Drug-eluting stents versus bare metal stents for angina or acute coronary syndromes (Review)

Transcription

1 Drug-eluting stents versus bare metal stents for angina or acute coronary syndromes (Review) Greenhalgh J, Hockenhull J, Rao N, Dundar Y, Dickson RC, Bagust A This is a reprint of a Cochrane review, prepared and maintained by The Cochrane Collaboration and published in The Cochrane Library 2011, Issue 1

2 T A B L E O F C O N T E N T S HEADER ABSTRACT PLAIN LANGUAGE SUMMARY BACKGROUND OBJECTIVES METHODS RESULTS Figure Figure DISCUSSION AUTHORS CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES CHARACTERISTICS OF STUDIES DATA AND ANALYSES Analysis 1.1. Comparison 1 All stents: event rate, Outcome 1 Event rate: 6 months Analysis 1.2. Comparison 1 All stents: event rate, Outcome 2 Event rate: 12 months Analysis 1.3. Comparison 1 All stents: event rate, Outcome 3 Event rate: 2 years Analysis 1.4. Comparison 1 All stents: event rate, Outcome 4 Event rate: 3 years Analysis 1.5. Comparison 1 All stents: event rate, Outcome 5 Event rate: 4 years Analysis 1.6. Comparison 1 All stents: event rate, Outcome 6 Event rate: 5 years Analysis 2.1. Comparison 2 All stents: mortality, Outcome 1 Mortality: 6 months Analysis 2.2. Comparison 2 All stents: mortality, Outcome 2 Mortality: 12 months Analysis 2.3. Comparison 2 All stents: mortality, Outcome 3 Mortality: 2 Years Analysis 2.4. Comparison 2 All stents: mortality, Outcome 4 Mortality: 3 years Analysis 2.5. Comparison 2 All stents: mortality, Outcome 5 Mortality: 4 years Analysis 2.6. Comparison 2 All stents: mortality, Outcome 6 Mortality: 5 years Analysis 3.1. Comparison 3 All stents: myocardial infarction, Outcome 1 MI any: 6 months Analysis 3.2. Comparison 3 All stents: myocardial infarction, Outcome 2 MI any: 12 months Analysis 3.3. Comparison 3 All stents: myocardial infarction, Outcome 3 MI any: 2 years Analysis 3.4. Comparison 3 All stents: myocardial infarction, Outcome 4 MI any: 3 years Analysis 3.5. Comparison 3 All stents: myocardial infarction, Outcome 5 MI any: 4 years Analysis 3.6. Comparison 3 All stents: myocardial infarction, Outcome 6 MI any: 5 years Analysis 4.1. Comparison 4 All stents: TLR, Outcome 1 TLR: 6 months Analysis 4.2. Comparison 4 All stents: TLR, Outcome 2 TLR: 12 months Analysis 4.3. Comparison 4 All stents: TLR, Outcome 3 TLR: 2 years Analysis 4.4. Comparison 4 All stents: TLR, Outcome 4 TLR: 3 years Analysis 4.5. Comparison 4 All stents: TLR, Outcome 5 TLR: 4 years Analysis 4.6. Comparison 4 All stents: TLR, Outcome 6 TLR: 5 years Analysis 5.1. Comparison 5 All stents: TVR, Outcome 1 TVR: 6 months Analysis 5.2. Comparison 5 All stents: TVR, Outcome 2 TVR: 12 months Analysis 5.3. Comparison 5 All stents: TVR, Outcome 3 TVR: 2 years Analysis 5.4. Comparison 5 All stents: TVR, Outcome 4 TVR: 3 years Analysis 5.5. Comparison 5 All stents: TVR, Outcome 5 TVR: 4 years Analysis 5.6. Comparison 5 All stents: TVR, Outcome 6 TVR: 5 years Analysis 6.1. Comparison 6 All stents: thrombosis, Outcome 1 Thrombosis: 6 months Analysis 6.2. Comparison 6 All stents: thrombosis, Outcome 2 Thrombosis: 12 months Analysis 6.3. Comparison 6 All stents: thrombosis, Outcome 3 Thrombosis: 2 years Analysis 6.4. Comparison 6 All stents: thrombosis, Outcome 4 Thrombosis: 3 years Analysis 6.5. Comparison 6 All stents: thrombosis, Outcome 5 Thrombosis: 4 years Analysis 6.6. Comparison 6 All stents: thrombosis, Outcome 6 Thrombosis: 5 years i

