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: Bonow RO, Maurer G, Lee KL, et al. Myocardial viability and survival in ischemic left ventricular dysfunction. N Engl J Med 2011;364: DOI: /NEJMoa

2 PROTOCOL for Surgical Treatment for Ischemic Heart (STICH) Failure Trial Authors: Document Type: Eric J. Velazquez, Kerry L. Lee, Daniel B. Mark, Christopher M. O Connor, Robert M. Califf, Robert H. Jones Clinical Study Protocol Document Status: Version 01 Release Date: January 18, 2002 Number of pages: 54 Property of DCRI Confidential May not be used, divulged, published or otherwise disclosed without the consent of DCRI

3 Signature Page for NHLBI STICH Surgical Treatment for Ischemic Heart Failure Study Number: 1RO1HL Approved by the following: name signature date name signature date name signature date name signature date DCRI/Confidential Page 2 01/18/2002

4 Signature Page for Investigators STICH Surgical Treatment for Ischemic Heart Failure Study Number: 1RO1HL I have read this protocol and agree to conduct this trial in accordance with all stipulations of the protocol and in accordance with the Declaration of Helsinki. name signature date DCRI/Confidential Page 3 01/18/2002

5 Table of Contents Signature Page for NHLBI 2 Signature Page for Investigators 3 List of Tables 6 List of Abbreviations 7 1. Overview of STICH Trial 8 2. Study Objectives Primary Objectives Secondary Objectives Other Key Parameters 9 3. Background Public Health Burden of Heart Failure Development of Surgical Ventricular Restoration Operation Need for a Randomized Trial of Medical and Surgical Therapies Evaluation of Noninvasive Studies Cost and Quality of Life Evaluation Neurohormonal, Cytokine, and Genetic Studies Significance Investigational Plan Overview Study Population Definition of Patient Strata Study Hypotheses Inclusion and Exclusion Criteria Study Subject Enrollment Screening Initial Evaluation of Stratum Specific Subject Eligibility Initial Evaluation Log Registry Randomization Baseline Evaluation of Randomized Patients Prior to Assigned Treatment Initiation of Study Treatments Investigational Therapies Treatment Groups 22 DCRI/Confidential Page 4 01/18/2002

6 4.5.2 Concomitant Therapies Treatment Compliance Interruption or Discontinuation of Treatment Study Subject Follow-up Assessment Schedule Overview of Study Phases Phase 1 - Patient Selection, Randomization, and Visit Phase 2 - Treatment Initiation and Visit Phase 3 - Follow-up Visits Through to Study End Efficacy Assessments Primary Efficacy Parameters Secondary Efficacy Parameters Other Key Efficacy Parmeters Safety Assessments Quality of Life and Health Status Assessments Economic Data/Hospital Billing Data (U.S. sites only) Echocardiographic Assessment Nuclear Cardiology Assessment End Systolic Volume Assessment by CMR Neurohormonal, Cytokine, and Genetic Testing Study Organization Monitoring Committees Operational Committees Data Management Data Collection Data Forms Manual of Operations STICH Trial Website Database Management and Quality Control Statistical Plan Sample Size Determination - Overview Sample Size and Power Considerations Sample Size - Summary Treatment Crossovers 40 DCRI/Confidential Page 5 01/18/2002

7 6.2. Statistical Analysis Background and Demographic Characteristics Concomitant Therapy Efficacy Evaluation Safety Evaluation Interim Analysis Ethics and Good Clinical Practice Institutional Review Board/Independent Ethics Committee Informed Consent References 52 List of Figures Figure 1. Treatment Regimen 16 Figure 2 Initial Evaluation, Eligibility Confirmation and Stratum Assignment 20 List of Tables Table 1. Visit Schedule 26 DCRI/Confidential Page 6 01/18/2002

8 List of Abbreviations ACE Angiotensin converting enzyme AE Adverse event AICD Automatic implantable cardioverter defibrillator AMI Acute myocardial infarction ARB Angiotensin receptor blocker CABG Coronary artery bypass graft CCS Canadian Cardiovascular Society CI Confidence interval CMR Cardiac magnetic resonance CRF Case Report Form DCRI Duke Clinical Research Institute DDCD Duke Databank for Cardiovascular Diseases DSMB Data and Safety Monitoring Board ECG Electrocardiogram ECHO Echocardiography EF Ejection fraction ESVI End-systolic volume index HF Heart failure ICH International conference on harmonization IEC Independent Ethics Committee IRB Institutional Review Board IVRS Interactive voice response system LVAD Left ventricular assist device LVEF Left ventricular ejection fraction MED Medical strategy of heart failure therapy MI Myocardial infarction mm Hg Millimeters of mercury OR Odds ratio PCI Percutaneous coronary intervention QOL Quality of life SAE Serious Adverse Event SPECT Single photon emission computed tomography SVR Surgical ventricular restoration WHO World Health Organization DCRI/Confidential Page 7 01/18/2002

9 1. Overview of the STICH Trial The Surgical Treatment for Ischemic Heart Failure (STICH) multicenter international randomized trial addresses two specific primary hypotheses in patients with clinical heart failure (HF) and left ventricular (LV) dysfunction who have coronary artery disease (CAD) amenable to surgical revascularization: 1) Coronary artery bypass grafting (CABG) with intensive medical therapy (MED) improves long-term survival compared to MED alone; 2) In patients with anterior LV dysfunction, surgical ventricular restoration (SVR) to a more normal LV size improves survival free of subsequent hospitalization for cardiac cause in comparison to CABG alone. Important secondary endpoints include morbidity, economics, and quality of life. Core laboratories for cardiac magnetic resonance (CMR), echocardiography (ECHO), neurohormonal/cytokine/genetic (NCG), and radionuclide (RN) studies will insure consistent testing practices and standardization of data necessary to identify eligible patients and to address specific questions related to the primary hypotheses. Over three years, 50 clinical sites will recruit 2,800 consenting patients with HF, LV ejection fraction (EF).35, and CAD amenable to CABG. These patients first will be characterized by angina intensity or presence of left main coronary stenosis as appropriate for only surgical therapy or either medical or surgical therapy. All patients will be evaluated further for appropriateness of SVR indicated by an end-systolic volume index (ESVI) 60 ml/m 2 and akinesia 35% of the anterior LV wall. The 600 patients estimated to be eligible for SVR but ineligible for randomization to medical therapy will be evenly randomized to CABG with or without SVR. Of the 2,200 consenting patients eligible for medical or surgical therapy, the 1,600 not SVR eligible will be evenly randomized between MED only and MED with CABG. The remaining 600 patients also eligible for SVR will be randomized between three treatments of MED only, or MED + CABG, or MED + CABG + SVR. Registries of clinical information will be maintained on eligible patients who decline trial entry. At fourmonth intervals for a minimum of three years, all randomized patients will be followed by a clinic visit and registry patients will be followed by telephone. Appropriate subgroups of randomized patients will have core laboratory studies repeated at specified follow-up intervals. In the patients randomized to MED only or to CABG with MED, CABG is hypothesized to demonstrate a 20% reduction in the primary endpoint of all-cause death with an 89% power from the projected 25% threeyear mortality for MED. In the SVR-eligible patients, CABG + SVR is hypothesized to show a 20% advantage with 90% power in the endpoint of survival free of hospitalization for cardiac cause projected to be 50% at three years in patients receiving CABG without SVR. Definition of efficacy of potential therapies and their mechanisms of benefit by the STICH Trial is certain to inform future choice of therapy and thereby extend and improve the quality of life of millions of patients who now suffer from ischemic HF. 2. Study Objectives 2.1 Primary Objectives 1. To test the hypothesis that improvement in myocardial perfusion by CABG combined with intensive medical therapy (MED) improves long-term survival compared to MED alone in patients with HF, LV dysfunction, and CAD amenable to CABG. 2. To test the hypothesis that in patients with anterior LV akinesia, LV shape and size optimization by SVR combined with CABG and MED improves long-term survival free of cardiac hospitalization compared to CABG and MED without SVR. DCRI/Confidential Page 8 01/18/2002

10 2.2 Secondary Objectives 1. To assess the clinical utility of the measurements by CMR of LV shape, size and function for predicting the benefit of a specific treatment strategy. 2. To assess the clinical utility of measurements by nuclear cardiology testing of myocardial ischemia, viability, and LV size and function for predicting preferential benefit associated with a specific treatment strategy. 3. To determine if the use of baseline ECHO derived assessments of LV systolic and diastolic function, hemodynamics, valvular regurgitation identify subgroups that derive either substantial benefit or harm from surgical intervention and how CABG compared to MED and CABG + SVR compared to CABG differentially effects these parameters over time. 4. To test the hypothesis that levels of neurohormonal and pro-inflammatory cytokine activation will be useful in identifying that group of patients who will most likely benefit from either CABG and/or CABG + SVR. 5. To compare health-related quality of life, total medical costs, and incremental cost effectiveness between CABG and MED and between CABG with or without SVR. 2.3 Other Key Parameters Mortality: 30-day all-cause; cardiovascular; all-cause death or LVAD or heart transplantation. Morbidity: All-cause hospitalization; days of all-cause hospitalization; hospitalization for HF; days of hospitalization for HF; emergency department visits for HF; AICD countershock. Cardiac Procedures: PCI; any cardiac surgery; LVAD; heart transplantation; AICD. Cost: total medical costs; cost-effectiveness. Functional Capacity: quality of life; six-minute walk distance (absolute and change); treadmill duration (absolute and change). Physiologic: LVEF (absolute and change); ESVI (absolute and change); neurohormonal levels (absolute and change); cytokine levels (absolute and change). 3. Background 3.1 Public Health Burden of Heart Failure American Heart Association statistics suggest that HF affects 4.7 million patients in the United States and is responsible for approximately one million hospitalizations and 300,000 deaths annually. 1 The total annual costs associated with this disorder have been estimated to exceed $22 billion. 2 CAD is the cause of HF in the majority of patients, and HF is the only mode of CAD presentation associated with increasing incidence and mortality. 3 Ischemic HF may develop after a known myocardial infarction or through a process of chronic ischemia that leads to cardiac dilatation, cellular hypertrophy, apoptosis, and extracellular matrix remodeling. 4 The increasing prevalence of HF largely derives from enhanced survival from improved therapies for other manifestations of CAD, especially acute myocardial infarction. However, survival is usually accompanied by some degree of myocardial injury that is the basis for development of HF. DCRI/Confidential Page 9 01/18/2002

11 3.2 Development of Surgical Ventricular Restoration Operation Reimer and Jennings showed that experimental coronary occlusion results in a wavefront of cell death moving from the subendocardial zone across the wall to involve progressively more of the transmural thickness of the ischemic zone. 5 Early reperfusion salvages subepicardial but not subendocardial myocardium. The decline in prevalence of thin-walled LV aneurysm that followed introduction of aggressive initial management of acute coronary syndromes suggests that the ischemic process also can be interrupted in patients before it reaches the transmural stage. With later healing of infarcted myocardium, scarring is maximal in the subendocardial region and is interlaced with normal myocardium in diminishing amounts toward the epicardial surface. Ventricular wall or chamber imaging commonly shows an akinetic zone that gradually blends with myocardium with increasingly normal function in contrast to the dyskinetic region with a discrete neck typical of a left ventricular aneurysm. LV ventricular aneurysmectomy appears to reverse HF by lowering LV wall stress, but these linear amputations of dyskinetic scar commonly deform the LV cavity into a box-like shape. Intra-cavitary reconstruction techniques have been developed for repairing defects left by aneurysm resection that reduced LV cavity size. 6 This surgical ventricular restoration (SVR) strategy has more recently been applied not only to patients with dyskinetic scar but also to those with only akinetic myocardial segments. 7 At the time of cardiac operation, the epicardium of these akinetic zones may appear normal and palpable thinning is often minimal in the arrested, decompressed heart. This appearance derives from preservation of a rim of normal myocardium covering the myocardial fibrosis and contrasts with the leather-like appearance and thinness typical of a LV aneurysm. Unlike the LV aneurysmectomy that removed myocardial scar or the Batista operation that reduced LV size by indiscriminate removal of portions of the LV wall, the SVR operation mechanically decreases the circumference of the zone of endocardial scar through an incision in normal epicardium. The surgical repair uses the intrinsic scar or an extrinsic patch to absorb excess linear wall tension from the adjacent myocardium. Decrease in wall stress appears to reduce the tendency for continued gradual expansion of the akinetic zone. The endocardial repair acutely decreases LV size, thereby acutely enhancing function in myocardial regions remote from the repair. 8 A recent registry report of 439 patients in 11 centers found a 5.1% operative mortality and 88% 18-month survival for patients who underwent CABG and SVR suggesting that this surgical strategy is safe and at least equivalent to CABG alone. 9 This operation is now sufficiently mature to justify this proposed randomized trial to evaluate whether appropriately performed SVR truly adds value to CABG in HF patients with regional dysfunction Need for a Randomized Trial of Medical and Surgical Therapies Along the broad spectrum of severity of ischemic HF, specific clinical information, such as severe angina or left main coronary artery stenosis, may clearly indicate the need for surgical therapy for some patients. However, a large number of patients fall into a gray zone without clear evidence for benefit from either medical or surgical therapy. For these patients, evidence supporting choice between therapies was never strong and has only been confused by recent studies showing improved outcomes with both therapies. Patients for whom equipoise of anticipated benefit now exists between modern medical and surgical therapy represent the broad population who are appropriate candidates for a randomized trial to provide the context for assessing the value of three therapeutic strategies: 1) MED alone; 2) MED + CABG; and 3) MED + CABG + SVR. No randomized trial has ever directly compared long-term benefits of surgical and medical treatment of patients with ischemic HF. In the developmental era of CABG, high surgical mortality rates were DCRI/Confidential Page 10 01/18/2002

12 observed in patients with HF. Therefore, initial randomized trials comparing CABG to medical treatment conducted from 1972 to 1978 excluded most of these patients from participation. In a subgroup of 160 CASS Trial patients with LVEF <.50, ten-year survival was 61% in the 82 medically-treated patients and 79% in the 78 patients who underwent CABG (p =.01). 11 This survival advantage of CABG was not related to the presence or severity of HF or angina symptoms. A meta-analysis by Yusuf 12 combined individual patient data from the CASS Trial with those enrolled in the six other early randomized trials. Only 191 (7.2%) of the 2,649 total patients had an EF <.40, and only 106 (4.0%) of these patients who were primarily symptomatic with angina also had HF symptoms. CABG improved survival among all patients with proximal LAD stenosis, three-vessel, or left main coronary artery disease regardless of LV function. In these patients with a survival benefit from CABG, a low EF increased the absolute benefit but did not change the relative benefit of CABG. A literature search of 326 published reports on results of CABG in patients with HF or LV dysfunction identified three well-designed cohort studies. 13 Mortality benefit of CABG over medical therapy was 10 and 20 lives per 100 patients at three years in two of the three studies and 29 lives per 100 patients at five years in the third study. Despite this apparent benefit, data from the Duke Databank for Cardiovascular Diseases (DDCD) show that more than 71% of patients with characteristics that would qualify for inclusion in the STICH Trial receive medical therapy, 22% receive CABG, and 7% receive PCI. Moreover, this distribution of treatments has remained constant over the past decade at an institution with an aggressive use of myocardial revascularization procedures. Perhaps CABG is not more widely used in patients with coronary angiograms that suggest survival benefit because of concern for postoperative complications, such as stroke, that detract from its survival advantage. In addition, the general medical community often over estimates surgical risks and medically manages the high-risk patients without thoroughly characterizing the presence and extent of CAD. The therapy called medical treatment in all randomized trials and most observational comparisons with CABG really only reflected the natural history of CAD since it rarely included drugs now known to be life prolonging, such as ACE inhibitors, beta blockers, lipid lowering, and anti-platelet agents. Refinement of operative and postoperative surgical management also has steadily improved CABG results over the past two decades. The paucity of modern data available to clinicians who must daily make high-risk management decisions in patients with HF emphasizes the need for a properly-designed randomized comparison of these therapies commonly used in clinical practice. 3.4 Evaluation of Noninvasive Studies Increasingly over the past three decades, information describing cardiac structure, perfusion, hemodynamics, and metabolism obtained from noninvasive cardiac imaging studies has been used to guide management decisions for patients with HF. Although this anatomic and physiologic information often adds value to clinical care, an accepted strategy has not evolved that tailors the testing sequence to specific presenting features of individual patients to efficiently identify the treatment strategy most likely to improve outcomes. Consensus on proper use of cardiac imaging studies has been hindered by absence of clinical studies that objectively evaluate the independent treatment-related prognostic information of these tests obtained using standardized methods on patients for whom the test has minimal influence on choice of therapy. In the absence of high-quality information, use of these tests by clinicians is commonly guided by mechanistic studies that use endpoints considered as surrogates for survival performed with a study design that does not control for test bias in treatment selection. For example, the widespread clinical use of myocardial viability studies to identify HF patients who would or would not benefit from myocardial revascularization is largely DCRI/Confidential Page 11 01/18/2002

13 based on more than 100 reports using improvement in EF after treatment as the study endpoint. No randomized and only ten observational studies are published describing results in 762 patients that compare the power of myocardial viability measurements for predicting patient survival after CABG or medical therapy In these studies CABG was associated with better survival than medical therapy in patients with viable but not in patients without viable myocardium. However, definition of the criteria used to define subgroups were derived on the same population used to assess outcome in these studies. Moreover, subgrouping these patients by treatment received insured that important baseline differences that influenced choice of treatment also influenced the observed outcome. The lack of survival benefit demonstrated for CABG in patients categorized as without myocardial viability derived from a low mortality in medicallytreated patients also categorized as without myocardial viability, and survival after CABG was not positively or negatively influenced by the preoperative amount of myocardial viability measured. The common clinical practice of not offering CABG to patients with LV dysfunction in regions found to be non-viable on noninvasive studies is not justified by published data. These criticisms of the quality of available information on the appropriate use of myocardial viability testing are equally true of all other noninvasive cardiac imaging data used to guide selection of therapy in patients with CAD. The optimal strategy for evaluation of these noninvasive tests would be to use multivariable modeling of all parameters of independently predictive test information to describe the total test value by a single continuous index. In this form, a myocardial viability index could readily be contextualized with all other relevant clinical and other test information to quantitate any survival advantage the myocardial viability index added by properly guiding the choice of the most effective HF therapy. The STICH Trial uses core laboratories for cardiac magnetic resonance (CMR), echocardiography (ECHO), and radionuclide (RN) studies to insure standardization of testing practices and of data analysis for operational use of these studies in the STICH Trial. These three core laboratories also will address specific questions that use anatomic or physiologic information from these noninvasive cardiac studies to elucidate mechanisms responsible for differences observed in clinical endpoints among the randomized treatment groups. This structured acquisition of cardiac imaging studies in a population with treatment assigned by randomization represents a unique opportunity to assess the individual value of these tests for guiding treatment selection. 3.5 Cost and Quality of Life Evaluation Any new therapeutic strategy that improves the health of HF patients also is likely to reduce a portion of the $22 billion costs of this condition by reducing hospitalizations and related expensive care. However, broader use of surgical therapies would increase initial cost, and the resulting economic change represents a complex issue measured both in terms of cost per added quality-adjusted life year and in total medical care cost to society. This study proposes careful measurement of resource use patterns and associated medical care costs to compare the randomized treatment strategies addressed in both primary study hypotheses. Additional comparisons will examine major subgroup effects. If the primary hypotheses are established for the two surgical strategies compared in this trial, cost effectiveness analyses will define whether these therapies are economically attractive relative to standard benchmarks. HF adversely impacts patient functional status and health-related quality of life (QOL). Modern medical therapy for HF modestly improves QOL, but a more aggressive approach with the surgical therapies being studied in this trial may produce even larger improvements. Careful serial measurements of QOL using wellvalidated instruments is crucial in contextualizing the impact of any invasive therapy and represent important secondary endpoints for the STICH Trial. DCRI/Confidential Page 12 01/18/2002

14 3.6 Neurohormonal, Cytokine, and Genetic Studies The progressive cardiac dilatation and diminished LV function that occur after myocardial damage are associated with over-expression of a variety of endogenous peptides that include: 1) neurohormonal mediators (eg. norepinephrine and endothelin), pro-inflammatory cytokines (eg. TNF and IL-6) and natriuretic peptides (eg. ANP and BNP). 4 These endogenous peptides appear to hasten progressive worsening of HF by cardiotoxic actions that may induce apoptosis and maladaptive remodeling of both cardiac myocytes and the extracellular matrix. Therapeutic strategies that attenuate production (ACE inhibition) or inhibit biologic activity (beta blockade) of a neurohormonal peptide have proven beneficial in the management of patients with HF. However, the ability of surgical interventions to attenuate neurohormonal and cytokine activation more than is achieved with MED alone has not been evaluated. Since myocardial stretch and ischemia may enhance induction and activation of these neurohormonal and cytokine pathways, the STICH Trial will test the hypothesis that rapid reduction of wall tension through the combination of CABG and SVR will be the most effective treatment for reducing peptide expression and that CABG with MED is associated with better clinical outcomes than MED alone. Conversely, high baseline levels of neurohormonal/cytokine peptides may identify a group of patients with irreversible extracellular matrix remodeling or significant apoptosis for whom CABG or CABG with SVR proves less effective than MED alone. Thus, measures of endogenous peptides may help to identify patient groups who will respond in a specific manner to all three randomized treatment strategies evaluated in the STICH Trial. Four recent observations suggest that outcome in patients with HF may be related to variations in genetic background: 1) dilated cardiomyopathy is often familial; 2) first-degree relatives of patients with HF have a high rate of abnormal echocardiograms; 24 3) specific loci are associated with adult-onset autosomal dominant dilated cardiomyopathy; 25 and 4) mutations affecting cytoskeletal proteins are associated with the development of HF. 26 Genetic determinants that might affect the progression of disease in individual patients appear to consist of either polymorphisms or allelic mutations that alter expression of proteins important in the pathophysiology of HF. Many of these genetic variations are directly related to alterations in the activity of the neurohormonal and/or pro-inflammatory pathways. The STICH Trial will provide an opportunity to evaluate the role of genotype determination for predicting the efficacy of subsequent therapy. Most of the polymorphisms that will be evaluated affect the constitutive and activated expression of neurohormonal/cytokine peptides that may be favorably altered by mechanical intervention directed at normalizing LV mechanics and myocardial perfusion. 3.7 Significance Should We Do CABG in Patients with Heart Failure? The design of the STICH Trial excludes only patients for whom MED is the only reasonable therapeutic alternative, those with coronary anatomy best suited for revascularization by PCI, and those who are heart transplant candidates. Therefore, results of the STICH Trial will be generalizable to almost all of the large number of patients with CAD, HF, and systolic LV dysfunction in this country. If surgical therapies prove not to significantly improve survival, MED without intensive evaluation for CAD may be the preferred strategy of care to be pursued in the more than 550,000 patients who first present with HF of unexplained etiology each year in this country. 4 A more aggressive search for the 70% of patients with HF and LV dysfunction who have CAD as the predominant etiology could await demonstrated failure of the MED strategy indicated by development of overt ischemic symptoms. Conversely, if surgical therapy is shown to have a survival advantage over medical therapy in the STICH Trial, early aggressive evaluation of CAD as a potentially correctable etiology of new onset HF would be the preferred strategy. Clinical testing then would become more widely employed to identify patients likely to benefit from CABG and those eligible for SVR. DCRI/Confidential Page 13 01/18/2002

15 When Do We Need Anatomic and Physiologic Studies to Guide the Choice of Therapy in HF Patients with CAD? This trial will relate continuous quantitative relationships between physiologic measurements of the magnitude of myocardial ischemia and viability and LV size to the outcome associated with the three treatment strategies. The resulting information is likely to decrease the cost and complexity of evaluation of patients with ischemic HF by eliminating use of tests that do not provide independent information. This trial will establish whether measurements of neurohormonal and cytokine levels and genetic profiling are useful for directing patient management decisions and for monitoring the effectiveness of therapy. Information obtained from physiologic studies in the STICH Trial will permit refinement of the optimal approach to efficiently evaluate and select the treatment strategy most likely to be effective for the many patients with CAD, HF, and LV dysfunction in this country. Does SVR Need to be Added to CABG in Patients with Regional Dysfunction and What Magnitude of Dysfunction Must be Present to Justify This Additional Surgery? This trial is designed to answer the question of whether mechanically decreasing LV size by SVR arrests or reverses LV enlargement and thereby enhances survival without hospitalization more than relief of chronic ischemia by CABG without SVR. The SVR operation is gaining popularity because promising observational reports confirm its safety and suggest its efficacy. Randomized comparison of CABG with SVR versus CABG alone will definitively test the superiority of adding SVR to CABG. In the event the two operations result in equivalent survival free of cardiac hospitalization, CABG will remain the dominant operation. However, if the addition of SVR to CABG in patients with akinesia of the anterior wall offers superior clinical outcomes, an important new therapeutic option will be added to the HF armamentarium. Moreover, information from CMR measurements of LV volume made before and after treatment in SVR-eligible patients is likely to guide optimal use of SVR and also provide a noninvasive method to assess the quality of operative result. 4. Investigational Plan 4.1. Overview The STICH trial is a three-arm randomized clinical trial that will enroll 2,800 patients with Class II-IV ischemic heart failure, left ventricular ejection fraction < 0.35, and CAD amenable to coronary artery bypass grafting (CABG). The three treatments include intensive, evidence-based medical therapy for heart failure (MED), MED plus revascularization with CABG, and MED plus CABG combined with surgical ventricular restoration (SVR). After eligible patients give informed consent, site personnel will use a 24-hour toll free telephone number with the interactive voice response system (IVRS) to initiate the randomization process. In addition to the baseline case report form (CRF), a Quality of Life (QOL) questionnaire will be administered by the coordinator at each site. Baseline studies will be completed and the randomized therapy will be initiated within two weeks. Randomized patients will be followed at clinic visits at four-month intervals for the duration of the trial. Follow-up echoes, treadmills, gated SPECTs, CMRs and neurohormonal/cytokine peptide collections will be performed at scheduled intervals throughout the remainder of the trial. Personnel from the Economics and Quality of Life (EQOL) Core Laboratory at the DCRI will perform QOL interviews by telephone at four months, one, two, and three years. In addition, EQOL Core Laboratory staff will collect hospital bills for all DCRI/Confidential Page 14 01/18/2002

16 hospitalizations and major outpatient tests and therapies for the length of the study. Follow-up will be a minimum of three years, an average of 4.5 years and a maximum of six years. An initial screening log will be maintained on patients who appear eligible to enter the STICH trial but for whom responsible physicians have selected a specific therapy. These patients will be followed annually using the National Death Index. Eligible patients who do not consent to enter the STICH trial will be approached for consent to enter the registry. Those consenting will have baseline information captured on an initial evaluation form in a similar format used for randomized patients. No baseline or follow-up testing will be required for registry patients. All registry patients will be followed by telephone contact at four-month intervals throughout the trial. 4.2 Study Population The patient population will consist of patients who have symptomatic HF defined as NYHA HF Class II-IV, an LVEF of.35, and coronary artery disease suitable for revascularization. A total of 2,800 patients are to be included in this study will be segregated into three strata depending on their MED and SVR eligibility. Upon randomization, 1300 patients will be allocated to receive CABG, 1000 allocated to MED and 500 patients to CABG + SVR Definition of Patient Strata The design will accommodate the fact that not all patients enrolled will be eligible for all three therapeutic approaches. Patients who meet trial eligibility criteria will first be stratified as eligible or ineligible for MED on the basis of CCS III angina or 50% left main coronary artery stenosis. Patients will then be further stratified based on eligibility for the SVR procedure as determined by end-systolic volume index (ESVI) 60 ml/m 2 and anterior akinesia or dyskinesia. Patients eligible for either MED or CABG but not eligible for the SVR procedure (Stratum A) will be randomized in equal proportions to MED vs CABG. Patients eligible for all three therapies (Stratum B) will be randomized in equal proportions to MED, CABG, and CABG plus SVR. Patients whose angina or coronary artery disease severity make them eligible for CABG and SVR but not appropriate for medical therapy alone (Stratum C) will be randomized in equal proportions to CABG vs. CABG + SVR. The allocation of patients to treatment within each Stratum are illustrated in Figure 1. DCRI/Confidential Page 15 01/18/2002

17 FIGURE 1. Distribution of STICH Patients Among Randomization Strata CAD EF.35 HF SVR Eligible? 2,200 Pts no CCS Class III angina yes SVR no Left main 50% stenosis Eligible? Not in Trial no yes yes R 1,600 Pts R R Stratum A 600 Pts Stratum B 600 Pts Stratum C 1 MED 2 CABG 3 MED 4 CABG 5 CABG + SVR 6 CABG 7 CABG + SVR 800 Pts 800 Pts 200 Pts 200 Pts 200 Pts 300 Pts 300 Pts Stratum A Stratum B Stratum C Primary Hypothesis Analysis Groups Patients Coronary revascularization MED/CABG 1 + 3/ /1000 LV restoration CABG/CABG + SVR 4 + 6/ /500 DCRI/Confidential Page 16 01/18/2002

18 4.2.2 Study Hypotheses There are two primary study hypotheses. Coronary revascularization hypothesis: Improvement in myocardial perfusion by CABG combined with intensive medical therapy (MED) improves long-term survival compared to MED alone. LV restoration hypothesis: In patients with anterior LV akinesia, LV shape and size optimization by SVR combined with CABG and MED improves long-term survival free of cardiac hospitalization compared to CABG and MED without SVR. The coronary revascularization hypothesis will be tested by comparing MED versus CABG in patients in Strata A and B. That is, MED patients, consisting of groups 1 and 3 in Figure 1 above, will be compared to CABG patients, consisting of groups 2 and 4, with respect to the primary endpoint of all-cause mortality. The LV restoration hypothesis will be tested by comparing CABG patients (without SVR) versus CABG + SVR patients in Strata B and C. That is, the CABG patients consisting of groups 4 and 6 will be compared to CABG + SVR patients, consisting of groups 5 and 7, with respect to the composite endpoint of mortality (all cause) or hospitalization for cardiac reasons Inclusion and Exclusion Criteria Inclusion Criteria The following patients may qualify for inclusion in the study: Men Women who are not of childbearing potential. Aged 18 years or above Who have symptomatic HF defined as NYHA HF Class II-IV within 3 months of enrollment. Who have evidence of left ventricular systolic dysfunction within 3 months of trial entry Who have coronary artery disease suitable for revascularization Left Ventricular Systolic Dysfunction A LVEF <.35 by at least one of the following modalities will be considered sufficient evidence of left ventricular systolic dysfunction to qualify for randomization: 1. Echocardiography: performed at the investigative site and confirmed by the echocardiography core laboratory. 2. Gated SPECT ventriculogram: performed and interpreted at the investigative site. 3. Contrast ventriculogram: performed and interpreted at the investigative site. 4. CMR ventriculogram: performed and interpreted at the investigative site. DCRI/Confidential Page 17 01/18/2002

19 Exclusion Criteria At the time of randomization, none of the following may exist: Failure to provide informed consent Valvular heart disease clearly indicating the need for valve repair or replacement Cardiogenic shock (within 72 hours of randomization) as defined by the need for intra-aortic balloon support or the requirement for intravenous inotropic support Plan for percutaneous intervention of coronary artery disease History of acute myocardial infarction within 30 days History of more than one prior cardiac operation Non-cardiac illness with a life expectancy of less than 3 years Non-cardiac illness imposing substantial operative mortality Refractory potentially lethal ventricular arrhythmia Previous major organ (e.g., lung, liver, heart, kidney) transplantation Other conditions/circumstances likely to lead to poor treatment adherence (e.g., history of poor compliance, alcohol or drug dependency, psychiatric illness, no fixed abode) Current participation in another clinical trial in which a patient is taking an investigational drug or receiving an investigational medical device. MED Therapy Ineligibility Criteria Left main coronary artery disease as defined by an intraluminal stenosis of 50% or greater Canadian Class III angina or greater (angina markedly limiting ordinary activity) SVR Eligibility Criteria ESVI > 60 ml/m2 Akinesia > 35% of the anterior LV wall Absence of combined anterior and inferior akinesia Absence of diffusely diseased coronary arteries not amenable to CABG supplying both the right coronary artery and left circumflex coronary artery territories. 4.3 Study Subject Enrollment Screening Screening of patients referred to HF clinics must address both new and established patients and those with and without known CAD. In these patients, screening will commonly begin with determination of clinical eligibility with attention to assessment of coronary anatomy and left ventricular function as information becomes available. Patients referred for diagnostic noninvasive testing or cardiac catheterization will usually be most efficiently screened first by CAD or LVEF criteria. The subset meeting these entry criteria will then be secondarily evaluated by clinical eligibility criteria. Baseline STICH Trial protocol tests that have been performed for clinical reasons in HF patients with CAD in compliance with the STICH core laboratory DCRI/Confidential Page 18 01/18/2002

