for the treatment of RA in patients with active disease refractory to methotrexate therapy. Triple therapy: methotrexate, sulfasalazine, hydroxychloroquine Anti-TNF therapy: etanercept (Enbrel) plus methotrexate As is noted elsewhere in the text of the protocol and this analysis plan, this trial is a comparison of two treatment strategies, not a direct comparison of the two different pharmaceutical therapies. Economic analysis will be central to this evaluation because anti-tnf therapy is more than 10 times as expensive as triple therapy; about $12,000 compared to about $1,000 per year for the VA, but costs for anti-tnf therapy may be considerably higher in other healthcare environments. Given that the existing trial data indicate marginal gains in effectiveness for anti-tnf therapy, compared to triple therapy, the added value of anti-tnf therapy is uncertain. 10

103 of 59 Title: Statistical Analysis Plan (SAP) These strategies will be evaluated using standard cost-effectiveness analysis methods. The two strategies will be evaluated relative to each other to identify which therapy is dominated by another (e.g., strategies that are more costly and less effective than available alternatives). In practice, however, most strategies are not dominant or dominated. Typically, more effective treatments are more expensive. In these situations, the cost-effectiveness analysis helps answer the question: is the incremental benefit worth the added cost? Costs will include direct treatment costs and costs associated with lost productivity (indirect costs). For a complete picture of direct treatment costs, it is necessary to collect utilization data from several sources. The analysis is complicated by the fact that that this is a bi-national study with each country having differing costs and practice patterns. To deal with issue we have developed the appropriate data collection and analysis strategy for each country respectively. In the VA, we have developed a method that first uses administrative data, and then uses Medicare data, provider bills, and self-report data to fill in any gaps. The indirect costs will be estimated from respondent s self-report. Alternate methods, appropriate for the context, will be used to capture costs and utilization for patients in the Canadian and non-va arms of the study. Outcomes will be expressed in Quality-Adjusted Life Years (QALYs). In the primary analysis, QALYs will be periodically measured using the Health Utilities Index (HUI) version 3. The HUI-3 is standardized multi-attribute utility instrument. Because many of the European studies have used the EuroQol (EQ-5D) to measure QALYs, we will also use the EQ-5D to facilitate comparisons with these other studies. We will also collect self-reported general health on a 10-cm visual analog scale (VAS). A VAS is usually included as part of the EQ-5D. Data from the VAS will provide an alternative measure of patient utility and we will use this in a sensitivity analysis. The VAS has also been used in prior studies of RA, and thus including this measure will enable us to compare our results to prior studies. The cost-effectiveness analysis will follow widely accepted and recommended methods. Costs and outcomes will be collected during the study. A societal perspective will be adopted so that all health care costs, including patient-incurred costs (e.g. caregiver costs) and the direct costs of health care, will be considered. The cost-effectiveness analysis will use the patients life span as the time horizon. Trial data will then be used as inputs into decision analytic models to project the trial results to obtain lifetime health care cost and QALYs for each treatment group. Stochastic analysis will be undertaken using cost-effectiveness acceptability curves to express uncertainty around costeffectiveness ratios based on mean costs and outcomes. Cost-effectiveness ratios will be calculated to compare trial participants lifetime cost and outcomes. Existing literature and VA clinical data will be used to identify starting points for the economic model, to determine necessary assumptions, and to identify parameters to be varied in the sensitivity analysis; when these data are ambiguous or incomplete, consensus from the clinical investigators will be sought. A sensitivity analysis will be done to reflect uncertainty of the input parameters and to test the robustness of the model. While cost-effectiveness analysis is the standard for evaluating medical technologies, VA managers and Canadian formulary authorities are usually more concerned with their own narrower perspectives and how a change in therapy will affect their budgets. Thus, we will conduct additional 11

104 of 59 Title: Statistical Analysis Plan (SAP) budget impact analyses. First, we will estimate the lifetime model, only considering VA costs. Second, we will estimate the short-term (one and two years) effect on VA expenditures. The later is especially important with respect to VA policy, as VA managers are most concerned about the immediate budget effects. Both of these analyses do not require any additional information; they just involve restricting the model to VA costs and limiting the time horizon. For both of these additional analyses, Canadian and RAIN costs will be converted to VA costs. Additional analyses will also be conducted from the Canadian perspective, converting U.S. costs to Canadian costs Safety Evaluation All safety summaries will be based on the Safety Population described in Section Analysis Populations Mortality The frequency and proportion of all deaths will be presented overall and by center. A by-subject listing will be provided Adverse Events (AEs) AEs will be coded using the Medical Dictionary for Regulatory Activities (MedDRA) ( 9.0) coding dictionary. AE summaries will be based upon the number of subjects reporting AEs and also on the number of events reported. Adverse events will be tabulated (i.e., those events that were not present at the baseline evaluation or worsened in severity following the start of treatment). Any AEs starting after study completion or withdrawal will not be tabulated. If the AE start date is not a complete date, the following rules will be applied to determine whether the AE will be tabulated or not. If the start date is completely missing: The AE will be tabulated unless the AE stop date is less than the baseline date. If part of the start date is missing, sites were instructed to enter 15 for missing day and 06 for missing month. The number and percentage of subjects with AEs through the end of the study (analysis period) will be summarized by system organ class and preferred term overall. Summaries of AEs by intensity and relation to study device will be presented by system organ class and preferred term overall. For each subject and each AE, the worst intensity/relation recorded will be attributed and used in the intensity/relation summaries. If intensity/relation is missing, the intensity/relation will be queried by data management. The number and percentage of subjects with serious adverse events (SAE) through the end of the study (or analysis period) will be summarized by system organ class and preferred term overall. A by-subject listing of all AEs will be provided. 12

105 of 59 Title: Statistical Analysis Plan (SAP) A by-subject listing of all serious AEs will be provided. 4.6 Determination of Sample Size The sample size for the study was based on the primary hypothesis. With a type 1 error of 5% and a sample size of N=300 per study arm, the one-sided comparison of etanercept with triple therapy has 92% power to reject the null hypothesis for the test of the difference between proportions. The study has 90% power allowing for up to an 8% dropout rate. In recent placebo-controlled trials, differences of 20% or more were shown, indicating 12.5% is an appropriate limit. 4.7 Missing Values Dropouts will not be included in the primary analysis but sensitivity analyses will address their potential to bias the results. Imputation of final outcomes requires questionable assumptions about treatment effects. We will consider a range of such assumptions in the confirmatory analyses. Relatively few dropouts are anticipated, so the expected bias is small. However, the confirmatory secondary analysis including the dropout data censored at the appropriate study visit will also be done to assess potential bias associated with the complete case analysis. A covariate adjusted primary analysis will be done in order to either confirm or not confirm the primary analysis that has no covariates. In either case, the confirmatory analysis will indicate the degree of bias in the complete case analysis. Practical measures will be taken to minimize missing values and losses to follow-up. One prominent conjecture about patients who drop out is that they tend to have poor outcomes. This clearly would violate the assumption of missing-at-random. 4.8 Multiplicity There will be no corrections done for multiple comparisons in this study. 4.9 Interim Analyses There is no interim analysis planned Data Monitoring Committee Data and study progress will be monitored by the study executive committee and by the study steering committee. The Data Monitoring Committee will review the study progress and safety on at least an annual basis with additional meetings and communications as needed. 5.0 Reporting Output All outputs will be produced using SAS 9.1 or a later version (if available at MAVERIC). Other applications (including Microsoft Word and Microsoft Office) will be used to aid in data presentation. Tables, listings, and figures will be arranged in the order that they appear in the textual sections of this analysis plan, i.e., disposition first, followed by baseline summaries, efficacy summaries, and safety summaries. 13

106 of 59 Title: Statistical Analysis Plan (SAP) 6.0 Lists of Tables, Figures and Listings All recorded data will be listed following the order and structure of the CRF. In addition, selected derived data used in the analyses will be listed. The listings will include all enrolled subjects (with data relevant to the particular listing) and will be ordered by center, subject, and visit. Lists of Tables, Figures and Listings A. Subject Disposition Charts A1. Description of Subject Disposition a. Overall recruitment by month b. Recruitment by month VA sites c. Recruitment by month RAIN sites d. Recruitment by month Canadian sites Tables A2. Recruitment a. Recruitment by site b. Recruitment to substudies Listing A3. Study Medication Discontinuation and Termination Listing A4. Protocol Deviations and Enrollment Waivers Tables A5. Forms Completion Status a. Forms Completion Status, by Form b. Forms Completion Status, by Site Table A6. Accrual of X-ray Data B. Baseline and Repeated Measures Table B1. Baseline Demographics Tables B2. Baseline Medical History a. Medical History b. Social History and Allergies Table B3. Baseline History of RA Tables B4. Baseline Health Status a. Physical Assessment (total and subset with 24 weeks followup) b. Labs (total and subset with 24 weeks followup) c. ECOG Toxicity Criteria (total and subset with 24 weeks followup) Table B5. Functional Assessments (total and subset with 24 weeks followup) Table B6. Baseline Medications Listing B7. Joint Procedures C. Outcome Measures Tables C1. RA Status (DAS28, ESR, Joint Count, Patient VASs) a. Total Study Population b. Subset with 24 weeks followup D. Adverse Events and Safety Table D1. Total Adverse events and serious Adverse events Table D2. Total Adverse events and serious Adverse events BY SITE Table D3. AEs by System Organ Class and Preferred terms (PT) Table D4. SAEs by System Organ Class and Preferred terms (PT) 14

107 of 59 Title: Statistical Analysis Plan (SAP) Table D5. Number of AEs per patient attributable and not attributable to treatment Table D6. Number of SAEs per patient attributable and not attributable to treatment Table D7. Number of AEs per patient (excluding SAEs) attributable and not attributable to treatment 15

108 of 59 Title: Statistical Analysis Plan (SAP) 7.0 Subject Disposition Shells 16