3 Analysis 7.1. Comparison 7 Available stents only: event rate, Outcome 1 Event rate: 6 months Analysis 7.2. Comparison 7 Available stents only: event rate, Outcome 2 Event rate: 12 months Analysis 7.3. Comparison 7 Available stents only: event rate, Outcome 3 Event rate: 2 years Analysis 7.4. Comparison 7 Available stents only: event rate, Outcome 4 Event rate: 3 years Analysis 7.5. Comparison 7 Available stents only: event rate, Outcome 5 Event rate: 4 years Analysis 7.6. Comparison 7 Available stents only: event rate, Outcome 6 Event rate: 5 years ADDITIONAL TABLES APPENDICES WHAT S NEW HISTORY CONTRIBUTIONS OF AUTHORS DECLARATIONS OF INTEREST SOURCES OF SUPPORT DIFFERENCES BETWEEN PROTOCOL AND REVIEW NOTES INDEX TERMS ii

4 [Intervention Review] Drug-eluting stents versus bare metal stents for angina or acute coronary syndromes Janette Greenhalgh 1, Juliet Hockenhull 1, Naveen Rao 2, Yenal Dundar 1, Rumona C Dickson 1, Adrian Bagust 1 1 Liverpool Reviews and Implementation Group, University of Liverpool, Liverpool, UK. 2 Abacus International, Bicester, UK Contact address: Janette Greenhalgh, Liverpool Reviews and Implementation Group, University of Liverpool, Sherrington Building, Ashton Street, Liverpool, L69 3GE, UK. janette.greenhalgh@liverpool.ac.uk. Editorial group: Cochrane Heart Group. Publication status and date: Edited (no change to conclusions), published in Issue 1, Review content assessed as up-to-date: 30 September Citation: Greenhalgh J, Hockenhull J, Rao N, Dundar Y, Dickson RC, Bagust A. Drug-eluting stents versus bare metal stents for angina or acute coronary syndromes. Cochrane Database of Systematic Reviews 2010, Issue 5. Art. No.: CD DOI: / CD pub2. Background A B S T R A C T Coronary artery stents are tiny tubular devices used to scaffold vessels open during percutaneous transluminal coronary angioplasty (PTCA). Restenosis (re-narrowing) of vessels treated with stents is a problem; in order to reduce restenosis, stents that elute drugs over time are now available. However these drug-eluting stents are more expensive and there is a need to assess their clinical benefits prior to recommending their use. Objectives To examine evidence from randomised controlled trials (RCTs) to assess the impact of drug eluting stents (DES) compared to bare metal stents (BMS) in the reduction of cardiac events. Search methods The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2008, Issue 4), MEDLINE ( April 2009) and EMBASE ( January 2009) were searched. We carried out handsearching (electronic and manual) up to January Selection criteria We included RCTs comparing DES with BMS used in conjunction with PTCA techniques in the review. Participants were adults with stable angina or acute coronary syndrome (ACS). We considered published and unpublished sources and included them if they reported outcome data of interest. Data collection and analysis Three review authors independently extracted data, assessed trial quality assessment and checked decisions within the group. Data extraction included composite event rates (major adverse cardiac event, target vessel failure); death; acute myocardial infarction (AMI); target lesion revascularisation (TLR); target vessel revascularisation (TVR) and thrombosis. Data synthesis included meta-analysis of composite event rate, death, AMI and revascularisation rates, presented as odds ratios with 95% confidence intervals (CI) using a fixedeffect model. We assessed heterogeneity between trials. 1