20 protocols will be used for defining trial eligibility or meeting baseline testing requirements, thereby decreasing inconvenience to the patient and unnecessary cost to the STICH Trial from duplicate testing. Regardless of the specific sequence of information acquisition and after the screening process identifies a patient who appears to meet STICH Trial entry criteria with a high probability, an initial evaluation form is used to document information needed to confirm eligibility of the patient for randomization Initial Evaluation of Stratum Specific Subject Eligibility Patients screened as eligible for initial evaluation for STICH Trial entry will have clinical information available from the initial work-up and coronary anatomy information from arteriography. The best available and most recently performed LV assessment by either contrast, gated SPECT, or CMR ventriculogram read at the clinical sites or a resting echocardiogram read by the ECHO Core Laboratory will identify patients meeting LVEF entry criteria and characterize the extent of anterior LV wall dysfunction needed to evaluate SVR eligibility. Patients will be approached for informed consent for randomization within Stratum A, B, or C. In consenting Stratum A patients, a gated SPECT ventriculogram must be sent or obtained and sent to the Radionuclide Core Laboratory. In consenting Stratum B or C patients, a CMR (or gated SPECT if CMR is not obtainable) must be sent to the appropriate core laboratory. FIGURE 2 LEGEND: Determination of LV function criteria for STICH Trial and SVR eligibility begins with evaluation of all LV function tests for which the acquired images are available. A If an LVEF.35 is confirmed by site reading of either a contrast, CMR, or gated SPECT LV study, the patient meets LVEF criterion for obtaining informed consent. B Patients without one of these three site-read LV studies meet LVEF entry criterion by an ECHO Core Laboratory reading of LVEF.35. J Patients meeting entry criteria for whom responsible physician agreement to trial entry cannot be obtained enter the Initial Evaluation Log if their physician and they consent to National Death Index follow-up. C The LV imaging study used to measure LVEF is also used to evaluate SVR eligibility by demonstration of a single zone of akinesia involving 35% of the anterior wall. D - Patients not meeting this criterion of SVR eligibility and who are MED eligible by absence of 50% left main stenosis and of CCS III angina criteria can be approached for consent to randomization in Stratum A (F). E Patients who are eligible for SVR are evaluated for medical therapy ineligibility by presence of 50% left main stenosis or CCS Class III angina and those eligible for medical therapy can be approached for informed consent to randomization in Stratum B (G). E Patients ineligible for randomization to medical therapy because of 50% left main stenosis or CCS Class III angina are approached for consent to randomization in Stratum C (H). I - Any patient declining randomization is approached for consent to enter the registry (I). Patients who consent enter the trial registry (J) and those who do not consent are excluded from the trial. L MED eligible patients who are ineligible for SVR enter the trial in Stratum A. M Patients meeting both MED and SVR eligibility enter the trial in Stratum B. N Patients meeting SVR eligibility criteria who are MED ineligible enter the trial in Stratum C with randomized treatment of CABG + MED vs CABG + SVR + MED. DCRI/Confidential Page 19 01/18/2002

21 FIGURE 2 Patients eligible by clinical and coronary angiographic criteria and screened for LVEF measurement.35 in past three months Patient consents to follow-up no yes Responsible physician agrees to consider trial entry no yes Not in Trial Is site reading of either a contrast, CMR, or gated SPECT LV study available? yes no A. Is LVEF.35 by any high-quality study? no Patient not eligible for STICH Trial no B. Is LVEF.35 by Echo Core Lab reading? yes yes C. Is single zone of akinesia involving 35% of the anterior wall present on echo or contrast, CMR, or gated SPECT LV study? yes no yes Not in Trial yes D. 50% left main stenosis or CCS III angina? E. 50% left main stenosis or CCS III angina? no no yes F. Obtain consent in Stratum A no G. Obtain consent in Stratum B H. Obtain consent in Stratum C no yes yes yes no I. Obtain consent for registry Not in Trial no yes yes yes yes J. Enter in Initial Evaluation Log if entry criteria met K. Enter study in registry Treatment not randomized L. Enter study in Stratum A MED vs CABG + MED M. Enter study in Stratum B MED vs CABG + MED vs SVR + CABG + MED N. Enter study in Stratum C CABG + MED vs CABG + SVR + MED 2000 Patients 3,400 Patients 1,600 Patients 600 Patients 600 Patients DCRI/Confidential Page 20 01/18/2002

22 4.3.3 Initial Evaluation Log To minimize and understand patient selection bias in the STICH Trial, an initial evaluation log and registry will be maintained for four specific purposes: 1) to monitor and encourage unbiased enrollment at each STICH Trial clinical site; 2) to determine how age, gender, and ethnicity appear to influence physician and patient treatment decisions for medical therapy, surgical therapy, and participation in research; 3) to compare outcomes of randomized and non-randomized patients; and 4) to place randomized patients in the context of the larger population with CAD, HF symptoms, and LVEF.35. The initial evaluation for STICH Trial entry will identify some patients who appear eligible to approach for consent to enter the STICH Trial but whose responsible physician declines to give study investigators permission to approach the patient for further evaluation for potential randomization. If physician permission can be obtained to approach patients only for consent for limited follow-up using primarily the National Death Index, those patients who consent will be assigned an initial evaluation log number through the IVRS. Baseline information will be entered into an initial evaluation log in the same format used for registry and randomized patients. Both physician and patient reasons for not entering the STICH Trial and choice of therapy will also be recorded. Patients in the initial evaluation log will be followed by telephone contact at four months to assess treatment status and thereafter only annually by the National Death Index. Preliminary clinical site estimates suggest 2,000 patients will be included in the Initial Evaluation Log Registry Patients who meet criteria for trial entry but when approached for consent decline randomization will be approached to consent to enter the STICH Trial registry. Patients consenting to enter the STICH Trial registry will be assigned a registry number through the IVRS and will have baseline clinical and process of care information recorded on the initial evaluation form. No baseline or follow-up testing will be required. All registry patients will be followed by telephone contact at four-month intervals throughout the trial. Preliminary clinical site estimates suggest 3,400 patients will be included in the STICH Trial Registry Randomization After eligible patients give informed consent, site personnel will initiate the randomization process by calling an always available toll-free telephone number to connect with the IVRS at the Coordinating Center that has a redundant voice back-up system. Once eligibility is established, a site-specific unique identifier will be assigned to the patient. A permuted block randomization, stratified by clinical site and by the three Strata A, B, or C shown in Figure 1, will be used with a block size not known by the investigators. Use of this randomization process insures proper trial enrollment and promptly provides the Coordinating Center with the basic patient enrollment information needed to monitor enrollment performance. Study design specifies enrollment of 2,800 patients into the STICH Trial Baseline Evaluation of Randomized Patients Prior to Assigned Treatment All patients eligible and consenting to be randomized will have had baseline clinical evaluation information entered on the initial evaluation form. An echocardiogram will be obtained and sent to the ECHO Core Laboratory if this test was not used to establish trial eligibility. A baseline quality of life assessment and sixminute walk test will be performed. A modified Bruce stress test will be obtained on patients in Strata A and B who are able to exercise. This usually will have been performed in conjunction with the stress sestamibi perfusion study. A baseline gated SPECT ventriculogram will be obtained in patients in Stratum A previously not evaluated by this modality. A baseline CMR will be obtained in Strata B and C patients if not previously performed during initial evaluation. DCRI/Confidential Page 21 01/18/2002

23 4.4 Initiation of Study Treatments Intensive medical treatment (MED) will be initiated after completion of baseline evaluation in all randomized patients. Patients randomized to treatment with CABG, with or without SVR, will undergo operation as soon as possible after completion of protocol-mandated baseline studies, but within 14 days in all patients. The standard of best practice for each of the three possible treatments will be reviewed in the context of relevant published literature at regular intervals throughout the study by the Medical and Surgical Therapy Committees, and the original treatment protocol may be modified by the Data Safety Monitoring Board (DSMB) if necessary. Definitions of the three therapies to be used at the start of the trial appear justified by current evidence. 4.5 Investigational Therapies Treatment Groups The three treatment groups are: 1. MED: MED for HF is now defined to include angiotensin-converting enzyme inhibitors (ACE-I), or angiotensin receptor blockers in patients who do not tolerate ACE inhibitors, beta blockers, spironolactone, and an antiplatelet agent such as aspirin or clopidogrel adjusted to optimal doses within 30 days of randomization unless contraindications are documented. HMG-CoA reductase inhibitors, diuretics, and digitalis use will be individualized to patient-specific indications. Physicians will be encouraged to manage concomitant conditions appropriately (diabetes, hypertension, hyperlipidemia, angina, atrial arrhythmias) and will be asked to avoid the use of non-steroidal anti-inflammatory drugs as well as most anti-arrhythmic agents and most calcium-channel blocking drugs. Electrophysiologic testing and the implantation of cardioverterdefibrillators may be used in compliance with standard guidelines. MED therapy will be monitored closely and modified when indicated at a frequency no less than the required four-month follow-up clinic visits. 2. CABG: CABG will be performed using at least one internal mammary graft in all patients unless this conduit is unavailable or has inadequate flow. Unequivocally needed reparative procedures, such as valvular repair or replacement for primary mitral valve disease or thoracic aortic aneurysmectomy, are exclusion criteria at entry. However, patients with secondary mitral regurgitation identified by echocardiography who are judged to need mitral valve repair may undergo this procedure combined with CABG with or without SVR. Therefore, use of CABG or CABG + SVR throughout this protocol implies that mitral repair for regurgitation secondary to ischemic HF etiology may be included in the operative strategy at the discretion of the operating surgeon. 3. CABG + SVR: The two criteria used by the Surgical Therapy Committee to define the acceptable range of specific operative maneuvers essential to be considered an acceptable technical SVR operation for the STICH Trial will be any ventricular reconstruction method that consistently results in: 1) a low operative mortality; and 2) an average EF increase of 10% and average LVESVI decrease of 30% as assessed on the four-month postoperative CMR measurement. SVR may be performed with or without cardioplegic arrest. In general, the SVR is begun after completion of the CABG and in a beating heart. The region of dysfunction identified by preoperative CMR will be incised regardless of the epicardial appearance. The border separating the dysfunctional from contracting myocardium identified by palpation will guide placement of sutures and a patch, if necessary, to pursestring the endocardial scar and thereby decrease LV size without distorting LV shape. For patients undergoing surgery, intensive medical treatment will resume as soon as permitted by post-procedure recovery. DCRI/Confidential Page 22 01/18/2002

24 4.5.2 Concomitant Therapy All study participants will receive optimal medical therapy as defined by the Medical Therapy Committee. Therapies directed at non-cardiac illness will be recorded on the CRF but will not be specified by the trial protocol Treatment Compliance Records of medication class, specific agents, total daily doses, and intervals between visits will be kept by the site during the study. CABG and SVR will be performed in STICH according to standards agreed to and described by the Surgical Therapy Committee. The field monitor during site visits and at the completion of the trial will review procedural details and the medication records for all study patients regardless of investigational arm to evaluate for violations to the protocol Interruption or Discontinuation of Treatment A patient is considered randomized when the patient identification number has been assigned by the IVRS. Every effort must be made to ensure that patients remain in the study and on medical therapy for the duration of the study. Each patient entering the study must be followed until study completion whether or not the patient receives the randomized treatment or medical therapy is temporarily interrupted or permanently discontinued. Interruption of Medical Therapy All patients regardless of investigational arm will receive optimal medical therapy as defined by the Medical Therapy Committee. The provision of optimal medical therapy to all study participants is independent of any potential expected or unexpected functional improvement provided by any of the three therapies throughout the course of the trial. An interruption of medical therapy may occasionally be required. If a temporary interruption occurs, the Coordinating Center Medical Hot Line is available and can be notified for any study related issues. Every attempt to reinitiate optimal medical therapy should be made throughout the duration of the study. The reinitiation of medical therapy is not subject to a time limit. Patients with a temporary or permanent discontinuation of any or all medications integral to optimal medical therapy should continue the visit schedule and undergo all protocol specified studies. 4.6 Study Subject Follow-up At each clinic visit at four-month intervals after study entry, a brief history and physical examination will be performed and follow-up forms will be completed. Follow-up data to be collected at each of these contacts will include an assessment of HF and angina symptoms and interval medical events including hospitalization, major cardiac procedures, and tests. Patients completing a treadmill test at baseline will have this test repeated at two years. Detailed quality of life assessment will be performed at the four-month visit, one year, and annually thereafter during follow-up. Table 1 below summarizes the schedule of follow-up studies in the STICH Trial. Telephone contact will be maintained with the patient unwilling or unable to return to the clinical site and their local care providers to assure that intensive medical therapy is continued. Discontinuation from the Study A patient will be considered lost to follow-up after exhausting all means of contact. The status of the patient at the last visit or contact will be used for the final analysis. The vital status of patients who withdraw consent or are lost to follow-up will be followed by the National Death Index until the end of the study DCRI/Confidential Page 23 01/18/2002

25 4.7 Assessment Schedule Overview of Study Phases The study consists of three phases. The objective of Phase 1 is proper patient selection leading to randomization. Most of Phase 1 will occur in advance of randomization and has no specific time course. It concludes with the proper identification and stratification of eligible patients into one of the three groups as presented in detail in Figure 2. The day of randomization will be noted as Day 0 of the visit schedule. The Visit 1 assessment will be completed and sent to the Coordinating Center within 5 days of Day 0. Phase 2, treatment initiation, begins on Day 0 and concludes no later than Day 14 with patients having received one of the three treatment regimens by no later than Day 14 and completion of Visit 2. The conclusion of Phase 2 will be Visit 2. Visit 2 will occur at hospital discharge or on Day 30 in MED patients treated as outpatients. Phase 3 follow-up Visits 3 16 are expected to occur most commonly as outpatient visits but may occur in hospital if required by patient status. Further visits every four months may be required if follow-up extends beyond Visit 15 and these visits will not differ in any way from those described for Visit 15. The acceptable visit window is up to 15 days before or after the protocol-defined date Phase 1 Patient Selection, Randomization, and Visit 1 Before Randomization Evaluate patient history and current status according to the study inclusion and exclusion criteria. If the patient is eligible for randomization: Obtain written informed consent. Record demographic data, medical history, current medications and LV function study results Record the heart rate, blood pressure, NYHA HF functional class, and CCS anginal class. Randomization Randomize eligible patients via the 24-hour IVRS. Record in the CRF the site number, the patient number, and the date and time of randomization. Visit 1 - After Randomization 1. Complete all protocol-required imaging studies not performed earlier. Stratum A echocardiography, gated SPECT ventriculogram, nuclear perfusion imaging, nuclear myocardial viability study, modified Bruce exercise treadmill test in patients able to exercise (combined with nuclear perfusion). Stratum B -- echocardiography, nuclear perfusion imaging, nuclear myocardial viability study, and CMR ventriculogram (or gated SPECT ventriculogram if CMR not feasible due to patient factors) and modified Bruce exercise treadmill test in patients able to exercise (combined with nuclear perfusion). Stratum C echocardiography and CMR ventriculogram (or gated SPECT ventriculogram if CMR not feasible due to patient factors). DCRI/Confidential Page 24 01/18/2002

26 2. Complete protocol mandated testing Collect NCG blood sample. Administer the baseline QOL questionnaire and ask the patient to complete the EuroQoL utility assessment. Perform the six-minute walk test. Imaging and blood collection standards are described in accompanying separate echocardiography, nuclear, CMR and NCG core lab instruction manuals Phase 2 Treatment Initiation and Visit 2 Treatment Initiation Study therapy allocated by randomization should be initiated as soon as feasible but no later than Day 14. Visit 2 (At hospital discharge or at Day 30) Assess and record all primary and secondary endpoints. Record CABG and/or SVR procedural descriptions and complications in surgical patients or complete initial medical care form in medical patients Record heart rate, blood pressure, and NHYA HF functional class. Record current medications Phase 3 Follow up Visits Through to Study End Follow-up visits 3 15 will occur every four months from Day 0 until the end of the trial. At each follow-up visit, the patient evaluation will consist of an assessment of heart rate, blood pressure, NYHA HF functional class, and CCS anginal class. At all four-month visits, current medications, any intercurrent SAEs, all primary and secondary endpoints, and any interval medical care will be recorded in the CRF. At the four-month visit (visit 3), repeat echocardiography will be performed in all patients regardless of strata and sent to the ECHO Core Laboratory, repeat NCG blood samples will be obtained and sent to the NCG Core Laboratory, and nuclear or CMR testing will be performed and sent to the core laboratory for Stratum B and C patients as detailed in the core laboratory manuals and in Table 1. Additionally, the EuroQoL questionnaire will be completed by all patients and sent to the CC and a six-minute walk will be performed and recorded in the CRF. At the yearly visits (visits 5,8,11, and 14), a six-minute walk will be repeated and all patients will be asked to complete the EuroQoL. At the two-year visit (visit 8), a repeat modified Bruce exercise test and echocardiogram will be obtained. Stratum B and C patients will have repeat gated SPECT nuclear imaging or CMR as previously described. If the study continues beyond 52 months, visit 15 will be repeated every four months until study completion. The final visit (visit 16) will occur at the time of study completion and require data collection similar to visit 15 and the study completion form. Table 1 details the schedule of all protocol mandated assessments. DCRI/Confidential Page 25 01/18/2002

27 Table 1 Visit Schedule Informed Consent Demographics Medical history Evaluation of Inclusion / Exclusion criteria Qualifying procedural data i.e. cardiac catheterization results Evaluation of left ventricular systolic dysfunction Screening echocardiography review by ECHO core (if necessary) Surgical procedural form or initial medical care form VISITS Day 0 Month (± 15 days) Month (± 15 days) 0 1 X X X X X X X (2) X Patient evaluation X X X X X X X X X X X X X X X Echocardiography X X X Gated SPECT nuclear imaging A# # # Nuclear myocardial viability imaging Nuclear myocardial perfusion imaging* Modified Bruce exercise protocol* X X CMR imaging B/C B/C B/C NCG blood sampling X X Baseline quality of life questionnaire A/B A/B X EuroQOL X X X X X X 6 minute walk X X X X X X Study completion form X (3) (1) If the study duration is longer or shorter than the four years presented as an example in this table, the Visit 15 schedule may be repeated every four months until the study is completed or deleted as required; (2) Visit at 1 month or at hospital discharge, whichever is sooner; (3) At the end of the study; # Gated SPECT nuclear imaging is to be performed in lieu of CMR in Strata B and C patients who are ineligible for CMR testing due to logistics, limitations or implanted devices; * It is anticipated that the majority of patients will undergo exercise testing as part of the nuclear perfusion imaging protocol at baseline. In those patients who are able to exercise but who have not undergone exercise testing using the Modified Bruce protocol as part of their perfusion imaging, a separate exercise test is required at baseline. All subjects who underwent baseline exercise testing and are able will undergo exercise testing at 2 years; X refers to all study participants; A Refers to subjects enrolled into Stratum A; B Refers to subjects enrolled into Stratum B; C refers to subjects enrolled into Stratum C. (1) Final DCRI/Confidential Page 26 01/18/2002

28 4.8 Efficacy Assessments Documentation for occurrences of potential primary or secondary efficacy endpoints will be required for submission to the Endpoint Committee. The Endpoint Committee will adjudicate causes of death and selected secondary endpoints based upon pre-defined definitions and procedures for this study. The process of endpoint adjudication, and the definitions and required documentation for the primary and secondary endpoints are included in the Endpoint Manual Primary Efficacy Parameters 1. All-cause mortality (time to death). 2. Survival free of hospitalization for cardiac cause Secondary Efficacy Parameters 1. All cause mortality at 30 days. 2. Survival free of hospitalization for heart failure. 3. Cardiovascular mortality (defined as sudden death, or death attributed to recurrent myocardial infarction, heart failure, a cardiovascular procedure, stroke, or other cardiovascular etiology). 4. Cardiovascular mortality and morbidity. 5. Survival free of revascularization. 6. Survival free of cardiac transplantation or LVAD implantation. 7. All-cause (unplanned and elective) hospitalization Other Key Efficacy Parameters Survival free of CABG Survival free of PCI Sudden death with or without resuscitation Fatal and non-fatal MI Fatal and non-fatal stroke Safety Assessments Safety assessments will consist of monitoring and recording the pre-defined safety endpoints, procedure complication rates, all serious adverse events, and the regular measurement of vital signs. Results of all safety assessments (e.g., physical examinations or laboratories) performed as part of the standard evaluation and care of the patient should be maintained in the patient s study chart (source documents). Safety Endpoints and Expected Procedure Complication Rates Upon request of the DSMB, the Coordinating Center, working with the Medical and Surgical Therapy Committees, will define Steering Committee approved safety endpoints and procedure complication rates with expected ranges and upper threshold limits for concern. Medical therapy will be monitored by safety endpoints, such as symptomatic bradycardia, angioedema, renal dysfunction, and drug intolerance. Procedure complications will include return to the operating room for bleeding, prolonged intubation, stroke, mediastinitis, and atrial fibrillation. DCRI/Confidential Page 27 01/18/2002

29 Serious Adverse Events (SAE) A serious adverse event is defined in general as an untoward (unfavorable) event which: 1. is fatal or life-threatening, 2. required or prolonged hospitalization, 3. was significantly or permanently disabling or incapacitating, 4. is medically significant (may jeopardize the subject and may require medical or surgical intervention to prevent one of the outcomes listed above). Hospitalizations planned before entry into the clinical study and not associated with any deterioration in condition, emergency, or outpatient visits that do not result in admission are not considered SAEs. SAEs will be reported to the Coordinating Center with the visit form. In this high-risk population, the rate of SAEs will be used to assess safety and, therefore, no individual event requires urgent reporting. Laboratory Evaluations Local, not central, laboratories will be used. Normal ranges of the local laboratory will serve as the reference for the patients of individual centers. If in the course of the study, a patient is hospitalized in a nonparticipating center, laboratory evaluations from that hospital will be considered in the same way. Vital Signs Heart rate and blood pressure, will be measured at each visit. Blood pressure and pulse is to be measured in a similar manner at each visit preferably in the sitting position after a equilibration period of ten minutes on the right arm. New York Heart Association (NYHA) functional HF class will be recorded at each visit Quality of Life Assessments Quality of life (QOL) questionnaire data will be collected at baseline by the site coordinators, and at four months, years one, two and three by trained telephone interviewers from the EQOL Core Laboratory at the DCRI who are blinded to treatment assignment. In addition, cardiac symptoms as reflected by the NYHA HF class and the CCS angina class, patient utilities, and interval medical care consumption will be assessed every four months at each clinic visit/contact by the sites. Content of Health-related Quality of Life Questionnaire The QOL questionnaire consists of a battery of validated instruments that build on a generic core supplemented by disease-specific measures where necessary to provide a comprehensive assessment of health-related quality of life. The generic core instrument to be used is the Medical Outcomes Study Short Form (SF-12). This instrument is a reduced version of the SF-36 that covers the broad range of health-related quality of life domains with two or three items per domain. To enhance sensitivity, we will include the full SF-36 items for the following domains: role function-physical, role function-emotional,, social function, psychological well-being/mental health, and vitality. The remaining scales in the SF-36 overlap with other items in our battery and will therefore be used in their reduced SF-12 form. A total of 22 items from the SF-12/SF-36 will be included on the battery. The primary disease-specific measure to be used is the Kansas City Cardiomyopathy Questionnaire (KCCQ), which contains 20 items covering the effects of heart failure on physical limitations, symptoms frequency and severity, social functioning and general quality of life. DCRI/Confidential Page 28 01/18/2002

30 The presence of anginal symptoms will be assessed with the six-item symptom scales from the Seattle Angina Questionnaire. Cardiac symptoms will be independently assessed by the sites at each four-month visit using the New York Heart Association (NYHA) congestive heart failure class and the Canadian Cardiovascular Society (CCS) class for angina. General psychological and well being/mental health will be measured using a five-item subscale of the SF-36 (as described above). We will also assess depressive symptoms in more detail using the 20-item Center for Epidemiologic Studies-Depression (CES-D) scale. Self-efficacy will be evaluated using the 13 item Cardiac Self-Efficacy scale. Role functioning, social functioning and vitality will be assessed from the SF-36 as described above. Employment details will be obtained using six questions adapted from the NHLBI Bypass Angioplasty Revascularization Investigation (BARI) Substudy in Economics and Quality of Life (SEQOL). In addition, the baseline questionnaire will ask four questions covering education/socioeconomic status (including income class) and social support (marital status, number in household). Measurement of Utilities Patient specific utilities will be assessed using the EuroQoL. The current version of this instrument consists of a five-dimension assessment of current health-related quality of life and a self-rating (0-100) "thermometer" of "your own health state today. This assessment will be made at baseline, four months, one year and yearly thereafter using a patient completed form. If the patient is unable to complete the form unassisted, the instrument will be administered as an interview. If the patient does not return for a clinic visit, the interview will be done over the telephone when the Coordinator does the follow-up contact Economic Data/Hospital Billing Data (U.S. sites only) Hospital bills (detailed, summary ledger and UB 92) will be collected by the EQOL Core Laboratory at the DCRI for all hospitalizations throughout the length of the study. They will include hospitalizations at clinical sites and at institutions not participating in STICH. In addition, cost to charge ratios (RCCs) will be obtained from each hospital where a STICH baseline or follow-up hospitalization is reported Echocardiography Assessment Core laboratory echocardiographic assessment of left ventricular function is one of four modalities that can be used to screen patients for STICH Trial eligibility. In all patients, echocardiograms will be used to evaluate the independent prognostic value of baseline echocardiographic variables and their interaction with treatment allocation and outcome on long-term follow-up. Screening for Trial Eligibility and SVR Eligibility Patients screened with an echocardiogram site reading of EF.35 require ECHO Core Laboratory confirmation of LV dysfunction entry criteria to be eligible for randomization. The magnitude of anterior akinesia or dyskinesia will be assessed to determine SVR eligibility prior to randomization. The ECHO Core Laboratory will support this critical operational step of the trial by prompt and accurate categorical reading of the two critical eligibility criteria returned to both the clinical site and the Coordinating Center, usually within one hour and always within 24 hours of receipt of the study. Echocardiographic analysis for STICH Trial and SVR eligibility will be limited to LVEF and anterior akinesia criteria. If LVEF is >.35, no further echocardiographic measurement will be made. If LVEF is.35, the patient will be considered eligible to be approached for consent to participate in the trial. DCRI/Confidential Page 29 01/18/2002

31 In patients with a LVEF.35, evaluation of the percent of akinetic segments will be performed by visual assessment of regional contractility. With the 16-segment model, 7 segments will be assigned to the distribution of the left anterior descending (LAD) coronary artery. Akinesia of 3 LAD segments will indicate 35% akinesia. On the basis of this visual assessment, the patient will be classified as: Akinesia 35% Akinesia < 35% All patients with an ECHO Core Laboratory LVEF assessment of.35 will be eligible to approach for informed consent in Stratum A. Patients with an ECHO Core Laboratory reading of 35% anterior akinesia will be eligible to approach for informed consent in Stratum B or Stratum C. Echocardiographic Measurements at Baseline, Four-month, and Two-year Follow-up A baseline echocardiogram for all patients who are randomized into any of the three strata of the trial will be required to address the echocardiographic specific secondary objective outlined in section 2.2 by quantitative measurements of cardiac structural, functional, and hemodynamic parameters. For patients whose echocardiographic study is used for eligibility, an additional baseline echocardiographic study may be necessary if all essential echocardiographic measurements are not available in the first study. The initial echocardiogram will provide baseline echocardiographic variables to be compared with those obtained from the 4 month and 2 year follow-up echocardiograms. Four-month and two year follow-up will be performed to compare the long-term effects of different treatment modalities in regional remodeling, MR, and diastolic function. The echocardiographic protocol for baseline and serial echocardiograms is provided in the echocardiography manual Nuclear Cardiology Assessment Baseline Nuclear Cardiology Testing Patients in Stratum A and B will undergo baseline rest/redistribution thallium SPECT nuclear imaging followed by stress sestamibi SPECT nuclear imaging. Patients will be brought to the nuclear cardiology laboratory in the fasting state. Thallium-201 will be administered intravenously under resting conditions at a dose of mci, followed 15 minutes later by SPECT imaging with the patient in the supine position. The SPECT acquisition will be accomplished using standardized methods, including imaging over a 180 degree anterior arc from 45 degree right anterior oblique to 45 degree left posterior oblique view at multiple steps. Patients will then rest and remain fasting for 3.5 to 4 hours, following which a second image acquisition will be performed to acquire the thallium redistribution data. Immediately after the redistribution acquisition, patients will undergo exercise stress imaging using the modified Bruce protocol (depending upon the patient s functional capacity). Heart rate and blood pressure will be monitored. Technetium-99m sestamibi will be administered intravenously at peak exercise, at a dose of mci, and the patient will be requested to continue to exercise for another 30 seconds to permit maximal myocardial extraction of the tracer. The initial resting thallium images will provide information regarding regional myocardial blood flow and myocardial viability, the thallium redistribution images will provide information regarding myocardial viability in regions with reduced perfusion at rest, and the exercise (or adenosine) sestamibi images will provide information regarding inducible myocardial ischemia. For assessment of the presence, extent, and severity of myocardial ischemia, it is essential to obtain a level of exercise stress capable of providing an estimate of regional disparities in vasodilator reserve. Exercise testing will be considered adequate for purposes of the nuclear cardiology components of STICH only if the patient is DCRI/Confidential Page 30 01/18/2002

32 able to achieve 85% (or greater) of maximal heart rate as predicted by age using the modified Bruce protocol. It is anticipated that many patients with heart failure will have limited exercise endurance because of deconditioning, and many others will not achieve a maximal heart rate response because of therapy with beta blockers. In such patients, stress imaging will be performed using pharmacologic stress with adenosine. This easily standardized approach will be employed at all STICH sites and will provide information regarding maximal coronary vasodilator reserve independent of concomitant medical therapy and exercise tolerance. Adenosine will be administered intravenously at a dose of 0.14 mg/kg/min for 6 minutes. Heart rate and blood pressure will be monitored. At minute 3 of the infusion, technetium-99m sestamibi will be administered intravenously at a dose of mci. Gated SPECT imaging will then begin 60 minutes after sestamibi administration. Many patients will have undergone SPECT imaging for clinical purposes prior to being enrolled in STICH. These prior tests may be used for the baseline nuclear tests, with the following stipulations. A measure of myocardial viability must be obtained. Acceptable previous viability protocols include rest/redistribution thallium imaging, stress/redistribution/reinjection thallium imaging, 27 or nitroglycerin-enhanced rest sestamibi SPECT imaging. 28 For example, if patients underwent a previous standard dual isotope SPECT study, with rest thallium imaging followed by stress sestamibi imaging, but without the thallium redistribution study, this will be considered inadequate in terms of viability information. The previous stress sestamibi information will be considered adequate for ischemia evaluation, but a separate study will be required to assess viability. The preferred viability test in this situation will be a nitroglycerin-enhanced rest sestamibi SPECT study, although a rest-redistribution thallium study will be acceptable. For nitroglycerin-enhanced rest sestamibi SPECT imaging, 30 mci sestamibi will be administered 5 minutes after sublingual nitroglycerin 0.4 mg. Imaging will commence 60 minutes after sestamibi injection to ensure that nitroglycerin effects have dissipated at the time that imaging is performed. In this manner, the perfusion pattern will reflect the nitroglycerin-enhanced perfusion at the time of sestamibi administration, but the functional information regarding wall motion and EF, obtained 60 minutes later, will not be affected by transient nitroglycerin effects on LV regional or global function. Nuclear Cardiology Testing for Evaluation of Left Ventricular Function Gated SPECT images will be obtained 60 minutes after sestamibi administration to assess regional and global LV function using standard software. At baseline, all patients in Stratum A will undergo gated SPECT left ventricular imaging. In addition, there will be a number of patients in Strata B and C who are not able to undergo CMR because of logistics, limitations, or because of metallic prostheses or implanted devices. All patients in Strata B and C who are unable to undergo CMR will be tested with resting technetium-99m sestamibi gated SPECT to determine LV EF, ESVI and extent of regional akinesis. Assessment of percent akinesis. The left ventricle will be evaluated quantitatively on a pixel by pixel basis for akinetic segments, defined as thickening less than 5% on gated SPECT images. The pixels will be summed to determine regional extent maps of the percent of myocardium that is akinetic for each of the LV regions: anterior, inferior, septal, lateral, and apical. Finally an overall quantitative extent will be obtained for the entire percentage of the left ventricle that is akinetic. In patients in Stratum B or C who had LV function assessed by gated SPECT ventriculography at baseline, rest sestamibi imaging also will be performed at 4 months and at 2 years to assess the effect of therapy on LV volumes and systolic function. DCRI/Confidential Page 31 01/18/2002

33 4.8.9 End-systolic Volume Assessment by CMR CMR Testing Patients in Stratum B and C who are willing and able will undergo baseline CMR testing to assess LV size, shape and function. Patients will complete questionnaires to determine CMR suitability in advance of the CMR examination. The detailed CMR protocol is described in the CMR manual Neurohormone, Cytokine and Genetic Assessments Samples for both neurohormonal/cytokine analysis and genotyping will be obtained from all patients enrolled in the STICH Trial at the time of study enrollment. A second blood sample will be obtained at the four month study visit for neurohormonal/cytokine analysis. Patients will be prepared prior to phlebotomy in order to minimize the influence of stress and other factors on neurohormone levels. The phlebotomy procedure necessary to minimize the influence of stress, sample preparation, sample storage, shipping and planned NCG analyses are detailed in the NCG manual. 4.9 Study Organization Monitoring Committees The following Monitoring Committees will be employed in this study: Executive Committee The Executive Committee leads the implementation of policies and practices approved by the Steering Committee. Executive Committee members will include the study chairman, principal investigator, the lead statistician, the director and associate director of trial operations, the CC project leader, the chairs and cochairs of the medical and surgical therapy committees, the chair of the core laboratory committee, the chair of the minority recruitment and retention committee, the chair and co-chair of the policy and publications committee, and the chair of the trial registry. The NHLBI project officer will be an ex officio member of the Executive Committee. All committees report through the Executive Committee to the Steering Committee. The main roles and responsibilities of the Executive Committee are: To lead the Steering Committee, to resolve differences of opinion within the Steering Committee, to resolve issues not requiring full Steering Committee input. To review operational aspects of the trial on an ongoing basis. To oversee the conduct of the study and initiate the approval process for any protocol changes needed to improve the quality of the study. To implement all protocol amendments approved by the DSMB. To propose membership of the other committees to the Steering Committee. To serve as a liaison between the NHLBI and the other committees. To serve as a liaison between the DSMB and the Steering Committee. To present the trial results to the steering committee in advance of national presentations. To approve all manuscripts prior to submission. DCRI/Confidential Page 32 01/18/2002