109 of 59 Title: Statistical Analysis Plan (SAP) Site Number Site Status CRRC Network 109 Recruiting 131 Recruiting 103 Recruiting 108 Recruiting 111 Recruiting Need 101 Training Need 138 Training Need 140 Training Awaiting 104 IRB RAIN Network 116 Recruiting 120 Recruiting 136 Recruiting 125 Recruiting 114 Recruiting 122 Recruiting Need 121 Training Need 134 Training Awaiting 130 IRB VA Network 119 Recruiting 118 Recruiting 128 Recruiting 133 Recruiting 112 Recruiting 127 Recruiting 129 Recruiting 113 Recruiting* 124 Recruiting* 105 Recruiting 115 Recruiting* 110 Capitated # of patients expected # patients recruited Percent of Expected Recruitment Actual Recruitment Rate Expected Recruitment Rate: 1.11 Expected Recruitment Rate: 0.56 Expected Recruitment Rate:

110 of 59 Title: Statistical Analysis Plan (SAP) 107 Withdrew 123 Withdrew Awaiting 102 IRB Awaiting 106 IRB Awaiting 135 IRB Awaiting 139 IRB 18

111 of 59 Title: Statistical Analysis Plan (SAP) Table A2b. Recruitment to substudies Study Number Percent Serum Bank Long Term Followup DNA Bank* 19

112 of 59 Title: Statistical Analysis Plan (SAP) Listing A3. Study Medication Discontinuation and Termination Subject Number Permanent Discontinuation of One Study Medication Description. Permanent Discontinuation of All Study Medications 20

113 of 59 Title: Statistical Analysis Plan (SAP) Listing A4. Protocol Deviations and Enrollment Waivers Subject Number Enrollment Waivers* Full Description Protocol Deviations 21

114 of 59 Title: Statistical Analysis Plan (SAP) Table A5a. Form Completion Status, by Form Form Number 01A-Eligibility_PartA 01B-Eligibility_PartB 01C-Consent Checklist 02-Demographics 03-History RA 04-Past Medical Hx 05-Social Hx & Allergies 06-Concomitant Medications 07-Physical Exam 08-Joint Count 09-Laboratory 10-Medication Dispensing 11-Medication Compliance 14-Health Services 15-Work Productivity 17-Inpatient Visit 18-Diagnostic Tests 24-Study Med Prescription DAS28-DAS28 PS1-PatientVAS PS2-HAQ PS3-HUI3 PS4-EQ-5D Expected Number of Forms Actual Number of Forms Number of Forms Missing Percent Complete 22

115 of 59 Title: Statistical Analysis Plan (SAP) Table A5b. Form Completion Status, by Site Actual Number Expected Number of Number of Forms of Forms Forms Missing Site Number Percent Complete 23

116 of 59 Title: Statistical Analysis Plan (SAP) Table A6. Accrual of X-Ray Data Expected number of X-rays Actual number of X-rays Percent Baseline Week 24 24

117 of 59 Title: Statistical Analysis Plan (SAP) 8.0 Baseline Characteristics Shells 25

118 of 59 Title: Statistical Analysis Plan (SAP) Table B1. Demographics Study Population (n= ) Age (yrs) Mean±S.D. Range Sex Male Female Ethnicity Hispanic or Latino Not Hispanic or Latino Race White Black or African American Asian Native Hawaiian or other Pacific Islander American Indian or Alaskan Native* First Nations** Other * American Indian or Alaskan Native asked of US patients only ** First Nations asked of Canadian patients only 26

119 of 59 Title: Statistical Analysis Plan (SAP) Table B2a. Baseline Medical History Condition/Diagnosis Head, Ears, Eyes, Nose, Throat Cataracts Seasonal Allergies Other Study Population (n=) Significant History Cardiovascular Myocardial infarction Cerebrovascular accident Unstable angina Coronary artery bypass graft Angioplasty Congestive heart failure Stent Other Respiratory Gastrointestinal Peptic Ulcer Disease Hepatitis Other Renal/GU Stones Other Endocrine Diabetes (Type 1) Diabetes (Type 2) Other Musculoskeletal Septic Arthritis Gout Other Dermatologic Nodules 27

120 of 59 Title: Statistical Analysis Plan (SAP) Skin Cancer Other Psychiatric 28

121 of 59 Title: Statistical Analysis Plan (SAP) Table B2b. Social History and Allergies Cigarette Smoking History Never Smoked Former Smoker Current Smoker Study Population (n=) Baseline Ever used cigars, pipes, or chewing tobacco Alcohol Use None One or less drink per week More than one per week Military Service None Active Service Reserves only Allergy to sulfa drugs Allergy to other medications 29

122 of 59 Title: Statistical Analysis Plan (SAP) Table B3. History of RA Study Population (n=) Age at Diagnosis (Mean Range SD) Medication History for RA Sulfasalazine Hydroxychloroquine Anakinra Leflunomide Cyclosporine Minocycline Gold Compound Penacillamine Azathioprine Chloroquine Joint Replacement Knee Hip MCP Shoulder Other Joint Fusion Wrist Ankle Other Joint Resection MTP Other 30

123 of 59 Title: Statistical Analysis Plan (SAP) Table B4a. Physical Assessment Number of Patients Baseline Week 24 Week 48 Vital Signs Mean Standard Deviation Systolic Blood Pressure (mmhg) Range Diastolic Blood Pressure (mmhg) Range Pulse Range Temperature (F) Range Weight (lbs) Range Height (inches) Range 31

124 of 59 Title: Statistical Analysis Plan (SAP) Table B4a - subgroup. Physical Assessment (in patients with 24 weeks followup) Baseline Week 24 Number of Patients Vital Signs Mean Standard Deviation Systolic Blood Pressure (mmhg) Range Diastolic Blood Pressure (mmhg) Range Pulse Range Temperature (F) Range Weight (lbs) Range Height (inches) Range 32

125 of 59 Title: Statistical Analysis Plan (SAP) Table B4b. Laboratory values Laboratory Value (Mean ± SD) Baseline Week 24 Week 48 Number of patients Erythrocyte Sedimentation Rate Minimum 5% Value Median 95% Value Maximum Complete Blood Count White Blood Cells (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Hemoglobin (g/dl) Minimum 5% Value Median 95% Value Maximum Platelets (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Neutrophils (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Creatinine (mg/dl) Minimum 5% Value 33

126 of 59 Title: Statistical Analysis Plan (SAP) Median 95% Value Maximum Liver Function Tests AST Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal ALT Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal Alk Phos Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal Albumin (g/dl) Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal ND = Not Done 34

127 of 59 Title: Statistical Analysis Plan (SAP) Table B4b - subgroup. Laboratory values (in patients with 24 weeks followup) Laboratory Value (Mean ± SD) Baseline Week 24 Number of patients Erythrocyte Sedimentation Rate Minimum 5% Value Median 95% Value Maximum Complete Blood Count White Blood Cells (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Hemoglobin (g/dl) Minimum 5% Value Median 95% Value Maximum Platelets (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Neutrophils (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Creatinine (mg/dl) Minimum 35

128 of 59 Title: Statistical Analysis Plan (SAP) 5% Value Median 95% Value Maximum Liver Function Tests AST Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal ALT Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal Alk Phos Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal Albumin (g/dl) Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal 36

129 of 59 Title: Statistical Analysis Plan (SAP) Table B4c. ECOG Toxicity Data Toxicity Criteria Baseline Week 24 Week 48 Number of patients Leukopenia White Blood Cells Stage 1 (3-3.9 thousand per mcl) Stage 2 (2-2.9 thousand per mcl) Stage 3 (1-1.9 thousand per mcl) Stage 4 (<1.0 thousand per mcl) Neutrophils (Granulocytes) Stage 1 ( thousand per mcl) Stage 2 ( thousand per mcl) Stage 3 ( thousand per mcl) Stage 4 (<0.5 thousand per mcl) Thrombocytopenia Platelets Stage 1 ( thousand per mcl) Stage 2 ( thousand per mcl) Stage 3 ( thousand per mcl) Stage 4 (<25 thousand per mcl) Anemia Hemoglobin Stage 1 (10-13 g/dl) Stage 2 (8-9.9 g/dl) Stage 3 ( g/dl) Stage 4 (<6.5 g/dl) Genitourinary Creatinine Stage 1 ( mg/dl) Stage 2 ( mg/dl) Stage 3 ( mg/dl) Stage 4 (>9.00 mg/dl) Liver AST Stage 1 ( ) Stage 2 ( ) Stage 3 ( ) 37

130 of 59 Title: Statistical Analysis Plan (SAP) Stage 4 (>700) ALT Stage 1 (50-125) Stage 2 ( ) Stage 3 ( ) Stage 4 (>1000) Alk Phos Stage 1 ( ) Stage 2 ( ) Stage 3 ( ) Stage 4 (>3000) Weight Gain Stage 1 (5-9.9%) Stage 2 ( %) Stage 3 (>20%) Weight Loss Stage 1 (5-9.9%) Stage 2 ( %) Stage 3 (>20%) 38

131 of 59 Title: Statistical Analysis Plan (SAP) Table B4c-subgroup. ECOG Toxicity Data (in patients with 24 weeks followup) Toxicity Criteria Baseline Week 24 Number of patients Leukopenia White Blood Cells Stage 1 (3-3.9 thousand per mcl) Stage 2 (2-2.9 thousand per mcl) Stage 3 (1-1.9 thousand per mcl) Stage 4 (<1.0 thousand per mcl) Neutrophils (Granulocytes) Stage 1 ( thousand per mcl) Stage 2 ( thousand per mcl) Stage 3 ( thousand per mcl) Stage 4 (<0.5 thousand per mcl) Thrombocytopenia Platelets Stage 1 ( thousand per mcl) Stage 2 ( thousand per mcl) Stage 3 ( thousand per mcl) Stage 4 (<25 thousand per mcl) Anemia Hemoglobin Stage 1 (10-13 g/dl) Stage 2 (8-9.9 g/dl) Stage 3 ( g/dl) Stage 4 (<6.5 g/dl) Genitourinary Creatinine Stage 1 ( mg/dl) Stage 2 ( mg/dl) Stage 3 ( mg/dl) Stage 4 (>9.00 mg/dl) Liver AST Stage 1 ( ) Stage 2 ( ) 39