5 Main results We included more than 14,500 patients in 47 RCTs. There were no statistically significant differences in death, AMI or thrombosis between DES and BMS. For composite events, TLR and TVR reductions were evident with use of sirolimus, paclitaxel, everolimus, dexamethasone, zotarolimus and (to a limited extent) tacrolimus-eluting stents. These effects are demonstrated in the longer term follow up. Subgroup analyses (e.g. diabetics) largely mirrored these findings. Authors conclusions Drug-eluting stents releasing sirolimus, paclitaxel, dexamethasone and zotarolimus reduce composite cardiac events. However, this reduction is due largely to reductions in repeat revascularisation rates as there is no evidence of a significant effect on rates of death, MI or thrombosis. The increased cost of drug-eluting stents and lack of evidence of their cost-effectiveness means that various health funding agencies are having to limit or regulate their use in relation to price premium. P L A I N L A N G U A G E S U M M A R Y Drug-eluting stents versus bare metal stents for angina or acute coronary syndromes Coronary artery disease generally results from the build-up of fatty material and other substances on the internal surface of the blood vessels which supply the heart, gradually reducing its supply of oxygen. There may or may not be associated pain (angina). The disease may cause sudden death and limit normal daily activity. One strategy used to control symptoms or restore blood supply is Percutaneous Transluminal Coronary Angioplasty (PTCA) in which a small elongated balloon is inflated at the site of the plaque, compacting the deposited material against the vessel wall. This review explores the effects of one category of interventional device used with PTCA: coronary artery stents. These are expandable devices resembling a tubular wire mesh used to scaffold vessels open during PTCA procedures to relieve coronary obstructions in patients. The success rates associated with these devices are high, complication rates low and most patients experience improvement in symptoms. Nonetheless, rates of restenosis (re-narrowing of the treated vessel) which may require a repeat intervention, are a significant limitation of the use of stents. An adaption of stent technology involves stents which release (elute) drugs over time in order to reduce restenosis; however these stents are expensive in comparison to their bare metal equivalent. Our review includes results from 47 studies (including more than 14,500 patients) and contains data up to five years following treatment. No statistically significant differences in death, myocardial infarction or vessel blockage were reported between drug-eluting stents (DES) and bare metal stents (BMS). Use of DES did result in decrease in the number of times patients had to be re-treated due to blockage of the blood vessel and/or stent. Thus, DES are effective in reducing rates of restenosis but are not superior to standard BMS in terms of decreasing death, myocardial infarction or thrombosis. The increased cost of DES and lack of evidence of their cost-effectiveness means that various health funding agencies are either limiting their use or attempting to regulate use in relation to their price. B A C K G R O U N D Description of the condition The current WHO 2004 report on the global burden of disease cites ischaemic heart diease (IHD; also known as coronary heart disease or CHD) as the number one cause of death (out of a total of 136) in the world, accounting for 12.2% of all deaths. This figure is predicted to rise to 14.2% by In the UK, IHD is also the leading cause of death; causing around 101,000 deaths each year; one in five men and one in six women. Moreover, ischaemic heart disease is the most common cause of premature death in the UK (BHF 2008). Coronary artery disease generally results from the build-up of fatty material and other substances on the internal surface of the blood vessels which supply the heart. This process can be gradual, reducing the flow of blood, resulting in a decrease to the supply of oxygen to the heart muscle. The disease may be silent (without symptoms) or result in symptoms such as angina (chest pain). Where the blood flow is cut off, a heart attack can result, damaging the heart and risking life. The burden of disease is significant in that it may lead to sudden 2