34 Steering Committee The Steering Committee consists of the Executive Committee and one representative investigator from each participating investigative site. The main roles and responsibilities of the Steering Committee are: To make ethical, scientific, and policy decisions regarding the overall conduct of the study. To approve committee membership. To serve as a liaison between all investigators and the Executive Committee. To review study progress. To act on the recommendations of the DSMB. To review and provide input to potential protocol amendments. To review and approve presentations and publications of the study. Data and Safety Monitoring Board (DSMB) The DSMB members are independent of the NHLBI, the study management organizations, and the investigators. A nationally recognized member of the cardiovascular community with experience in oversight of cardiovascular clinical trials will chair the Data and Safety Monitoring Board (DSMB) comprised of experts in relevant biomedical fields including cardiology, cardiovascular surgery, biostatistics, and bioethics appointed by the NHLBI. No member of the DSMB shall be located at an institution that participates in any direct way in the STICH Trial. The main roles and responsibilities of the DSMB are: To initially review the study protocol and commission a specific regular process for evaluation of STICH trial data. To meet no less than twice yearly for the duration of the trial to monitor the trial progress using confidential data summarizing trial safety and outcomes and to prepare a summary report of their review to the Executive Committee. To advise the Executive Committee on the implications of HF treatment advances on the conduct of the STICH trial. To approve protocol modifications if warranted. To advise the NHLBI directly regarding recommendations for trial modification or early trial cessation. The efficacy and safety analyses will be performed semi-blinded (i.e., treatments A, B and C) by the Statistical Data Coordinating Center. The DSMB statistician will possess a copy of the treatment codes for unblinding purposes if deemed necessary by the DSMB. The study may be amended, or stopped early, or a treatment arm may be discontinued should any of these be deemed necessary based upon DSMB recommendation. The chairman of the DSMB will discuss the recommendations of the DSMB with the NHLBI project officer who will inform the Executive Committee. The chairman of the Executive Committee will notify the Steering Committee of any significant proposed amendment to the study or recommended discontinuation of the study. The roles, responsibilities and procedures of the DSMB are summarized in the DSMB Manual Operational Committees Ancillary Studies Committee An Ancillary Study Committee will be led by the Director of Clinical Trial Operations and populated by leading physicians at investigative sites chosen by the Executive Committee. The main roles and responsibilities of the Ancillary Studies Committee are: To review, prioritize, and approve all ancillary study applications. To facilitate approved ancillary study applications for funding from external sources. To serve as a liaison between ancillary study investigators and the steering committee. To review and report on the conduct of approved ancillary studies. DCRI/Confidential Page 33 01/18/2002

35 Core Laboratory Committee This Core Laboratory Committee will be chaired by Dr. Robert Bonow and consist of the directors of the Cardiac Magnetic Resonance, Echocardiography, Economics and Quality of Life, Neurohormonal, and Radionuclide core laboratories. The main role and responsibilities of the Core Laboratory Committee will be: To oversee clinical site training and write guidelines for the acquisition of core laboratory data at the outset of the trial. To review core laboratory processes on an ongoing basis. To regularly review the quality of data acquisition at STICH investigative sites and inform the Executive Committee of consistent variance from established guidelines at any clinical site. Endpoints Committee The Endpoints Classification Committee members will be independent of the NHLBI and will not have direct contact with patients randomized into this study. The Executive Committee will choose the members of this committee. The main roles and responsibilities of the Endpoints Classification Committee are: To agree on standard definitions for the primary and secondary endpoints To standardize adjudication procedures for assessing these endpoints. To summarize above definitions and procedures in an Endpoint Manual. To provide an independent and blinded assessment of mode of death, and hospitalizations according to the Endpoint Manual definitions and procedures. The decisions of the Endpoint Committee will be used for health authority submissions and publications. The Endpoint Committee will also adjudicate the eligibility criteria on a quality control basis. Medical Therapy Committee Dr. Milton Packer will chair the Medical Therapy Committee. Membership will be available to participating cardiologists at leading trial centers. The main role and responsibilities of the Medical Therapy Committee will be: To initially define the standard of care for medical strategies in all arms of the trial. To provide guidelines on the appropriate application of all non-surgical pharmacological and nonpharmacological therapies for the management of HF, LV systolic dysfunction and CAD exclusive of cardiac surgery. To review Coordinating Center data reflecting clinical site compliance with these standards of all medical therapy given during the trial in all three treatment groups. To recommend corrective action for any consistent variance from medical therapy standards at any clinical site to the Executive Committee for corrective action. To review advances in medical therapy that might necessitate modification of the standardized MED regimen during the trial and recommend adoption of changes that appear appropriate. Minority Recruitment and Retention Committee This committee will be chaired by Dr. Clyde Yancy and consists of the associate director of trial operations, the trial project leader, selected steering committee members and interested site personnel. The main role and responsibilities of this committee will be: To optimize opportunities for participation by minority investigators To propose remedies to overcome barriers to minority subject inclusion encountered during the trial To explore scientific opportunities relative to minority status available in the STICH trial data DCRI/Confidential Page 34 01/18/2002

36 Policy and Publications Committee The Policy and Publications Committee, chaired by Dr. Eric Rose and co-chaired by Dr. Jean Rouleau, will have the overall objective of insuring all interested STICH Trial investigators receive equal access and fairness in performing all scientific work for which they are qualified. Specific tasks include: Review all operational policies regarding trial conduct that require Steering Committee approval. Create and monitor a manuscript publication policy. Recommend authorship of publications and presentations. Review and approve all manuscripts prior to publication. Surgical Therapy Committee Dr. Lorenzo Menicanti and Dr. Andrew Wechsler will chair and Dr. Myron Weisfeldt will co-chair the Surgical Therapy Committee. Membership will be available to interested and experienced participating surgeons representing all clinical sites. The overall objective of this committee is to assure that patients in the STICH Trial receive the highest quality surgical therapy throughout the trial. These objectives will be discharged by site certification and monitoring, education, and regular updates of the standard of therapy throughout the seven-year trial duration. The main role and responsibilities of the Surgical Therapy Committee will be: To determine guidelines for initial surgical site certification. To review data reflecting clinical site compliance with these standards of all surgical therapy provided during the trial in all three randomization strata. To recommend corrective action to the Executive Committee for any consistent variance from surgical therapy standards at any clinical site. To regularly review published surgical refinements and recommend modifications of the standard STICH surgical therapy to the Executive Committee as needed. 5. Data Management 5.1 Data Collection Data Forms Data collection forms used in this study are tabulated in the manual of operations with summaries of their content and purpose. The front page of each form will identify the date and version of the form, and provide brief summary instructions for data collection. The subject s study number and initials will be included on every page of each form. The study subject number will reflect both a site specific code, the subject's sequence number for that site and a check digit. Each form will require the signature of the physician investigator and the study coordinator with the date of completion Manual of Operations The manual of operations will provide more detail regarding study objectives, scope, design and operational policies and procedures than the clinical protocol. The manual will include detailed instructions for completing study forms, including baseline health status assessments, and instructions for handling and transferring data forms to the Coordinating Center. The follow-up schedule and description of follow-up data and endpoint data will also be covered in the manual. DCRI/Confidential Page 35 01/18/2002

37 The operations manual will also describe the data management system, the flow of data forms, quality control measures, data queries, who and where to call for assistance, and other operational procedures employed at the Coordinating Center. The manual will also include a specific section on the appropriate procedures for echocardiography, nuclear testing, CMR image acquisition and transfer, and NCG blood sampling and transfer STICH Trial Website The STICH Trial website will provide printable updated versions of the protocol and manual of operations. Also included on the website will be a complete directory of participating physician investigators and study coordinators at all clinical sites. The website will provide a rich source of current information to investigators with secure access. Free access will be available to the public to a portion of the site containing general information. 5.2 Database Management and Quality Control Database management and quality control for this study are the responsibility of Duke Clinical Research Institute, Durham, NC, USA. Structured data elements from the CRFs will be entered into the STICH database and reviewed using double data entry for verification. Coexistent diseases and adverse events will be coded using a standard coding dictionary, MEDDRA 1.5. Concomitant medications will be coded using a standard medication dictionary, WHO DRL. Information entered into the database will be systematically checked by data management staff, using error messages generated from validation programs and database listings. Obvious errors will be corrected by data management center personnel. Other errors, omissions or questions will be entered on data query forms, which will be returned to the investigational site for resolution. After the investigator response is received at the data management center, the resolutions will be entered into the database. A copy of the signed data query form will be kept with the CRFs. Quality control audits of all key safety and efficacy data in the database will be made at designated times during the study. All clinical site patient-related reimbursement will be prompted by completion of data forms with appropriate responses to all data elements. 6. Statistical Plan 6.1 Sample Size Determination -- Overview Several design factors and research objectives have been considered in developing appropriate sample size estimates for the study. First, patient enrollment has been determined so there would be a sufficient number of endpoints to provide a high degree of confidence (power > 90%) for testing both primary hypotheses. Second, important secondary endpoints, including cardiovascular mortality, short-term (30-day) mortality, and measures of quality-of-life have also been considered. Third, we considered it important for the overall sample to be large enough to permit a prudent examination of treatment effects in selected subgroups of patients where surgical intervention might be particularly advantageous, or where the question of a treatment benefit from surgical intervention is particularly relevant. Important pre-specified subgroups of interest in this study include those defined by age, gender, race, HF class, ejection fraction, and the strata outlined in Figure 1. Finally, the sample size has been determined to provide a reasonable level of confidence of detecting clinically important therapeutic effects even in the event that current projections of event rates and treatment differences prove to be optimistic. DCRI/Confidential Page 36 01/18/2002

38 6.1.1 Sample Size and Power Considerations A combination of survival data from recent HF trials that reflect patient outcomes when treated with intensive medical therapy (including beta blockers and ACE inhibitors) and historical data from the Duke Cardiovascular Database provide useful information on the range of patient outcomes that would be expected among the HF patients enrolled in STICH. Based on this information, we project that for testing the coronary revascularization hypothesis, the mortality rate in the medically treated arm of STICH will be approximately 25% after 3 years of follow-up. For the LV restoration hypothesis, we project that mortality or cardiac hospitalization will occur in the CABG (without SVR) arm with a 3-year rate of 50% or higher. Recognizing, however, that the actual event rates in STICH may vary somewhat from these estimates, we have calculated sample size requirements for several different combinations of event rates and power levels in order to examine the sensitivity of the required sample size to different event rates and outcome scenarios that might conceivably arise in this trial. In the absence of detailed life-table follow-up data on patients similar to those who will be enrolled in the trial and treated with intensive medical therapy, the following approach was used for calculating sample size requirements for the coronary revascularization hypothesis. (A similar approach was used for the LV restoration hypothesis, although the event rates differed because the two primary hypotheses are based on different endpoints). For a specified benchmark outcome rate in the medically treated (control) group (e.g., an overall event rate at 3 years of 25%), event rates at other time points during follow-up were estimated by assuming an exponential survival distribution. We then postulated that in the CABG arm, there would be a specific reduction of this benchmark rate (e.g., a 25% reduction). Event rates for the CABG group at various time points during follow-up were also estimated assuming an exponential survival distribution. Since the two groups will be compared with respect to length of survival using the log-rank test, 29 or equivalently, the Cox proportional hazards model, the one-tailed formula of Schoenfeld 32 was modified to allow calculating sample size requirements based on the use of two-tailed tests. Schoenfeld's method is based on first calculating the total number of events required in the combined treatment groups. By estimating the proportion of patients who will have an event by the end of the study, the total number of patients required for the trial can be determined. The necessary number of events depends on the level of power desired, the significance level (which we have chosen to be α=0.05), and the ratio of the hazard function for patients receiving surgery to the hazard function for patients in the medically treated arm. Using exponential distributions to approximate survival in the two treatment groups (as described above), the hazard functions are particularly simple (constant over time) and thus an estimate of the hazard ratio is easily obtained. For example, if the mortality rate at 3 years was 25% in the medical arm, and if this rate was reduced by 25% in the CABG arm (to 18.75%), characterizing survival over time in the two groups using exponential distributions produces a CABG versus MED hazard ratio of Once the hazard ratio, the level of significance, the desired power, and the proportion of patients allocated to each treatment are specified, equation 1 in Schoenfeld 32 can be used to calculate the required number of events. If we specify α=0.05, power=0.90, equal allocation of patients to treatment groups, and use the hazard ratio of 0.72 from the example above, the required number of events for a two-group comparison is 396. The proportion of patients who will experience an event during the patient follow-up period can be estimated using the method in Section 2.1 of Schoenfeld. 32 If we designate S MED (t) and S CABG (t) to represent survival at time t in the medical and CABG groups respectively, and if we assume patient accrual will occur over a period of 3 years with follow-up extending 3 more years after enrollment is completed, the proportion of patients who will die by the end of the study is estimated by the following expression: 1/2 {1-1/6[S MED (3) + 4 S MED (4.5) + S MED (6)]} + 1/2 {1-1/6[S CABG (3) + 4 S CABG (4.5) + S CABG (6)]}. DCRI/Confidential Page 37 01/18/2002

39 With the combination of parameters used in the example above, the value of this expression (and hence the expected proportion of events) is The total number of patients required to provide a high level of power (0.90) for detecting a 25% improvement of CABG relative to MED would thus be 1,290, or 645 in each group. If the magnitude of the treatment benefit is smaller, the sample size requirements are substantially increased. For instance, if the event rate at 3 years in the control patients is 25% and the magnitude of the treatment benefit is only 20%, the required number of events increases to 652, and 1,033 patients per arm are required to provide 90% power for detecting a 20% reduction, and 883 patients per arm would provide 85% power. (1,000 patients per arm will provide 89% power for detecting a 20% reduction.) A key issue in determining an adequate sample size for this trial is the magnitude of the mortality reduction we could expect to achieve in the CABG arm compared to the medical arm. It may be optimistic to assume that CABG will reduce mortality by 30% or more. But based on what is known about the effects of CABG in other patient populations, it is reasonable (and relatively conservative) to postulate that surgery will be sufficiently effective to reduce overall mortality by 20-25%. And a mortality reduction of this magnitude would be highly important from a clinical and public health standpoint, given the large population of patients in this country and throughout the world who suffer from moderate to severe HF secondary to CAD. Compared to the medically-treated HF patients, we hypothesize, therefore, that in the CABG arm, the mortality rate at 3 years will be reduced by at least 20%. For the LV restoration hypothesis where the primary endpoint is death (all cause) or hospitalization for a cardiac reason, the three-year event rate in the control arm (patients treated with CABG without SVR) is expected to be in the range of 50% or higher. If the control rate at 3 years is 50%, 387 patients per arm will provide 90% power for detecting a 20% reduction in the incidence of the primary endpoint, and 460 patients per arm will provide 90% power for detecting a 20% reduction if the control rate should be as low as 45%. To provide an adequate number of patients in the trial that will be robust even under conservative assumptions about the control event rates for the primary hypotheses and the magnitude of the benefit of the interventions, we propose to enroll 2,000 patients for testing the coronary revascularization hypothesis and 1,000 patients for testing the LV restoration hypothesis. Because the patients in Stratum B who are treated with CABG without SVR (estimated to be 200 patients) can be used in assessing both primary hypotheses (see Figure 1), the total number of patients required for the trial is estimated to be 2,800 rather than 3,000. Below is a summary of what this number of patients will provide the study. Coronary revascularization hypothesis (all-cause mortality): 1. Power > 90% for detecting a 25% reduction in mortality with CABG if the event rate at 3 years in medically-treated patients is 25% or higher. (Thus we have high power for detecting a moderate benefit if the control arm event rate is consistent with what is expected based on other studies.) 2. Power = 89% for detecting only a 20% reduction in mortality with CABG if the 3-year event rate in medically treated patients is 25% or higher. (Thus we have very good power for detecting a conservative estimate of the mortality reduction if events occur in the medical arm at the rate expected.) DCRI/Confidential Page 38 01/18/2002

40 3. Power = 80% for detecting a 20% improvement if the 3-year mortality rate in medically treated patients should be as low as 20%. (Thus we have good power for detecting a conservative estimate of the benefit if the control arm event rate is somewhat lower than expected.) 4. Power > 85% for detecting a 30% or greater reduction in mortality in any subgroup consisting of at least 40% of the patients, and for detecting a reduction in mortality by one-third in any subgroup consisting of at least 30% of the patients if the control group event rate at 3 years is 25% or higher. (Thus we even have power for detecting benefit in selected subgroups if the treatment benefit in those subgroups is rather sizable.) LV restoration hypothesis (mortality or hospitalization for a cardiac reason): 1. Power > 90% for detecting a 20% reduction in the primary endpoint with CABG + SVR if the event rate at 3 years in patients treated with CABG (without SVR) is 45% or higher. (Thus we have high power for detecting a conservative estimate of benefit if the control arm event rate is consistent with what is expected based on other studies (or even somewhat less than expected based on other studies). 2. Power > 85% for detecting a 20% improvement if the 3-year control event rate should be as low as 40%. (Thus we have high power for detecting a conservative estimate of the benefit if the control arm event rate is lower than expected.) 3. Similar power for detecting effects in subgroups as outlined above for the coronary revascularization hypothesis. Secondary Endpoints: Cardiac Mortality Since a very high proportion of the deaths in this patient population will be cardiovascular, the power for detecting differences in this secondary endpoint will differ little from that outlined above for the primary endpoint of total mortality in assessing the coronary revascularization hypothesis. Economics and Quality of Life Endpoints All patients enrolled in STICH will participate in the quality of life component of the trial, and the economic component will be based on patients enrolled in the United States. With the projection that a high proportion of the patients enrolled in STICH will come from the U.S. (at least 75%) and thus an even higher percentage from North America, the patient numbers that will be available for each of the major treatment comparisons will provide approximately 90% power for detecting a difference between treatment arms of $2,000 or greater in cumulative total medical costs and 80% power for detecting treatment differences in total cost of $1,200. For quality of life, the study will have 90% power for detecting a treatment difference of ¼ standard deviation in the Kansas City Cardiomyopathy scale and 90% power for detecting a clinically meaningful difference in patient utilities (based on the EuroQol) and in the other scales outlined in the EQOL Core Laboratory manual, even taking into account the loss of data due to patient death or incapacity. DCRI/Confidential Page 39 01/18/2002

41 6.1.2 Sample Size -- Summary In summary, 2,800 patients will provide excellent power for detecting clinically relevant and realistic treatment benefits for both of the primary hypotheses as well as for secondary endpoints. Furthermore, this number is robust (maintains good power) even under conservative assumptions about the control group event rates, the magnitude of the benefit from the interventions, and the effects of crossovers or dropouts. Therefore 2,800 patients will be the target enrollment for STICH. Approximately 1,000 of the patients will be randomized to the arm receiving intensive medical therapy, approximately 1,300 will be allocated to CABG without SVR (although only patients in Strata A and B will be used for assessing the coronary revascularization hypothesis), and a total of approximately 500 patients will be assigned to receive CABG + SVR (Figure 1). It should be noted that the total enrollment is predicated on the assumption that approximately 200 patients will be enrolled in the CABG arm in Stratum B, and hence can be used for testing both of the primary hypotheses. The numbers shown in Figure 1 for each stratum, however, are only estimates. The main consideration for testing the coronary revascularization hypothesis is that we enroll a total of 2,000 patients for comparing MED vs. CABG (without SVR). The relative proportion of the 2,000, which comes from Stratum A versus Stratum B, is not particularly important for testing the revascularization hypothesis. If there should be more than 200 CABG patients in Stratum B, we could enroll correspondingly fewer patients in Stratum A. If there are fewer than 200 CABG patients in Stratum B, we would expect to enroll correspondingly more in Stratum A. Under this latter scenario (fewer than 200 CABG patients in Stratum B), the number of patients from Stratum B available for assessing the LV restoration hypothesis would be reduced. If one considers the unlikely but most extreme case of no patients in Stratum B [i.e., no overlap of CABG patients for testing the two primary hypotheses (which would basically amount to considering the study as two separate non-overlapping trials)], the study would enroll 2,000 patients in Stratum A for the MED vs. CABG comparison, and the remainder of the 2,800 patients would of necessity come from Stratum C. This would mean there would be only 800 patients for assessing the LV restoration hypothesis. Even with only 800 patients, however, the power for detecting a 20% reduction in the primary endpoint of mortality or cardiac hospitalization would still be approximately 90% if the three-year event rate is 50% or higher. Thus, the choice of 2,800 patients is robust also to perturbations in the relative numbers that might emerge from each of the three major strata, as long as the combined enrollment of MED and CABG patients (without SVR) from Stratum A and Stratum B is 2,000. To enroll 2,800 patients over a period of three years will require randomizing an average of 78 patients/month. With the number of high volume sites committed to participate in STICH, the enrollment goals of this trial are achievable. Additional sites will be added if needed in order to maintain the pace of recruitment that will be required to complete patient recruitment within the three-year enrollment period Treatment Crossovers The sample size calculations previously outlined do not include any adjustment to compensate for the fact that some patients enrolled in STICH may crossover from one treatment arm to another. Although participating clinical centers will be instructed concerning the importance of treatment compliance, and strict guidelines will be established regarding any changes to a patient s assigned therapy, we recognize that in certain instances a downstream change may be clinically indicated. Although all therapy changes and reasons for them will be documented, they are considered part of the initial treatment strategy, and groups will be compared with respect to the initially assigned treatment. Nonetheless, we have attempted to estimate the extent to which crossovers or dropouts might occur, and how such changes in therapy could affect the sample size/power figures cited above. From personal communication with investigators from the Bypass Angioplasty Revascularization Investigation (BARI), and from reviewing data in the Duke Cardiovascular Database, there is ample experience to suggest that the percent of consenting patients who are randomized to CABG, but DCRI/Confidential Page 40 01/18/2002

42 don t actually receive CABG surgery, will be low (1-2%). We expect there will be almost no crossovers from CABG without SVR to CABG with SVR. We anticipate there will be no more than 5% of the patients randomized to SVR who do not actually undergo the SVR procedure. Crossover is most likely to occur among patients randomized to MED who, because of deteriorating symptoms or evidence of ischemia during follow-up, are felt to need revascularization with CABG. We project that up to 10% of MED patients may crossover to CABG within the first year following randomization, and up to 15% may crossover to CABG within three years. Some of the MED patients will receive percutaneous intervention. Indeed there may be patients in all three arms of the trial who undergo percutaneous intervention during follow-up, although this is more likely to occur among patients randomized to MED. Percutaneous intervention is not regarded as a treatment crossover, but rather as part of the downstream care associated with any of the treatment strategies in the trial. Even if we assume that dropouts and treatment crossovers would reduce the effective sample size by as much as 20% (the impact is not expected to be that large), the randomization of 2,000 patients for testing the coronary revascularization hypothesis will still provide > 90% power for detecting a 25% reduction in mortality and > 80% power for detecting a 20% improvement if the event rate in MED patients is consistent with our projection (25% at three years) based on other recent studies involving intensive medical care. For the LV restoration hypothesis, the planned sample size of 1,000 patients will provide 90% power for detecting a 20% reduction in the primary endpoint even with liberal estimates for the amount of crossover if the threeyear event rate for mortality or cardiac hospitalization is 50% or higher. 6.2 Statistical Analysis Statistical analysis will be performed at the Statistical Data Coordinating Center at Duke University. Although the methodologic approaches and operational details of the data analysis will be coordinated by the study biostatisticians, the major analyses of the study data will be highly collaborative, involving both statisticians and physicians to ensure appropriate interpretation of the data. All major treatment comparisons between the randomized groups in this trial will be performed according to the principle of "intention-to-treat;" that is, subjects will be analyzed (and endpoints attributed) according to the treatment arm to which patients were randomized, regardless of subsequent crossover. Statistical comparisons will be performed using twosided significance tests. Additional perspective regarding the interpretation of the data will be provided through extensive use of confidence intervals 33 and graphical displays Background and Demographic Characteristics Baseline demographic and clinical variables, including relevant descriptors from the history and physical examination; risk factors; comorbidity; HF functional class, angiographic assessment of the extent of CAD, left ventricular function, and descriptors obtained from the various noninvasive tests will be summarized for each randomized arm of the study. Descriptive summaries of the distribution of continuous baseline variables will be presented in terms of percentiles (e.g., median, 25th and 75th percentiles), while discrete variables will be summarized in terms of frequencies and percentages. Statistical comparisons of treatment groups with respect to baseline characteristics will be limited to selected variables and disease factors known to influence prognosis. These variables will include age; sex; race; previous myocardial infarction; prior revascularization; descriptors of comorbidity; HF functional class; anatomic severity of disease; and left ventricular function (ejection fraction). For comparisons of treatment groups with respect to continuous baseline variables, emphasis will be given to nonparametric procedures such as the Wilcoxon rank sum test, or Kruskal-Wallis nonparametric analysis of variance. 34 Group comparisons with respect to discrete baseline variables will use the conventional chi-square test. DCRI/Confidential Page 41 01/18/2002

43 6.2.2 Concomitant Therapy Summary statistics will be provided as appropriate. No formal analyses are planned Efficacy Evaluation Coronary Revascularization Hypothesis For the coronary revascularization hypothesis, the major statistical assessment will involve a comparison of the CABG arm (without SVR) versus the medically treated arm with respect to all-cause mortality. The logrank test 29 (sometimes called the Mantel-Haenszel test for survival data 35 ), which is a special case of the more general Cox proportional hazards regression model, will be the primary analytic tool for assessing mortality differences between these two treatment arms. The log-rank test (and the Cox model) can accommodate varying lengths of patient follow-up, and not only uses information for each patient as to whether an event occurred, but also takes into account how long from study entry the patient survived before the endpoint occurred. The log-rank test also accommodates "censored" survival times, which arise because many patients will be alive when the analyses are performed, and the length of time they will survive without an event is known only to be greater than the length of their current follow-up. The log-rank test is a wellestablished approach for providing an overall comparison of the entire survival curves. Using this procedure, the analysis strategy will be to perform the treatment comparison stratifying by whether or not patients are SVR eligible (i.e., by whether patients are in Stratum A, or in Stratum B; see Figure 1). The analysis is thus reflective of the fact that the randomization will be stratified according to whether patients are in Stratum A versus Stratum B. This comparison will constitute the primary analysis to assess the effect of CABG on overall mortality. Kaplan-Meier survival estimates 36 based on the primary endpoint of all-cause mortality will be calculated for each treatment group to display the outcome results graphically. As previously noted, the coronary revascularization hypothesis will be assessed by comparing MED versus CABG (without SVR). We note that an alternative approach would have been to include all CABG patients from Stratum A and Stratum B in this comparison (including those assigned to SVR), and test the hypothesis by performing a stratified analysis of MED versus CABG using groups 1 and 3 versus groups 2, 4, and 5 (Figure 1). However, because CABG + SVR may have a different outcome than CABG without SVR (either better or worse), the approach that is planned will provide a cleaner, clearer test of the coronary revascularization hypothesis without the ambiguities that could result from mixing SVR and non-svr CABG patients together. Even though the primary comparison outlined above will contrast treatment groups that are determined by randomization, supplementary analysis involving other covariate adjustment will be performed with the Cox model. Such adjustment will be limited to a relatively small, prospectively defined set of patient characteristics that are known a priori to have a strong prognostic relationship with mortality. This adjustment will serve as a prelude to supplementary analyses examining differential treatment effects. The adjustment variables will include (in addition SVR eligibility) age, sex, race, HF class, history of myocardial infarction, previous revascularization, number of significantly diseased major coronary arteries, and ejection fraction. Cox model analyses may also be performed using study site as a stratification factor. If the data provide evidence of an overall difference in outcome between treatment groups, we will further examine whether the therapeutic effect is similar for all patients, or whether it varies according to specific patient characteristics. In particular, we will focus on whether the relative therapeutic benefit differs according to patient age, sex, race, HF class, ejection fraction, and whether patients are eligible for SVR. These issues will be addressed formally with the Cox model by testing for interactions between treatment and these specific baseline variables. DCRI/Confidential Page 42 01/18/2002

44 Although the analysis will include an examination for treatment interactions as indicated above, treatment effects for the primary endpoint as characterized by the CABG:MED hazard ratio (with 95% confidence intervals) will be calculated and displayed for several prospectively-defined subgroups of patients defined by age, gender, race, SVR eligibility, HF class, and ejection fraction. These comparisons will be carefully interpreted in conjunction with the formal interaction tests. Because the final core laboratory assessment of ejection fraction will not always be available when a patient is randomized, there may be a small number of patients enrolled where the final core laboratory assessment of ejection fraction is greater than.35. A secondary analysis of the coronary revascularization hypothesis (i.e., an assessment of mortality differences in MED vs. CABG) will be performed in the patients whose ejection fraction is confirmed by core laboratory review to be.35. We note that the core laboratory will have no knowledge of the randomized treatment assignment and hence cannot possibly be biased in their assessment as a result of knowing which treatment the patient has been assigned. To supplement the conventional significance testing and confidence interval approaches that will constitute the major analyses for this trial, we will provide additional perspective on the assessment of treatment effects using some established Bayesian approaches to the analysis of clinical trial data. The application of Bayesian methods to clinical trials has recently received considerable attention in the statistical and clinical trials literature. In part, its appeal stems from the fact that what consumers of clinical trial results often want to know is the likelihood (probability) that the treatment is actually beneficial, or the likelihood that the treatment has a clinically important benefit. Such probability assessments are directly obtainable from a Bayesian analysis. Spiegelhalter et al. 37 demonstrated how one can derive clinically useful information such as an estimate of the probability that the hazard ratio (CABG: medical therapy) is less than some specified value (e.g., 0.90) and an estimate of, for example, the probability that CABG therapy is clinically equivalent to intensive medical therapy, i.e., that the hazard ratio is within some interval close to 1.0 (such as 0.90 to 1.10). We will supplement the conventional statistical presentations discussed above by computing Bayesian probabilities that CABG therapy is beneficial and that it has a clinically important benefit. For these computations, a flat (non-informative) prior distribution for the CABG:MED hazard ratio will be assumed. We will use readily available S-Plus functions to perform the Bayesian calculations, or the Cambridge group s BUGS program (Bayesian Inference Using Gibbs Sampling 38 ) to derive a posterior distribution from a full Bayesian analysis. LV Restoration Hypothesis For the LV restoration hypothesis, the key statistical assessment will involve a comparison of CABG plus SVR versus CABG without SVR with respect to the composite endpoint of death (all cause) or hospitalization for a cardiac reason. The log-rank test will also be the primary analytic tool for this assessment. In this case, the test will be performed stratifying by whether the patients were eligible for either medical or surgical therapy versus whether surgery was indicated (i.e., by whether the patients were in Stratum B or Stratum C as depicted in Figure 1. Again, Kaplan-Meier estimates will be calculated and plotted for each treatment group to display the outcome results graphically. In this case, the Kaplan-Meier curves will display event-free survival, i.e., survival free of any cardiac hospitalization. Again, the primary analysis will be supplemented by Cox model analyses involving other covariate adjustment. In addition to the stratification factors specified above for the primary LV restoration hypothesis, the small, prospectively-defined set of patient characteristics listed above as covariates for the coronary revascularization hypothesis will also be used in supportive analyses of the LV restoration hypothesis. DCRI/Confidential Page 43 01/18/2002

45 If the data provide evidence of an overall difference in outcome between treatment groups, we will determine whether the therapeutic effect is similar for all patients, or whether it varies according to specific patient characteristics. In particular, we will assess whether the relative therapeutic benefit differs according to patient age, sex, race, HF class, ejection fraction, and whether or not the patients are eligible for either medical or surgical therapy. This will be addressed formally with the Cox model by testing for interactions between treatment and these specific baseline variables. A secondary analysis of the LV restoration hypothesis (i.e., an assessment of CABG with SVR vs. CABG without SVR) will be performed in the patients where final core laboratory review confirms the ejection fraction to be.35, ESVI to be 60 ml/m 2, and akinesia 35% of the anterior LV wall. We will provide additional perspective on the assessment of treatment effects with respect to the LV restoration hypothesis through calculating the hazard ratio for CABG with SVR relative to CABG without SVR and 95% confidence intervals (overall and in prospectively defined subgroups), and also through performing supplementary Bayesian analysis as outlined above for the coronary revascularization hypothesis. Analysis of Secondary Endpoints Secondary endpoints that will be evaluated in this trial include (1) cardiovascular mortality; (2) all-cause mortality within 30 days; (3) quality of life; (4) cost of care and cost effectiveness; (5) morbidity; (6) functional measures, including six minute walk distance and exercise duration on a treadmill test; and (7) physiologic measures from noninvasive testing, neurohormonal levels, and cytokine levels. Data analyses for each of these endpoints are discussed below. Cardiovascular Mortality This endpoint is most relevant for the assessment of the effects of coronary revascularization, and therefore will be used in comparisons of medical therapy vs. CABG (without SVR). The analysis will be similar to that outlined for the primary endpoint, using time from enrollment until death from a cardiovascular cause or censoring as the response variable, and assessing treatment differences using the Cox proportional hazards model. Kaplan-Meier survival curves will be computed to graphically display the survival experience of the treatment groups as a function of time from randomization. Mortality within 30 days The analysis for this endpoint will also be similar to that outlined for the primary mortality endpoint except that 30-day outcome will be treated as a binary endpoint (due to its short-term nature), and the logistic regression model will be used for treatment comparisons rather than the Cox proportional hazards model. These analyses will be performed for both MED versus CABG and for CABG + SVR versus CABG (without SVR) in order to fully characterize and compare procedural mortality. Thirty-day mortality rates will be reported for each treatment arm along with odds ratios and 95% confidence intervals for CABG:MED and for CABG + SVR:CABG (without SVR). Quality of Life and Cost of Care The important Quality of Life and cost of care endpoints for the quality of life and economic analyses and the analysis plans are addressed in detail in the EQOL Core Laboratory manual. These important endpoints include health status measures from the Medical Outcomes Study 36 Item Short Form (SF-36); quality of life as assessed by the disease-specific Kansas City Cardiomyopathy Questionnaire; a measure of depressive symptoms based on the Center for Epidemiologic Studies-Depression (CES-D) scale; and utilities assessed by patient interview using the EuroQol. DCRI/Confidential Page 44 01/18/2002