132 of 59 Title: Statistical Analysis Plan (SAP) Stage 3 ( ) Stage 4 (>700) ALT Stage 1 (50-125) Stage 2 ( ) Stage 3 ( ) Stage 4 (>1000) Alk Phos Stage 1 ( ) Stage 2 ( ) Stage 3 ( ) Stage 4 (>3000) Weight Gain Stage 1 (5-9.9%) Stage 2 ( %) Stage 3 (>20%) Weight Loss Stage 1 (5-9.9%) Stage 2 ( %) Stage 3 (>20%) 40

133 of 59 Title: Statistical Analysis Plan (SAP) Table B5. Functional Assessments Number of Patients Baseline Week 24 Week 48 Morning Stiffness (in minutes) Mean SD Range Current Pain (VAS, 10=Excruciating) Mean SD Range Patient Global Score (VAS, 10=Abnormal) Mean SD Range Fatigue Scale Score (VAS, 10=Exhausted) Mean SD Range Health Assessment Questionnaire (HAQ) Not adjusted for aids and devices 0 (Best) (Worst) Adjusted for aids and devices 0 (Best) (Worst) Euro-Quol 5 Dimensions (EQ-5D) Mobility No problems in walking about Some problems in walking about Confined to bed Self Care No problems with self care 41

134 of 59 Title: Statistical Analysis Plan (SAP) Some problems with washing or dressing Unable to wash or dress myself Usual Activities No problems with performing usual activities Some problems performing usual activities Unable to perform usual activities Pain/Discomfort No pain or discomfort Moderate pain or discomfort Extreme pain or discomfort Anxiety/Depression Not anxious or depressed Moderately anxious or depressed Extremely anxious or depressed Visual Analog Scale (100=Best Health State Imaginable) Mean SD Range 42

135 of 59 Title: Statistical Analysis Plan (SAP) Table B5-subgroup. Functional Assessments (in patients with 24 weeks followup) Number of Patients Baseline Week 24 Morning Stiffness (in minutes) Mean SD Range Current Pain (VAS, 10=Excruciating) Mean SD Range Patient Global Score (VAS, 10=Abnormal) Mean SD Range Fatigue Scale Score (VAS, 10=Exhausted) Mean SD Range Health Assessment Questionnaire (HAQ) Not adjusted for aids and devices 0 (Best) (Worst) Adjusted for aids and devices 0 (Best) (Worst) Euro-Quol 5 Dimensions (EQ-5D) Mobility No problems in walking about Some problems in walking about Confined to bed Self Care 43

136 of 59 Title: Statistical Analysis Plan (SAP) No problems with self care Some problems with washing or dressing Unable to wash or dress myself Usual Activities No problems with performing usual activities Some problems performing usual activities Unable to perform usual activities Pain/Discomfort No pain or discomfort Moderate pain or discomfort Extreme pain or discomfort Anxiety/Depression Not anxious or depressed Moderately anxious or depressed Extremely anxious or depressed Visual Analog Scale (100=Best Health State Imaginable) Mean SD Range 44

137 of 59 Title: Statistical Analysis Plan (SAP) Table B6. Baseline Medications Study Population (n=) Medication Category Concomitant Methotrexate 10 or 12.5 mg/week (intolerant of higher dose) 15 or 17.5 mg/week 20 or 22.5 mg/week 25 or more mg/week Other Concomitant Medications DMARDs (minocycline or doxycycline ONLY) Folate Steroids (e.g. prednisone oral or injected, hydrocortisone) Osteoporosis drugs (e.g. aledronate, calcitonin) Narcotics and combination painkillers (e.g. oxycodone, pentazocine) NSAIDs (e.g. aspirin, sulindac) Statins (e.g. simvastatin) Neutraceuticals (e.g. glucosamine, SAMe) Other (includes all medications not included in the above lists) 45

138 of 59 Title: Statistical Analysis Plan (SAP) Listing B7. Joint Procedures Study Population (n=) Patient Number Study Week Description R Steroid injection, both wrists 46

139 of 59 Title: Statistical Analysis Plan (SAP) 9.0 Efficacy Shells 47

140 of 59 Title: Statistical Analysis Plan (SAP) Table C1. DAS28 and Components Number of Patients Baseline Week 24 Week 48 DAS28 Mean SD Range Components Tender Joint Count Mean SD Range Swollen Joint Count Mean SD Range Erythrocyte Sedimentation Rate (mm/hr) Mean SD Range Patient Global Health Assessment (mm) Mean SD Range Number achieving DAS

141 of 59 Title: Statistical Analysis Plan (SAP) Table C1-subgroup. DAS28 and Components (in patients with 24 weeks followup) Number of Patients Baseline Week 24 DAS28 Mean SD Range Components Tender Joint Count Mean SD Range Swollen Joint Count Mean SD Range Erythrocyte Sedimentation Rate (mm/hr) Mean SD Range Patient Global Health Assessment (mm) Mean SD Range Number achieving DAS

142 of 59 Title: Statistical Analysis Plan (SAP) 10.0 Safety Shells 50

143 of 59 Title: Statistical Analysis Plan (SAP) Table D1. Total Adverse Events (AEs) and Serious Adverse Events (SAEs) Total Events # Adverse Events # Patients with AEs # Serious Adverse Events # Patients with SAEs 51

144 of 59 Title: Statistical Analysis Plan (SAP) Table D2. Total Adverse Events (AEs) and Serious Adverse Events (SAEs) by Site # SAE # of Patients # AE events # Patients events (% # Patients Site Enrolled (% Total) with AE Total) with SAE 105-VA-Omaha 108-CRRC-Winnipeg 112-VA-St Louis 113-VA-Dallas 114-RAIN-Minneapolis 115-VA-Washington DC 118-VA-Salt Lake City 119-VA Pittsburgh 122-RAIN-Omaha 124-VA-Los Angeles 125-RAIN-Bismarck 127-VA-Philadelphia 128-VA-Minneapolis 129-VA-White River Jct 133-VA-Tampa Total 52

145 of 59 Title: Statistical Analysis Plan (SAP) Table D3. Adverse Events (AEs) by System Organ Class (SOC) and Preferred Term (PT) System Organ Class (SOC) Blood and lymphatic system disorders Blood and lymphatic system disorders Total Ear and labyrinth disorders Ear and labyrinth disorders Total Eye disorders Eye disorders Total Gastrointestinal disorders Gastrointestinal disorders Total General disorders and administration site conditions Preferred Term (PT) Lymphadenopathy Tinnitus Miosis Abdominal pain Abdominal pain lower Abdominal pain upper Abdominal tenderness Colitis ulcerative Constipation Diarrhoea Gastrointestinal pain Gastrointestinal sounds abnormal Gastrointestinal ulcer haemorrhage Gastrooesophageal reflux disease Nausea Stomach discomfort Vomiting Adverse drug reaction Asthenia Chest pain Fatigue Injection site pain Malaise Non-cardiac chest pain Pyrexia Sensation of pressure General disorders and administration site conditions Total Immune system disorders Immune system disorders Total Infections and infestations Hypersensitivity Eye infection Total # Events Total # Patients 53

146 of 59 Title: Statistical Analysis Plan (SAP) Infections and infestations Total Herpes zoster Infection Influenza Nasopharyngitis Sinusitis Upper respiratory tract infection Injury, poisoning and procedural complications Contusion Corneal abrasion Injury, poisoning and procedural complications Total Investigations Investigations Total Metabolism and nutrition disorders Metabolism and nutrition disorders Total Musculoskeletal and connective tissue disorders Musculoskeletal and connective tissue disorders Blood glucose decreased Decreased appetite Arthralgia Arthritis Back pain Myalgia Neck pain Rheumatoid arthritis Musculoskeletal and connective tissue disorders Total Nervous system disorders Nervous system disorders Total Psychiatric disorders Psychiatric disorders Total Renal and urinary disorders Renal and urinary disorders Total Dizziness Headache Paraesthesia Sciatica Depression Insomnia Mood swings Panic attack Suicidal ideation Haematuria Pollakiuria Reproductive system and breast disorders Erectile dysfunction Reproductive system and breast disorders Total 54

147 of 59 Title: Statistical Analysis Plan (SAP) Respiratory, thoracic and mediastinal disorders Respiratory, thoracic and mediastinal disorders Total Skin and subcutaneous tissue disorders Skin and subcutaneous tissue disorders Total Vascular disorders Vascular disorders Total Total Chronic obstructive pulmonary disease Cough Dysphonia Nasal congestion Pharyngolaryngeal pain Pulmonary congestion Respiratory tract congestion Rhinorrhoea Sinus congestion Sneezing Hyperhidrosis Night sweats Rash Flushing 55

148 of 59 Title: Statistical Analysis Plan (SAP) Table D4. Serious Adverse Events (SAEs) by System Organ Class (SOC) and Preferred Term (PT) System Organ Class (SOC) Preferred Term (PT) Total # Events Total # Patients Total 56

149 of 59 Title: Statistical Analysis Plan (SAP) Table D5. Adverse Events (AEs) and Serious Adverse Events (SAEs) per patient Attributable to Study Therapy # Adverse Events or more Total # patients with AEs # of patients with events "Not attributed" # of patients with events "Possibly attributed" # of patients with events "Yes attributed" 57

150 of 59 Title: Statistical Analysis Plan (SAP) Table D6. Serious Adverse Events (SAEs) Attributable to Study Therapy # Serious Adverse Events 0 1 Total # patients with SAEs # of patients with events "Not attributed" # of patients with events "Possibly attributed" # of patients with events "Yes attributed" 58

151 of 59 Title: Statistical Analysis Plan (SAP) Table D7. Adverse Events (AEs) Attributable to Study Therapy (Excludes SAEs) # of patients with # Adverse # of patients with events "Possibly # of patients with Events events "Not attributed" attributed" events "Yes attributed" or more Total # patients with AEs 59

152 2.0 1 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) MAVERIC Study Specific Documents DATE OF ISSUE DOCUMENT DESCRIPTION This document describes the statistical analysis plan for the RACAT randomized clinical trial. AUTHOR: Robert A. Lew, PhD SUBMITTED BY: Robert A. Lew, PhD DOCUMENT/RECORD HISTORY: DATE VERSION RECORD HISTORY 09/09/ Initial 05/17/ Reflecting changes from Protocol V4.0 AUTHORIZATION SIGNATURES NAME FUNCTION TITLE SIGNATURE DATE James O Dell MD, Principal Investigator Mary Brophy, MD, MPH Study Director Robert Lew, PhD Keri Hannagan, MPH Ryan Ferguson, MPH Erika Holmberg, MPH Study Biostatistician Project Manager Associate Center Director for Science and Technology Associate Center Director for Quality 1

153 2.0 2 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) 1. INTRODUCTION STUDY OBJECTIVES INVESTIGATIONAL PLAN OVERALL STUDY DESIGN AND PLAN EFFICACY ENDPOINTS SAFETY VARIABLES NOT DEFINED IN PROTOCOL STATISTICAL METHODS STUDY SUBJECTS DISPOSITION OF SUBJECTS ANALYSIS POPULATIONS: DEMOGRAPHICS AND BASELINE CHARACTERISTICS DEMOGRAPHICS MEDICAL HISTORY AND COMORBIDITIES GENERAL CONSIDERATIONS EFFICACY EVALUATION PRIMARY EFFICACY ANALYSIS SECONDARY EFFICACY ANALYSIS SAFETY EVALUATION MORTALITY ADVERSE EVENTS DETERMINATION OF SAMPLE SIZE MISSING VALUES MULTIPLICITY INTERIM ANALYSES DATA MONITORING COMMITTEE REPORTING OUTPUT LISTS OF TABLES, FIGURES AND LISTINGS SUBJECT DISPOSITION SHELLS BASELINE CHARACTERISTICS SHELLS EFFICACY SHELLS SAFETY SHELLS

154 2.0 3 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) 1. Introduction This document outlines the statistical methods for the display, summary and analysis of data collected in the Department of Veterans Affairs Cooperative Studies s study #551 entitled Rheumatoid Arthritis: Comparison of Active Therapies in Patients with Active Disease Despite Methotrexate Therapy. The purpose of this document is to provide guidelines from which the analysis will proceed. Deviations from these guidelines will be documented. The following documents were used in preparation of this statistical analysis plan (SAP): Clinical Study Protocol CSP #551 Case Report Forms (CRF) from above entitled Protocol MAVERIC SOP: Drafting and Approving Statistical Analysis Plans 2. Study Objectives The primary objective of the study is to determine in subjects with rheumatoid arthritis who have active disease despite methotrexate whether two successful treatment strategies are clinically equivalent as measured by the 48-week change in Disease Activity Score for 28 joints (DAS28). Treatment Strategies: (1) Strategy A (Triple Therapy): the addition of hydroxychloroquine and sulfasalazine to methotrexate, with a switch over at 24 weeks to the addition of etanercept to methotrexate in non-responders, compared to (2) Strategy B (Biologic Therapy): initial therapy with etanercept and methotrexate and switch to the addition of hydroxychloroquine and sulfasalazine to methotrexate in non-responders at 24 weeks. Secondary Objectives: (1) To determine whether radiographic progression at 48 weeks will be equivalent in patients randomized to Strategy A compared to Strategy B. (2) To determine whether functional outcomes are equivalent in patients randomized to Strategy A compared to Strategy B. This is particularly important because the DAS28 does not incorporate functional outcomes. (3) To determine the cost-effectiveness of Strategy A compared to Strategy B. Other Objectives: To identify genetic factors and biomarkers that predict disease progression and success or toxicity of the different strategies, serum, plasma and DNA will be collected and banked for future use. To carry out other descriptive and exploratory analyses; analyses directly related to the primary hypothesis, analyses of other endpoints of interest, and cost analyses. 3. Investigational Plan 3.1. Overall Study Design and Plan This is a double-blind, randomized, 48-week, two-arm trial. The trial will test the hypothesis that the two strategies are equivalent with respect to the 48-week change in DAS28. Patients with rheumatoid arthritis who have active disease despite 15 to 25 mg of methotrexate weekly will be 3

155 2.0 4 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) offered an opportunity to participate. They will be randomized to receive either triple therapy (sulfasalazine and hydroxychloroquine added to their baseline methotrexate), or etanercept 50 mg subcutaneously weekly in addition to baseline methotrexate. Clinical response will be measured using a DAS28, which is a well validated measure of composite clinical disease activity. If, at the end of 24 weeks, patients have not achieved a DAS28 improvement of 1.2 units they will, in a blinded fashion, be switched to the other therapy. Improvement of 1.2 or greater is recognized as clinically meaningful. Thus, patients randomized to receive triple therapy at the onset who fail to achieve a DAS28 improvement of at least 1.2 will receive methotrexate plus etanercept for the second 24 weeks and similarly patients randomized to methotrexate plus etanercept who fail to achieve this level of improvement will receive triple therapy for the second 24 weeks. The timing of the switch (24 weeks) is important since most rheumatologists are unwilling to wait longer than this to obtain clinically meaningful improvement for their patients. Since both groups will be switched if the 24 week DAS28 goal (improvement of 1.2) is not achieved, the blind will be protected. Importantly, for the purpose of the intention to treat analysis of the primary clinical endpoint, outcomes of patients who switch will be attributed to initial treatment assignment. Furthermore, because patients may have a clinical response but still demonstrate radiologic progression, an important component of this trial is to determine whether the two strategies have similar effects on radiologic outcomes at 48 weeks. At the end of 48 weeks, the subject will exit the blinded portion of the trial and the code will be broken so that the clinicians will know what the subject has been receiving. If patients have achieved a DAS28 of 3.2, indicating good clinical response, clinicians will be encouraged to continue the current therapy. Descriptive data will continue to be collected on all these patients in an open fashion until all 600 patients have completed the 48 week portion of the trial. Since RA is a life-long disease this observational period will allow us to collect long term toxicity data as well as important effectiveness (clinical and radiographic), tolerability and economic data. Importantly, the radiographic data on this population will continue to be read in a blinded manner. Study Design Randomization DAS28 Improved 1.2? Primary Outcome: **DAS28 3.2? *SSZ + HCQ *Etanercept **No Yes Yes SSZ + HCQ Etanercept Yes Yes Start 24 Weeks 48 Weeks Open Continuation Open Continuation *All patients continue to receive methotrexate **Patients who switch at 24 weeks remain with their randomization group for analysis of primary outcome In summary, this unique design compares two treatment strategies instead of the traditional trial design where one treatment is compared to another in a static, inflexible way. The design is dynamic and mirrors the way patients are treated in practice, allowing therapy to be changed at 24 4

156 2.0 5 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) weeks if patients have not had clinically significant improvement. This innovative design is critical for the question we are addressing. Patients may respond more rapidly to the addition of etanercept to methotrexate compared to sulfasalazine, hydroxychloroquine, and methotrexate. However, a substantial number of patients will respond to initial therapy with the three drug combination. If this trial shows that the two strategies are equivalent at 48 weeks, a significant percentage of patients will have been spared expensive therapies and health agencies can ethically recommend the strategy of using sulfasalazine, hydroxychloroquine, and methotrexate initially, which is much less expensive. Study procedures related to unblinding the study During the study, the accumulative study data will encode the two treatment strategies as A and B. All analyses by treatment will be blind; that is, the treatments will only be identified as A and B. When the study ends and subsequently the final dataset is locked, then the treatments will be unblinded so that the analysts and the principal investigator can write the primary paper. In the unusual circumstance that the DMC requests the data be unblinded before the study ends, the only MAVERIC staff members to be unblinded will be the 1) statistician responsible for the primary analysis, and 2) as few other persons needed to interpret the results of analyses requested by the DMC. Staff who directly deal with the day-to-day data will not be unblinded. The director of MAVERIC will determine whom to unblind as appropriate to the specific DMC request. A record to file will list those persons and the date they were unblinded. 3.2 Efficacy and Safety Variables Efficacy endpoints The efficacy endpoints for this study are as follows Primary Efficacy endpoints Change in DAS28 at 48 weeks Secondary Efficacy endpoints Proportion of subjects achieving DAS at 48 weeks. Radiographic progression will be assessed by obtaining standard PA x-rays of the hand and wrist and AP x-rays of the feet, which will be graded according to the Van der Heijde-modified Sharp score Functional outcomes will be measured using the Health Assessment Questionnaire (HAQ) Patient self reported quality of life will be measured at baseline and at follow-up using the Health Utilities Index (HUI) and the EQ-5D (Euro QOL). The cost-effectiveness of Strategy A compared to Strategy B Safety Variables While the safety variables for this study are not defined in the protocol, the following will be evaluated in keeping with the VA s mission to protect the safety of our patients. Overall mortality rate at 48 weeks The number of adverse events at 48 weeks and percent of subjects with adverse events 5

157 2.0 6 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Number of serious adverse events at 48 weeks and percent of subjects with serious adverse events. Any unresolved adverse events will be followed for up to 30 days after the end of the study. The study executive committee will determine whether continued follow-up after the end of study is required. 4. Statistical Methods 4.1 Study Subjects Disposition of Subjects The number of subjects screened, enrolled, and completing the study will be presented overall and by center subdivided by demographic categories (i.e., gender, race, etc.) Reasons for the early termination of subjects as noted on Form 23 will be summarized. Subjects that are lost to follow-up will be summarized as one category. The following listings will be provided: A by-subject listing of completion/termination details A by-subject listing of protocol violations and deviations Analysis Populations: Three datasets will be constructed: the ITT analysis set for the definitive analyses, validation analysis sets (primarily the per protocol analysis subset), and the safety analysis dataset. All enrolled subjects included in the subject listings (by study ID number) will be included in the Intention-To-Treat (ITT) analysis provided they were randomized to the study and received at least one dose of study treatment (either Triple or Biologic). The safety analysis population includes all subjects enrolled. A variety of datasets will be used for confirmatory analyses. Foremost among these is the Evaluable Per-Protocol (EPP) analysis. Patients in this analysis subset of the ITT analysis set will have had no major protocol violations and have completed follow up (48 week visit for end point assessment). We will repeat the primary analyses on the EPP subset to confirm or not confirm the results found in the analysis of the ITT patients. Other confirmatory analyses will be carried out on subsets of the ITT patients, stratified by patient characteristics such as age, sex, and disease history. 4.2 Demographics and Baseline Characteristics Demographics Demographic details (e.g., age, sex, ethnicity, race, marital status, and smoking status) will be summarized overall and by center Medical History and Comorbidities The pertinent medical histories (i.e., comorbidities), and history of RA will be summarized overall and by center. Frequencies and percentages will be reported. 6