46 Morbidity Morbidity in this trial will be assessed by examining the incidence of several different complications and clinical events that occur during the follow-up period. These events will include complications associated with each treatment arm such as the frequency and duration of hospitalization for HF, additional cardiac procedures (heart transplantation, LVAD, ICD, and stroke). We outline below the approach that we will take for analyzing one of these important complications, for example hospitalization for HF, noting that specific components of this approach can also be applied to the other non-fatal complications. The approach involves several related yet different analyses to comprehensively assess and fully characterize differences among the treatment arms with respect to this outcome. One characterization of this secondary endpoint will be in terms of the time from randomization until a patient experiences their first hospitalization for HF during their follow-up in the trial. The methodology for analyzing censored failure-time data outlined previously for the primary endpoint will thus be applied to compare treatment groups with respect to this outcome. We point out that this analysis must be interpreted cautiously, however, because hospitalization as a non-fatal event may appear to occur with a lower incidence in one treatment group simply because that arm had a higher death rate. Obviously after patients die, they can no longer be hospitalized for their HF. To deal with this complexity, we will perform further analyses of this adverse complication by considering hospitalization for HF and death as a combined endpoint. We will thus consider the time until the first occurrence of either death (any cause) or hospitalization for HF, and treatment differences will be assessed using the Cox proportional hazards model. In support of this analysis, Kaplan- Meier event-free survival curves for each treatment arm will be calculated to graphically display event rates over time for this composite endpoint. The third approach for analyzing this endpoint will take advantage of the fact that the two components of the composite endpoint (death or hospitalization) have a natural rank-ordering in terms of their severity. The treatments can thus be compared with an approach that not only uses the time until the first occurrence of one of these component outcomes, but also factors in the severity of the outcome. The approach we will employ, developed by Berridge and Whitehead, 39 combines the Cox proportional hazards model in conjunction with an ordinal severity of event model to potentially increase power for the treatment comparisons. The severity of event portion is a continuation ratio-type ordinal logistic model. An extended hazard function is defined by multiplying the regular Cox hazard function by the probability of the event being of severity j from the ordinal model, where j references the various severity categories. In this analysis, death will obviously be considered as the most severe event, and hospitalization for HF as least severe. Within this framework, one can test whether one of the interventions under study in this trial prolongs the time to the first occurrence of either of these events, whether the therapy decreases the severity of the first event that occurs, or (with 2 degrees of freedom), whether the therapy has either effect. Finally, since some patients may experience hospitalization multiple times, a comprehensive comparison of the treatment arms with respect to this outcome should take into account the longitudinal pattern in the repeated occurrences of such an event. In clinical situations where there may be multiple events per subject, it is desirable to be able to apply time-to-event methods such as the proportional hazards model which explicitly allow for varying lengths of patient follow-up and appropriately handle censored observations. A major issue in extending proportional hazards regression models to this situation, however, is intra-subject correlation (i.e., where multiple events occur within the same patient, those events will be correlated). Fortunately, there is active, ongoing methodological research on the application of survival models to this situation In order to provide a more complete and comprehensive analysis of the hospitalization data, the approach we will use in STICH is based on extensions of the proportional hazards model that accommodate multiple events per patient. The specific approach makes use of cluster modifications of so-called sandwich estimates of the DCRI/Confidential Page 45 01/18/2002

47 variance-covariance matrix of the regression coefficients, thus providing standard errors of the regression coefficients that take into account the correlations among multiple event times within a given patient Specialized software functions for performing such analyses are available in S-Plus. 43 Because of the problems already mentioned in analyzing hospitalization by itself, we will use the multipleevent methodology outlined above, which not only allows multiple events per patient of the same type (e.g., multiple hospitalizations), but also accommodates events of different types. Thus we will model both death and the multiple episodes of rehospitalization using multiple-events Cox model analysis By synthesizing the results from these different but complementary approaches (that is, by considering hospitalization in terms of a single event per patient, in terms of multiple events per patient, and hospitalization combined with death), we will be able to provide a comprehensive assessment of treatment differences with respect to hospitalization, and as part of the analyses, identify and assess other clinical factors that are associated with this important secondary outcome. The approaches outlined above for analyzing hospitalization for HF can also be applied to other clinical outcomes and adverse events that will be of interest in this trial, such as stroke and other morbidity outcomes enumerated above. Functional Capacity Outcomes These outcomes will include distance covered in a six minute walk test and exercise duration on a treadmill test. Of interest will be both the absolute magnitude of these measures as obtained during follow-up and also the change from baseline to the follow-up measurement. Treatment group differences will be statistically assessed using the nonparametric Wilcoxon rank sum test. Physiologic Measures from Noninvasive Tests (Core Laboratory Data) There will be considerable noninvasive test data acquired in the trial, including cardiac magnetic resonance imaging, echocardiography, and nuclear cardiology (radionuclide) studies. The schedule for obtaining these various tests is outlined in Table 1 and generally involves the acquisition of data at baseline, at four months following study entry, and after two years of follow-up. This information will be valuable in addressing the mechanisms responsible for differences or similarities in patient outcomes observed among the randomized arms in the trial. Furthermore, with treatment groups defined by randomization and tests systematically performed and uniformly interpreted, bias in the choice of therapy and in the choice of which tests are obtained is eliminated. The data from these various tests may be very important for identifying subsets of patients where there are clear benefits of myocardial revascularization, and perhaps other subsets where intensive medical therapy is an attractive choice. Even where revascularization appears to be a superior strategy, there may be some patients who do and others who do not benefit from SVR. The noninvasive information acquired in the trial will therefore be very useful in defining a pretreatment strategy for performing a sequence of tests that will optimize the choice of therapy. The analysis of the data from these tests will consist of several parts: (1) a detailed assessment of the prognostic relationships of the baseline test data to clinical outcomes, and of the relative prognostic importance of each test; (2) an examination of differential treatment effects as a function of the baseline noninvasive test data; (3) a comparison between treatments of the changes in test results from baseline to four months as a characterization of early mechanistic effects of therapy; (4) an assessment of the four month measures and the change from baseline to four months as predictors of subsequent outcomes; (5) a comparison of treatments with respect to the two year test outcomes and the change from baseline to two years as viewed from the perspective of an endpoint, and also as it might relate to late clinical outcomes; (6) DCRI/Confidential Page 46 01/18/2002

48 an analysis comparing treatment arms with respect to the trajectory of key noninvasive test measures, considering all three measurement times (baseline, four months, two years). For the analyses in part (1), we will use the Cox proportional hazards model to examine individual and joint relations between baseline test results and length of survival (or event-free survival in the case of the LV restoration endpoint). For continuous baseline measures such as ejection fraction and ESVI, we will examine the shape and strength of the relationship by use of a flexible model-fitting approach involving cubic spline functions (cubic polynomials) The resulting functions will be graphically and statistically examined to assess the model assumptions and to judge the linearity of the relationships. Where relations are nonlinear, their shape will be characterized using spline functions. Determining how variables should be modeled will be an important step in characterizing prognostic relations and identifying which variables are most strongly related to long-term outcome. Once the important predictors from each individual test are identified, we will contrast the relative prognostic importance and the independent contribution of the various tests by jointly considering multiple tests. The modeling strategies to be used and the approach to evaluating the yield of each test are similar to those we have employed in previous studies. 47,48 Because there will likely be over 600 deaths occurring in the study population during the period of follow-up, the sample size (number of events) will adequately support these extensive analyses. For step (2), we will use the key features identified in step (1) from each noninvasive test, and statistically test for a differential treatment effect using the Cox model. This will involve testing for interactions between treatment and specific baseline test measures (e.g., ESVI, measures of myocardial viability and myocardial perfusion). For step (3), the four-month test results and the change from baseline to four months in the prognostic factors identified in step (1) above will be descriptively summarized, and treatments will be compared using the Wilcoxon rank sum test in the case of continuous or ordinal measures, and the chi-square test for categorical variables. For step (4) where we are dealing with post-randomization information and attempting to judge its usefulness for predicting subsequent outcomes, the analysis becomes more complex. We will use two approaches. First, we will assess the relations of the four-month noninvasive test data to subsequent outcomes (outcomes after four months) in a fashion similar to that described above for the analysis of the baseline test data. Second, we will include events starting from randomization, but consider the four-month test results as intervening or time-dependent data, and factor this into the Cox model using time-dependent covariates. In both of these analyses, we will seek to address the issue of whether the four-month test results or the change from baseline to four months provides independent predictive information beyond that contained in the baseline test data, and thereby assess the utility of doing the follow-up studies. For step (5), treatment arms will be compared with respect to two-year test outcomes (considering the test results as an endpoint) using a similar approach as described for the four-month data in step (3). We will also attempt to judge the utility of the two-year test data in predicting subsequent outcomes as in step (4) above, recognizing, however, that the complexities of this analysis will require very cautious interpretation. For step (6), we will characterize the trajectory of key prognostic measures from the tests by considering all three measurement times. In considering the longitudinal nature of these data, there are three major statistical issues that must be considered in making treatment comparisons: (a) the data may not be normally distributed, (b) the longitudinal measures within subjects will be correlated, and (3) there will be missing data, resulting from patients in whom it is not possible to obtain all three measurements (because they die or DCRI/Confidential Page 47 01/18/2002

49 fail to return for other reasons to enable the follow-up measurements to be made). To address the first two issues, we will use the generalized estimating equation (GEE) approach to model estimation and testing proposed by Liang and Zeger. 49 The software for implementing this method is now available in a SAS macroroutine accessible to the CC. For the third issue above (missing data), we will, of course, make every effort to minimize the amount of missing data. However, because there will inherently be missing data, we will incorporate several strategies and assess the sensitivity of our conclusions to the missing data approach employed. The alternatives that will be considered include (1) analyze only the patients with complete data; (2) assign worst case scores to missing observations; (3) carry forward previously observed data to observation periods that are missed later; and (4) use likelihood-based methods to impute missing data. A description of these methods and their limitations is given in the review by Heyting. 50 By synthesizing the results from these different but complementary approaches, we will be able to provide a comprehensive assessment of treatment differences with respect to the key mechanistic noninvasive test data, judge whether there is a preference for one particular treatment depending on the results of the noninvasive tests, and develop a recommendation for performing the sequence of tests that will optimize the choice of therapy. With regard to physiologic measures such as neurohormonal levels and cytokine levels provided by the Neurohormone/Cytokine/Genetic Core Laboratory, there will be interest in both the absolute levels at followup and also in the change from baseline to follow-up. Treatment group differences will be statistically assessed using the Wilcoxon rank sum test applied to these various measures. Multiple Comparisons Although there are two primary aims, and there will be some overlap of the patients used for testing the two primary hypotheses [the overlapping patients will be those in Stratum B who are randomized to CABG (without SVR), projected to be approximately 200 patients], the statistical test for each primary hypothesis will be performed at the 0.05 level of significance. We justify this choice on the basis that the two hypotheses are addressing two very distinct questions. Indeed, the endpoints for the two questions are different, and were it not for the overlapping patients, the study would basically resemble two separate trials. With two primary hypotheses and the various secondary endpoints that have been outlined, we recognize that there is a multiplicity of analyses to be performed, which leads to an increased probability that at least one of the comparisons could be "significant" by chance. There are adjustments (e.g., based on the Bonferroni inequality) that can be used to preserve the overall type I error level. To attempt to adjust for the effects of the repeated significance testing that will occur as part of the interim monitoring (discussed below), plus adjust for the multiplicity of analyses, would require that small significance levels be used for every comparison. Instead of formally adjusting the significance level for the fact that there are two primary hypotheses and numerous secondary comparisons, we will be conservative in the interpretation of the analyses, taking into account the degree of significance, and looking for consistency across endpoints. The actual p-value for each comparison will be reported (not simply whether significance is achieved) to aid in the overall interpretation. Also, the Bayesian interpretations discussed above (along with so-called credible intervals) will assist in providing an appropriate interpretation of the study results. We have also prespecified the primary and secondary outcome variables to help guard against the multiple testing problem Safety Evaluation The assessment of safety is based mainly on the frequency of the pre-defined safety parameters and serious adverse events suspected by the investigator to be related to study treatment. Other safety data (e.g., vital signs) will be considered and summarized as appropriate. DCRI/Confidential Page 48 01/18/2002

50 Serious adverse events suspected by the investigator to be related to study treatment will be summarized for each treatment group by presenting the number and percentage of patients having any serious related adverse event, having a serious related event in each body system and having each individual serious related adverse event Interim Analysis For ethical reasons, an interim examination of key safety and endpoint data will be performed at regular intervals during the course of the trial. The primary objective of these analyses will be to evaluate the accumulating data for an unacceptably high frequency of negative clinical outcomes in any of the treatment arms. In addition, however, the interim monitoring will also involve a review of the control arm event rates, patient recruitment, compliance with the study protocol, submission of data forms, and other factors which reflect the overall progress and integrity of the study. The results of the interim analyses will be carefully and confidentially reviewed by a Data and Safety Monitoring Board (DSMB). It is anticipated that the DSMB will meet at approximately six-month intervals to review the accumulating data. Prior to each meeting, the Statistical Data Coordinating Center will conduct the desired statistical analyses and prepare a summary report that will be carefully reviewed by the DSMB. The extracted data files and analysis programs for each DSMB report will be archived and maintained at the Statistical Data Coordinating Center for the life of the study. Reports will be presented describing the progress of patient enrollment, the rates of compliance with therapy, and the frequency of protocol violations. The Statistical Data Coordinating Center will also report on the number of forms received, the number of forms entered in the computer, the number of forms queried, the number of queries completed, the number of forms reviewed through on-site monitoring, and the number of forms receiving the final, completed entry in the double dataentry system. These interim safety and efficacy reports introduce well-recognized statistical problems related to the multiplicity of statistical tests performed on an accumulating set of data. 51,52 As a solution to the problem of repeated tests, we propose to adopt for use in STICH a group sequential method similar to that proposed by O'Brien and Fleming 53 as a guide in interpreting interim analyses. This procedure requires large critical values early in the study, but relaxes (i.e., decreases) the critical value as the trial progresses. Because of the conservatism early in the trial, the critical value at the final analysis is near the "nominal" critical value. Hence the sample size requirements with this group sequential procedure remain essentially the same as the conventional fixed sample size estimate. The actual method for this interim monitoring that will be employed in STICH is the general approach to group sequential testing developed by Lan and DeMets 54 for which neither the number of looks nor the increments between looks must be pre-specified. Rather, the Lan- DeMets approach only requires specification of the rate at which the Type I error (which in this trial will be chosen to be α=0.05) will be "spent". This procedure allows "spending" a little of α at each interim analysis in such a way that at the end of the study, the total Type I error does not exceed One such spending function generates boundaries that are nearly identical to the O'Brien-Fleming boundaries. It is this approach that we propose to use in STICH, namely two-sided, symmetric O'Brien-Fleming 53 type boundaries generated using the flexible Lan-DeMets 54 approach to group sequential testing. Since the number of looks and the increments between looks need not be prespecified, it allows considerable flexibility in the monitoring process for accommodating additional comparative examinations of the data in response to concerns of the DSMB that may arise during the course of the trial. Assuming that the DSMB will conduct its first formal data review in the latter half of the first year of recruitment, and then continue those reviews approximately every 6 months thereafter through the patient recruitment period (three years) and the follow-up phase (three years), there will be approximately eight to ten reviews of the data. With nine interim analyses approximately equally spaced in time, the Lan and DeMets "spending function" that approximates the O'Brien-Fleming DCRI/Confidential Page 49 01/18/2002

51 stopping boundaries involves a very stringent alpha level ( ) for declaring significance at the first interim analysis. At the subsequent interim analyses, the required significance levels will be somewhat less stringent. The requirements for significance at each interim analysis, depending on exactly when the analysis occurs, can be computed with the Lan-DeMets methodology, for which we have suitable computer software. The final analysis can be undertaken with a significance level of approximately 0.04, relatively close to the nominal 0.05 level. This approach to interim monitoring will be carried out in parallel for both of the primary study hypotheses. The analytic approach that will be used at the interim analyses for assessing treatment differences will be the time-to-event analysis methods described above, except that interpretation of statistical significance associated with treatment comparisons of key study endpoints will be guided using the group sequential stopping boundaries outlined above The appropriateness of using the log-rank test (or equivalently the Cox model) in the group sequential framework has previously been well established For each of these interim analyses, the critical value of the test statistic and the corresponding p-value required for significance in that particular analysis will be presented so that significance can be assessed precisely. If significantly large and important treatment differences are observed at any of the interim analyses, the DSMB may recommend that randomization of patients be stopped, or that the design and conduct of the trial be appropriately modified. Judgment concerning the continuation or termination of the study will involve not only the degree of statistical significance observed at the interim analysis, but also the likelihood of achieving significance should enrollment continue to the originally projected sample size. As an aid in this latter assessment, the Statistical Data Coordinating Center will supplement the group sequential analyses outlined above with calculations of conditional power based on the method of stochastic curtailment This procedure evaluates the conditional probability that a particular statistical comparison will or will not be significant at the end of the trial at the α level used in the design, given the hypothesized treatment difference and the data obtained to date. Conditional power for the primary comparisons associated with both primary hypotheses will be computed and provided to the DSMB as part of the interim study reports. The approach to interim monitoring outlined above will be carried out in parallel for both of the primary study hypotheses. In addition, for the LV restoration hypothesis where the primary endpoint is a composite of death or cardiac hospitalization, it will also be important to carefully monitor the mortality component of this endpoint. Thus mortality rates and associated confidence intervals for each arm (CABG and CABG + SVR) will also be monitored at the interim reviews to ensure that the safety of patients enrolled in the trial is not compromised. 7. Ethics and Good Clinical Practice This study must be carried out in compliance with the protocol and in accordance with STICH standard operating procedures. These procedures are designed to ensure adherence to Good Clinical Practice, as described in the following documents: DCRI/Confidential Page 50 01/18/2002

52 1. ICH Harmonized Tripartite Guidelines for Good Clinical Practice Directive 91/507/EEC, The Rules Governing Medicinal Products in the European Community. 3. US 21 Code of Federal Regulations dealing with clinical studies (including parts 50 and 56 concerning informed consent and IRB regulations). 4. Declaration of Helsinki, concerning medical research in humans (Recommendations Guiding Physicians in Biomedical Research Involving Human Subjects, Helsinki 1964, amended Tokyo 1975, Venice 1983, Hong Kong 1989, Somerset West 1996). The investigator agrees, when signing the protocol, to adhere to the instructions and procedures described in it and thereby to adhere to the principles of Good Clinical Practice that it conforms to. 7.1 Institutional Review Board/Independent Ethics Committee Before implementing this study, the protocol, the proposed informed consent form and other information to subjects, must be reviewed by a properly constituted Institutional Review Board/Independent Ethics Committee (IRB/IEC). A signed and dated statement that the protocol and informed consent have been approved by the IRB/IEC must be given to the Coordinating Center before study initiation. The name and occupation of the chairman and the members of the IRB/IEC must be supplied to the Coordinating Center. Any amendments to the protocol, other than administrative ones, must be approved by this committee Informed Consent The investigator must explain to each subject (or legally authorized representative) the nature of the study, its purpose, the procedures involved, the expected duration, the potential risks and benefits involved and any discomfort it may entail. Each subject must be informed that participation in the study is voluntary and that he/she may withdraw from the study at any time and that withdrawal of consent will not affect his/her subsequent medical treatment or relationship with the treating physician. This informed consent should be given by means of a standard written statement, written in non-technical language. The subject should read and consider the statement before signing and dating it, and should be given a copy of the signed document. If written consent is not possible, oral consent can be obtained if witnessed by a signed statement from one or more persons not involved in the study, mentioning why the patient was unable to sign the form. No patient can enter the study before his/her informed consent has been obtained. Separate informed consents will be required for obtaining and sending an ECHO to the core laboratory to confirm eligibility prior to randomization if a site reading of LV EF.35 by contrast ventriculography, CMR and/or gated SPECT is unavailable, and for National Death Index follow-up in eligible patients who can not be approached for randomization consent due to treating physician preference. Eligible patients will be approached for consent to enter a specific Stratum determined by MED and SVR eligibility. If a patient refuses to provide consent for randomization, they will be offered entry into the registry and a separate registry informed consent will be provided. Once randomized, subjects will be asked to provide a separate informed consent for genotyping. The informed consent forms are part of the protocol, and must be submitted by the investigator with it for IRB/IEC approval. The Coordinating Center will supply proposed informed consent forms, which comply with regulatory requirements and are considered appropriate for the study. Any changes to the proposed consent form suggested by the Investigator must be agreed to by the Coordinating Center before submission to the IRB/IEC, and a copy of the approved version must be provided to the Coordinating Center after IRB/IEC approval. DCRI/Confidential Page 51 01/18/2002

53 8. References O Connell JB, Bristow MR. Economic impact of heart failure in the United States: Time for a different approach. Heart Lung Transplant 1993; 13:S107-S Kaesemeyer WH. Holding smokers accountable for heart disease costs. Circulation 1994; 90: Braunwald E, Bristow MR. Congestive heart failure: Fifty years of progress. Circulation 2000; 102:IV- 14 IV Reimer KA, Jennings RB. The Wavefront Phenomenon of myocardial ischemic cell death. II. Transmural progression of necrosis within the framework of ischemic bed size (myocardium at risk) and collateral flow. Lab Invest 1979; 40: Jatene AD. Left ventricular aneurysmectomy. Resection or reconstruction. J Thorac Cardiovasc Surg 1985; 89(3): Dor V, Saab M, Coste P, Kornaszewska M, Montiglio F. Left ventricular aneurysm: a new surgical approach. J Thorac Cardiovasc Surg 1989; 37(1): Di Donato M, Sabatier M, Dor V, Gensini GF, Toso A, Maioli M, Stanley AWH, Athanasuleas C, Buckberg G. Effects of the Dor procedure on left ventricular dimension and shape and geometric correlates of mitral regurgitation one year after surgery. J Thorac Cardiovasc Surg 2001; 121: Athanasuleas CL, Stanley AWH, Buckberg GD, Dor V, DiDonato M, Blackstone EH, and the RESTORE Group. Surgical anterior ventricular endocardial restoration (SAVER) in the dilated remodeled ventricle following anterior myocardial infarction. J Am Coll Cardiol 2001; 37: Jones RH. Is it time for a randomized trial of surgical treatment of ischemic heart failure? Invited Editorial. J Am Coll Cardiol 2001: Alderman EL, Bourassa MG, Cohen LS, Davis KB, Kaiser GG, Killip T, Mock MB, Pettinger M, Robertson TL for the CASS Investigators. Ten-year follow-up of survival and myocardial infarction in the randomized Coronary Artery Surgery Study. Circulation 1990; 82: Yusuf S, Zucker D, Peduzzi P, Fisher LD, Takaro T, Kennedy JW, Davis K, Killip T, Passamani E, Norris R, Morris C, Mathur V, Varnauskas E, Chalmers TC. Effect of coronary artery bypass grafting on survival: Overview of 10-year results from randomised trials by the Coronary Artery Bypass Graft Surgery Trialists Collaboration. Lancet 1994; 344: Baker DW, Jones R, Hodges J, Massie BM, Konstam MA, Rose EA. Management of Heart Failure. III. The role of revascularization in the treatment of patients with moderate or severe left ventricular systolic dysfunction. JAMA 1994; 272: Eitzman D, al-aouar Z, Kanter HL, et al. Clinical outcome of patients with advanced coronary artery disease after viability studies with positron emission tomography. J Am Coll Cardiol 1992; 20: Di Carli MF, Davidson M, Little R, et al. Value of metabolic imaging with positron emission tomography for evaluating prognosis in patients with coronary artery disease and left ventricular dysfunction. Am J Cardiol 1994; 73: Lee KS, Marwick TH, Cook SA, et al. Prognosis of patients with left ventricular dysfunction with and without viable myocardium after myocardial infarction: relative efficacy of medical therapy and revascularization. Circulation 1994; 90: Gioia G, Powers J, Heo J, Iskandrian AS. Prognostic value of rest-redistribution tomographic thallium-201 imaging in ischemic cardiomyopathy. Am J Cardiol 1995; 75: Vom Dahl J, Altehoefer C, Sheehan FH, et al. Effect of myocardial viability assessed by technetium-99m sestamibi SPECT and fluorine-18-fdg PET on clinical outcome in coronary artery disease. J Nucl Med 1997; 38: Afridi I, Grayburn PA, Panza JA, Oh JK, Zoghbi WA, Marwick TH. Myocardial viability during dobutamine echocardiography predicts survival in patients with coronary artery disease and severe left ventricular systolic dysfunction. J Am Coll Cardiol 1998; 32: DCRI/Confidential Page 52 01/18/2002

54 20. Cuocolo A, Petretta M, Nicolai E, et al. Successful coronary revascularization improves prognosis in patients with previous myocardial infarction and evidence of viable myocardium at thallium-201 imaging. Eur J Nucl Med 1998; 25: Pasquet A, Robert A, D Hondt AM, Dion R, Melin JA, Vanoverschelde JL. Prognostic value of myocardial ischemia and viability in patients with chronic left ventricular ischemia. Circulation 1999; 100: Senior R, Kaul S, Lahiri A. Myocardial viability on echocardiography predicts long-term survival after revascularization in patients with ischemic congestive heart failure. J Am Coll Cardiol 1999; 33: Chaudhry FA, Tauke JT, Alessandrini RS, Vardi G, Parker MA, Bonow RO. Prognostic implications of myocardial contractile reserve in patients with coronary artery disease and left ventricular dysfunction. J Am Coll Cardiol 1999; 34: Michels VV, Moll PP, Miller FA. The frequency of familial dilated cardiomyopathy in a series of patients with idiopathic dilated cardiomyopathy. N Engl J Med 1992; 326: Leiden JM. The genetics of dilated cardiomyopathy: emerging clues to the puzzle. N Engl J Med 1997; 337: Olson TM, Michels VV, Thibodeau SN, Tai YS, Keating MT. Actin mutations in dilated cardiomyopathy, a heritable form of heart failure. Science 1998; 280: Dilsizian V, Rocco TP, Freedman NMT, Leon MB, Bonow RO. Enhanced detection of ischemic but viable myocardium by the reinjection of thallium after stress-redistribution imaging. N Engl J Med 1990;323: Sciagra R, Bisi G, Santoro GM, et al. Comparison of baseline-nitrate technetium-99m sestamibi with restredistribution thallium-201 tomography in detecting viable hibernating myocardium and predicting postrevascularization recovery. J Am Coll Cardiol 1997;30: Kalbfleisch JD, Prentice RL. The Statistical Analysis of Failure Time Data. New York, NY: John Wiley & Sons, Inc; Cox DR. Regression models and life-tables (with discussion). J Royal Statist Soc B 1972;34: Breslow NE. Covariance analysis of censored survival data. Biometrics 1974;30: Schoenfeld DA. Sample-size formula for the proportional-hazards regression model. Biometrics 1983;39: Borenstein M. The case for confidence intervals in controlled clinical trials. Contro Clin Trials 1994;15: Lehmann EL. Nonparametrics: Statistical Methods Based on Ranks. San Francisco: Holden-Day Inc.; Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 1966;50: Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Statist Assn 1958;53: Spiegelhalter DJ, Freedman LS, Parmar MKB. Bayesian approaches to randomized trials. J Royal Stat Soc 1994;157: Thomas A, Spiegelhalter DJ, Gilks WR. BUGS: A program to perform Bayesian inference using Gibbs sampling. In: Bernardo JM, Berger JO, Dawid AP, Smith AFM. Bayesian Statistics. Oxford: Clarendon Press; 1992: Berridge DM, Whitehead J. Analysis of failure time data with ordinal categories of response. Stat Med 1991;10: Pepe MS, Cai J. Some graphical displays and marginal regression analyses for recurrent failure times and time dependent covariates. J Am Statist Assoc 1993;88: Wei LJ, Lin DY, Weissfeld L. Regression analysis of multivariate incomplete failure time data by modeling marginal distributions. J Am Statist Assoc 1989;84: Lee EW, Wei LJ, Amato D. Cox-type regression analysis for large numbers of small groups of correlated failure time observations. In: Klein JP, Goel PK, eds. Survival Analysis, State of the Art. Rotterdam: Kluwer Academic Publishers; DCRI/Confidential Page 53 01/18/2002

55 43. Therneau TM, Grambsch P. Modeling Survival Data: Extending the Cox Model. New York: Springer- Verlag; Harrell FE Jr, Pollock BG, Lee KL. Graphical methods for the analysis of survival data. Proc 12th Annual Conference. SAS Users Group International 1987;12: Harrell FE Jr, Lee KL, Pollock BG. Regression models in clinical studies: determining relationships between predictors and response. J Nat Cancer Inst 1988;80: Smith PL. Splines as a useful and convenient statistical tool. Am Statistician 1979;33: Harrell FE Jr, Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluating the yield of medical tests. JAMA 1982;247: Califf RM, Pieper KS, Lee KL, Van de Werf F, Simes RJ, Armstrong PW, Topol EJ for the GUSTO-I Investigators. Prediction of 1-year survival after thrombolysis for acute myocardial infarction in the Global Utilization of Streptokinase and TPA for Occluded Coronary Arteries Trial. Circulation 2000;101: Liang K-Y, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73: Heyting A, Tolboom JTBM, Essers JGA. Statistical handling of drop-outs in longitudinal clinical trials. Statistics in Medicine 1992;11: Armitage P, McPherson CK, Rowe BC. Repeated significance tests on accumulating data. J R Stat Soc Series A 1969;132: McPherson K. Statistics: the problem of examining accumulating data more than once. N Engl J Med 1974;290: O'Brien PC, Fleming TR. A multiple testing procedure for clinical trials. Biometrics 1979;35: Lan KKG, DeMets DL. Discrete sequential boundaries for clinical trials. Biometrika 1983;70: Geller NL, Pocock SJ. Interim analyses in randomized clinical trials: Ramifications and guidelines for practitioners. Biometrics 1987; Tsiatis, AA. The asymptotic joint distribution of the efficient scores tests for the proportional hazards model calculated over time. Biometrika 1981; 68: Tsiatis AA. Repeated significance testing for a general class of statistics used in censored survival analysis. J Am Stat Assoc 1982;77: Gail MH, DeMets DL and Stud EV. Simulation studies on increments of the two-sample log rank score test for survival data, with application to group sequential boundaries. Edited by R. Johnson and J. Crowley. In Survival Analysis. Monograph Series 2. Hayward, California: IMS Lecture Notes, 1982; DeMets DL, Gail MH. Use of logrank test and group sequential methods at fixed calendar times. Biometrics 1985;41: Halperin M, Lan KKG, Ware JH et al. An aid to data monitoring in long-term clinical trials. Controlled Clin Trials 1982;3: Lan KKG, Simon R and Halperin M. Stochastically curtailed tests in long-term clinical trials. Commun Stat., Sequential Analysis. 1982;1: Ware JH, Muller JE, Braunwald E. The futility index, an approach to the cost-effective termination of randomized clinical trials. Am J Cardiol 1985;78: DCRI/Confidential Page 54 01/18/2002

56 PROTOCOL Surgical Treatment for Ischemic Heart (STICH) Failure Trial U01 HL69015 Authors: Eric J. Velazquez, Kerry L. Lee, Daniel B. Mark, Christopher M. O Connor, Robert M. Califf, Robert H. Jones Document Type: Clinical Study Protocol Document Status: Version 04 Date: June 16, 2006 Number of Pages: 41 Property of DCRI Confidential May not be used, divulged, published or otherwise disclosed without the consent of DCRI

57 Table of Contents List of Tables 4 List of Abbreviations 5 1. Overview of STICH Trial 6 2. Study Objectives Primary Objectives Secondary Objectives Other Key Parameters 7 3. Background Public Health Burden of Heart Failure Development of Surgical Ventricular Reconstruction Operation Need for a Randomized Trial of Medical and Surgical Therapies Evaluation of Noninvasive Studies Cost and Quality of Life Evaluation Neurohormonal, Cytokine, and Genetic Studies Significance Investigational Plan Overview Study Population Definition of Patient Strata Study Hypotheses Inclusion and Exclusion Criteria Study Subject Enrollment Screening Randomization Baseline Evaluation of Randomized Patients Prior to Assigned Treatment Initiation and Completion of Initial Treatment Treatment Groups Interruption or Discontinuation of Treatment Study Subject Follow-up Efficacy Assessments Primary Efficacy Parameters Secondary Efficacy Parameters 19 DCRI/Confidential Protocol 4 06/16/2006 Page 2

58 4.6.3 Other Key Efficacy Parameters Safety Assessments Quality of Life Assessments Economic Data/Hospital Billing Data (U.S. sites only) Echocardiographic Assessment Nuclear Cardiology Assessment CMR Assessment Neurohormonal, Cytokine, and Genetic Assessments Study Organization Monitoring Committees Operational Committees Data Management Data Collection Data Forms Manual of Operations STICH Trial Website Database Management and Quality Control Statistical Plan Sample Size Determination - Overview Sample Size and Power Considerations Treatment Crossovers Statistical Analysis Background and Demographic Characteristics Concomitant Therapy Efficacy Evaluation Safety Evaluation Interim Analysis Ethics and Good Clinical Practice Institutional Review Board/Independent Ethics Committee Informed Consent References 39 DCRI/Confidential Protocol 4 06/16/2006 Page 3

59 List of Figures Figure 1. Patient Stratum and Treatment Assignment at Time of Completion of 14 Hypothesis 2 Enrollment on January 24, 2006 List of Tables Table 1. Eligibility Criteria 13 Table 2. Baseline Evaluation Studies 17 Table 3. Follow-up Evaluation Studies 19 Table 4. Hypothesis 1 - Statistical Power in Relation to Number of Patients and Length 27 of Follow-up Table 5. Hypothesis 1 - Enrollment, Follow-up, and Statistical Power 27 DCRI/Confidential Protocol 4 06/16/2006 Page 4

60 List of Abbreviations ACE Angiotensin converting enzyme AE Adverse event AICD Automatic implantable cardioverter defibrillator AMI Acute myocardial infarction ARB Angiotensin receptor blocker CABG Coronary artery bypass graft CCS Canadian Cardiovascular Society CI Confidence interval CMR Cardiac magnetic resonance CRF Case report form DCRI Duke Clinical Research Institute DDCD Duke Databank for Cardiovascular Diseases DSMB Data and Safety Monitoring Board ECG Electrocardiogram ECHO Echocardiography EF Ejection fraction EQOL Economics and quality of life ESVI End-systolic volume index HF Heart failure ICH International conference on harmonization IEC Independent Ethics Committee IRB Institutional Review Board IVRS Interactive voice response system LVAD Left ventricular assist device LVEF Left ventricular ejection fraction MED Medical strategy of heart failure therapy MI Myocardial infarction mmhg Millimeters of mercury OR Odds ratio PCI Percutaneous coronary intervention QOL Quality of life SAE Serious adverse event SPECT Single photon emission computed tomography SVR Surgical ventricular reconstruction WHO World Health Organization DCRI/Confidential Protocol 4 06/16/2006 Page 5

61 1. Overview of the STICH Trial The Surgical Treatment for Ischemic Heart Failure (STICH) multicenter international randomized trial addresses two specific primary hypotheses in patients with left ventricular (LV) dysfunction who have coronary artery disease (CAD) amenable to surgical revascularization: 1) Coronary artery bypass grafting (CABG) with intensive medical therapy (MED) improves long-term survival compared to MED alone; 2) In patients with anterior LV dysfunction, surgical ventricular reconstruction (SVR) to a more normal LV size improves survival free of subsequent hospitalization for cardiac cause in comparison to CABG alone. Important secondary endpoints include morbidity, economics, and quality of life. Core laboratories for cardiac magnetic resonance (CMR), echocardiography (ECHO), neurohormonal/cytokine/genetic (NCG), and radionuclide (RN) studies will insure consistent testing practices and standardization of data necessary to identify eligible patients and to address specific questions related to the primary hypotheses. Over 4.5 years, 125 clinical sites will recruit approximately 2,000 consenting patients with LV ejection fraction (EF).35 and CAD amenable to CABG. Patients with ongoing CCS III angina intensity or presence of left main coronary stenosis will be classified as ineligible for medical therapy. Patients eligible for SVR but ineligible for randomization to medical therapy will be evenly randomized to CABG with or without SVR. Patients eligible for medical or surgical therapy, but not SVR eligible, will be evenly randomized between MED only and MED with CABG. The remaining patients eligible for both MED and SVR will be randomized between three treatments of MED only, or MED + CABG, or MED + CABG + SVR. Registries of clinical information will be maintained on a subset of eligible patients who decline trial entry. At four-to six-month intervals for a minimum of two years, all randomized patients will be followed by a clinic visit or by telephone. Appropriate subgroups of randomized patients will have core laboratory studies repeated at specified follow-up intervals. In the patients randomized to MED only or to CABG with MED, CABG is hypothesized to demonstrate a 25% reduction in the primary endpoint of all-cause death from the projected 25% three-year mortality for MED. In the SVR-eligible patients, CABG + SVR is hypothesized to show a 20% advantage in the endpoint of survival free of hospitalization for cardiac cause projected to be 50% at three years in patients receiving CABG without SVR. The trial is designed to test both hypotheses with 90% statistical power. Definition of efficacy of potential therapies and their mechanisms of benefit by the STICH Trial is certain to inform future choice of therapy and thereby extend and improve the quality of life of millions of patients who now suffer from ischemic heart failure (HF). 2. Study Objectives 2.1 Primary Objectives 1. To test the hypothesis that improvement in myocardial perfusion by CABG combined with intensive medical therapy (MED) improves long-term survival compared to MED alone in patients with LV dysfunction and CAD amenable to CABG. 2. To test the hypothesis that in patients with anterior LV akinesia, LV shape and size optimization by SVR combined with CABG and MED improves long-term survival free of cardiac hospitalization compared to CABG and MED without SVR. DCRI/Confidential Protocol 4 06/16/2006 Page 6

62 2.2 Secondary Objectives 1. To assess the clinical utility of the measurements by CMR of LV shape, size and function for predicting the benefit of a specific treatment strategy. 2. To assess the clinical utility of measurements by nuclear cardiology and/or echocardiography testing of myocardial ischemia, viability, and LV size and function for predicting preferential benefit associated with a specific treatment strategy. 3. To determine if the use of baseline ECHO derived assessments of LV systolic and diastolic function, hemodynamics, and valvular regurgitation identify subgroups that derive either substantial benefit or harm from surgical intervention and how CABG compared to MED and CABG + SVR compared to CABG differentially effects these parameters over time. 4. To test the hypothesis that levels of neurohormonal and pro-inflammatory cytokine activation will be useful in identifying that group of patients who will most likely benefit from either CABG and/or CABG + SVR. 5. To compare health-related quality of life, total medical costs, and incremental cost effectiveness between CABG and MED and between CABG with or without SVR. 2.3 Other Key Parameters Mortality: 30-day all-cause; cardiovascular; all-cause death or LVAD or heart transplantation. Morbidity: All-cause hospitalization; days of all-cause hospitalization; hospitalization for HF; days of hospitalization for HF; emergency department visits for HF. Cardiac Procedures: PCI; any cardiac surgery; LVAD; heart transplantation; ICD. Cost: total medical costs; cost-effectiveness. Functional Capacity: quality of life; six-minute walk distance (absolute and change); treadmill duration (absolute and change). Physiologic: LVEF (absolute and change); ESVI (absolute and change); neurohormonal levels (absolute and change); cytokine levels (absolute and change). 3. Background 3.1 Public Health Burden of Heart Failure American Heart Association statistics suggest that HF affects 4.7 million patients in the United States and is responsible for approximately one million hospitalizations and 300,000 deaths annually. 1 The total annual costs associated with this disorder have been estimated to exceed $22 billion. 2 CAD is the cause of HF in the majority of patients, and HF is the only mode of CAD presentation associated with increasing incidence and mortality. 3 Ischemic HF may develop after a known myocardial infarction or through a process of chronic ischemia that leads to cardiac dilatation, cellular hypertrophy, apoptosis, and extracellular matrix remodeling. 4 The increasing prevalence of HF largely derives from enhanced survival from improved therapies for other manifestations of CAD, especially acute myocardial infarction. However, survival is usually accompanied by some degree of myocardial injury that is the basis for development of HF. DCRI/Confidential Protocol 4 06/16/2006 Page 7

63 3.2 Development of Surgical Ventricular Reconstruction Operation Reimer and Jennings showed that experimental coronary occlusion results in a wavefront of cell death moving from the subendocardial zone across the wall to involve progressively more of the transmural thickness of the ischemic zone. 5 Early reperfusion salvages subepicardial but not subendocardial myocardium. The decline in prevalence of thin-walled LV aneurysm that followed introduction of aggressive initial management of acute coronary syndromes suggests that the ischemic process also can be interrupted in patients before it reaches the transmural stage. With later healing of infarcted myocardium, scarring is maximal in the subendocardial region and is interlaced with normal myocardium in diminishing amounts toward the epicardial surface. Ventricular wall or chamber imaging commonly shows an akinetic zone that gradually blends with myocardium with increasingly normal function in contrast to the dyskinetic region with a discrete neck typical of a left ventricular aneurysm. LV ventricular aneurysmectomy appears to reverse HF by lowering LV wall stress, but these linear amputations of dyskinetic scar commonly deform the LV cavity into a box-like shape. Intra-cavitary reconstruction techniques have been developed for repairing defects left by aneurysm resection that reduced LV cavity size. 6 This surgical ventricular reconstruction (SVR) strategy has more recently been applied not only to patients with dyskinetic scar but also to those with only akinetic myocardial segments. 7 At the time of cardiac operation, the epicardium of these akinetic zones may appear normal and palpable thinning is often minimal in the arrested, decompressed heart. This appearance derives from preservation of a rim of normal myocardium covering the myocardial fibrosis and contrasts with the leather-like appearance and thinness typical of a LV aneurysm. Unlike the LV aneurysmectomy that removed myocardial scar or the Batista operation that reduced LV size by indiscriminate removal of portions of the LV wall, the SVR operation mechanically decreases the circumference of the zone of endocardial scar through an incision in normal epicardium. The surgical repair uses the intrinsic scar or an extrinsic patch to absorb excess linear wall tension from the adjacent myocardium. Decrease in wall stress appears to reduce the tendency for continued gradual expansion of the akinetic zone. The endocardial repair acutely decreases LV size, thereby acutely enhancing function in myocardial regions remote from the repair. 8 A recent registry report of 439 patients in 11 centers found a 5.1% operative mortality and 88% 18-month survival for patients who underwent CABG and SVR suggesting that this surgical strategy is safe and at least equivalent to CABG alone. 9 This operation is now sufficiently mature to justify this proposed randomized trial to evaluate whether appropriately performed SVR truly adds value to CABG in HF patients with regional dysfunction Need for a Randomized Trial of Medical and Surgical Therapies Along the broad spectrum of severity of ischemic HF, specific clinical information, such as severe angina or left main coronary artery stenosis, may clearly indicate the need for surgical therapy for some patients. However, a large number of patients fall into a gray zone without clear evidence for benefit from either medical or surgical therapy. For these patients, evidence supporting choice between therapies was never strong and has only been confused by recent studies showing improved outcomes with both therapies. Patients for whom equipoise of anticipated benefit now exists between modern medical and surgical therapy represent the broad population who are appropriate candidates for a randomized trial to provide the context for assessing the value of three therapeutic strategies: 1) MED alone; 2) MED + CABG; and 3) MED + CABG + SVR. DCRI/Confidential Protocol 4 06/16/2006 Page 8

64 No randomized trial has ever directly compared long-term benefits of surgical and medical treatment of patients with ischemic HF. In the developmental era of CABG, high surgical mortality rates were observed in patients with HF. Therefore, initial randomized trials comparing CABG to medical treatment conducted from 1972 to 1978 excluded most of these patients from participation. In a subgroup of 160 CASS Trial patients with LVEF <.50, ten-year survival was 61% in the 82 medically-treated patients and 79% in the 78 patients who underwent CABG (p =.01). 11 This survival advantage of CABG was not related to the presence or severity of HF or angina symptoms. A meta-analysis by Yusuf 12 combined individual patient data from the CASS Trial with those enrolled in the six other early randomized trials. Only 191 (7.2%) of the 2,649 total patients had an EF <.40, and only 106 (4.0%) of these patients who were primarily symptomatic with angina also had HF symptoms. CABG improved survival among all patients with proximal LAD stenosis, three-vessel, or left main coronary artery disease regardless of LV function. In these patients with a survival benefit from CABG, a low EF increased the absolute benefit but did not change the relative benefit of CABG. A literature search of 326 published reports on results of CABG in patients with HF or LV dysfunction identified three well-designed cohort studies. 13 Mortality benefit of CABG over medical therapy was 10 and 20 lives per 100 patients at three years in two of the three studies and 29 lives per 100 patients at five years in the third study. Despite this apparent benefit, data from the Duke Databank for Cardiovascular Diseases (DDCD) show that more than 71% of patients with characteristics that would qualify for inclusion in the STICH Trial receive medical therapy, 22% receive CABG, and 7% receive PCI. Moreover, this distribution of treatments has remained constant over the past decade at an institution with an aggressive use of myocardial revascularization procedures. Perhaps CABG is not more widely used in patients with coronary angiograms that suggest survival benefit because of concern for postoperative complications, such as stroke, that detract from its survival advantage. In addition, the general medical community often over estimates surgical risks and medically manages the high-risk patients without thoroughly characterizing the presence and extent of CAD. The therapy called medical treatment in all randomized trials and most observational comparisons with CABG really only reflected the natural history of CAD since it rarely included drugs now known to be life prolonging, such as ACE inhibitors, beta blockers, lipid lowering, and anti-platelet agents. Refinement of operative and postoperative surgical management also has steadily improved CABG results over the past two decades. The paucity of modern data available to clinicians who must daily make high-risk management decisions in patients with HF emphasizes the need for a properly-designed randomized comparison of these therapies commonly used in clinical practice. 3.4 Evaluation of Noninvasive Studies Increasingly over the past three decades, information describing cardiac structure, perfusion, hemodynamics, and metabolism obtained from noninvasive cardiac imaging studies has been used to guide management decisions for patients with HF. Although this anatomic and physiologic information often adds value to clinical care, an accepted strategy has not evolved that tailors the testing sequence to specific presenting features of individual patients to efficiently identify the treatment strategy most likely to improve outcomes. Consensus on proper use of cardiac imaging studies has been hindered by absence of clinical studies that objectively evaluate the independent treatment-related prognostic information of these tests obtained using standardized methods on patients for whom the test has minimal influence on choice of therapy. DCRI/Confidential Protocol 4 06/16/2006 Page 9

65 In the absence of high-quality information, use of these tests by clinicians is commonly guided by mechanistic studies that use endpoints considered as surrogates for survival performed with a study design that does not control for test bias in treatment selection. For example, the widespread clinical use of myocardial viability studies to identify HF patients who would or would not benefit from myocardial revascularization is largely based on more than 100 reports using improvement in EF after treatment as the study endpoint. No randomized and only ten observational studies are published describing results in 762 patients that compare the power of myocardial viability measurements for predicting patient survival after CABG or medical therapy In these studies CABG was associated with better survival than medical therapy in patients with viable but not in patients without viable myocardium. However, definition of the criteria used to define subgroups was derived on the same population used to assess outcome in these studies. Moreover, subgrouping these patients by treatment received insured that important baseline differences that influenced choice of treatment also influenced the observed outcome. The lack of survival benefit demonstrated for CABG in patients categorized as without myocardial viability derived from a low mortality in medicallytreated patients also categorized as without myocardial viability, and survival after CABG was not positively or negatively influenced by the preoperative amount of myocardial viability measured. The common clinical practice of not offering CABG to patients with LV dysfunction in regions found to be non-viable on noninvasive studies is not justified by published data. These criticisms of the quality of available information on the appropriate use of myocardial viability testing are equally true of all other noninvasive cardiac imaging data used to guide selection of therapy in patients with CAD. The optimal strategy for evaluation of these noninvasive tests would be to use multivariable modeling of all parameters of independently predictive test information to describe the total test value by a single continuous index. In this form, a myocardial viability index could readily be contextualized with all other relevant clinical and other test information to quantitate any survival advantage the myocardial viability index added by properly guiding the choice of the most effective HF therapy. The STICH Trial uses core laboratories for cardiac magnetic resonance (CMR), echocardiography (ECHO), and radionuclide (RN) studies to insure standardization of testing practices and of data analysis for operational use of these studies in the STICH Trial. These three core laboratories also will address specific questions that use anatomic or physiologic information from these noninvasive cardiac studies to elucidate mechanisms responsible for differences observed in clinical endpoints among the randomized treatment groups. This structured acquisition of cardiac imaging studies in a population with treatment assigned by randomization represents a unique opportunity to assess the individual value of these tests for guiding treatment selection. 3.5 Cost and Quality of Life Evaluation Any new therapeutic strategy that improves the health of HF patients also is likely to reduce a portion of the $22 billion costs of this condition by reducing hospitalizations and related expensive care. However, broader use of surgical therapies would increase initial cost, and the resulting economic change represents a complex issue measured both in terms of cost per added quality-adjusted life year and in total medical care cost to society. This study proposes careful measurement of resource use patterns and associated medical care costs to compare the randomized treatment strategies addressed in both primary study hypotheses. Additional comparisons will examine major subgroup effects. If the primary hypotheses are established for the two surgical strategies compared in this trial, cost effectiveness analyses will define whether these therapies are economically attractive relative to standard benchmarks. DCRI/Confidential Protocol 4 06/16/2006 Page 10

66 HF adversely impacts patient functional status and health-related quality of life (QOL). Modern medical therapy for HF modestly improves QOL, but a more aggressive approach with the surgical therapies being studied in this trial may produce even larger improvements. Careful serial measurements of QOL using wellvalidated instruments are crucial in contextualizing the impact of any invasive therapy and represent important secondary endpoints for the STICH Trial. 3.6 Neurohormonal, Cytokine, and Genetic Studies The progressive cardiac dilatation and diminished LV function that occur after myocardial damage are associated with over-expression of a variety of endogenous peptides that include: 1) neurohormonal mediators (eg. norepinephrine and endothelin), pro-inflammatory cytokines (eg. TNF and IL-6) and natriuretic peptides (eg. ANP and BNP). 4 These endogenous peptides appear to hasten progressive worsening of HF by cardiotoxic actions that may induce apoptosis and maladaptive remodeling of both cardiac myocytes and the extracellular matrix. Therapeutic strategies that attenuate production (ACE inhibition) or inhibit biologic activity (beta blockade) of a neurohormonal peptide have proven beneficial in the management of patients with HF. However, the ability of surgical interventions to attenuate neurohormonal and cytokine activation more than is achieved with MED alone has not been evaluated. Since myocardial stretch and ischemia may enhance induction and activation of these neurohormonal and cytokine pathways, the STICH Trial will test the hypothesis that rapid reduction of wall tension through the combination of CABG and SVR will be the most effective treatment for reducing peptide expression and that CABG with MED is associated with better clinical outcomes than MED alone. Conversely, high baseline levels of neurohormonal/cytokine peptides may identify a group of patients with irreversible extracellular matrix remodeling or significant apoptosis for whom CABG or CABG with SVR proves less effective than MED alone. Thus, measures of endogenous peptides may help to identify patient groups who will respond in a specific manner to all three randomized treatment strategies evaluated in the STICH Trial. Four recent observations suggest that outcome in patients with HF may be related to variations in genetic background: 1) dilated cardiomyopathy is often familial; 2) first-degree relatives of patients with HF have a high rate of abnormal echocardiograms; 24 3) specific loci are associated with adult-onset autosomal dominant dilated cardiomyopathy; 25 and 4) mutations affecting cytoskeletal proteins are associated with the development of HF. 26 Genetic determinants that might affect the progression of disease in individual patients appear to consist of either polymorphisms or allelic mutations that alter expression of proteins important in the pathophysiology of HF. Many of these genetic variations are directly related to alterations in the activity of the neurohormonal and/or pro-inflammatory pathways. The STICH Trial will provide an opportunity to evaluate the role of genotype determination for predicting the efficacy of subsequent therapy. Most of the polymorphisms that will be evaluated affect the constitutive and activated expression of neurohormonal/cytokine peptides that may be favorably altered by mechanical intervention directed at normalizing LV mechanics and myocardial perfusion. 3.7 Significance Should We Do CABG in Patients with Heart Failure? The design of the STICH Trial excludes only patients for whom MED is the only reasonable therapeutic alternative, those with coronary anatomy best suited for revascularization by PCI, those severely symptomatic patients not eligible for SVR, and those who are heart transplant candidates. Therefore, results of the STICH Trial will be generalizable to almost all of the large number of patients with CAD and systolic LV dysfunction in this country. If surgical therapies prove not to significantly improve survival, MED without intensive evaluation for CAD may be the preferred strategy of care to be pursued in the more than 550,000 patients who first present with HF of unexplained etiology each year in the United States. 4 A more DCRI/Confidential Protocol 4 06/16/2006 Page 11

67 aggressive search for the 70% of patients with HF and LV dysfunction who have CAD as the predominant etiology could await demonstrated failure of the MED strategy indicated by development of overt ischemic symptoms. Conversely, if surgical therapy is shown to have a survival advantage over medical therapy in the STICH Trial, early aggressive evaluation of CAD as a potentially correctable etiology of new onset HF would be the preferred strategy. Clinical testing then would become more widely employed to identify patients likely to benefit from CABG and those eligible for SVR. When Do We Need Anatomic and Physiologic Studies to Guide the Choice of Therapy in HF Patients with CAD? This trial will relate continuous quantitative relationships between physiologic measurements of the magnitude of myocardial ischemia and viability and LV size to the outcome associated with the three treatment strategies. The resulting information is likely to decrease the cost and complexity of evaluation of patients with ischemic HF by eliminating use of tests that do not provide independent information. This trial will establish whether measurements of neurohormonal and cytokine levels and genetic profiling are useful for directing patient management decisions and for monitoring the effectiveness of therapy. Information obtained from physiologic studies in the STICH Trial will permit refinement of the optimal approach to efficiently evaluate and select the treatment strategy most likely to be effective for the many patients with CAD, HF, and LV dysfunction in the United States. Does SVR Need to be Added to CABG in Patients with Regional Dysfunction and What Magnitude of Dysfunction Must be Present to Justify This Additional Surgery? This trial is designed to answer the question of whether mechanically decreasing LV size by SVR arrests or reverses LV enlargement and thereby enhances survival without hospitalization more than relief of chronic ischemia by CABG without SVR. The SVR operation is gaining popularity because promising observational reports confirm its safety and suggest its efficacy. Randomized comparison of CABG with SVR versus CABG alone will definitively test the superiority of adding SVR to CABG. In the event the two operations result in equivalent survival free of cardiac hospitalization, CABG will remain the dominant operation. However, if the addition of SVR to CABG in patients with akinesia of the anterior wall offers superior clinical outcomes, an important new therapeutic option will be added to the HF armamentarium. Moreover, information from CMR measurements of LV volume made before and after treatment in SVR-eligible patients is likely to guide optimal use of SVR and also provide a noninvasive method to assess the quality of operative result. 4. Investigational Plan 4.1. Overview The STICH Trial is a three-arm randomized clinical trial that will enroll approximately 2,000 patients with left ventricular ejection fraction <.35 and CAD amenable to coronary artery bypass grafting (CABG). The three treatments include intensive, evidence-based medical therapy for heart failure (MED), MED plus revascularization with CABG, and MED plus CABG combined with surgical ventricular reconstruction (SVR). After eligible patients give informed consent, site personnel will use a 24-hour toll free telephone number with an interactive voice response system (IVRS) to initiate the randomization process. In addition to the baseline case report form (CRF), a Quality of Life (QOL) questionnaire will be administered by the coordinator at each site. Baseline studies will be completed and the randomized therapy will be initiated within two weeks. DCRI/Confidential Protocol 4 06/16/2006 Page 12

68 Randomized patients will be followed at clinic visits at four-month intervals for the first year and a minimum of six-month intervals for the duration of the trial. Follow-up echoes, treadmills, gated SPECTs, CMRs and neurohormonal/cytokine peptide collections will be performed at scheduled intervals throughout the remainder of the trial. Personnel from the Economics and Quality of Life (EQOL) Core Laboratory at the DCRI will perform QOL interviews by telephone at four months, one, two, and three years for clinical sites in the United States. In addition, EQOL Core Laboratory staff will collect hospital bills from all United States clinical sites for all hospitalizations and major outpatient tests and therapies for the length of the study. Follow-up of Hypothesis 2 patients will be a minimum of two years, an average of 3.5 years and a maximum of six years. Hypothesis 1 patients will be followed until 400 deaths occur. 4.2 Study Population The patient population will consist of patients who have an LVEF of.35, and coronary artery disease suitable for revascularization. Patients included in this study will be segregated into three strata depending on their MED and SVR eligibility Definition of Patient Strata The design will accommodate the fact that not all patients enrolled will be eligible for all three therapeutic approaches. TABLE 1. Eligibility Criteria Stratum A Stratum B Stratum C MED eligible MED eligible MED ineligible SVR ineligible SVR eligible SVR eligible Patients who meet trial eligibility criteria will first be stratified as eligible or ineligible for MED on the basis of CCS III angina or 50% left main coronary artery stenosis. Patients will then be further stratified based on eligibility for the SVR procedure. Patients eligible for either MED or CABG but not eligible for the SVR procedure (Stratum A) will be randomized in equal proportions to MED vs CABG. Patients eligible for all three therapies (Stratum B) will be randomized in equal proportions to MED, CABG, and CABG plus SVR. Patients whose angina or coronary artery disease severity make them ineligible for medical therapy but who are appropriate candidates for both CABG and SVR (Stratum C) will be randomized in equal proportions to CABG vs. CABG + SVR. The allocation of patients to treatment within each stratum at the conclusion of Hypothesis 2 enrollment on January 24, 2006 is illustrated in Figure 1. At this time, 635 total Hypothesis 1 patients had been enrolled with a treatment assignment of MED in 310 patients (235 Group 1 and 75 Group 3) and of CABG in 325 patients (249 Group 2 and 76 Group 4). The 365 or more patients remaining to be recruited after January 24, 2006 will all be entered into Stratum A. DCRI/Confidential Protocol 4 06/16/2006 Page 13

69 FIGURE 1. Patient Stratum and Treatment Assignment at Time of Completion of Hypothesis 2 Enrollment on January 24, 2006 CAD EF MED 235 Pts No R 484 Pts 2 CABG + MED 249 Pts SVR Eligible? 700 Pts Yes Yes R 216 Pts 3 MED 75 Pts Medical Eligibility 4 CABG + MED 76 Pts 5 CABG + SVR + MED 65 Pts No SVR Eligible? 6 CABG + MED 423 Pts Yes No 7 CABG + SVR + MED 436 Pts Not in Trial These strata R NOW CLOSED 859 Pts Hypothesis 2 enrollment by Stratum A Stratum B Stratum C treatment assigned CABG + MED = 499 CABG + SVR + MED = 501 DCRI/Confidential STICH Protocol 06/16/2006 Page 14

70 4.2.2 Study Hypotheses There are two primary study hypotheses. Coronary revascularization hypothesis: Improvement in myocardial perfusion by CABG combined with intensive medical therapy (MED) improves long-term survival compared to MED alone. LV reconstruction hypothesis: In patients with dominant anterior LV akinesia or dyskinesia, LV shape and size optimization by SVR combined with CABG and MED improves long-term survival free of cardiac hospitalization compared to CABG and MED without SVR Inclusion and Exclusion Criteria Inclusion Criteria - The following patients may qualify for inclusion in the study: Men. Women who are not of childbearing potential. Age 18 years or above. Who have a LVEF <.35 measured by CMR ventriculogram, gated SPECT ventriculogram, echocardiography, or contrast ventriculogram within three months of trial entry. Who have coronary artery disease suitable for revascularization. Exclusion Criteria - At the time of randomization, none of the following may exist: Failure to provide informed consent. Aortic valvular heart disease clearly indicating the need for aortic valve repair or replacement. Cardiogenic shock (within 72 hours of randomization) as defined by the need for intra-aortic balloon support or the requirement for intravenous inotropic support. Plan for percutaneous intervention of coronary artery disease. Recent acute myocardial infarction judged to be an important cause of left ventricular dysfunction. History of more than one prior coronary bypass operation. Non-cardiac illness with a life expectancy of less than 3 years. Non-cardiac illness imposing substantial operative mortality. Conditions/circumstances likely to lead to poor treatment adherence (e.g., history of poor compliance, alcohol or drug dependency, psychiatric illness, no fixed abode). Previous heart, kidney, liver, or lung transplantation. Current participation in another clinical trial in which a patient is taking an investigational drug or receiving an investigational medical device. MED Therapy Eligibility Criteria Absence of left main coronary artery disease as defined by an intraluminal stenosis of 50% or greater. Absence of Canadian Class III angina or greater (angina markedly limiting ordinary activity). SVR Eligibility Criterion Dominant akinesia or dyskinesia of the anterior left ventricular wall amenable to SVR. DCRI/Confidential Protocol 4 06/16/2006 Page 15

71 4.3 Study Subject Enrollment Screening Screening of patients referred must address clinical eligibility criteria and assess coronary anatomy and left ventricular function. The best available and most recently performed LV assessment by either contrast, gated SPECT, echocardiogram, or CMR ventriculogram read at the clinical sites will identify patients meeting LVEF entry criteria and be used to evaluate SVR eligibility. Baseline STICH Trial protocol tests that have been performed for clinical reasons in potential STICH patients in compliance with the STICH core laboratory protocols may be used for defining trial eligibility or meeting baseline testing requirements, thereby decreasing inconvenience to the patient and unnecessary cost to the STICH Trial from duplicate testing. After the screening process identifies a patient who meets STICH Trial entry criteria, an initial evaluation form is used to document information needed to confirm eligibility of the patient and to define the appropriate stratum for randomization Randomization After eligible patients give informed consent, site personnel will initiate the randomization process by calling an always available toll-free telephone number to connect with the IVRS at the Coordinating Center that has a redundant voice back-up system. Once eligibility is established, a site-specific unique identifier will be assigned to the patient. A permuted block randomization, stratified by clinical site and by the three Strata A, B, or C shown in Figure 1, will be used with a block size not known by the investigators. Use of this randomization process insures proper trial enrollment and promptly provides the Coordinating Center with the basic patient enrollment information needed to monitor enrollment performance. Study design specifies enrollment of at least 2,000 randomized patients into the STICH Trial Baseline Evaluation of Randomized Patients Prior to Assigned Treatment All patients eligible and consenting to be randomized will have had baseline clinical evaluation information entered on the initial evaluation form. Table 2 summarizes baseline evaluation studies for each stratum. Where logistically feasible, a blood sample for neurohormone, cytokine, and genetic studies is strongly encouraged. An echocardiogram will be sent to the ECHO Core Laboratory on all patients. A myocardial viability assessment by either or both radionuclide or echocardiographic techniques is strongly encouraged. A baseline quality of life assessment and six-minute walk test will be performed. A modified Bruce or bicycle stress test may be obtained on patients in Strata A and B who are able to exercise. This may be performed in conjunction with a stress sestamibi perfusion study. When possible, a baseline CMR or gated SPECT ventriculogram will be obtained in all patients. In situations when obtaining one of these two studies would preclude patient entry into the trial, the echocardiogram may serve as the baseline assessment of LV function. DCRI/Confidential Protocol 4 06/16/2006 Page 16

72 Table 2. Baseline Evaluation Studies Patient Characteristics All Consenting Patients When Feasible Patients Able Stratum A Stratum B Stratum C ECHO to Core QOL to Core EuroQOL to Core CMR, Gated SPECT, or ECHO for LV function NCG Blood to Core Myocardial Viability to Core 6-minute Walk Exercise Stress ECHO to Core QOL to Core EuroQOL to Core CMR, Gated SPECT, or ECHO for LV function NCG Blood to Core Myocardial Viability to Core 6-minute Walk Exercise Stress ECHO to Core QOL to Core EuroQOL to Core CMR, Gated SPECT, or ECHO for LV function NCG Blood to Core 6-minute Walk Exercise Stress 4.4 Initiation and Completion of Initial Treatment Intensive medical treatment (MED) will be initiated after completion of baseline evaluation in all randomized patients. Patients randomized to CABG, with or without SVR, will undergo operation as soon as possible after completion of protocol-mandated baseline studies, but within 14 days in all patients. The standard of best practice for each of the three possible treatments will be reviewed in the context of relevant published literature at regular intervals throughout the study by the Medical and Surgical Therapy Committees, and the original treatment protocol may be modified by the Data and Safety Monitoring Board (DSMB) if necessary. Definitions of the three therapie s to be used at the start of the trial appear justified by current evidence. In hospitalized patients, initial treatment will be considered to be completed at hospital discharge or at 30 days if hospitalization continues. In patients randomized to medical therapy and treated as outpatients, initial treatment will be considered to be completed when the medical regimen is considered stable or at 30 days. The Initial Treatment Form (Medical or Surgical) should reflect all events between initiation and completion of the initial therapy period Treatment Groups The three treatment groups are: 1. MED: MED for HF is now defined to include angiotensin-converting enzyme inhibitors (ACE-I), or angiotensin receptor blockers in patients who do not tolerate ACE inhibitors, beta blockers, spironolactone, and an antiplatelet agent such as aspirin or clopidogrel adjusted to optimal doses within 30 days of randomization unless contraindications are documented. HMG-CoA reductase inhibitors, diuretics, and digitalis use will be individualized to patient-specific indications. Physicians will be encouraged to manage concomitant conditions appropriately (diabetes, hypertension, hyperlipidemia, angina, atrial arrhythmias) and will be asked to avoid the use of non-steroidal anti-inflammatory drugs as well as most anti-arrhythmic agents and most calcium-channel blocking drugs. Electrophysiologic testing and the implantation of cardioverterdefibrillators and biventricular pacing should be used in compliance with standard guidelines. MED therapy will be monitored closely and modified when indicated at a frequency no less than the required follow-up clinic visits. DCRI/Confidential Protocol 4 06/16/2006 Page 17