158 2.0 7 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) 4.3 General Considerations In general, the number of observations, mean, median, standard deviation, minimum, and maximum will be calculated for continuous variables. The number of decimals places will be as follows: minimum and maximum to the same number of decimal places as present in the raw data, mean and median to plus one decimal place, standard deviation to plus two decimal places. Frequencies and percentages will be calculated for categorical data. Baseline is defined as the randomization visit. Missing endpoint data will not be imputed. Missing covariate data will be imputed in selected circumstances as described in Little and Rubin (2002). 1 All planned analyses for the final clinical study report will be performed using a two-sided nominal significance level of Efficacy Evaluation Primary Efficacy Analysis The primary hypothesis compares the 48-week mean change in DAS28 scores between treatment regimens. Equivalence tests comparing two treatment strategies wherein one strategy is postulated to be not worse than the standard are referred to as tests of non-inferiority and therefore typically set up as one-sided tests. All the proposed tests for the primary and secondary hypotheses are tests of non-inferiority since it is expected that initial treatment with etanercept will have the higher success rate. The proposed test is a non-inferiority test, a one-sided equivalence test with an upper bound limit of 0.6. The null hypothesis that the difference between two mean 48-week changes is not zero is rejected (considered equivalent ) when the upper bound on a one-sided 95% confidence interval for the difference between mean changes is 0.6 or less. The mean change is the change in the DAS28 score from baseline to 48 weeks, averaged for each treatment strategy. A clinically meaningful change in the DAS28 score is 1.2. Thus, for the equivalence test, we have set the upper bound to 0.6, one half of 1.2. With a continuous endpoint for the primary outcome variable, the proposed test is a one-sided equivalence test of the null hypothesis that the upper limit on the 95% confidence interval on the difference between the change in two DAS28 scores 0.60, the boundary on equivalence. Using ΔAand ΔB to denote the change in disease activity state for Strategy A and Strategy B respectively, the hypotheses to be tested can be stated as H 0 : ΔB > ΔA H a : ΔB = ΔA In addition to the formal test comparing the upper confidence limit on the difference between change in DAS28 to 0.60, we will carry out a descriptive analysis. We will report the means at 1 Little, Roderick JA, and Rubin, Donald B (2002), Statistical analysis with missing data. John Wiley & Sons (New York; Chichester). 7

159 2.0 8 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) each time point, the change over time, and the difference between 48-week change in DAS28 for biological treatment and triple therapy treatment. We will also report the proportion and number of subjects who switched treatment regimens at week 24. We will compare the two treatment strategies in the primary hypothesis using a one-sided test based on an ANOVA model. We will repeat the descriptive analysis, stratified by 4 groups accounting for treatment switch, with the groups being: (1) triple therapy throughout the entire study, (2) triple therapy switched to etanercept, (3) etanercept throughout the entire study, and (4) etanercept switched to triple therapy. The formal model equation can be expressed as follows: Change in DAS28 at 48 weeks = α + β 1 X 1 + β 2 X 2 + γ X 1 *X 2 Where X 1 is the indicator variable for Triple therapy (1=Triple, 0=Biologic) and X 2 is the indicator variable for a switch in therapy (1= switched, 0=not switched). The parameter β 1 will estimate the independent effect of Triple therapy compared to Biologic therapy, the parameter β 2 will estimate the independent effect of switching therapies, and the parameter γ will estimate the additional effect of switching from Triple therapy to Biologic therapy. In addition, further ANOVA modeling will be undertaken to include all treatment by baseline covariate interactions to assess whether there may be relevant treatment-covariate interactions that merit consideration of subgroup analyses. SAS code for the analysis Proc glm; model change_das = X1 X2; This analysis will be followed by further, confirmatory analyses using multivariable ANCOVA modeling to take account of the effects of baseline covariates on the primary hypotheses. Our analytic strategy is to use PROC MIXED in the SAS statistical package to model the change in DAS28 with biologic treatment compared to triple therapy. This method will estimate the average difference of change in DAS28 48 weeks among those who started on etanercept (biologic) therapy compared to those who started on triple regimen, after adjusting for following baseline covariates: age, sex, prior treatment with sulfasalazine or hydroxychloroquine, duration of treatment with methotrexate, disease duration, baseline ESR, and concurrent prednisone use. The multivariable ANCOVA model can be expressed as follows: Change in DAS28 at 48 weeks = α + β 1 X 1 + β 2 X 2 + γ X 1 *X 2 + [Z] Where [Z] is the vector of baseline covariates adjusted in the model. The mixed model treats the switch variable, X2, and hence the interaction term, X1*X2, as random effects. Sample Size and Statistical Power Considerations A one-sided equivalence comparison test will be used for the primary hypothesis test. This test will compare the 48-week change in DAS28 in each arm of the study. The null hypothesis is that the upper limit on the 95% one-sided confidence interval on the difference between the two 48-week changes in DAS28 exceeds 0.6. Because it is expected that etanercept response will be higher than the triple therapy response the proposed test is one-sided equivalence. The boundary or limit on equivalence is The power calculation is based on a simplification of the ANOVA model. 8

160 2.0 9 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) We obtain the power of the non-inferiority test for the total sample sizes (2N) of 270, 360, and 450 for values of the difference between the two groups, ΔB- ΔA, for values ΔB- ΔA = 0.0, 0.1, 0.2, and 0.3. The usual test considers only ΔB- ΔA = 0, but because a meaningful clinical difference is 1.2 units, we will allow for a small difference; namely ΔB- ΔA = 0.3. POWER CALCULATIONS with the upper bound U = 0.6 with Type I error of 5% Total Sample Group Size N ΔB- ΔA =0.0 ΔB- ΔA = 0.1 ΔB- ΔA = 0.2 ΔB- ΔA = With a type I error of 5% and a total sample size of 450, 225 subjects per study arm, the one-sided comparison of Strategy Biologic with Strategy Triple has 93% power to reject the null hypothesis if ΔB- ΔA = 0.3. Smaller values of Dch increase the power. With ΔB- ΔA = 0.3, the study has 87% power allowing for up to a 20% dropout-unrelated-to-treatment rate that reduces the total sample size to 360 subjects. We designed this equivalence trial by choosing a difference between the two treatments that is small enough that clinicians believe it is not clinically meaningful. Typically such a choice depends on the effect sizes observed in relevant clinical trials. Our equivalence point of 0.60 was based on discussions with the seven rheumatologists on our Planning Committee, selected for their expertise in rheumatoid arthritis and clinical trials. Losses to Follow-up This two-armed randomized clinical trial seeks to enroll a total of 450 subjects. With an intent-totreat analysis (see section 4), the final sample sizes remain 225 per arm because all subjects who stop study medications for reasons related to treatment are regarded as treatment failures (side effects and withdrawal of consent) even though they are no longer on blinded medications at week 48. About 5% of the subjects are expected to stop study medications due to adverse treatment effects before week 48. Among patients receiving the active treatment in the 24-week ARMADA trial, only 2.3% withdrew for adverse reactions attributed to the treatment. This is a rate of 4.6% per year. For this trial these subjects will be regarded as treatment failures. At most another 5% per year will drop out for reasons unrelated to treatment such as a severe episode of an unrelated condition, death or moving away. These subjects will be regarded as study dropouts. Hence, the overall dropout rate for any reason should be significantly less than 8% Secondary Efficacy Analysis (1) The radiologic hypothesis will be tested with a one-sided equivalence test of the null hypothesis that the difference in the two mean Sharp scores for the two treatment strategies is not zero. We will reject the null hypothesis of non-equivalence if the upper limit of the one-sided 95% confidence interval on the difference is 1.4 Sharp score units or less. A consensus of our Planning Committee rheumatologists was that differences in Sharp score of 2.0 units or less are clinically 9

161 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) unremarkable. We shall also use the independent samples t-test. Further analysis will account for baseline Sharp score as a covariate. (2) We will compare the proportion of subjects in each arm of the study who reach the endpoint of DAS28 of 3.2. This target value is a validated measure and represents a very low disease activity (Appendix A). The null hypothesis is that the upper limit on the 95% one-sided confidence interval on the difference between two proportions of patients achieving a DAS28 of 3.2 exceeds 12.5%. Because it is expected that Strategy B response will be higher than the Strategy A response the proposed test is one-sided equivalence test with the boundary or limit on equivalence of 12.5%. With a type 1 error of 5% the study has 90% power allowing for up to an 8% dropout rate. In recent placebo-controlled trials, differences of 20% or more were shown, indicating 12.5% is an appropriate limit. (12-14). See Appendix F for further detail on sample size calculations and statistical analysis. (3) We will assess the level of disability at baseline and after 12 months using the Disability Index of the Health Assessment Questionnaire (HAQ). The primary hypothesis tests the equivalence of the two treatment strategies. In particular, the proposed test is a one-sided equivalence test of the null hypothesis that the difference between the two means is not zero. The two means are the mean 12 month change in HAQ score for patients initially on triple therapy and the mean 12 month change in HAQ score for patients initially on Etanercept. The boundary or limit on equivalence is a difference between the means of 0.2 HAQ score units where the HAQ score ranges from 0 to 3. We will reject the null hypothesis of non-equivalence and accept the alternative hypothesis of equivalence if the upper limit of the one-sided 95% confidence interval on the difference between the means is 0.2 units or less. To test equivalence, the upper limit of the 95% confidence interval for the difference between the means will be compared to 0.2 HAQ score units. To confirm that confounding has not produced a misleading result, the hypotheses will be retested using the mixed effects model multiple regression done with PROC GENMOD in the SAS statistical package. Fixed covariates will include age, sex, and baseline severity of disease. Study site and subject will be random effects. We propose to carry out other descriptive and exploratory analyses; analyses directly related to the primary hypothesis, analyses of other endpoints of interest, and cost analyses. These analyses will, for the most part, be hypothesis-generating. Analyses of this type include, but are not limited to: DAS28 responses at 24 weeks; DAS28 responses 3.2 at 48 weeks; DAS28 responses 3.2 at 48 weeks with switchers counted as failures; proportion of patients achieving an American College of Rheumatology Criteria (ACR20, ACR50, ACR70) response at 24 and 48 weeks; percentages of responders to each therapy after failure to respond at 24 weeks; European League Against Rheumatism (EULAR) response (those that improve by at least 1.2 and have a DAS28 of 5.1); correlation of DAS28 changes with radiographic outcomes; correlations between radiographic and quality of life changes; percent of subjects not progressing radiographically; percent achieving major clinical response; and proportion receiving steroid injections. Other analyses will be confirmatory in nature such as treatment comparisons of the continuous components of the DAS28 and ACR responses. The difference of 12.5% is small compared to differences between treatments observed in randomized clinical trials. In the 2003 ARMADA trial the group of patients receiving 20-mg adalimumab and the groups receiving 40 mg and 80 mg adalimumab differed by 18% and 19% in 10

162 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) terms of the proportion of ACR20 responders after 24 weeks. Similar results were observed for the ACR50 endpoint. In the 2004 trial comparing the costimulation blocker CTLA4Ig to placebo the groups differed by 25% in terms of the proportion of ACR20 responders after 24 weeks. Assuming these proportions have a common standard deviation of 0.5, the corresponding effect sizes are double these differences. In these terms the proposed trial has a limit on effect size of 0.25 compared to trials reported in the literature with effects sizes near Note that the endpoint of the proposed trial is the proportion of patients achieving a specific DAS28 score, rather than an ACR20. Hence, our assertions assume that these distinct endpoints have very similar effect sizes. Effect size is a unit-free measurement Cost analyses The health economics analysis for this trial will evaluate the cost effectiveness of the following strategies for the treatment of RA in patients with active disease refractory to methotrexate therapy. Triple therapy: methotrexate, sulfasalazine, hydroxychloroquine Anti-TNF therapy: etanercept (Enbrel) plus methotrexate As is noted elsewhere in the text of the protocol and this analysis plan, this trial is a comparison of two treatment strategies, not a direct comparison of the two different pharmaceutical therapies. Economic analysis will be central to this evaluation because anti-tnf therapy is more than 10 times as expensive as triple therapy; about $12,000 compared to about $1,000 per year for the VA, but costs for anti-tnf therapy may be considerably higher in other healthcare environments. Given that the existing trial data indicate marginal gains in effectiveness for anti-tnf therapy, compared to triple therapy, the added value of anti-tnf therapy is uncertain. These strategies will be evaluated using standard cost-effectiveness analysis methods. The two strategies will be evaluated relative to each other to identify which therapy is dominated by another (e.g., strategies that are more costly and less effective than available alternatives). In practice, however, most strategies are not dominant or dominated. Typically, more effective treatments are more expensive. In these situations, the cost-effectiveness analysis helps answer the question: is the incremental benefit worth the added cost? Costs will include direct treatment costs and costs associated with lost productivity (indirect costs). For a complete picture of direct treatment costs, it is necessary to collect utilization data from several sources. The analysis is complicated by the fact that that this is a bi-national study with each country having differing costs and practice patterns. To deal with issue we have developed the appropriate data collection and analysis strategy for each country respectively. In the VA, we have developed a method that first uses administrative data, and then uses Medicare data, provider bills, and self-report data to fill in any gaps. The indirect costs will be estimated from respondent s self-report. Alternate methods, appropriate for the context, will be used to capture costs and utilization for patients in the Canadian and non-va arms of the study. Outcomes will be expressed in Quality-Adjusted Life Years (QALYs). In the primary analysis, QALYs will be periodically measured using the Health Utilities Index (HUI) version 3. The HUI-3 is standardized multi-attribute utility instrument. Because many of the European studies have used the EuroQol (EQ-5D) to measure QALYs, we will also use the EQ-5D to facilitate comparisons with these other studies. We will also collect self-reported general health on a 10-cm visual analog scale (VAS). A VAS is usually included as part of the EQ-5D. Data from the VAS will provide an 11

163 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) alternative measure of patient utility and we will use this in a sensitivity analysis. The VAS has also been used in prior studies of RA, and thus including this measure will enable us to compare our results to prior studies. The cost-effectiveness analysis will follow widely accepted and recommended methods. Costs and outcomes will be collected during the study. A societal perspective will be adopted so that all health care costs, including patient-incurred costs (e.g. caregiver costs) and the direct costs of health care, will be considered. The cost-effectiveness analysis will use the patients life span as the time horizon. Trial data will then be used as inputs into decision analytic models to project the trial results to obtain lifetime health care cost and QALYs for each treatment group. Stochastic analysis will be undertaken using cost-effectiveness acceptability curves to express uncertainty around costeffectiveness ratios based on mean costs and outcomes. Cost-effectiveness ratios will be calculated to compare trial participants lifetime cost and outcomes. Existing literature and VA clinical data will be used to identify starting points for the economic model, to determine necessary assumptions, and to identify parameters to be varied in the sensitivity analysis; when these data are ambiguous or incomplete, consensus from the clinical investigators will be sought. A sensitivity analysis will be done to reflect uncertainty of the input parameters and to test the robustness of the model. While cost-effectiveness analysis is the standard for evaluating medical technologies, VA managers and Canadian formulary authorities are usually more concerned with their own narrower perspectives and how a change in therapy will affect their budgets. Thus, we will conduct additional budget impact analyses. First, we will estimate the lifetime model, only considering VA costs. Second, we will estimate the short-term (one and two years) effect on VA expenditures. The later is especially important with respect to VA policy, as VA managers are most concerned about the immediate budget effects. Both of these analyses do not require any additional information; they just involve restricting the model to VA costs and limiting the time horizon. For both of these additional analyses, Canadian and RAIN costs will be converted to VA costs. Additional analyses will also be conducted from the Canadian perspective, converting U.S. costs to Canadian costs Safety Evaluation All safety summaries will be based on the Safety Population described in Section Analysis Populations Mortality The frequency and proportion of all deaths will be presented overall and by center. A by-subject listing will be provided Adverse Events (AEs) AEs will be coded using the Medical Dictionary for Regulatory Activities (MedDRA) ( 9.0) coding dictionary. AE summaries will be based upon the number of subjects reporting AEs and also on the number of events reported. Adverse events will be tabulated (i.e., those events that were not present at the baseline evaluation or worsened in severity following the start of treatment). Any AEs starting after study completion or withdrawal will not be tabulated. 12

164 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) If the AE start date is not a complete date, the following rules will be applied to determine whether the AE will be tabulated or not. If the start date is completely missing: The AE will be tabulated unless the AE stop date is less than the baseline date. If part of the start date is missing, sites were instructed to enter 15 for missing day and 06 for missing month. The number and percentage of subjects with AEs through the end of the study (analysis period) will be summarized by system organ class and preferred term overall. Summaries of AEs by intensity and relation to study device will be presented by system organ class and preferred term overall. For each subject and each AE, the worst intensity/relation recorded will be attributed and used in the intensity/relation summaries. If intensity/relation is missing, the intensity/relation will be queried by data management. The number and percentage of subjects with serious adverse events (SAE) through the end of the study (or analysis period) will be summarized by system organ class and preferred term overall. A by-subject listing of all AEs will be provided. A by-subject listing of all serious AEs will be provided. 4.6 Determination of Sample Size The sample size for the study was based on the primary hypothesis. With a type I error of 5% and a total sample size of 450, 225 subjects per study arm, the one-sided comparison of Strategy Biologic with Strategy Triple has 93% power to reject the null hypothesis if ΔB- ΔA = 0.3. Smaller values of difference in 48-week change in DAS28 increase the power. With ΔB- ΔA = 0.3, the study has 87% power allowing for up to a 20% dropout-unrelated-to-treatment rate that reduces the total sample size to 360 subjects. 4.7 Missing Values Dropouts will not be included in the primary analysis but sensitivity analyses will address their potential to bias the results. Imputation of final outcomes requires questionable assumptions about treatment effects. We will consider a range of such assumptions in the confirmatory analyses. Relatively few dropouts are anticipated, so the expected bias is small. However, the confirmatory secondary analysis including the dropout data censored at the appropriate study visit will also be done to assess potential bias associated with the complete case analysis. A covariate adjusted primary analysis will be done in order to either confirm or not confirm the primary analysis that has no covariates. In either case, the confirmatory analysis will indicate the degree of bias in the complete case analysis. 13

165 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Practical measures will be taken to minimize missing values and losses to follow-up. One prominent conjecture about patients who drop out is that they tend to have poor outcomes. This clearly would violate the assumption of missing-at-random. 4.8 Multiplicity There will be no corrections done for multiple comparisons in this study. 4.9 Interim Analyses There is no interim analysis planned Data Monitoring Committee Data and study progress will be monitored by the study executive committee and by the study steering committee. The Data Monitoring Committee will review the study progress and safety on at least an annual basis with additional meetings and communications as needed. 5.0 Reporting Output All outputs will be produced using SAS 9.1 or a later version (if available at MAVERIC). Other applications (including Microsoft Word and Microsoft Office) will be used to aid in data presentation. Tables, listings, and figures will be arranged in the order that they appear in the textual sections of this analysis plan, i.e., disposition first, followed by baseline summaries, efficacy summaries, and safety summaries. 6.0 Lists of Tables, Figures and Listings All recorded data will be listed following the order and structure of the CRF. In addition, selected derived data used in the analyses will be listed. The listings will include all enrolled subjects (with data relevant to the particular listing) and will be ordered by center, subject, and visit. Lists of Tables, Figures and Listings A. Subject Disposition Charts A1. Description of Subject Disposition a. Overall recruitment by month b. Recruitment by month VA sites c. Recruitment by month RAIN sites d. Recruitment by month Canadian sites Tables A2. Recruitment a. Recruitment, by Site b. Recruitment to Sub-studies Listing A3. Study Medication Discontinuation Listing A4. Study Medication Termination Listing A5. Enrollment Waivers Listing A6. Protocol Deviations Tables A7. Forms Completion Status a. Forms Completion Status, by Form b. Forms Completion Status, by Site Table A8. X-rays Obtained by Sites 14