73 2. CABG: CABG will be performed using at least one internal mammary graft in all patients unless this conduit is unavailable or has inadequate flow. Patients with secondary mitral regurgitation identified by echocardiography who are judged to need mitral valve repair may undergo this procedure combined with CABG with or without SVR. Therefore, use of CABG or CABG + SVR throughout this protocol implies that mitral repair for regurgitation secondary to ischemic HF etiology may be included in the operative strategy at the discretion of the operating surgeon. 3. CABG + SVR: The two criteria used by the Surgical Therapy Committee to define the acceptable range of specific operative maneuvers essential to be considered an acceptable technical SVR operation for the STICH Trial will be any ventricular reconstruction method that consistently results in: 1) a low operative mortality; and 2) an average EF increase of 10% and average LVESVI decrease of 30% as assessed on the four-month postoperative CMR measurement. SVR may be performed with or without cardioplegic arrest. In general, the SVR is begun after completion of the CABG and in a beating heart. The region of dysfunction identified by preoperative CMR will be incised regardless of the epicardial appearance. The border separating the dysfunctional from contracting myocardium identified by palpation will guide placement of sutures and a patch, if necessary, to pursestring the endocardial scar and thereby decrease LV size without distorting LV shape. For patients undergoing surgery, intensive medical treatment will resume as soon as permitted by post-procedure recovery Interruption or Discontinuation of Treatment A patient is considered randomized when the patient identification number has been assigned by the IVRS. Every effort must be made to ensure that patients remain in the study and on medical therapy for the duration of the study. An interruption of medical therapy may occasionally be required. If a temporary interruption occurs, the Coordinating Center Clinical Hot Line is available through the IVRS for consultation or notification about any study-related issues. Every attempt to reinitiate optimal medical therapy should be made throughout the duration of the study. The reinitiation of medical therapy is not subject to a time limit. Patients with a temporary or permanent discontinuation of any or all medications integral to optimal medical therapy should continue the visit schedule and undergo all protocol specified studies. Each patient entering the study must be followed until study completion whether or not the patient receives the randomized treatment or medical therapy is temporarily interrupted or permanently discontinued. A patient will be considered lost to follow-up only after exhausting all means of contact. The status of the patient at the last visit or contact will be used for the final analysis. The vital status of patients who withdraw consent or are lost to follow-up will be followed by a national death registry when possible. 4.5 Study Subject Follow-up At each clinic visit at four-month intervals for the first year after study entry and at six-month intervals thereafter, a brief history and physical examination will be performed and follow-up forms will be completed. Follow-up data to be collected at each of these contacts will include an assessment of HF and angina symptoms and interval medical events including hospitalization, major cardiac procedures, and tests. Where logistically feasible, a follow-up blood sample is strongly encouraged at the four-month follow-up visit. Patients completing a treadmill test at baseline will have this test repeated at two years. Detailed quality of life assessment will be performed at the four-month visit, one year, and annually thereafter during follow-up. Table 3 summarizes the schedule of follow-up studies in the STICH Trial. Telephone contact will be maintained with the patient unwilling or unable to return to the clinical site and their local care providers to assure that intensive medical therapy is continued. DCRI/Confidential Protocol 4 06/16/2006 Page 18

74 Interval 4 Months 12 Months 24 Months 36 Months 48 Months Patient Characteristics Table 3. Follow-up Evaluation Studies Stratum A Stratum B Stratum C All Patients ECHO to Core EuroQOL to Core ECHO to Core EuroQOL to Core V-gram to Core (CMR, SPECT, or Echo) ECHO to Core EuroQOL to Core V-gram to Core (CMR, SPECT, or Echo) When Feasible NCG Blood to Core NCG Blood to Core NCG Blood to Core Patients Able 6-minute Walk 6-minute Walk 6-minute Walk All Patients EuroQOL to Core EuroQOL to Core EuroQOL to Core Patients Able 6-minute Walk 6-minute Walk 6-minute Walk All Patients ECHO to Core EuroQOL to Core ECHO to Core EuroQOL to Core V-gram to Core (CMR, SPECT, or Echo) 6-minute Walk Exercise Stress ECHO to Core EuroQOL to Core V-gram to Core (CMR, SPECT, or Echo) 6-minute Walk Exercise Stress Patients Able 6-minute Walk Exercise Stress All Patients EuroQOL to Core EuroQOL to Core EuroQOL to Core Patients Able 6-minute Walk 6-minute Walk 6-minute Walk All Patients EuroQOL to Core EuroQOL to Core EuroQOL to Core Patients Able 6-minute Walk 6-minute Walk 6-minute Walk 4.6 Efficacy Assessments Primary Efficacy Parameters Documentation for occurrences of death or hospitalization will be required for submission to the Endpoint Committee. The Endpoint Committee will adjudicate causes of death and hospitalization based upon prespecified definitions and procedures for this study. 1. All-cause mortality (time to death). 2. Survival free of hospitalization for cardiac cause Secondary Efficacy Parameters 1. All cause mortality at 30 days. 2. Survival free of hospitalization for heart failure. 3. Cardiovascular mortality (defined as sudden death, or death attributed to recurrent myocardial infarction, heart failure, a cardiovascular procedure, stroke, or other cardiovascular etiology). 4. Cardiovascular mortality and morbidity. 5. Survival free of revascularization. 6. Survival free of cardiac transplantation or LVAD implantation. 7. All-cause (unplanned and elective) hospitalization. DCRI/Confidential Protocol 4 06/16/2006 Page 19

75 Other Key Efficacy Parameters 1. Survival free of CABG 2. Survival free of PCI 3. Sudden death with or without resuscitation 4. Fatal and non-fatal MI 5. Fatal and non-fatal stroke Safety Assessments Safety assessments will consist of monitoring and recording the pre-defined safety endpoints, procedure complication rates and all serious adverse events. A serious adverse event is defined in general as an untoward (unfavorable) event which: 1. is fatal or life-threatening, 2. required or prolonged hospitalization, 3. was significantly or permanently disabling or incapacitating, 4. is medically significant (may jeopardize the patient and may require medical or surgical intervention to prevent one of the outcomes listed above). Hospitalizations planned before entry into the clinical study and not associated with any deterioration in condition, emergency, or outpatient visits that do not result in admission are not considered SAEs. Source documents of all safety assessments (e.g., physical examinations or laboratory data) performed as part of the standard evaluation and care of the patient should be maintained in the patient study chart. Upon request of the DSMB, the Coordinating Center, working with the Medical and Surgical Therapy Committees, will define Steering Committee approved safety endpoints and procedure complication rates with expected ranges and upper threshold limits for concern. Medical therapy will be monitored by safety endpoints, such as symptomatic bradycardia, angioedema, renal dysfunction, and drug intolerance. Procedure complications will include return to the operating room for bleeding, prolonged intubation, stroke, mediastinitis, and atrial fibrillation. SAEs will be reported to the Coordinating Center with the visit form. In this high-risk population, the rate of SAEs will be used to assess safety. All deaths and unexpected protocol-related major SAEs will be reported to the Coordinating Center within hours of investigator knowledge of the event Quality of Life Assessments Quality of life (QOL) questionnaire data will be collected at baseline by the site coordinators. Englishspeaking patients from the United States and Canada will have telephone assessment at four months, one year, two years, and three years by trained telephone interviewers from the EQOL Core Laboratory at the DCRI. All other non-english speaking patients will be interviewed at the same intervals by the clinical site. In addition, cardiac symptoms as reflected by the NYHA HF class and the CCS angina class, patient utilities, and interval medical care consumption will be assessed at each clinic visit/contact by the sites. Content of Health-related Quality of Life Questionnaire The QOL questionnaire consists of a battery of validated instruments that build on a generic core supplemented by disease-specific measures where necessary to provide a comprehensive assessment of health-related quality of life. DCRI/Confidential Protocol 4 06/16/2006 Page 20

76 The generic core instrument to be used is the Medical Outcomes Study Short Form (SF-12). This instrument is a reduced version of the SF-36 that covers the broad range of health-related quality of life domains with two or three items per domain. To enhance sensitivity, we will include the full SF-36 items for the following domains: role function-physical, role function-emotional, social function, psychological well-being/mental health, and vitality. The remaining scales in the SF-36 overlap with other items in our battery and will therefore be used in their reduced SF-12 form. A total of 22 items from the SF-12/SF-36 will be included on the battery. The primary disease-specific measure to be used is the Kansas City Cardiomyopathy Questionnaire (KCCQ), which contains 20 items covering the effects of heart failure on physical limitations, symptoms frequency and severity, social functioning and general quality of life. The presence of anginal symptoms will be assessed with the six-item symptom scales from the Seattle Angina Questionnaire. Cardiac symptoms will be independently assessed by the sites at each four-month visit using the New York Heart Association (NYHA) congestive heart failure class and the Canadian Cardiovascular Society (CCS) class for angina. General psychological and well being/mental health will be measured using a five-item subscale of the SF-36 (as described above). We will also assess depressive symptoms in more detail using the 20-item Center for Epidemiologic Studies-Depression (CES-D) scale. Self-efficacy will be evaluated using the 13 item Cardiac Self-Efficacy scale. Role functioning, social functioning and vitality will be assessed from the SF-36 as described above. Employment details will be obtained using six questions adapted from the NHLBI Bypass Angioplasty Revascularization Investigation (BARI) Substudy in Economics and Quality of Life (SEQOL). In addition, the baseline questionnaire will ask four questions covering education/socioeconomic status and social support (marital status, number in household). Measurement of Utilities Patient specific utilities will be assessed using the EuroQoL. The current version of this instrument consists of a five-dimension assessment of current health-related quality of life and a self-rating (0-100) "thermometer" of "your own health state today. This assessment will be completed by the patient at baseline, four months, and yearly for four years. If the patient is unable to complete the form unassisted, the instrument will be administered as an interview. If the patient does not return for a clinic visit, the interview will be done over the telephone when the Coordinator does the follow-up contact Economic Data/Hospital Billing Data (U.S. sites only) Hospital bills (detailed, summary ledger and UB 92) will be collected by the EQOL Core Laboratory at the DCRI for all hospitalizations throughout the length of the study. They will include hospitalizations at clinical sites and at institutions not participating in STICH. In addition, cost to charge ratios (RCCs) will be obtained from each hospital where a STICH baseline or follow-up hospitalization is reported Echocardiography Assessment Echocardiographic assessment of left ventricular function is one of four modalities that can be used to evaluate patients for STICH Trial eligibility. In all patients, echocardiograms will be used to evaluate the independent prognostic value of baseline echocardiographic variables and their interaction with treatment allocation and outcome on long-term follow-up. Dobutamine echocardiography may be used in conjunction with or instead of radionuclide techniques to assess myocardial viability. Detailed protocols for these studies are described in the Echocardiography Core Laboratory manual on the STICH Trial website ( DCRI/Confidential Protocol 4 06/16/2006 Page 21

77 A baseline echocardiogram will be required for all patients who are randomized into any of the three strata of the trial to address the echocardiographic specific secondary objective outlined in section 2.2 by quantitative measurements of cardiac structural, functional, and hemodynamic parameters. The initial echocardiogram will provide baseline echocardiographic variables to be compared with those obtained from the 4 month and 2 year follow-up echocardiograms. Four-month and two year follow-up will be performed to compare the longterm effects of different treatment modalities in regional remodeling, MR, and diastolic function. The echocardiographic protocol for baseline and serial echocardiograms is provided in the echocardiography manual Nuclear Cardiology Assessment Myocardial Viability and Perfusion Myocardial viability and perfusion studies are strongly encouraged, but not required, at baseline. A preferred protocol for these studies is defined in the Radionuclide Core Laboratory manual on the STICH Trial website ( However, studies performed by all generally used techniques will be acceptable as long as data quality is acceptable to the Radionuclide Core Laboratory. Left Ventricular Function Assessment Gated SPECT perfusion studies may be used to assess left ventricular function at baseline and during follow-up. Details of performing this study are outlined in the Radionuclide Core Laboratory manual on the STICH Trial website ( CMR Assessment Patients will undergo baseline CMR testing to assess LV size, shape and function. Patients will complete questionnaires to determine CMR suitability in advance of the CMR examination. The detailed CMR protocol is described in the CMR manual on the STICH Trial website ( Neurohormone, Cytokine and Genetic Assessments When feasible, blood samples for both neurohormonal/cytokine analysis and genotyping will be obtained from patients at the time of study enrollment, and a blood sample will be obtained at the four-month study visit for neurohormonal/cytokine analysis only. Patients will be prepared prior to phlebotomy in order to minimize the influence of stress and other factors on neurohormone levels. The phlebotomy procedure necessary to minimize the influence of stress, sample preparation, sample storage, shipping and planned NCG analyses are detailed in the NCG manual on the STICH Trial website ( 4.7 Study Organization Monitoring Committees The following Monitoring Committees will be employed in this study: DCRI/Confidential Protocol 4 06/16/2006 Page 22

78 Executive Committee The Executive Committee leads the implementation of policies and practices approved by the Steering Committee. Executive Committee members will include the study chairman, principal investigator, the lead statistician, the director and associate director of trial operations, the CC project leader, the chairs and cochairs of the medical and surgical therapy committees, the chair of the core laboratory committee, the chair of the minority recruitment and retention committee, the chair and co-chair of the policy and publications committee, and the chair of the trial registry. The NHLBI project officer will be an ex officio member of the Executive Committee. All committees report through the Executive Committee to the Steering Committee. All committees may appoint working subgroups called councils to facilitate leadership activities. The main roles and responsibilities of the Executive Committee are: To lead the Steering Committee, to resolve differences of opinion within the Steering Committee, to resolve issues not requiring full Steering Committee input. To review operational aspects of the trial on an ongoing basis. To oversee the conduct of the study and initiate the approval process for any protocol changes needed to improve the quality of the study. To implement all protocol amendments approved by the DSMB. To propose membership of the other committees to the Steering Committee. To serve as a liaison between the NHLBI and the other committees. To serve as a liaison between the DSMB and the Steering Committee. To present the trial results to the steering committee in advance of national presentations. To approve all manuscripts prior to submission. Steering Committee The Steering Committee consists of the Executive Committee and one representative investigator from each participating investigative site. The main roles and responsibilities of the Steering Committee are: To make ethical, scientific, and policy decisions regarding the overall conduct of the study. To approve committee membership. To serve as a liaison between all investigators and the Executive Committee. To review study progress. To act on the recommendations of the DSMB. To review and provide input to potential protocol amendments. To review and approve presentations and publications of the study. Data and Safety Monitoring Board (DSMB) The DSMB members are independent of the NHLBI, the study management organizations, and the investigators. A nationally recognized member of the cardiovascular community with experience in oversight of cardiovascular clinical trials will chair the Data and Safety Monitoring Board (DSMB) comprised of experts in relevant biomedical fields including cardiology, cardiovascular surgery, biostatistics, and bioethics appointed by the NHLBI. No member of the DSMB shall be located at an institution that participates in any direct way in the STICH Trial. The main roles and responsibilities of the DSMB are: DCRI/Confidential Protocol 4 06/16/2006 Page 23

79 To initially review the study protocol and commission a specific regular process for evaluation of STICH trial data. To meet no less than twice yearly for the duration of the trial to monitor the trial progress using confidential data summarizing trial safety and outcomes and to prepare a summary report of their review to the Executive Committee. To advise the Executive Committee on the implications of HF treatment advances on the conduct of the STICH trial. To approve protocol modifications if warranted. To advise the NHLBI directly regarding recommendations for trial modification or early trial cessation. The efficacy and safety analyses will be performed semi-blinded (i.e., treatments A, B and C) by the Statistical Data Coordinating Center. The DSMB statistician will possess a copy of the treatment codes for unblinding purposes if deemed necessary by the DSMB. The study may be amended, or stopped early, or a treatment arm may be discontinued should any of these be deemed necessary based upon DSMB recommendation. The chairman of the DSMB will discuss the recommendations of the DSMB with the NHLBI project officer who will inform the Executive Committee. The chairman of the Executive Committee will notify the Steering Committee of any significant proposed amendment to the study or recommended discontinuation of the study. The roles, responsibilities and procedures of the DSMB are summarized in the DSMB Manual Operational Committees The Executive Committee, with approval of the Steering Committee, is empowered to appoint operational committees to assist with trial work. Active committees are described in the Manual of Operations. 5. Data Management 5.1 Data Collection Data Forms Data collection forms used in this study are tabulated in the manual of operations with summaries of their content and purpose. The front page of each form will identify the date and version of the form, and provide brief summary instructions for data collection. The subject s study number and initials will be included on every page of each form. The study subject number will reflect both a site specific code, the subject's sequence number for that site and a check digit. Each form will require the signature of the physician investigator and the study coordinator with the date of completion Manual of Operations The manual of operations provides more detail regarding study objectives, scope, design and operational policies and procedures than the clinical protocol. The manual can be accessed on the STICH Trial website ( STICH Trial Website The STICH Trial website provides printable updated versions of the protocol and manual of operations. The website provides access to a complete directory of participating physician investigators and study coordinators at all clinical sites. The website provides a rich source of current information to investigators with secure access. Free access is available to the public to a portion of the site containing general information. DCRI/Confidential Protocol 4 06/16/2006 Page 24

80 5.2 Database Management and Quality Control Database management and quality control for this study are the responsibility of Duke Clinical Research Institute, Durham, NC, USA. Structured data elements from the CRFs are entered into the STICH database and reviewed using double data entry for verification. Coexistent diseases and adverse events are coded using a standard coding dictionary, MEDDRA 1.5. Concomitant medications are coded using a standard medication dictionary, WHO DRL. Information entered into the database is systematically checked by data management staff, using error messages generated from validation programs and database listings. Obvious errors are corrected by data management center personnel. Other errors, omissions or questions are entered on data query forms, which will be returned to the investigational site for resolution. After the investigator response is received at the data management center, the resolutions will be entered into the database. A copy of the signed data query form is kept with the CRFs. Quality control audits of all key safety and efficacy data in the database are made at designated times during the study. All clinical site patient-related reimbursement is prompted by completion of data forms with appropriate responses to all data elements. 6. Statistical Plan 6.1 Sample Size Determination - Overview Several design factors and research objectives have been considered in developing appropriate sample size estimates for the study. First, patient enrollment has been determined so there would be a sufficient number of endpoints to provide a high degree of confidence (power > 90%) for testing both primary hypotheses. Second, important secondary endpoints, including cardiovascular mortality, short-term (30-day) mortality, and measures of quality-of-life have also been considered. Third, we considered it important for the overall sample to be large enough to permit a prudent examination of treatment effects in selected subgroups of patients where surgical intervention might be particularly advantageous, or where the question of a treatment benefit from surgical intervention is particularly relevant. Important pre-specified subgroups of interest in this study include those defined by age, gender, race, HF class, ejection fraction, and the strata outlined in Figure 1. Finally, the sample size has been determined to provide a reasonable level of confidence of detecting clinically important therapeutic effects even in the event that current projections of event rates and treatment differences prove to be optimistic Sample Size and Power Considerations A combination of survival data from recent HF trials that reflect patient outcomes when treated with intensive medical therapy (including beta blockers and ACE inhibitors) and historical data from the Duke Cardiovascular Database provide useful information on the range of patient outcomes that would be expected among the HF patients enrolled in STICH. Based on this information, we project that for testing the coronary revascularization hypothesis, the mortality rate in the medically treated arm of STICH will be approximately 25% after 3 years of follow-up. For the LV reconstruction hypothesis, we project that mortality or cardiac hospitalization will occur in the CABG (without SVR) arm with a 3-year rate of 50% or higher. Recognizing, however, that the actual event rates in STICH may vary somewhat from these estimates, we calculated sample size requirements for several different combinations of event rates and power levels in order to examine the sensitivity of the required sample size to different event rates and outcome scenarios that might conceivably arise in this trial. DCRI/Confidential Protocol 4 06/16/2006 Page 25

81 In the absence of detailed life-table follow-up data on patients similar to those who will be enrolled in the trial and treated with intensive medical therapy, the following approach was used for calculating sample size requirements for the coronary revascularization hypothesis. (A similar approach was used for the LV reconstruction hypothesis, although the event rates differed because the two primary hypotheses are based on different endpoints). For a specified benchmark outcome rate in the medically treated (control) group (e.g., an overall event rate at 3 years of 25%), event rates at other time points during follow-up were estimated by assuming an exponential survival distribution. We then postulated that in the CABG arm, there would be a specific reduction of this benchmark rate (e.g., a 25% reduction). Event rates for the CABG group at various time points during follow-up were also estimated assuming an exponential survival distribution. Since the two groups will be compared with respect to length of survival using the log-rank test, 29 or equivalently, the Cox proportional hazards model, the one-tailed formula of Schoenfeld 32 was modified to allow calculating sample size requirements based on the use of two-tailed tests. Schoenfeld's method is based on first calculating the total number of events required in the combined treatment groups. By estimating the proportion of patients who will have an event by the end of the study, the total number of patients required for the trial can be determined. The necessary number of events depends on the level of power desired, the significance level (which we have chosen to be α=0.05), and the ratio of the hazard function for patients receiving surgery to the hazard function for patients in the medically treated arm. Once the number of events has been determined, the proportion of patients who will experience an event during the patient follow-up period can be estimated using the method in Section 2.1 of Schoenfeld. 32 The total number of patients required is simply the number of events divided by the proportion of patients who will experience an event. A key issue in determining an adequate sample size for this trial is the magnitude of the event reduction we could expect to achieve in the CABG arm compared to the medical arm. Based on what is known about the effects of CABG in other patient populations, it is reasonable to postulate that surgery will be sufficiently effective to reduce overall mortality by 25%. A mortality reduction of this magnitude would be highly important from a clinical and public health standpoint, given the large population of patients in this country and throughout the world who suffer from moderate to severe HF secondary to CAD. Sample Size for Hypothesis 1: Coronary Revascularization (all-cause mortality) The three-year mortality rate was assumed to be 25% in the medically-treated arm in the design of the Hypothesis 1 component of STICH. The ischemic cohort of SCD-HeFT randomized to medical therapy (the control arm of that trial) had a three-year mortality of 28.5%. Although some differences exist between the SCD-HeFT and STICH populations (e.g., SCD-HeFT enrolled all NYHA class II or class III heart failure patients, whereas 85% of STICH patients are class II or class III), the differences are not expected to result in markedly lower mortality among the STICH medical cohort. A 25% reduction in mortality with CABG is hypothesized in the design of STICH. The level of statistical power chosen for assessing treatment differences in mortality is 90%. Although STICH was originally designed to enroll a fixed number of patients, the number of patients (the sample size) and the length of time they would be followed were driven by the number of events required to achieve a statistical power of 90%. Due to the slower than anticipated enrollment in Hypothesis 1, the optimal use of available resources for the trial mandated that we carefully choose the number of patients and the duration of their follow-up to accrue the number of events required to achieve the desired level of statistical power. Approximately 400 events are required to achieve 90% power for detecting the hypothesized mortality reduction. DCRI/Confidential Protocol 4 06/16/2006 Page 26

82 Statistical Power Tradeoffs. We have examined the tradeoff between the number of patients enrolled and the length of follow-up required to achieve a given level of statistical power, as reflected in Table 4. TABLE 4. Hypothesis 1 Statistical Power in Relation to Number of Patients and Length of Follow-up Total Average Length of Follow-up Patients 4.0 yrs 4.5 yrs 5.0 yrs 5.5 yrs 6.0 yrs % 73% 77% 80% 83% 1,000 73% 77% 81% 84% 87% 1,100 77% 81% 84% 87% 90% 1,200 81% 84% 87% 90% >90% For the primary mortality endpoint in Hypothesis 1, 80% and 90% power can be achieved under any of the scenarios outlined in Table 5. TABLE 5. Hypothesis 1 Enrollment, Follow-up, and Statistical Power Total Average Follow-up Required to Achieve Patients 80% Power 90% Power yrs ( Dec. 2010) 7.0 yrs (June 2012) 1, yrs (June 2010) 6.5 yrs (Dec. 2011) 1, yrs (Dec. 2009) 6.0 yrs (June 2011) 1, yrs (June 2009) 5.5 yrs (Dec. 2010) If only 900 patients are enrolled, follow-up that is long enough to achieve 90% power will extend the study beyond what the budget will allow. An enrollment of 1,000 patients, our minimum acceptable sample size, prolongs the study rather substantially in order to achieve 90% power. A more desirable option as far as achieving the desired 90% level of power while not unduly extending follow-up is to enroll 1,200 patients and extend follow-up by 2 additional years (through 2010). Sample Size for Hypothesis 2: LV Reconstruction (mortality or hospitalization for a cardiac reason): For the LV reconstruction hypothesis where the primary endpoint is death (all cause) or hospitalization for a cardiac reason, the three-year event rate in the control arm (patients treated with CABG without SVR) is expected to be in the range of 50% or higher. If the control rate at 3 years is 50%, 387 patients per arm will provide 90% power for detecting a 20% reduction in the incidence of the primary endpoint, and 460 patients per arm will provide 90% power for detecting a 20% reduction if the control rate should be as low as 45%. To achieve a robust sample size for testing the LV reconstruction hypothesis, 1,000 patients will be enrolled. Below is a summary of what this number of patients will provide the study. DCRI/Confidential Protocol 4 06/16/2006 Page 27

83 1. Power >90% for detecting a 20% reduction in the primary endpoint with CABG + SVR if the event rate at 3 years in patients treated with CABG (without SVR) is 45% or higher. (Thus we have high power for detecting a conservative estimate of benefit if the control arm event rate is consistent with what is expected based on other studies (or even somewhat less than expected based on other studies). 2. Power > 85% for detecting a 20% improvement if the 3-year control event rate should be as low as 40%. (Thus we have high power for detecting a conservative estimate of the benefit if the control arm event rate is lower than expected.) 3. Adequate power for detecting clinically important effects in subgroups of interest. Secondary Endpoints: Cardiac Mortality Since a very high proportion of the deaths in this patient population will be cardiovascular, the power for detecting differences in this secondary endpoint will differ little from that outlined above for the primary endpoint of total mortality in assessing the coronary revascularization hypothesis Treatment Crossovers The sample size calculations previously outlined do not include any adjustment to compensate for the fact that some patients enrolled in STICH may crossover from one treatment arm to another. Although participating clinical centers will be instructed concerning the importance of treatment compliance, and strict guidelines will be established regarding any changes to a patient s assigned therapy, we recognize that in certain instances a downstream change may be clinically indicated. Although all therapy changes and reasons for them will be documented, they are considered part of the initial treatment strategy, and groups will be compared with respect to the initially assigned treatment. Nonetheless, we have attempted to estimate the extent to which crossovers or dropouts might occur, and how such changes in therapy could affect the sample size/power figures cited above. From personal communication with investigators from the Bypass Angioplasty Revascularization Investigation (BARI), and from reviewing data in the Duke Cardiovascular Database, there is ample experience to suggest that the percent of consenting patients who are randomized to CABG, but don t actually receive CABG surgery, will be low (1-2%). We expect 5% crossovers from CABG without SVR to CABG with SVR and 5% of the patients randomized to SVR to not actually undergo the SVR procedure. Crossover is most likely to occur among patients randomized to MED who, because of deteriorating symptoms or evidence of ischemia during follow-up, are felt to need revascularization with CABG. We project that up to 10% of MED patients may crossover to CABG within the first year following randomization, and up to 15% may crossover to CABG within three years. Some of the MED patients will receive percutaneous intervention. Indeed there may be patients in all three arms of the trial who undergo percutaneous intervention during follow-up, although this is more likely to occur among patients randomized to MED. Percutaneous intervention is not regarded as a treatment crossover, but rather as part of the downstream care associated with any of the treatment strategies in the trial. Even if we assume that dropouts and treatment crossovers would reduce the effective sample size by as much as 15-20% (the impact is not expected to be that large), the randomization of 1,000 patients for testing the coronary revascularization hypothesis will still provide > 80% power for detecting a 25% reduction in mortality if the event rate in MED patients is consistent with our projection (25% at three years). For the LV DCRI/Confidential Protocol 4 06/16/2006 Page 28

84 reconstruction hypothesis, the planned sample size of 1,000 patients will provide 90% power for detecting a 20% reduction in the primary endpoint even with liberal estimates for the amount of crossover if the threeyear event rate for mortality or cardiac hospitalization is 50% or higher. 6.2 Statistical Analysis Statistical analysis will be performed at the Statistical and Data Coordinating Center at Duke University. Although the methodologic approaches and operational details of the data analysis will be coordinated by the study biostatisticians, the major analyses of the study data will be highly collaborative, involving both statisticians and physicians to ensure appropriate interpretation of the data. All major treatment comparisons between the randomized groups in this trial will be performed according to the principle of "intention-to-treat;" that is, subjects will be analyzed (and endpoints attributed) according to the treatment arm to which patients were randomized, regardless of subsequent crossover. Statistical comparisons will be performed using twosided significance tests. Additional perspective regarding the interpretation of the data will be provided through extensive use of confidence intervals 33 and graphical displays Background and Demographic Characteristics Baseline demographic and clinical variables, including relevant descriptors from the history and physical examination; risk factors; comorbidity; HF functional class, angiographic assessment of the extent of CAD, left ventricular function, and descriptors obtained from the various noninvasive tests will be summarized for each randomized arm of the study. Descriptive summaries of the distribution of continuous baseline variables will be presented in terms of percentiles (e.g., median, 25th and 75th percentiles), while discrete variables will be summarized in terms of frequencies and percentages. Statistical comparisons of treatment groups with respect to baseline characteristics will be limited to selected variables and disease factors known to influence prognosis. These variables will include age; sex; race; previous myocardial infarction; prior revascularization; descriptors of comorbidity; HF functional class; anatomic severity of disease; and left ventricular function (ejection fraction). For comparisons of treatment groups with respect to continuous baseline variables, emphasis will be given to nonparametric procedures such as the Wilcoxon rank sum test, or Kruskal-Wallis nonparametric analysis of variance. 34 Group comparisons with respect to discrete baseline variables will use the conventional chi-square test Concomitant Therapy Summary statistics will be provided as appropriate. No formal analyses are planned Efficacy Evaluation Coronary Revascularization Hypothesis For the coronary revascularization hypothesis, the major statistical assessment will involve a comparison of the CABG arm (without SVR) versus the medically treated arm with respect to all-cause mortality. The logrank test 29 (sometimes called the Mantel-Haenszel test for survival data 35 ), which is a special case of the more general Cox proportional hazards regression model, will be the primary analytic tool for assessing mortality differences between these two treatment arms. The log-rank test (and the Cox model) can accommodate varying lengths of patient follow-up, and not only uses information for each patient as to whether an event occurred, but also takes into account how long from study entry the patient survived before the endpoint occurred. The log-rank test also accommodates "censored" survival times, which arise because many patients will be alive when the analyses are performed, and the length of time they will survive DCRI/Confidential Protocol 4 06/16/2006 Page 29

85 without an event is known only to be greater than the length of their current follow-up. The log-rank test is a well-established approach for providing an overall comparison of the entire survival curves. Using this procedure, the analysis strategy will be to perform the treatment comparison stratifying by whether or not patients are SVR eligible (i.e., by whether patients are in Stratum A, or in Stratum B; see Figure 1). The analysis is thus reflective of the fact that the randomization will be stratified according to whether patients are in Stratum A versus Stratum B. This comparison will constitute the primary analysis to assess the effect of CABG on overall mortality. Kaplan-Meier survival estimates 36 based on the primary endpoint of all-cause mortality will be calculated for each treatment group to display the outcome results graphically. Even though the primary comparison outlined above will contrast treatment groups that are determined by randomization, supplementary analysis involving other covariate adjustment will be performed with the Cox model. Such adjustment will be limited to a relatively small, prospectively defined set of patient characteristics that are known a priori to have a strong prognostic relationship with mortality. This adjustment will serve as a prelude to supplementary analyses examining differential treatment effects. The adjustment variables will include (in addition SVR eligibility) age, sex, race, HF class, history of myocardial infarction, previous revascularization, number of significantly diseased major coronary arteries, and ejection fraction. Cox model analyses may also be performed using geographic region as a stratification factor. If the data provide evidence of an overall difference in outcome between treatment groups, we will further examine whether the therapeutic effect is similar for all patients, or whether it varies according to specific patient characteristics. In particular, we will focus on whether the relative therapeutic benefit differs according to patient age, sex, race, HF class, ejection fraction, and whether patients are eligible for SVR. These issues will be addressed formally with the Cox model by testing for interactions between treatment and these specific baseline variables. Although the analysis will include an examination for treatment interactions as indicated above, treatment effects for the primary endpoint as characterized by the CABG:MED hazard ratio (with 95% confidence intervals) will be calculated and displayed for several prospectively-defined subgroups of patients defined by age, gender, race, SVR eligibility, HF class, and ejection fraction. These comparisons will be carefully interpreted in conjunction with the formal interaction tests. To supplement the conventional significance testing and confidence interval approaches that will constitute the major analyses for this trial, we will provide additional perspective on the assessment of treatment effects using some established Bayesian approaches to the analysis of clinical trial data. The application of Bayesian methods to clinical trials has recently received considerable attention in the statistical and clinical trials literature. In part, its appeal stems from the fact that what consumers of clinical trial results often want to know is the likelihood (probability) that the treatment is actually beneficial, or the likelihood that the treatment has a clinically important benefit. Such probability assessments are directly obtainable from a Bayesian analysis. Spiegelhalter et al. 37 demonstrated how one can derive clinically useful information such as an estimate of the probability that the hazard ratio (CABG: medical therapy) is less than some specified value (e.g., 0.90) and an estimate of, for example, the probability that CABG therapy is clinically equivalent to intensive medical therapy, i.e., that the hazard ratio is within some interval close to 1.0 (such as 0.90 to 1.10). We will supplement the conventional statistical presentations discussed above by computing Bayesian probabilities that CABG therapy is beneficial and that it has a clinically important benefit. For these computations, a flat (non-informative) prior distribution for the CABG:MED hazard ratio will be assumed. We will use readily available S-Plus functions to perform the Bayesian calculations, or the DCRI/Confidential Protocol 4 06/16/2006 Page 30

86 Cambridge group s BUGS program (Bayesian Inference Using Gibbs Sampling 38 ) to derive a posterior distribution from a full Bayesian analysis. LV Reconstruction Hypothesis For the LV reconstruction hypothesis, the key statistical assessment will involve a comparison of CABG plus SVR versus CABG without SVR with respect to the composite endpoint of death (all cause) or hospitalization for a cardiac reason. The log-rank test will also be the primary analytic tool for this assessment. In this case, the test will be performed stratifying by whether the patients were eligible for either medical or surgical therapy versus whether surgery was indicated (i.e., by whether the patients were in Stratum B or Stratum C as depicted in Figure 1. Again, Kaplan-Meier estimates will be calculated and plotted for each treatment group to display the outcome results graphically. In this case, the Kaplan-Meier curves will display event-free survival, i.e., survival free of any cardiac hospitalization. Again, the primary analysis will be supplemented by Cox model analyses involving other covariate adjustment. In addition to the stratification factor specified above for the primary LV reconstruction hypothesis, the small, prospectively-defined set of patient characteristics listed above as covariates for the coronary revascularization hypothesis will also be used in supportive analyses of the LV reconstruction hypothesis. If the data provide evidence of an overall difference in outcome between treatment groups, we will determine whether the therapeutic effect is similar for all patients, or whether it varies according to specific patient characteristics. In particular, we will assess whether the relative therapeutic benefit differs according to patient age, sex, race, HF class, ejection fraction, and whether or not the patients are eligible for either medical or surgical therapy. This will be addressed formally with the Cox model by testing for interactions between treatment and these specific baseline variables. We will provide additional perspective on the assessment of treatment effects with respect to the LV reconstruction hypothesis through calculating the hazard ratio for CABG with SVR relative to CABG without SVR and 95% confidence intervals (overall and in prospectively defined subgroups), and also through performing supplementary Bayesian analysis as outlined above for the coronary revascularization hypothesis. Analysis of Secondary Endpoints Secondary endpoints that will be evaluated in this trial include (1) cardiovascular mortality; (2) all-cause mortality within 30 days; (3) quality of life; (4) cost of care and cost effectiveness; (5) morbidity; (6) functional measures, including six minute walk distance and exercise duration on a treadmill test; and (7) physiologic measures from noninvasive testing, neurohormonal levels, and cytokine levels. Data analyses for each of these endpoints are discussed below. Cardiovascular Mortality This endpoint is most relevant for the assessment of the effects of coronary revascularization, and therefore will be used in comparisons of medical therapy vs. CABG (without SVR). The analysis will be similar to that outlined for the primary endpoint, using time from enrollment until death from a cardiovascular cause or censoring as the response variable, and assessing treatment differences using the Cox proportional hazards model. Kaplan-Meier survival curves will be computed to graphically display the survival experience of the treatment groups as a function of time from randomization. DCRI/Confidential Protocol 4 06/16/2006 Page 31

87 Mortality within 30 days The analysis for this endpoint will also be similar to that outlined for the primary mortality endpoint except that 30-day outcome will be treated as a binary endpoint (due to its short-term nature), and the logistic regression model will be used for treatment comparisons rather than the Cox proportional hazards model. These analyses will be performed for both MED versus CABG and for CABG + SVR versus CABG (without SVR) in order to fully characterize and compare procedural mortality. Thirty-day mortality rates will be reported for each treatment arm along with odds ratios and 95% confidence intervals for CABG:MED and for CABG + SVR:CABG (without SVR). Quality of Life and Cost of Care The important quality of life and cost of care endpoints analysis plans are addressed in detail in the EQOL Core Laboratory manual. These important endpoints include health status measures from the Medical Outcomes Study 36 Item Short Form (SF-36); quality of life and cardiac symptoms as assessed by the disease-specific Kansas City Cardiomyopathy Questionnaire and Seattle Angina Questionnaire; a measure of depressive symptoms based on the Center for Epidemiologic Studies-Depression (CES-D) scale; and utilities assessed by patient interview using the EuroQol. Morbidity Morbidity in this trial will be assessed by examining the incidence of several different complications and clinical events that occur during the follow-up period. These events will include complications associated with each treatment arm such as the frequency and duration of hospitalization for HF, and additional cardiac procedures (heart transplantation, LVAD, ICD, and stroke). We outline below the approach that we will take for analyzing one of these important complications, for example hospitalization for HF, noting that specific components of this approach can also be applied to the other non-fatal complications. The approach involves several related yet different analyses to comprehensively assess and fully characterize differences among the treatment arms with respect to this outcome. One characterization of this secondary endpoint will be in terms of the time from randomization until a patient experiences their first hospitalization for HF during their follow-up in the trial. The methodology for analyzing censored failure-time data outlined previously for the primary endpoint will thus be applied to compare treatment groups with respect to this outcome. We point out that this analysis must be interpreted cautiously, however, because hospitalization as a non-fatal event may appear to occur with a lower incidence in one treatment group simply because that arm had a higher death rate. Obviously after patients die, they can no longer be hospitalized for their HF. To deal with this complexity, we will perform further analyses of this adverse complication by considering hospitalization for HF and death as a combined endpoint. We will thus consider the time until the first occurrence of either death (any cause) or hospitalization for HF, and treatment differences will be assessed using the Cox proportional hazards model. In support of this analysis, Kaplan- Meier event-free survival curves for each treatment arm will be calculated to graphically display event rates over time for this composite endpoint. The third approach for analyzing this endpoint will take advantage of the fact that the two components of the composite endpoint (death or hospitalization) have a natural rank-ordering in terms of their severity. The treatments can thus be compared with an approach that not only uses the time until the first occurrence of one of these component outcomes, but also factors in the severity of the outcome. The approach we will employ, developed by Berridge and Whitehead, 39 combines the Cox proportional hazards model in conjunction with an ordinal severity of event model to potentially increase power for the treatment comparisons. The severity of event portion is a continuation ratio-type ordinal logistic model. An extended hazard function is defined by multiplying the regular Cox hazard function by the probability of the event being of severity j from the ordinal model, where j references the various severity categories. In this analysis, death will obviously be DCRI/Confidential Protocol 4 06/16/2006 Page 32

88 considered as the most severe event, and hospitalization for HF as least severe. Within this framework, one can test whether one of the interventions under study in this trial prolongs the time to the first occurrence of either of these events, whether the therapy decreases the severity of the first event that occurs, or (with 2 degrees of freedom), whether the therapy has either effect. Finally, since some patients may experience hospitalization multiple times, a comprehensive comparison of the treatment arms with respect to this outcome should take into account the longitudinal pattern in the repeated occurrences of such an event. In clinical situations where there may be multiple events per subject, it is desirable to be able to apply time-to-event methods such as the proportional hazards model which explicitly allow for varying lengths of patient follow-up and appropriately handle censored observations. A major issue in extending proportional hazards regression models to this situation, however, is intra-subject correlation (i.e., where multiple events occur within the same patient, those events will be correlated). Fortunately, there is active, ongoing methodological research on the application of survival models to this situation In order to provide a more complete and comprehensive analysis of the hospitalization data, the approach we will use in STICH is based on extensions of the proportional hazards model that accommodate multiple events per patient. The specific approach makes use of cluster modifications of so-called sandwich estimates of the variance-covariance matrix of the regression coefficients, thus providing standard errors of the regression coefficients that take into account the correlations among multiple event times within a given patient Specialized software functions for performing such analyses are available in S-Plus. 43 Because of the problems already mentioned in analyzing hospitalization by itself, we will use the multipleevent methodology outlined above, which not only allows multiple events per patient of the same type (e.g., multiple hospitalizations), but also accommodates events of different types. Thus we will model both death and the multiple episodes of rehospitalization using multiple-events Cox model analysis By synthesizing the results from these different but complementary approaches (that is, by considering hospitalization in terms of a single event per patient, in terms of multiple events per patient, and hospitalization combined with death), we will be able to provide a comprehensive assessment of treatment differences with respect to hospitalization, and as part of the analyses, identify and assess other clinical factors that are associated with this important secondary outcome. The approaches outlined above for analyzing hospitalization for HF can also be applied to other clinical outcomes and adverse events that will be of interest in this trial, such as stroke and other morbidity outcomes enumerated above. Functional Capacity Outcomes These outcomes will include distance covered in a six-minute walk test and exercise duration on a treadmill test. Of interest will be both the absolute magnitude of these measures as obtained during follow-up and also the change from baseline to the follow-up measurement. Treatment group differences will be statistically assessed using the nonparametric Wilcoxon rank sum test. Physiologic Measures from Noninvasive Tests (Core Laboratory Data) There will be considerable noninvasive test data acquired in the trial, including cardiac magnetic resonance imaging, echocardiography, and nuclear cardiology (radionuclide) studies. The schedule for obtaining these various tests is outlined in Tables 2 and 3 and generally involves the acquisition of data at baseline, at four months following study entry, and after two years of follow-up. This information will be valuable in addressing the mechanisms responsible for differences or similarities in patient outcomes observed among the randomized arms in the trial. Furthermore, with treatment groups defined by randomization and tests systematically performed and uniformly interpreted, bias in the choice of therapy and in the choice of which tests are obtained is eliminated. The data from these various tests may be very important for identifying DCRI/Confidential Protocol 4 06/16/2006 Page 33

89 subsets of patients where there are clear benefits of myocardial revascularization, and perhaps other subsets where intensive medical therapy is an attractive choice. Even where revascularization appears to be a superior strategy, there may be some patients who do and others who do not benefit from SVR. The noninvasive information acquired in the trial will therefore be very useful in defining a pretreatment strategy for performing a sequence of tests that will optimize the choice of therapy. The analysis of the data from these tests will consist of several parts: (1) a detailed assessment of the prognostic relationships of the baseline test data to clinical outcomes, and of the relative prognostic importance of each test; (2) an examination of differential treatment effects as a function of the baseline noninvasive test data; (3) a comparison between treatments of the changes in test results from baseline to four months as a characterization of early mechanistic effects of therapy; (4) an assessment of the fourmonth measures and the change from baseline to four months as predictors of subsequent outcomes; (5) a comparison of treatments with respect to the two year test outcomes and the change from baseline to two years as viewed from the perspective of an endpoint, and also as it might relate to late clinical outcomes; (6) an analysis comparing treatment arms with respect to the trajectory of key noninvasive test measures, considering all three measurement times (baseline, four months, two years). For the analyses in part (1), we will use the Cox proportional hazards model to examine individual and joint relations between baseline test results and length of survival (or event-free survival in the case of the LV reconstruction endpoint). For continuous baseline measures such as ejection fraction and ESVI, we will examine the shape and strength of the relationship by use of a flexible model-fitting approach involving cubic spline functions (cubic polynomials) The resulting functions will be graphically and statistically examined to assess the model assumptions and to judge the linearity of the relationships. Where relations are nonlinear, their shape will be characterized using spline functions. Determining how variables should be modeled will be an important step in characterizing prognostic relations and identifying which variables are most strongly related to long-term outcome. Once the important predictors from each individual test are identified, we will contrast the relative prognostic importance and the independent contribution of the various tests by jointly considering multiple tests. The modeling strategies to be used and the approach to evaluating the yield of each test are similar to those we have employed in previous studies. 47,48 Because there will be over 400 deaths occurring in the study population during the period of follow-up, the sample size (number of events) will adequately support these extensive analyses. For step (2), we will use the key features identified in step (1) from each noninvasive test, and statistically test for a differential treatment effect using the Cox model. This will involve testing for interactions between treatment and specific baseline test measures (e.g., ESVI, measures of myocardial viability and myocardial perfusion). For step (3), the four-month test results and the change from baseline to four months in the prognostic factors identified in step (1) above will be descriptively summarized, and treatments will be compared using the Wilcoxon rank sum test in the case of continuous or ordinal measures, and the chi-square test for categorical variables. For step (4) where we are dealing with post-randomization information and attempting to judge its usefulness for predicting subsequent outcomes, the analysis becomes more complex. We will use two approaches. First, we will assess the relations of the four-month noninvasive test data to subsequent outcomes (outcomes after four months) in a fashion similar to that described above for the analysis of the baseline test data. Second, we will include events starting from randomization, but consider the four-month test results as intervening or time-dependent data, and factor this into the Cox model using time-dependent covariates. In both of these analyses, we will seek to address the issue of whether the four-month test results DCRI/Confidential Protocol 4 06/16/2006 Page 34

90 or the change from baseline to four months provides independent predictive information beyond that contained in the baseline test data, and thereby assess the utility of doing the follow-up studies. For step (5), treatment arms will be compared with respect to two-year test outcomes (considering the test results as an endpoint) using a similar approach as described for the four-month data in step (3). We will also attempt to judge the utility of the two-year test data in predicting subsequent outcomes as in step (4) above, recognizing, however, that the complexities of this analysis will require very cautious interpretation. For step (6), we will characterize the trajectory of key prognostic measures from the tests by considering all three measurement times. In considering the longitudinal nature of these data, there are three major statistical issues that must be considered in making treatment comparisons: (a) the data may not be normally distributed, (b) the longitudinal measures within subjects will be correlated, and (3) there will be missing data, resulting from patients in whom it is not possible to obtain all three measurements (because they die or fail to return for other reasons to enable the follow-up measurements to be made). To address the first two issues, we will use the generalized estimating equation (GEE) approach to model estimation and testing proposed by Liang and Zeger. 49 The software for implementing this method is now available in a SAS macroroutine accessible to the CC. For the third issue above (missing data), we will, of course, make every effort to minimize the amount of missing data. However, because there will inherently be missing data, we will incorporate several strategies and assess the sensitivity of our conclusions to the missing data approach employed. The alternatives that will be considered include (1) analyze only the patients with complete data; (2) assign worst case scores to missing observations; (3) carry forward previously observed data to observation periods that are missed later; and (4) use likelihood-based methods to impute missing data. A description of these methods and their limitations is given in the review by Heyting. 50 By synthesizing the results from these different but complementary approaches, we will be able to provide a comprehensive assessment of treatment differences with respect to the key mechanistic noninvasive test data, judge whether there is a preference for one particular treatment depending on the results of the noninvasive tests, and develop a recommendation for performing the sequence of tests that will optimize the choice of therapy. With regard to physiologic measures such as neurohormonal levels and cytokine levels provided by the Neurohormone/Cytokine/Genetic Core Laboratory, there will be interest in both the absolute levels at followup and also in the change from baseline to follow-up. Treatment group differences will be statistically assessed using the Wilcoxon rank sum test applied to these various measures. Multiple Comparisons Although there are two primary aims, and there will be some overlap of the patients used for testing the two primary hypotheses [the overlapping patients will be those in Stratum B who are randomized to CABG (without SVR), projected to be approximately 75 patients], the statistical test for each primary hypothesis will be performed at the 0.05 level of significance. We justify this choice on the basis that the two hypotheses are addressing two very distinct questions. Indeed, the endpoints for the two questions are different, and were it not for the overlapping patients, the study would basically resemble two separate trials. With two primary hypotheses and the various secondary endpoints that have been outlined, we recognize that there is a multiplicity of analyses to be performed, which leads to an increased probability that at least one of the comparisons could be "significant" by chance. There are adjustments (e.g., based on the Bonferroni inequality) that can be used to preserve the overall type I error level. To attempt to adjust for the effects of the repeated significance testing that will occur as part of the interim monitoring (discussed below), plus adjust for the multiplicity of analyses, would require that small significance levels be used for DCRI/Confidential Protocol 4 06/16/2006 Page 35

91 every comparison. Instead of formally adjusting the significance level for the fact that there are two primary hypotheses and numerous secondary comparisons, we will be conservative in the interpretation of the analyses, taking into account the degree of significance, and looking for consistency across endpoints. The actual p-value for each comparison will be reported (not simply whether significance is achieved) to aid in the overall interpretation. Also, the Bayesian interpretations discussed above (along with so-called credible intervals) will assist in providing an appropriate interpretation of the study results. We have also pre-specified the primary and secondary outcome variables to help guard against the multiple testing problems Safety Evaluation The assessment of safety is based mainly on the frequency of the pre-defined safety parameters and serious adverse events suspected by the investigator to be related to study treatment. Other safety data (e.g., vital signs) will be considered and summarized as appropriate. Serious adverse events suspected by the investigator to be related to study treatment will be summarized for each treatment group by presenting the number and percentage of patients having any serious related adverse event, having a serious related event in each body system and having each individual serious related adverse event Interim Analysis For ethical reasons, an interim examination of key safety and endpoint data will be performed at regular intervals during the course of the trial. The primary objective of these analyses will be to evaluate the accumulating data for an unacceptably high frequency of negative clinical outcomes in any of the treatment arms. In addition, however, the interim monitoring will also involve a review of the control arm event rates, patient recruitment, compliance with the study protocol, submission of data forms, and other factors which reflect the overall progress and integrity of the study. The results of the interim analyses will be carefully and confidentially reviewed by a Data and Safety Monitoring Board (DSMB). It is anticipated that the DSMB will meet at approximately six-month intervals to review the accumulating data. Prior to each meeting, the Statistical Data Coordinating Center will conduct the desired statistical analyses and prepare a summary report that will be carefully reviewed by the DSMB. The extracted data files and analysis programs for each DSMB report will be archived and maintained at the Statistical Data Coordinating Center for the life of the study. Reports will be presented describing the progress of patient enrollment, the rates of compliance with therapy, and the frequency of protocol violations. The Statistical Data Coordinating Center will also report on the number of forms received, the number of forms entered in the computer, the number of forms queried, the number of queries completed, the number of forms reviewed through on-site monitoring, and the number of forms receiving the final, completed entry in the double dataentry system. These interim safety and efficacy reports introduce well-recognized statistical problems related to the multiplicity of statistical tests performed on an accumulating set of data. 51,52 As a solution to the problem of repeated tests, we propose to adopt for use in STICH a group sequential method similar to that proposed by O'Brien and Fleming 53 as a guide in interpreting interim analyses. This procedure requires large critical values early in the study, but relaxes (i.e., decreases) the critical value as the trial progresses. Because of the conservatism early in the trial, the critical value at the final analysis is near the "nominal" critical value. Hence the sample size requirements with this group sequential procedure remain essentially the same as the conventional fixed sample size estimate. The actual method for this interim monitoring that will be employed in STICH is the general approach to group sequential testing developed by Lan and DeMets 54 for which neither the number of looks nor the increments between looks must be pre-specified. Rather, the Lan- DeMets approach only requires specification of the rate at which the Type I error (which in this trial will be DCRI/Confidential Protocol 4 06/16/2006 Page 36

92 chosen to be α=0.05) will be "spent". This procedure allows "spending" a little of α at each interim analysis in such a way that at the end of the study, the total Type I error does not exceed One such spending function generates boundaries that are nearly identical to the O'Brien-Fleming boundaries. It is this approach that we propose to use in STICH, namely two-sided, symmetric O'Brien-Fleming 53 type boundaries generated using the flexible Lan-DeMets 54 approach to group sequential testing. Since the number of looks and the increments between looks need not be pre-specified, it allows considerable flexibility in the monitoring process for accommodating additional comparative examinations of the data in response to concerns of the DSMB that may arise during the course of the trial. Assuming that the DSMB will conduct its first formal data review in the latter half of the first year of recruitment, and then continue those reviews approximately every 6 months thereafter through the patient recruitment period (three years) and the follow-up phase (three years), there will be approximately eight to ten reviews of the data. With nine interim analyses approximately equally spaced in time, the Lan and DeMets "spending function" that approximates the O'Brien-Fleming stopping boundaries involves a very stringent alpha level ( ) for declaring significance at the first interim analysis. At the subsequent interim analyses, the required significance levels will be somewhat less stringent. The requirements for significance at each interim analysis, depending on exactly when the analysis occurs, can be computed with the Lan-DeMets methodology, for which we have suitable computer software. The final analysis can be undertaken with a significance level of approximately 0.04, relatively close to the nominal 0.05 level. This approach to interim monitoring will be carried out in parallel for both of the primary study hypotheses. The analytic approach that will be used at the interim analyses for assessing treatment differences will be the time-to-event analysis methods described above, except that interpretation of statistical significance associated with treatment comparisons of key study endpoints will be guided using the group sequential stopping boundaries outlined above The appropriateness of using the log-rank test (or equivalently the Cox model) in the group sequential framework has previously been well established For each of these interim analyses, the critical value of the test statistic and the corresponding p-value required for significance in that particular analysis will be presented so that significance can be assessed precisely. If significantly large and important treatment differences are observed at any of the interim analyses, the DSMB may recommend that randomization of patients be stopped, or that the design and conduct of the trial be appropriately modified. Judgment concerning the continuation or termination of the study will involve not only the degree of statistical significance observed at the interim analysis, but also the likelihood of achieving significance should enrollment continue to the originally projected sample size. As an aid in this latter assessment, the Statistical Data Coordinating Center will supplement the group sequential analyses outlined above with calculations of conditional power based on the method of stochastic curtailment This procedure evaluates the conditional probability that a particular statistical comparison will or will not be significant at the end of the trial at the α level used in the design, given the hypothesized treatment difference and the data obtained to date. Conditional power for the primary comparisons associated with both primary hypotheses will be computed and provided to the DSMB as part of the interim study reports. The approach to interim monitoring outlined above will be carried out in parallel for both of the primary study hypotheses. In addition, for the LV reconstruction hypothesis where the primary endpoint is a composite of death or cardiac hospitalization, it will also be important to carefully monitor the mortality component of this endpoint. Thus mortality rates and associated confidence intervals for each arm (CABG and CABG + SVR) will also be monitored at the interim reviews to ensure that the safety of patients enrolled in the trial is not compromised. DCRI/Confidential Protocol 4 06/16/2006 Page 37

93 7. Ethics and Good Clinical Practice This study must be carried out in compliance with the protocol and in accordance with STICH standard operating procedures. These procedures are designed to ensure adherence to Good Clinical Practice, as described in the following documents: 1. ICH Harmonized Tripartite Guidelines for Good Clinical Practice Directive 91/507/EEC, The Rules Governing Medicinal Products in the European Community. 3. US 21 Code of Federal Regulations dealing with clinical studies (including parts 50 and 56 concerning informed consent and IRB regulations). 4. Declaration of Helsinki, concerning medical research in humans (Recommendations Guiding Physicians in Biomedical Research Involving Human Subjects, Helsinki 1964, amended Tokyo 1975, Venice 1983, Hong Kong 1989, Somerset West 1996). The investigator agrees, when signing the contract with the Coordinating Center, to adhere to the instructions and procedures described in the protocol and thereby to adhere to the principles of Good Clinical Practice that it conforms to. 7.1 Institutional Review Board/Independent Ethics Committee Before implementing this study, the protocol, the proposed informed consent form and other information to subjects, must be reviewed by a properly constituted Institutional Review Board/Independent Ethics Committee (IRB/IEC). A signed and dated statement that the protocol and informed consent have been approved by the IRB/IEC must be given to the Coordinating Center before study initiation. The name and occupation of the chairman and the members of the IRB/IEC must be supplied to the Coordinating Center. Any amendments to the protocol, other than administrative ones, must be approved by this committee Informed Consent The investigator must explain to each subject (or legally authorized representative) the nature of the study, its purpose, the procedures involved, the expected duration, the potential risks and benefits involved and any discomfort it may entail. Each subject must be informed that participation in the study is voluntary and that he/she may withdraw from the study at any time and that withdrawal of consent will not affect his/her subsequent medical treatment or relationship with the treating physician. Eligible patients will be approached for consent to enter a specific stratum determined by MED and SVR eligibility. This informed consent should be given by means of a standard written statement, written in nontechnical language. The patient should read and consider the statement before signing and dating it, and should be given a copy of the signed document. If written consent is not possible, oral consent can be obtained if witnessed by a signed statement from one or more persons not involved in the study, mentioning why the patient was unable to sign the form. No patient can enter the study before his/her informed consent has been obtained. Once randomized, patients will be asked to provide a separate informed consent for genotyping. The informed consent forms are part of the protocol, and must be submitted by the investigator with it for IRB/IEC approval. The Coordinating Center will supply proposed informed consent forms, which comply with regulatory requirements and are considered appropriate for the study. Any changes to the proposed consent form suggested by the investigator must be agreed to by the Coordinating Center before submission to the IRB/IEC, and a copy of the approved version must be provided to the Coordinating Center after IRB/IEC approval. DCRI/Confidential Protocol 4 06/16/2006 Page 38

94 8. References O Connell JB, Bristow MR. Economic impact of heart failure in the United States: Time for a different approach. Heart Lung Transplant 1993; 13:S107-S Kaesemeyer WH. Holding smokers accountable for heart disease costs. Circulation 1994; 90: Braunwald E, Bristow MR. Congestive heart failure: Fifty years of progress. Circulation 2000; 102:IV- 14 IV Reimer KA, Jennings RB. The Wavefront Phenomenon of myocardial ischemic cell death. II. Transmural progression of necrosis within the framework of ischemic bed size (myocardium at risk) and collateral flow. Lab Invest 1979; 40: Jatene AD. Left ventricular aneurysmectomy. Resection or reconstruction. J Thorac Cardiovasc Surg 1985; 89(3): Dor V, Saab M, Coste P, Kornaszewska M, Montiglio F. Left ventricular aneurysm: a new surgical approach. J Thorac Cardiovasc Surg 1989; 37(1): Di Donato M, Sabatier M, Dor V, Gensini GF, Toso A, Maioli M, Stanley AWH, Athanasuleas C, Buckberg G. Effects of the Dor procedure on left ventricular dimension and shape and geometric correlates of mitral regurgitation one year after surgery. J Thorac Cardiovasc Surg 2001; 121: Athanasuleas CL, Stanley AWH, Buckberg GD, Dor V, DiDonato M, Blackstone EH, and the RESTORE Group. Surgical anterior ventricular endocardial restoration (SAVER) in the dilated remodeled ventricle following anterior myocardial infarction. J Am Coll Cardiol 2001; 37: Jones RH. Is it time for a randomized trial of surgical treatment of ischemic heart failure? Invited Editorial. J Am Coll Cardiol 2001: Alderman EL, Bourassa MG, Cohen LS, Davis KB, Kaiser GG, Killip T, Mock MB, Pettinger M, Robertson TL for the CASS Investigators. Ten-year follow-up of survival and myocardial infarction in the randomized Coronary Artery Surgery Study. Circulation 1990; 82: Yusuf S, Zucker D, Peduzzi P, Fisher LD, Takaro T, Kennedy JW, Davis K, Killip T, Passamani E, Norris R, Morris C, Mathur V, Varnauskas E, Chalmers TC. Effect of coronary artery bypass grafting on survival: Overview of 10-year results from randomised trials by the Coronary Artery Bypass Graft Surgery Trialists Collaboration. Lancet 1994; 344: Baker DW, Jones R, Hodges J, Massie BM, Konstam MA, Rose EA. Management of Heart Failure. III. The role of revascularization in the treatment of patients with moderate or severe left ventricular systolic dysfunction. JAMA 1994; 272: Eitzman D, al-aouar Z, Kanter HL, et al. Clinical outcome of patients with advanced coronary artery disease after viability studies with positron emission tomography. J Am Coll Cardiol 1992; 20: Di Carli MF, Davidson M, Little R, et al. Value of metabolic imaging with positron emission tomography for evaluating prognosis in patients with coronary artery disease and left ventricular dysfunction. Am J Cardiol 1994; 73: Lee KS, Marwick TH, Cook SA, et al. Prognosis of patients with left ventricular dysfunction with and without viable myocardium after myocardial infarction: relative efficacy of medical therapy and revascularization. Circulation 1994; 90: Gioia G, Powers J, Heo J, Iskandrian AS. Prognostic value of rest-redistribution tomographic thallium-201 imaging in ischemic cardiomyopathy. Am J Cardiol 1995; 75: Vom Dahl J, Altehoefer C, Sheehan FH, et al. Effect of myocardial viability assessed by technetium-99m sestamibi SPECT and fluorine-18-fdg PET on clinical outcome in coronary artery disease. J Nucl Med 1997; 38: Afridi I, Grayburn PA, Panza JA, Oh JK, Zoghbi WA, Marwick TH. Myocardial viability during dobutamine echocardiography predicts survival in patients with coronary artery disease and severe left ventricular systolic dysfunction. J Am Coll Cardiol 1998; 32: DCRI/Confidential Protocol 4 06/16/2006 Page 39

95 20. Cuocolo A, Petretta M, Nicolai E, et al. Successful coronary revascularization improves prognosis in patients with previous myocardial infarction and evidence of viable myocardium at thallium-201 imaging. Eur J Nucl Med 1998; 25: Pasquet A, Robert A, D Hondt AM, Dion R, Melin JA, Vanoverschelde JL. Prognostic value of myocardial ischemia and viability in patients with chronic left ventricular ischemia. Circulation 1999; 100: Senior R, Kaul S, Lahiri A. Myocardial viability on echocardiography predicts long-term survival after revascularization in patients with ischemic congestive heart failure. J Am Coll Cardiol 1999; 33: Chaudhry FA, Tauke JT, Alessandrini RS, Vardi G, Parker MA, Bonow RO. Prognostic implications of myocardial contractile reserve in patients with coronary artery disease and left ventricular dysfunction. J Am Coll Cardiol 1999; 34: Michels VV, Moll PP, Miller FA. The frequency of familial dilated cardiomyopathy in a series of patients with idiopathic dilated cardiomyopathy. N Engl J Med 1992; 326: Leiden JM. The genetics of dilated cardiomyopathy: emerging clues to the puzzle. N Engl J Med 1997; 337: Olson TM, Michels VV, Thibodeau SN, Tai YS, Keating MT. Actin mutations in dilated cardiomyopathy, a heritable form of heart failure. Science 1998; 280: Dilsizian V, Rocco TP, Freedman NMT, Leon MB, Bonow RO. Enhanced detection of ischemic but viable myocardium by the reinjection of thallium after stress-redistribution imaging. N Engl J Med 1990;323: Sciagra R, Bisi G, Santoro GM, et al. Comparison of baseline-nitrate technetium-99m sestamibi with restredistribution thallium-201 tomography in detecting viable hibernating myocardium and predicting postrevascularization recovery. J Am Coll Cardiol 1997;30: Kalbfleisch JD, Prentice RL. The Statistical Analysis of Failure Time Data. New York, NY: John Wiley & Sons, Inc; Cox DR. Regression models and life-tables (with discussion). J Royal Statist Soc B 1972;34: Breslow NE. Covariance analysis of censored survival data. Biometrics 1974;30: Schoenfeld DA. Sample-size formula for the proportional-hazards regression model. Biometrics 1983;39: Borenstein M. The case for confidence intervals in controlled clinical trials. Contro Clin Trials 1994;15: Lehmann EL. Nonparametrics: Statistical Methods Based on Ranks. San Francisco: Holden-Day Inc.; Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 1966;50: Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Statist Assn 1958;53: Spiegelhalter DJ, Freedman LS, Parmar MKB. Bayesian approaches to randomized trials. J Royal Stat Soc 1994;157: Thomas A, Spiegelhalter DJ, Gilks WR. BUGS: A program to perform Bayesian inference using Gibbs sampling. In: Bernardo JM, Berger JO, Dawid AP, Smith AFM. Bayesian Statistics. Oxford: Clarendon Press; 1992: Berridge DM, Whitehead J. Analysis of failure time data with ordinal categories of response. Stat Med 1991;10: Pepe MS, Cai J. Some graphical displays and marginal regression analyses for recurrent failure times and time dependent covariates. J Am Statist Assoc 1993;88: Wei LJ, Lin DY, Weissfeld L. Regression analysis of multivariate incomplete failure time data by modeling marginal distributions. J Am Statist Assoc 1989;84: Lee EW, Wei LJ, Amato D. Cox-type regression analysis for large numbers of small groups of correlated failure time observations. In: Klein JP, Goel PK, eds. Survival Analysis, State of the Art. Rotterdam: Kluwer Academic Publishers; DCRI/Confidential Protocol 4 06/16/2006 Page 40

96 43. Therneau TM, Grambsch P. Modeling Survival Data: Extending the Cox Model. New York: Springer- Verlag; Harrell FE Jr, Pollock BG, Lee KL. Graphical methods for the analysis of survival data. Proc 12th Annual Conference. SAS Users Group International 1987;12: Harrell FE Jr, Lee KL, Pollock BG. Regression models in clinical studies: determining relationships between predictors and response. J Nat Cancer Inst 1988;80: Smith PL. Splines as a useful and convenient statistical tool. Am Statistician 1979;33: Harrell FE Jr, Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluating the yield of medical tests. JAMA 1982;247: Califf RM, Pieper KS, Lee KL, Van de Werf F, Simes RJ, Armstrong PW, Topol EJ for the GUSTO-I Investigators. Prediction of 1-year survival after thrombolysis for acute myocardial infarction in the Global Utilization of Streptokinase and TPA for Occluded Coronary Arteries Trial. Circulation 2000;101: Liang K-Y, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73: Heyting A, Tolboom JTBM, Essers JGA. Statistical handling of drop-outs in longitudinal clinical trials. Statistics in Medicine 1992;11: Armitage P, McPherson CK, Rowe BC. Repeated significance tests on accumulating data. J R Stat Soc Series A 1969;132: McPherson K. Statistics: the problem of examining accumulating data more than once. N Engl J Med 1974;290: O'Brien PC, Fleming TR. A multiple testing procedure for clinical trials. Biometrics 1979;35: Lan KKG, DeMets DL. Discrete sequential boundaries for clinical trials. Biometrika 1983;70: Geller NL, Pocock SJ. Interim analyses in randomized clinical trials: Ramifications and guidelines for practitioners. Biometrics 1987; Tsiatis, AA. The asymptotic joint distribution of the efficient scores tests for the proportional hazards model calculated over time. Biometrika 1981; 68: Tsiatis AA. Repeated significance testing for a general class of statistics used in censored survival analysis. J Am Stat Assoc 1982;77: Gail MH, DeMets DL and Stud EV. Simulation studies on increments of the two-sample log rank score test for survival data, with application to group sequential boundaries. Edited by R. Johnson and J. Crowley. In Survival Analysis. Monograph Series 2. Hayward, California: IMS Lecture Notes, 1982; DeMets DL, Gail MH. Use of logrank test and group sequential methods at fixed calendar times. Biometrics 1985;41: Halperin M, Lan KKG, Ware JH et al. An aid to data monitoring in long-term clinical trials. Controlled Clin Trials 1982;3: Lan KKG, Simon R and Halperin M. Stochastically curtailed tests in long-term clinical trials. Commun Stat., Sequential Analysis. 1982;1: Ware JH, Muller JE, Braunwald E. The futility index, an approach to the cost-effective termination of randomized clinical trials. Am J Cardiol 1985;78: DCRI/Confidential Protocol 4 06/16/2006 Page 41