166 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) B. Baseline and Repeated Measures Table B1. Baseline Demographics Tables B2. Baseline History a. Medical History b. Social History and Allergies Table B3. Baseline History of RA Tables B4. Physical Assessment a. Total Population b. Subset with 24 Weeks Follow-up c. Subset with 48 Weeks Follow-up Tables B5. Laboratory Values with Normals a. Total Population b. Subset with 24 Weeks Follow-up c. Subset with 48 Weeks Follow-up Tables B6. ECOG Toxicity Criteria a. Total Population b. Subset with 24 Weeks Follow-up c. Subset with 48 Weeks Follow-up Tables B7. Functional Assessments a. Total Population b. Subset with 24 Weeks Follow-up c. Subset with 48 Weeks Follow-up Table B8. Baseline Medications Listing B9. Joint Procedures C. Outcome Measures Tables C1. DAS28 and Components a. Total Population b. Subset with 24 Weeks Follow-up c. Subset with 48 Weeks Follow-up D. Adverse Events and Safety Table D1. Total Non-Serious Adverse Events (NAE) and Serious Adverse Events (SAE) Table D2. Total Non-Serious Adverse Events (NAE) and Serious Adverse Events (SAE), by Site Table D3. Non-Serious Adverse Events (NAE), by System Organ Class and Preferred Terms Table D4. Serious Adverse Events (SAE), by System Organ Class and Preferred Terms Table D5. Non-Serious Adverse Events (NAE), by Attributable Class Table D6. Serious Adverse Events (SAE), by Attributable Class 15

167 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) 7.0 Subject Disposition Shells Site Number Site Status CRRC Network 109 Recruiting 131 Recruiting 103 Recruiting 108 Recruiting 111 Recruiting Need 101 Training Need 138 Training Need 140 Training Awaiting 104 IRB RAIN Network 116 Recruiting 120 Recruiting 136 Recruiting 125 Recruiting 114 Recruiting 122 Recruiting Need 121 Training Need 134 Training Awaiting 130 IRB VA Network 119 Recruiting 118 Recruiting 128 Recruiting 133 Recruiting 112 Recruiting 127 Recruiting 129 Recruiting 113 Recruiting* 124 Recruiting* 105 Recruiting 115 Recruiting* 110 Capitated # of patients expected # patients recruited Percent of Expected Recruitment Actual Recruitment Rate Expected Recruitment Rate: 1.11 Expected Recruitment Rate: 0.56 Expected Recruitment Rate:

168 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) 107 Withdrew 123 Withdrew Awaiting 102 IRB Awaiting 106 IRB Awaiting 135 IRB Awaiting 139 IRB 17

169 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table A2b. Recruitment to substudies Study Number Percent Serum Bank Long Term Followup DNA Bank* Listing A3. Study Medication Discontinuation and Termination Subject Number Permanent Discontinuation of One Study Medication Description. Permanent Discontinuation of All Study Medications 18

170 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Listing A4. Protocol Deviations and Enrollment Waivers Subject Number Enrollment Waivers* Full Description Protocol Deviations 19

171 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table A5a. Form Completion Status, by Form Form Number 01A-Eligibility_PartA 01B-Eligibility_PartB 01C-Consent Checklist 02-Demographics 03-History RA 04-Past Medical Hx 05-Social Hx & Allergies 06-Concomitant Medications 07-Physical Exam 08-Joint Count 09-Laboratory 10-Medication Dispensing 11-Medication Compliance 14-Health Services 15-Work Productivity 17-Inpatient Visit 18-Diagnostic Tests 24-Study Med Prescription DAS28-DAS28 PS1-PatientVAS PS2-HAQ PS3-HUI3 PS4-EQ-5D Expected Number of Forms Actual Number of Forms Number of Forms Missing Percent Complete 20

172 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table A5b. Form Completion Status, by Site Actual Number Expected Number of Number of Forms of Forms Forms Missing Site Number Percent Complete Table A6. Accrual of X-Ray Data Baseline Week 24 Expected number of X-rays Actual number of X-rays Percent 21

173 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) 8.0 Baseline Characteristics Shells Table B1. Demographics Study Population (n= ) Age (yrs) Mean±S.D. Range Sex Male Female Ethnicity Hispanic or Latino Not Hispanic or Latino Race White Black or African American Asian Native Hawaiian or other Pacific Islander American Indian or Alaskan Native* First Nations** Other * American Indian or Alaskan Native asked of US patients only ** First Nations asked of Canadian patients only 22

174 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B2a. Baseline Medical History Condition/Diagnosis Head, Ears, Eyes, Nose, Throat Cataracts Seasonal Allergies Other Study Population (n=) Significant History Cardiovascular Myocardial infarction Cerebrovascular accident Unstable angina Coronary artery bypass graft Angioplasty Congestive heart failure Stent Other Respiratory Gastrointestinal Peptic Ulcer Disease Hepatitis Other Renal/GU Stones Other Endocrine Diabetes (Type 1) Diabetes (Type 2) Other Musculoskeletal Septic Arthritis Gout Other 23

175 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Dermatologic Nodules Skin Cancer Other Psychiatric 24

176 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B2b. Social History and Allergies Cigarette Smoking History Never Smoked Former Smoker Current Smoker Study Population (n=) Baseline Ever used cigars, pipes, or chewing tobacco Alcohol Use None One or less drink per week More than one per week Military Service None Active Service Reserves only Allergy to sulfa drugs Allergy to other medications 25

177 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B3. History of RA Study Population (n=) Age at Diagnosis (Mean Range SD) Medication History for RA Sulfasalazine Hydroxychloroquine Anakinra Leflunomide Cyclosporine Minocycline Gold Compound Penacillamine Azathioprine Chloroquine Joint Replacement Knee Hip MCP Shoulder Other Joint Fusion Wrist Ankle Other Joint Resection MTP Other 26

178 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B4a. Physical Assessment Number of Patients Baseline Week 24 Week 48 Vital Signs Mean Standard Deviation Systolic Blood Pressure (mmhg) Range Diastolic Blood Pressure (mmhg) Range Pulse Range Temperature (F) Range Weight (lbs) Range Height (inches) Range 27

179 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B4a - subgroup. Physical Assessment (in patients with 24 weeks followup) Baseline Week 24 Number of Patients Vital Signs Mean Standard Deviation Systolic Blood Pressure (mmhg) Range Diastolic Blood Pressure (mmhg) Range Pulse Range Temperature (F) Range Weight (lbs) Range Height (inches) Range 28

180 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B4b. Laboratory values Laboratory Value (Mean ± SD) Baseline Week 24 Week 48 Number of patients Erythrocyte Sedimentation Rate Minimum 5% Value Median 95% Value Maximum Complete Blood Count White Blood Cells (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Hemoglobin (g/dl) Minimum 5% Value Median 95% Value Maximum Platelets (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Neutrophils (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Creatinine (mg/dl) Minimum 5% Value Median 29

181 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) 95% Value Maximum Liver Function Tests AST Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal ALT Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal Alk Phos Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal Albumin (g/dl) Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal ND = Not Done 30

182 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B4b - subgroup. Laboratory values (in patients with 24 weeks followup) Laboratory Value (Mean ± SD) Baseline Week 24 Number of patients Erythrocyte Sedimentation Rate Minimum 5% Value Median 95% Value Maximum Complete Blood Count White Blood Cells (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Hemoglobin (g/dl) Minimum 5% Value Median 95% Value Maximum Platelets (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Neutrophils (thousand per mcl) Minimum 5% Value Median 95% Value Maximum Creatinine (mg/dl) Minimum 5% Value 31

183 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Median 95% Value Maximum Liver Function Tests AST Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal ALT Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal Alk Phos Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal Albumin (g/dl) Minimum 5% Value Median 95% Value Maximum % Above Upper Limit of Normal 32

184 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B4c. ECOG Toxicity Data Toxicity Criteria Baseline Week 24 Week 48 Number of patients Leukopenia White Blood Cells Stage 1 (3-3.9 thousand per mcl) Stage 2 (2-2.9 thousand per mcl) Stage 3 (1-1.9 thousand per mcl) Stage 4 (<1.0 thousand per mcl) Neutrophils (Granulocytes) Stage 1 ( thousand per mcl) Stage 2 ( thousand per mcl) Stage 3 ( thousand per mcl) Stage 4 (<0.5 thousand per mcl) Thrombocytopenia Platelets Stage 1 ( thousand per mcl) Stage 2 ( thousand per mcl) Stage 3 ( thousand per mcl) Stage 4 (<25 thousand per mcl) Anemia Hemoglobin Stage 1 (10-13 g/dl) Stage 2 (8-9.9 g/dl) Stage 3 ( g/dl) Stage 4 (<6.5 g/dl) Genitourinary Creatinine Stage 1 ( mg/dl) Stage 2 ( mg/dl) Stage 3 ( mg/dl) Stage 4 (>9.00 mg/dl) Liver AST Stage 1 ( ) Stage 2 ( ) Stage 3 ( ) Stage 4 (>700) 33

185 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) ALT Stage 1 (50-125) Stage 2 ( ) Stage 3 ( ) Stage 4 (>1000) Alk Phos Stage 1 ( ) Stage 2 ( ) Stage 3 ( ) Stage 4 (>3000) Weight Gain Stage 1 (5-9.9%) Stage 2 ( %) Stage 3 (>20%) Weight Loss Stage 1 (5-9.9%) Stage 2 ( %) Stage 3 (>20%) 34