97 PROTOCOL Surgical Treatment for Ischemic Heart (STICH) Failure Trial Author: Eric J. Velazquez, Kerry L. Lee, Daniel B. Mark, Christopher M. O Connor, Robert M. Califf, Robert H. Jones Document Type: Document Status: Clinical Study Protocol Version 5, Surgical Treatment for Ischemic Heart Failure Extended Study (STICHES) Date: July 29, 2010 Number of Pages: 47 Property of Duke Clinical Research Institute (DCRI) Confidential May not be used, divulged, published or otherwise disclosed without the consent of DCRI

98 Table of Changes to Version 4 Page Section Version 4 Version 5 Rationale 9 List of Abbreviations Removed abbreviations not used in text and added abbreviations used but omitted from the list Added: Surviving patients from Hypothesis 1 will be followed for an additional 5 years, thereby acquiring critically important longer-term (10-year average) information on patients with HF and LV systolic dysfunction treated with MED with and without CABG. 18 Figure 1 Modified table to reflect actual enrollment of STICH Table 3.1 summarizes the schedule of follow-up studies in the STICH Trial Table 3.1 STICH Follow-up Evaluation Studies A serious adverse event is defined in general as an untoward (unfavorable) event which: Tables 3.1 and 3.2 summarize the schedule of follow-up studies in the STICH trial and STICH Extended Study (STICHES), which is an additional 5-year follow up for surviving patients from Hypothesis 1. Table 3.1 STICH Follow-up Evaluation Studies Table 3.2 STICHES Follow-up Evaluation Studies An SAE is defined in general as an untoward (unfavorable) event that meets any of the following criteria: Added: For the extension study, an ECHO assessment is not required. However, if an ECHO assessment is performed as standard of care during the 5-year extension study, the performing site should forward a copy of the study to the DCRI The following Monitoring Committees will be employed in this study: The following Monitoring Committees will be employed in this study and will continue through the 5-year extension: To match the text. To add follow-up information for STICHES. Enrollment has ended. To add follow-up information for STICHES study. To add follow-up information for STICHES. Clarification. Clarification of the requirements for patients entering the 5-year extension study. Clarification of the requirements for the 5-year extension study. DCRI/Confidential Protocol 5 07/29/2010 Page 2

99 Page Section Version 4 Version 5 Rationale Sample Size for STICHES: Coronary Revascularization (allcause mortality) The sample size for STICHES is the 1212 patients who were enrolled in the coronary revascularization hypothesis (Hypothesis 1) of STICH. There will be no new patients enrolled as part of the extended followup. By the time the planned period of follow-up of the Hypothesis 1 patient cohort is completed, it is expected that 400 of these 1212 patients will have died. Thus the number of patients remaining for the extended follow-up will be approximately 800. The additional follow-up of this cohort planned in the extended study will provide very important information for enhancing the precision with which we can estimate the true long-term effect of CABG in this patient population over an average 10- year period. The follow-up of these patients for an additional 5 years is projected to add 300 or more deaths, which will substantially increase the statistical power for detecting mortality differences if the reduction in STICH should prove to be less than originally hypothesized. Of critical importance, however, is that the additional 300+ events will improve considerably the precision of our estimate of the true treatment effect. This is information that is vitally important for clinicians who care for these patients, guideline committees who are writing recommendations for treating these patients, and policy makers. For example, with the 400 deaths in STICH, we can project that the width of the 95% confidence interval (CI) for the hazard ratio that will be Clarification of the statistical analysis planned for the extended followup cohort. DCRI/Confidential Protocol 5 07/29/2010 Page 3

100 Page Section Version 4 Version 5 Rationale computed to describe the effect of CABG vs. MED will be approximately This means, for example, that if the observed hazard ratio is 0.85 (a 15% relative reduction), the CI would range from approximately 0.70 to 1.05, which of course means the result would not be statistically significant. We can easily project the width of the CI with 700 (or any other number of deaths) because the standard error of the regression coefficient for the treatment variable in a Cox model analysis (the approach planned for STICH) decreases in proportion to the square root of the number of events. The hazard ratio is obtained through exponentiation of the estimated regression coefficient. With 700 deaths, the width of the CI will narrow to approximately 0.25, and the 15% reduction in the example above would become statistically significant. An important benefit of the additional events is the added power and precision for assessments of treatment effects in subgroups of patients that are of great interest, including subgroups defined by age, sex, race, HF class, and LV function (both global and regional function). With the additional events that will accrue with the extended follow-up, we can project the degree of improvement in the precision of the estimated treatment effect in subgroups, which would be considerable. For example, with 300 additional events, the statistical power for detecting a reduction in mortality of 25% (or even slightly less) in any subgroup consisting of one-third or more of the randomized patient cohort would exceed 80%, depending on the level of risk represented in the subgroup. Furthermore, the DCRI/Confidential Protocol 5 07/29/2010 Page 4

101 Page Section Version 4 Version 5 Rationale power for detecting treatment by subgroup interactions will be considerably enhanced (ie, for assessing whether there is a statistically different effect of CABG compared to MED depending on patient age, sex, HF class, or other subgroups defined by the variables listed above). In summary, the additional follow-up and the increased number of events will provide a robust assessment of the longer-term effects of CABG vs. MED in this important patient population. throughout document Corrected grammatical, abbreviation, and formatting errors. To ensure the correctness of these elements of the document. DCRI/Confidential Protocol 5 07/29/2010 Page 5

102 Table of Contents Table of Changes to Version 4 2 List of Figures 8 List of Tables 8 List of Abbreviations 9 1. Overview of STICH Trial Study Objectives Primary Objectives Secondary Objectives Other Key Parameters Background Public Health Burden of Heart Failure Development of Surgical Ventricular Reconstruction Operation Need for a Randomized Trial of Medical and Surgical Therapies Evaluation of Noninvasive Studies Cost and Quality of Life Evaluation Neurohormonal, Cytokine, and Genetic Studies Significance Investigational Plan Overview Study Population Definition of Patient Strata Study Hypotheses Inclusion and Exclusion Criteria Study Subject Enrollment Screening Randomization Baseline Evaluation of Randomized Patients Prior to Assigned Treatment Initiation and Completion of Initial Treatment Treatment Groups Interruption or Discontinuation of Treatment Study Subject Follow-up Efficacy Assessments Primary Efficacy Parameters Secondary Efficacy Parameters Other Key Efficacy Parameters 25 DCRI/Confidential Protocol 5 07/29/2010 Page 6

103 4.6.4 Safety Assessments Quality of Life Assessments Economic Data/Hospital Billing Data (U.S. sites only) Echocardiographic Assessment Nuclear Cardiology Assessment CMR Assessment Neurohormonal, Cytokine, and Genetic Assessments Study Organization Monitoring Committees Operational Committees Data Management Data Collection Data Forms Manual of Operations STICH Trial Website Database Management and Quality Control Statistical Plan Sample Size Determination - Overview Sample Size and Power Considerations Treatment Crossovers Statistical Analysis Background and Demographic Characteristics Concomitant Therapy Efficacy Evaluation Safety Evaluation Interim Analysis Ethics and Good Clinical Practice Institutional Review Board/Independent Ethics Committee Informed Consent References 45 DCRI/Confidential Protocol 5 07/29/2010 Page 7

104 List of Figures Figure 1. Patient Stratum and Treatment Assignment at Time of Completion of 18 Hypothesis 2 Enrollment on January 24, 2006 List of Tables Table 1. Eligibility Criteria 17 Table 2. Baseline Evaluation Studies 21 Table 3.1 STICH Follow-up Evaluation Studies 23 Table 3.2 STICHES Follow-up Evaluation Studies 23 Table 4. Hypothesis 1 Statistical Power in Relation to Number of Patients and Length 32 of Follow-up Table 5. Hypothesis 1 Enrollment, Follow-up, and Statistical Power 32 DCRI/Confidential Protocol 5 07/29/2010 Page 8

105 List of Abbreviations ACE AE CABG CAD CC CCS CI CMR CRF DCRI DDCD DSMB ECHO EF EQOL ESVI HF IEC IRB IVRS LV LVAD LVEF MED MI NCG NHLBI NYHA PCI QOL RN SAE SCD-HeFT SPECT STICH STICHES SVR Angiotensin converting enzyme Adverse event Coronary artery bypass graft Coronary artery disease Coordinating Center Canadian Cardiovascular Society Confidence interval Cardiac magnetic resonance Case report form Duke Clinical Research Institute Duke Databank for Cardiovascular Diseases Data and Safety Monitoring Board Echocardiography Ejection fraction Economics and quality of life End systolic volume index Heart failure Independent Ethics Committee Institutional Review Board Interactive voice-response system Left ventricular Left ventricular assist device Left ventricular ejection fraction Medical strategy of heart failure therapy Myocardial infarction Neurohormonal/cytokine/genetic National Heart Lung Blood Institute New York Heart Association Percutaneous coronary intervention Quality of life Radionuclide Serious adverse event Sudden Cardiac Death in Heart Failure Trial Single photon emission computed tomography Surgical Treatment for Ischemic Heart Failure Surgical Treatment for Ischemic Heart Failure Extended Study Surgical ventricular reconstruction DCRI/Confidential Protocol 5 07/29/2010 Page 9

106 1. Overview of the STICH Trial The Surgical Treatment for Ischemic Heart Failure (STICH) multicenter international randomized trial addresses 2 specific primary hypotheses in patients with left ventricular (LV) dysfunction who have coronary artery disease (CAD) amenable to surgical revascularization: 1) Coronary artery bypass grafting (CABG) with intensive medical therapy (MED) improves long-term survival compared with MED alone; 2) In patients with anterior LV dysfunction, surgical ventricular reconstruction (SVR) to a more normal LV size improves survival free of subsequent hospitalization for cardiac cause in comparison with CABG alone. Important secondary endpoints include morbidity, economics, and quality of life (QOL). Core laboratories for cardiac magnetic resonance (CMR), echocardiography (ECHO), neurohormonal/cytokine/genetic (NCG), and radionuclide (RN) studies will insure consistent testing practices and standardization of data necessary to identify eligible patients and to address specific questions related to the primary hypotheses. Over 4.5 years, 125 clinical sites will recruit approximately 2,000 consenting patients with LV ejection fraction (LVEF) less than or equal to.35 and CAD amenable to CABG. Patients with ongoing Canadian Cardiovascular Society (CCS) III angina intensity or presence of left main coronary stenosis will be classified as ineligible for MED. Patients eligible for SVR but ineligible for randomization to MED will be evenly randomized to CABG with or without SVR. Patients eligible for MED or surgical therapy, but not SVR eligible, will be evenly randomized between MED only and MED with CABG. The remaining patients eligible for both MED and SVR will be randomized between 3 treatments of MED only, or MED plus CABG, or MED plus CABG plus SVR. Registries of clinical information will be maintained on a subset of eligible patients who decline trial entry. At 4- to 6-month intervals for a minimum of 2 years, all randomized patients will be followed by a clinic visit or by telephone. Appropriate subgroups of randomized patients will have core laboratory studies repeated at specified follow-up intervals. In the patients randomized to MED only or to CABG with MED, CABG is hypothesized to demonstrate a greater than or equal to 25% reduction in the primary endpoint of all-cause death from the projected 25% 3-year mortality for MED. In the SVR-eligible patients, CABG plus SVR is hypothesized to show a 20% advantage in the endpoint of survival free of hospitalization for cardiac cause projected to be 50% at 3 years in patients receiving CABG without SVR. The trial is designed to test both hypotheses with 90% statistical power. Definition of efficacy of potential therapies and their mechanisms of benefit by the STICH trial is certain to inform future choice of therapy and thereby extend and improve the QOL of millions of patients who now suffer from ischemic heart failure (HF). 2. Study Objectives 2.1 Primary Objectives 1. To test the hypothesis that improvement in myocardial perfusion by CABG combined with MED improves long-term survival compared with MED alone in patients with LV dysfunction and CAD amenable to CABG. 2. To test the hypothesis that in patients with anterior LV akinesia, LV shape and size optimization by SVR combined with CABG and MED improves long-term survival free of cardiac hospitalization compared with CABG and MED without SVR. DCRI/Confidential Protocol 5 07/29/2010 Page 10

107 2.2 Secondary Objectives 1. To assess the clinical utility of the measurements by CMR of LV shape, size, and function for predicting the benefit of a specific treatment strategy. 2. To assess the clinical utility of measurements by nuclear cardiology and/or ECHO testing of myocardial ischemia, viability, and LV size and function for predicting preferential benefit associated with a specific treatment strategy. 3. To determine if the use of baseline ECHO-derived assessments of LV systolic and diastolic function, hemodynamics, and valvular regurgitation identify subgroups that derive either substantial benefit or harm from surgical intervention and how CABG compared with MED and CABG plus SVR compared with CABG differentially affects these parameters over time. 4. To test the hypothesis that levels of neurohormonal and proinflammatory cytokine activation will be useful in identifying that group of patients who will most likely benefit from CABG and/or CABG plus SVR. 5. To compare health-related QOL, total medical costs, and incremental cost effectiveness between CABG and MED and between CABG with or without SVR. 2.3 Other Key Parameters Mortality: 30-day all-cause; cardiovascular; all-cause death or LV assist device (LVAD) or heart transplantation. Morbidity: all-cause hospitalization; days of all-cause hospitalization; hospitalization for HF; days of hospitalization for HF; emergency department visits for HF. Cardiac procedures: percutaneous coronary intervention (PCI); any cardiac surgery; LVAD; heart transplantation; implantable cardioverter defibrillator (ICD). Cost: total medical costs; cost-effectiveness. Functional capacity: QOL; 6-minute walk distance (absolute and change); treadmill duration (absolute and change). Physiologic: LVEF (absolute and change); end systolic volume index (ESVI) (absolute and change); neurohormonal levels (absolute and change); cytokine levels (absolute and change). 3. Background 3.1 Public Health Burden of Heart Failure American Heart Association statistics suggest that HF affects 4.7 million patients in the United States and is responsible for approximately 1 million hospitalizations and 300,000 deaths annually. 1 The total annual costs associated with this disorder have been estimated to exceed $22 billion. 2 Coronary artery disease is the cause of HF in the majority of patients, and HF is the only mode of CAD presentation associated with increasing incidence and mortality. 3 Ischemic HF may develop after a known myocardial infarction (MI) or through a process of chronic ischemia that leads to cardiac dilatation, cellular hypertrophy, apoptosis, and extracellular matrix remodeling. 4 The increasing prevalence of HF largely derives from enhanced survival from improved therapies for other manifestations of CAD, especially acute MI. However, survival is usually accompanied by some degree of myocardial injury that is the basis for development of HF. DCRI/Confidential Protocol 5 07/29/2010 Page 11

108 3.2 Development of Surgical Ventricular Reconstruction Operation Reimer and Jennings showed that experimental coronary occlusion results in a wavefront of cell death moving from the subendocardial zone across the wall to involve progressively more of the transmural thickness of the ischemic zone. 5 Early reperfusion salvages subepicardial but not subendocardial myocardium. The decline in prevalence of thin-walled LV aneurysm that followed introduction of aggressive initial management of acute coronary syndromes suggests that the ischemic process also can be interrupted in patients before it reaches the transmural stage. With later healing of infarcted myocardium, scarring is maximal in the subendocardial region and is interlaced with normal myocardium in diminishing amounts toward the epicardial surface. Ventricular wall or chamber imaging commonly shows an akinetic zone that gradually blends with myocardium with increasingly normal function in contrast to the dyskinetic region with a discrete neck typical of an LV aneurysm. Left ventricular aneurysmectomy appears to reverse HF by lowering LV wall stress, but these linear amputations of dyskinetic scar commonly deform the LV cavity into a box-like shape. Intracavitary reconstruction techniques have been developed for repairing defects left by aneurysm resection that reduced LV cavity size. 6 This surgical ventricular reconstruction (SVR) strategy has more recently been applied not only to patients with dyskinetic scar but also to those with only akinetic myocardial segments. 7 At the time of cardiac operation, the epicardium of these akinetic zones may appear normal, and palpable thinning is often minimal in the arrested, decompressed heart. This appearance derives from preservation of a rim of normal myocardium covering the myocardial fibrosis and contrasts with the leather-like appearance and thinness typical of a LV aneurysm. Unlike the LV aneurysmectomy that removed myocardial scar or the Batista operation that reduced LV size by indiscriminate removal of portions of the LV wall, the SVR operation mechanically decreases the circumference of the zone of endocardial scar through an incision in normal epicardium. The surgical repair uses the intrinsic scar or an extrinsic patch to absorb excess linear wall tension from the adjacent myocardium. Decrease in wall stress appears to reduce the tendency for continued gradual expansion of the akinetic zone. The endocardial repair acutely decreases LV size, thereby acutely enhancing function in myocardial regions remote from the repair. 8 A recent registry report of 439 patients in 11 centers found a 5.1% operative mortality and 88% 18-month survival for patients who underwent CABG and SVR, suggesting that this surgical strategy is safe and at least equivalent to CABG alone. 9 This operation is now sufficiently mature to justify this proposed randomized trial to evaluate whether appropriately performed SVR truly adds value to CABG in HF patients with regional dysfunction Need for a Randomized Trial of Medical and Surgical Therapies Along the broad spectrum of severity of ischemic HF, specific clinical information, such as severe angina or left main coronary artery stenosis, may clearly indicate the need for surgical therapy for some patients. However, a large number of patients fall into a gray zone without clear evidence for benefit from either medical or surgical therapy. For these patients, evidence supporting choice between therapies was never strong and has only been confused by recent studies showing improved outcomes with both therapies. Patients for whom equipoise of anticipated benefit now exists between modern medical and surgical therapy represent the broad population who are appropriate candidates for a randomized trial to provide the context for assessing the value of 3 therapeutic strategies: 1) MED alone; 2) MED plus CABG; and 3) MED plus CABG plus SVR. DCRI/Confidential Protocol 5 07/29/2010 Page 12

109 No randomized trial has ever directly compared long-term benefits of surgical and medical treatment of patients with ischemic HF. In the 1969 to 1972 developmental era of CABG, high surgical mortality rates were observed in patients with HF. Therefore, initial randomized trials comparing CABG with medical treatment conducted from 1972 to 1978 excluded most of these patients from participation. In a subgroup of 160 CASS Trial patients with LVEF less than.50, 10-year survival was 61% in the 82 medically treated patients and 79% in the 78 patients who underwent CABG (p =.01). 11 This survival advantage of CABG was not related to the presence or severity of HF or angina symptoms. A meta-analysis by Yusuf 12 combined individual patient data from the CASS Trial with those enrolled in the 6 other early randomized trials. Only 191 (7.2%) of the 2,649 total patients had an ejection fraction (EF) less than.40, and only 106 (4.0%) of these patients who were primarily symptomatic with angina also had HF symptoms. Coronary artery bypass graft surgery improved survival among all patients with proximal left anterior descending stenosis, 3-vessel, or left main coronary artery disease regardless of LV function. In these patients with a survival benefit from CABG, a low EF increased the absolute benefit but did not change the relative benefit of CABG. A literature search of 326 published reports on results of CABG in patients with HF or LV dysfunction identified 3 welldesigned cohort studies. 13 Mortality benefit of CABG over MED was 10 and 20 lives per 100 patients at 3 years in 2 of the 3 studies and 29 lives per 100 patients at 5 years in the third study. Despite this apparent benefit, data from the Duke Databank for Cardiovascular Diseases (DDCD) show that more than 71% of patients with characteristics that would qualify for inclusion in the STICH trial receive MED, 22% receive CABG, and 7% receive PCI. Moreover, this distribution of treatments has remained constant over the past decade at an institution with an aggressive use of myocardial revascularization procedures. Perhaps CABG is not more widely used in patients with coronary angiograms that suggest survival benefit because of concern for postoperative complications, such as stroke, that detract from its survival advantage. In addition, the general medical community often overestimates surgical risks and medically manages the high-risk patients without thoroughly characterizing the presence and extent of CAD. The therapy called medical treatment in all randomized trials and most observational comparisons with CABG really only reflected the natural history of CAD since it rarely included drugs now known to be life prolonging, such as angiotensin-converting enzyme (ACE) inhibitors, beta blockers, lipid lowering, and antiplatelet agents. Refinement of operative and postoperative surgical management also has steadily improved CABG results over the past 2 decades. The paucity of modern data available to clinicians who must daily make high-risk management decisions in patients with HF emphasizes the need for a properly designed randomized comparison of these therapies commonly used in clinical practice. 3.4 Evaluation of Noninvasive Studies Increasingly over the past 3 decades, information describing cardiac structure, perfusion, hemodynamics, and metabolism obtained from noninvasive cardiac imaging studies has been used to guide management decisions for patients with HF. Although this anatomic and physiologic information often adds value to clinical care, an accepted strategy has not evolved that tailors the testing sequence to specific presenting features of individual patients to efficiently identify the treatment strategy most likely to improve outcomes. Consensus on proper use of cardiac imaging studies has been hindered by absence of clinical studies that objectively evaluate the independent treatment-related prognostic information of these tests obtained using standardized methods on patients for whom the test has minimal influence on choice of therapy. In the absence of high-quality information, use of these tests by clinicians is commonly guided by mechanistic studies that use endpoints considered as surrogates for survival performed with a study design that does not control for test bias in treatment selection. For example, the widespread clinical use of myocardial viability studies to identify HF patients who would or would not benefit from myocardial revascularization is largely based on more than 100 reports using improvement in EF after treatment as the DCRI/Confidential Protocol 5 07/29/2010 Page 13

110 study endpoint. No randomized and only 10 observational studies are published describing results in 762 patients that compare the power of myocardial viability measurements for predicting patient survival after CABG or MED In these studies CABG was associated with better survival than MED in patients with viable myocardium but not in patients without viable myocardium. However, definition of the criteria used to define subgroups was derived on the same population used to assess outcome in these studies. Moreover, subgrouping these patients by treatment received insured that important baseline differences that influenced choice of treatment also influenced the observed outcome. The lack of survival benefit demonstrated for CABG in patients categorized as without myocardial viability derived from a low mortality in medically treated patients also categorized as without myocardial viability, and survival after CABG was not positively or negatively influenced by the preoperative amount of myocardial viability measured. The common clinical practice of not offering CABG to patients with LV dysfunction in regions found to be nonviable on noninvasive studies is not justified by published data. These criticisms of the quality of available information on the appropriate use of myocardial viability testing are equally true of all other noninvasive cardiac imaging data used to guide selection of therapy in patients with CAD. The optimal strategy for evaluation of these noninvasive tests would be to use multivariable modeling of all parameters of independently predictive test information to describe the total test value by a single continuous index. In this form, a myocardial viability index could readily be contextualized with all other relevant clinical and other test information to quantitate any survival advantage the myocardial viability index added by properly guiding the choice of the most effective HF therapy. The STICH trial uses core laboratories for CMR, ECHO, and RN studies to insure standardization of testing practices and of data analysis for operational use of these studies in the STICH trial. These 3 core laboratories also will address specific questions that use anatomic or physiologic information from these noninvasive cardiac studies to elucidate mechanisms responsible for differences observed in clinical endpoints among the randomized treatment groups. This structured acquisition of cardiac imaging studies in a population with treatment assigned by randomization represents a unique opportunity to assess the individual value of these tests for guiding treatment selection. 3.5 Cost and Quality of Life Evaluation Any new therapeutic strategy that improves the health of HF patients also is likely to reduce a portion of the $22 billion costs of this condition by reducing hospitalizations and related expensive care. However, broader use of surgical therapies would increase initial cost, and the resulting economic change represents a complex issue measured both in terms of cost per added quality-adjusted life year and in total medical care cost to society. This study proposes careful measurement of resource use patterns and associated medical care costs to compare the randomized treatment strategies addressed in both primary study hypotheses. Additional comparisons will examine major subgroup effects. If the primary hypotheses are established for the 2 surgical strategies compared in this trial, cost effectiveness analyses will define whether these therapies are economically attractive relative to standard benchmarks. Heart failure adversely impacts patient functional status and health-related QOL. Modern medical therapy for HF modestly improves QOL, but a more aggressive approach with the surgical therapies being studied in this trial may produce even larger improvements. Careful serial measurements of QOL using well-validated instruments are crucial in contextualizing the impact of any invasive therapy and represent important secondary endpoints for the STICH trial. DCRI/Confidential Protocol 5 07/29/2010 Page 14

111 3.6 Neurohormonal, Cytokine, and Genetic Studies The progressive cardiac dilatation and diminished LV function that occur after myocardial damage are associated with overexpression of a variety of endogenous peptides that include: 1) neurohormonal mediators (eg, norepinephrine and endothelin), proinflammatory cytokines (eg, tumor necrosis factor and interleukin-6) and natriuretic peptides (eg, ANP and BNP). 4 These endogenous peptides appear to hasten progressive worsening of HF by cardiotoxic actions that may induce apoptosis and maladaptive remodeling of both cardiac myocytes and the extracellular matrix. Therapeutic strategies that attenuate production (ACE inhibition) or inhibit biologic activity (beta blockade) of a neurohormonal peptide have proven beneficial in the management of patients with HF. However, the ability of surgical interventions to attenuate neurohormonal and cytokine activation more than is achieved with MED alone has not been evaluated. Since myocardial stretch and ischemia may enhance induction and activation of these neurohormonal and cytokine pathways, the STICH trial will test the hypothesis that rapid reduction of wall tension through the combination of CABG and SVR will be the most effective treatment for reducing peptide expression and that CABG with MED is associated with better clinical outcomes than MED alone. Conversely, high baseline levels of neurohormonal/cytokine peptides may identify a group of patients with irreversible extracellular matrix remodeling or significant apoptosis for whom CABG or CABG with SVR proves less effective than MED alone. Thus, measures of endogenous peptides may help to identify patient groups who will respond in a specific manner to all 3 randomized treatment strategies evaluated in the STICH trial. Four recent observations suggest that outcome in patients with HF may be related to variations in genetic background: 1) dilated cardiomyopathy is often familial; 2) first-degree relatives of patients with HF have a high rate of abnormal ECHOs 24 ; 3) specific loci are associated with adult-onset autosomal dominant dilated cardiomyopathy 25 ; and 4) mutations affecting cytoskeletal proteins are associated with the development of HF. 26 Genetic determinants that might affect the progression of disease in individual patients appear to consist of either polymorphisms or allelic mutations that alter expression of proteins important in the pathophysiology of HF. Many of these genetic variations are directly related to alterations in the activity of the neurohormonal and/or proinflammatory pathways. The STICH trial will provide an opportunity to evaluate the role of genotype determination for predicting the efficacy of subsequent therapy. Most of the polymorphisms that will be evaluated affect the constitutive and activated expression of neurohormonal/cytokine peptides that may be favorably altered by mechanical intervention directed at normalizing LV mechanics and myocardial perfusion. 3.7 Significance Should We Do CABG in Patients with Heart Failure? The design of the STICH trial excludes only patients for whom MED is the only reasonable therapeutic alternative, those with coronary anatomy best suited for revascularization by PCI, those severely symptomatic patients not eligible for SVR, and those who are heart transplant candidates. Therefore, results of the STICH trial will be generalizable to almost all of the large number of patients with CAD and systolic LV dysfunction in this country. If surgical therapies prove not to significantly improve survival, MED without intensive evaluation for CAD may be the preferred strategy of care to be pursued in the more than 550,000 patients who first present with HF of unexplained etiology each year in the United States. 4 A more aggressive search for the 70% of patients with HF and LV dysfunction who have CAD as the predominant etiology could await demonstrated failure of the MED strategy indicated by development of overt ischemic symptoms. Conversely, if surgical therapy is shown to have a survival advantage over medical therapy in the STICH trial, early aggressive evaluation of CAD as a potentially correctable etiology of new-onset HF would be the preferred strategy. Clinical testing then would become more widely employed to identify patients likely to benefit from CABG and those eligible for SVR. DCRI/Confidential Protocol 5 07/29/2010 Page 15

112 When Do We Need Anatomic and Physiologic Studies to Guide the Choice of Therapy in HF Patients with CAD? This trial will relate continuous quantitative relationships between physiologic measurements of the magnitude of myocardial ischemia and viability and LV size to the outcome associated with the 3 treatment strategies. The resulting information is likely to decrease the cost and complexity of evaluation of patients with ischemic HF by eliminating use of tests that do not provide independent information. This trial will establish whether measurements of neurohormonal and cytokine levels and genetic profiling are useful for directing patient-management decisions and for monitoring the effectiveness of therapy. Information obtained from physiologic studies in the STICH trial will permit refinement of the optimal approach to efficiently evaluate and select the treatment strategy most likely to be effective for the many patients with CAD, HF, and LV dysfunction in the United States. Does SVR Need to be Added to CABG in Patients with Regional Dysfunction and What Magnitude of Dysfunction Must be Present to Justify This Additional Surgery? This trial is designed to answer the question of whether mechanically decreasing LV size by SVR arrests or reverses LV enlargement and thereby enhances survival without hospitalization more than relief of chronic ischemia by CABG without SVR. The SVR operation is gaining popularity because promising observational reports confirm its safety and suggest its efficacy. Randomized comparison of CABG with SVR versus CABG alone will definitively test the superiority of adding SVR to CABG. In the event the 2 operations result in equivalent survival free of cardiac hospitalization, CABG will remain the dominant operation. However, if the addition of SVR to CABG in patients with akinesia of the anterior wall offers superior clinical outcomes, an important new therapeutic option will be added to the HF armamentarium. Moreover, information from CMR measurements of LV volume made before and after treatment in SVR-eligible patients is likely to guide optimal use of SVR and also provide a noninvasive method to assess the quality of operative result. 4. Investigational Plan 4.1. Overview The STICH Trial is a 3-arm randomized clinical trial that will enroll approximately 2,000 patients with LVEF less than or equal to.35 and CAD amenable to CABG. The 3 treatments include intensive, evidencebased MED, MED plus revascularization with CABG, and MED plus CABG combined with SVR. After eligible patients give informed consent, site personnel will use a 24-hour toll free telephone number with an interactive voice-response system (IVRS) to initiate the randomization process. In addition to the baseline case report form (CRF), a QOL questionnaire will be administered by the coordinator at each site. Baseline studies will be completed and the randomized therapy will be initiated within 2 weeks. Randomized patients will be followed at clinic visits at 4-month intervals for the first year and a minimum of 6-month intervals for the duration of the trial. Follow-up echoes, treadmills, gated single photon emission computed tomography (SPECT), CMRs, and neurohormonal/cytokine peptide collections will be performed at scheduled intervals throughout the remainder of the trial. Personnel from the Economics and Quality of Life (EQOL) Core Laboratory at the Duke Clinical Research Institute (DCRI) will perform QOL interviews by telephone at 4 months and 1, 2, and 3 years for clinical sites in the United States. In addition, EQOL Core Laboratory staff will collect hospital bills from all United States clinical sites for all hospitalizations and major outpatient tests and therapies for the length of the study. Follow-up of Hypothesis 2 patients will be a minimum of 2 years, an average of 3.5 years and a maximum of 6 years. Hypothesis 1 patients will be followed until 400 deaths occur. Surviving patients from Hypothesis 1 will be followed for an additional DCRI/Confidential Protocol 5 07/29/2010 Page 16

113 5 years, thereby acquiring critically important longer-term (10-year average) information on patients with HF and LV systolic dysfunction treated with MED with and without CABG. 4.2 Study Population The patient population will consist of patients who have an LVEF of less than or equal to.35 and CAD suitable for revascularization. Patients included in this study will be segregated into 3 strata depending on their MED and SVR eligibility Definition of Patient Strata The design will accommodate the fact that not all patients enrolled will be eligible for all 3 therapeutic approaches. TABLE 1. Eligibility Criteria Stratum A Stratum B Stratum C MED eligible MED eligible MED ineligible SVR ineligible SVR eligible SVR eligible MED=medical strategy of heart failure therapy; SVR=surgical ventricular reconstruction Patients who meet trial eligibility criteria will first be stratified as eligible or ineligible for MED on the basis of CCS greater than or equal to III angina or greater than or equal to 50% left main coronary artery stenosis. Patients will then be further stratified based on eligibility for the SVR procedure. Patients eligible for either MED or CABG but not eligible for the SVR procedure (Stratum A) will be randomized in equal proportions to MED or CABG. Patients eligible for all 3 therapies (Stratum B) will be randomized in equal proportions to MED, CABG, or CABG plus SVR. Patients whose angina or CAD severity make them ineligible for MED but who are appropriate candidates for both CABG and SVR (Stratum C) will be randomized in equal proportions to CABG or CABG plus SVR. The allocation of patients to treatment within each stratum at the conclusion of Hypothesis 2 enrollment on January 24, 2006, is illustrated in Figure 1. At this time, 635 total Hypothesis 1 patients had been enrolled with a treatment assignment of MED in 310 patients (235 Group 1 and 75 Group 3) and of CABG in 325 patients (249 Group 2 and 76 Group 4). The 365 or more patients remaining to be recruited after January 24, 2006, will all be entered into Stratum A Study Hypotheses There are 2 primary study hypotheses. Coronary revascularization hypothesis: Improvement in myocardial perfusion by CABG combined with MED improves long-term survival compared with MED alone. LV reconstruction hypothesis: In patients with dominant anterior LV akinesia or dyskinesia, LV shape and size optimization by SVR combined with CABG and MED improves long-term survival free of cardiac hospitalization compared with CABG and MED without SVR. DCRI/Confidential Protocol 5 07/29/2010 Page 17

114 FIGURE 1. Patient Stratum and Treatment Assignment at Time of Completion of STICH Enrollment on May 05, 2007 CAD EF 0.35 SVR Eligible? Yes Medical Eligibility No SVR Eligible? No Not in Trial No R 1061 Pts Yes R 216 Pts Yes R 859 Pts 1 MED 527 Pts 2 CABG + MED 534 Pts 3 MED 75 Pts 4 CABG + MED 76 Pts 5 CABG + SVR + MED 65 Pts 6 CABG + MED 423 Pts 7 CABG + SVR + MED 436 Pts Stratum A Stratum B Stratum C DCRI/Confidential STICH, Amendment /29/2010 Page 14 DCRI/Confidential Protocol 5 07/29/2010 Page 18