186 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B4c-subgroup. ECOG Toxicity Data (in patients with 24 weeks followup) Toxicity Criteria Baseline Week 24 Number of patients Leukopenia White Blood Cells Stage 1 (3-3.9 thousand per mcl) Stage 2 (2-2.9 thousand per mcl) Stage 3 (1-1.9 thousand per mcl) Stage 4 (<1.0 thousand per mcl) Neutrophils (Granulocytes) Stage 1 ( thousand per mcl) Stage 2 ( thousand per mcl) Stage 3 ( thousand per mcl) Stage 4 (<0.5 thousand per mcl) Thrombocytopenia Platelets Stage 1 ( thousand per mcl) Stage 2 ( thousand per mcl) Stage 3 ( thousand per mcl) Stage 4 (<25 thousand per mcl) Anemia Hemoglobin Stage 1 (10-13 g/dl) Stage 2 (8-9.9 g/dl) Stage 3 ( g/dl) Stage 4 (<6.5 g/dl) Genitourinary Creatinine Stage 1 ( mg/dl) Stage 2 ( mg/dl) Stage 3 ( mg/dl) Stage 4 (>9.00 mg/dl) Liver AST Stage 1 ( ) Stage 2 ( ) Stage 3 ( ) 35

187 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Stage 4 (>700) ALT Stage 1 (50-125) Stage 2 ( ) Stage 3 ( ) Stage 4 (>1000) Alk Phos Stage 1 ( ) Stage 2 ( ) Stage 3 ( ) Stage 4 (>3000) Weight Gain Stage 1 (5-9.9%) Stage 2 ( %) Stage 3 (>20%) Weight Loss Stage 1 (5-9.9%) Stage 2 ( %) Stage 3 (>20%) 36

188 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B5. Functional Assessments Number of Patients Baseline Week 24 Week 48 Morning Stiffness (in minutes) Mean SD Range Current Pain (VAS, 10=Excruciating) Mean SD Range Patient Global Score (VAS, 10=Abnormal) Mean SD Range Fatigue Scale Score (VAS, 10=Exhausted) Mean SD Range Health Assessment Questionnaire (HAQ) Not adjusted for aids and devices 0 (Best) (Worst) Adjusted for aids and devices 0 (Best) (Worst) Euro-Quol 5 Dimensions (EQ-5D) Mobility No problems in walking about Some problems in walking about Confined to bed Self Care No problems with self care Some problems with washing or dressing 37

189 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Unable to wash or dress myself Usual Activities No problems with performing usual activities Some problems performing usual activities Unable to perform usual activities Pain/Discomfort No pain or discomfort Moderate pain or discomfort Extreme pain or discomfort Anxiety/Depression Not anxious or depressed Moderately anxious or depressed Extremely anxious or depressed Visual Analog Scale (100=Best Health State Imaginable) Mean SD Range 38

190 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B5-subgroup. Functional Assessments (in patients with 24 weeks followup) Number of Patients Baseline Week 24 Morning Stiffness (in minutes) Mean SD Range Current Pain (VAS, 10=Excruciating) Mean SD Range Patient Global Score (VAS, 10=Abnormal) Mean SD Range Fatigue Scale Score (VAS, 10=Exhausted) Mean SD Range Health Assessment Questionnaire (HAQ) Not adjusted for aids and devices 0 (Best) (Worst) Adjusted for aids and devices 0 (Best) (Worst) Euro-Quol 5 Dimensions (EQ-5D) Mobility No problems in walking about Some problems in walking about Confined to bed Self Care No problems with self care 39

191 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Some problems with washing or dressing Unable to wash or dress myself Usual Activities No problems with performing usual activities Some problems performing usual activities Unable to perform usual activities Pain/Discomfort No pain or discomfort Moderate pain or discomfort Extreme pain or discomfort Anxiety/Depression Not anxious or depressed Moderately anxious or depressed Extremely anxious or depressed Visual Analog Scale (100=Best Health State Imaginable) Mean SD Range 40

192 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table B6. Baseline Medications Study Population (n=) Medication Category Concomitant Methotrexate 10 or 12.5 mg/week (intolerant of higher dose) 15 or 17.5 mg/week 20 or 22.5 mg/week 25 or more mg/week Other Concomitant Medications DMARDs (minocycline or doxycycline ONLY) Folate Steroids (e.g. prednisone oral or injected, hydrocortisone) Osteoporosis drugs (e.g. aledronate, calcitonin) Narcotics and combination painkillers (e.g. oxycodone, pentazocine) NSAIDs (e.g. aspirin, sulindac) Statins (e.g. simvastatin) Neutraceuticals (e.g. glucosamine, SAMe) Other (includes all medications not included in the above lists) Listing B7. Joint Procedures Study Population (n=) Patient Number Study Week Description R Steroid injection, both wrists 41

193 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) 9.0 Efficacy Shells Table C1. DAS28 and Components Number of Patients Baseline Week 24 Week 48 DAS28 Mean SD Range Components Tender Joint Count Mean SD Range Swollen Joint Count Mean SD Range Erythrocyte Sedimentation Rate (mm/hr) Mean SD Range Patient Global Health Assessment (mm) Mean SD Range Number achieving DAS

194 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table C1-subgroup. DAS28 and Components (in patients with 24 weeks followup) Number of Patients Baseline Week 24 DAS28 Mean SD Range Components Tender Joint Count Mean SD Range Swollen Joint Count Mean SD Range Erythrocyte Sedimentation Rate (mm/hr) Mean SD Range Patient Global Health Assessment (mm) Mean SD Range Number achieving DAS

195 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) 10.0 Safety Shells Table D1. Total Adverse Events (AEs) and Serious Adverse Events (SAEs) Total Events # Adverse Events # Patients with AEs # Serious Adverse Events # Patients with SAEs Table D2. Total Adverse Events (AEs) and Serious Adverse Events (SAEs) by Site # SAE Site # of Patients Enrolled # AE events (% Total) # Patients with AE events (% Total) # Patients with SAE 105-VA-Omaha 108-CRRC-Winnipeg 112-VA-St Louis 113-VA-Dallas 114-RAIN-Minneapolis 115-VA-Washington DC 118-VA-Salt Lake City 119-VA Pittsburgh 122-RAIN-Omaha 124-VA-Los Angeles 125-RAIN-Bismarck 127-VA-Philadelphia 128-VA-Minneapolis 129-VA-White River Jct 133-VA-Tampa Total 44

196 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table D3. Adverse Events (AEs) by System Organ Class (SOC) and Preferred Term (PT) System Organ Class (SOC) Blood and lymphatic system disorders Blood and lymphatic system disorders Total Ear and labyrinth disorders Ear and labyrinth disorders Total Eye disorders Eye disorders Total Gastrointestinal disorders Gastrointestinal disorders Total General disorders and administration site conditions Preferred Term (PT) Lymphadenopathy Tinnitus Miosis Abdominal pain Abdominal pain lower Abdominal pain upper Abdominal tenderness Colitis ulcerative Constipation Diarrhoea Gastrointestinal pain Gastrointestinal sounds abnormal Gastrointestinal ulcer haemorrhage Gastrooesophageal reflux disease Nausea Stomach discomfort Vomiting Adverse drug reaction Asthenia Chest pain Fatigue Injection site pain Malaise Non-cardiac chest pain Pyrexia Sensation of pressure General disorders and administration site conditions Total Immune system disorders Immune system disorders Total Infections and infestations Hypersensitivity Eye infection Herpes zoster Total # Events Total # Patients 45

197 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Infections and infestations Total Infection Influenza Nasopharyngitis Sinusitis Upper respiratory tract infection Injury, poisoning and procedural complications Contusion Corneal abrasion Injury, poisoning and procedural complications Total Investigations Investigations Total Metabolism and nutrition disorders Metabolism and nutrition disorders Total Musculoskeletal and connective tissue disorders Musculoskeletal and connective tissue disorders Blood glucose decreased Decreased appetite Arthralgia Arthritis Back pain Myalgia Neck pain Rheumatoid arthritis Musculoskeletal and connective tissue disorders Total Nervous system disorders Nervous system disorders Total Psychiatric disorders Psychiatric disorders Total Renal and urinary disorders Renal and urinary disorders Total Dizziness Headache Paraesthesia Sciatica Depression Insomnia Mood swings Panic attack Suicidal ideation Haematuria Pollakiuria Reproductive system and breast disorders Erectile dysfunction Reproductive system and breast disorders Total 46

198 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Respiratory, thoracic and mediastinal disorders Respiratory, thoracic and mediastinal disorders Total Skin and subcutaneous tissue disorders Skin and subcutaneous tissue disorders Total Vascular disorders Vascular disorders Total Total Chronic obstructive pulmonary disease Cough Dysphonia Nasal congestion Pharyngolaryngeal pain Pulmonary congestion Respiratory tract congestion Rhinorrhoea Sinus congestion Sneezing Hyperhidrosis Night sweats Rash Flushing 47

199 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table D4. Serious Adverse Events (SAEs) by System Organ Class (SOC) and Preferred Term (PT) System Organ Class (SOC) Preferred Term (PT) Total # Events Total # Patients Total Table D5. Adverse Events (AEs) and Serious Adverse Events (SAEs) per patient Attributable to Study Therapy # Adverse Events or more Total # patients with AEs # of patients with events "Not attributed" # of patients with events "Possibly attributed" # of patients with events "Yes attributed" 48

200 of 49 Title: CSP#551 Statistical Analysis Plan (SAP) Table D6. Serious Adverse Events (SAEs) Attributable to Study Therapy # Serious Adverse Events 0 1 Total # patients with SAEs # of patients with events "Not attributed" # of patients with events "Possibly attributed" # of patients with events "Yes attributed" Table D7. Adverse Events (AEs) Attributable to Study Therapy (Excludes SAEs) # of patients with # Adverse # of patients with events "Possibly # of patients with Events events "Not attributed" attributed" events "Yes attributed" or more Total # patients with AEs 49