6 death or cause pain and limit an individual s ability to go about their normal daily activities. Description of the intervention A number of strategies have been developed to control symptoms or restore blood supply in people with narrowed vessels, including medication and surgical interventions. Less invasive methods utilise special equipment which is introduced into a patient s vessels and guided to the site of obstruction on a fine catheter. In the case of percutaneous transluminal coronary angioplasty (PTCA), a small elongated balloon is inflated at the site of the plaque, effectively compacting the deposited material against the vessel wall. Other methods which involve introducing cutting devices, radiation or lasers are also available in some medical facilities. This review will explore the effects of one category of interventional device used with PTCA: coronary artery stents. Coronary artery stents are expandable devices resembling a tubular wire mesh used to scaffold vessels open during PTCA procedures to relieve coronary obstructions in patients. The first of these were metal and are referred to as bare metal stents (BMS). Procedural success rates for deploying these devices are high, complication rates are low and most patients experience improvement in symptoms after the intervention (Gunn 2003). However, rates of restenosis (re-narrowing of the treated vessel) which may require a repeat intervention, are a significant limitation of PTCA with the use of stents. Restenosis following stent placement is a complex process and results predominantly from a proliferation of smooth muscle cells into and around implanted stents. Rates of restenosis in PTCA with stents are recorded as ranging between 20 and 50 per cent, depending on the size, location and complexity of the lesion (Meads 2000; Hill 2007). How the intervention might work In order to improve results and reduce restenosis, developments in stent design have been augmented by new drug-eluting technologies. Drug-eluting stents (DES) release anti-proliferative agents from their surface with the objective of limiting cell growth around the stent using cytotoxic, cytostatic and other agents. The three key components of a DES are the stent design, the drug eluted and the type of polymer used to coat the stent. The focus of this review is the evaluation of the drugs eluted from a stent. Stents coated with a drug may elute at different rates, depending on the presence or absence of additional polymer coatings on the stent.the most widely evaluated DES are those impregnated with either sirolimus or paclitaxel. Sirolimus (previously known as rapamycin) is an immunosuppressive agent that reduces inflammation and inhibits cellular proliferation. There are a number of sirolimus-type analogues including, but not limited to: everolimus, biolimus A9, tacrolimus and zotarolimus (ABT-578). Paclitaxel is a commonly used chemotherapeutic agent that inhibits cell division and has minimal anti-inflammatory properties. The molecular analogue of paclitaxel is QP2/7-Hexanoyltaxol, which functions in a similar way. Other drugs include dexamethasone, a synthetic adrenocortical steroid that reduces inflammation; actinomycin, which inhibits cellular proliferation; and 17 beta-estradiol, a derivative of the oestrogen hormone which reduces intimal hyperplasia and accelerates endothelial regrowth and vessel healing after injury. Criteria for use of drug-eluting stents Each DES has an instruction for use document that includes the indications for how the specific device can be used. The indications for each DES vary, although the majority specify the sizes of vessels (diameter and length) to be treated and are in accordance with their Conformité Européene (CE) marking. Also included are details of side effects and specific contraindications for DES. In general, coronary artery stenting is not advised for use in patients in whom anti-platelet and/or anticoagulant therapy is contraindicated and in patients judged to have a lesion that prevents complete inflation of an angioplasty balloon or proper placement of the stent or delivery device. In the USA the first DES received approval from the Federal Drug Administration (FDA) in 2003; this was awarded to the Cordis CYPHER sirolimus-eluting stent. Since then three additional stents have received approval: the Boston Scientific TAXUS paclitaxel-eluting stent system, the MedTronic Endeavour zotarolimus stent and the Xience V everolimus-eluting system. In Europe, DES receive CE marking and, up to and including June 2008 we are aware of a number that have received such approval. The approval process for medical devices is somewhat different to that used for drugs and therefore there is no easily accessible information source that lists all DES that have been approved. Table 1 shows the status of those DES which appear in the trials and are included in this review. Table 2 shows other CE marked DES that we are aware of but that do not appear in the trials in this review. Recent issues relating to DES In late 2006, presentations were made at a variety of international professional meetings reporting the possibility of an increase in thrombosis risk for patients who had received a DES (FDA 2006). In response to this, the FDA held a special meeting in December The initial response was the release of new guidelines related to the duration of antiplatelet therapy following DES insertion. The American College of Cardiologists/American Heart Association percutaneous coronary intervention (PCI) guidelines (King 2007) (also endorsed by the Society for Cardiovascular Angiography Interventions and the British Cardiovascular Intervention Society) have recommended that for patients receiving a DES, the 3

7 duration of antiplatelet therapy should be increased to at least 12 months. After this time continuation should be reviewed, taking into account the risk for further events on an individual patient basis. For BMS patients, antiplatelet therapy should be given for a minimum of one month and ideally up to 12 months. Similarly, the European Society of Cardiology 2005 guidelines (Silber 2005) recommend 6-12 months of antiplatelet therapy for post- DES patients and 3-4 weeks for BMS. The optimal duration of treatment with antiplatelet therapy remains an unknown and DES thrombosis may still occur despite its use. In new draft guidance on research, the FDA have also called for long-term monitoring of safety outcomes with all clinical trials reporting 12 month outcomes, with data at 24 months for the majority of patients at the time of marketing application submission. The length of trials assessing a DES should be viewed in terms of the entire follow up, which should extend through a fiveyear clinical follow up per protocol. (FDA 2008). (The FDA panel were highly critical of Johnson and Johnson for failing to provide post-market registry data - a post-approval condition attached its Cypher stent - see Lenzer 2008) A further development arose from the investigation of the thrombosis risk. Seeking to clarify the evidence, the Academic Research Consortium (ARC) found that a variety of definitions of thrombosis were employed in clinical trials. Consequently, ARC proposed a new definition as a standard which would provide consistency in the reporting of future trials. The proposed ARC definition includes patients with definite/confirmed stent thrombosis, probable stent thrombosis, and possible stent thrombosis (Cutlip 2007). Cost-effectiveness Early economic analyses (Cohen 2004; van Hout 2005) showed that specific DES, examined with data from clinical trials, appeared to be both clinically effective and cost-effective in certain economies. However, alternative views exist which analyse the cost effectiveness of DES outside clinical trials and in other healthcare systems. Drug-eluting stents may not be cost-effective in different funding contexts and in real life, natural case mixes. An analysis conducted by the Technology Assessment Unit, McGill University Health Centre, Quebec (TAU 2003) found that DES were not proven to be cost-effective and the report authors suggested that introduction of DES would impact heavily on the hospital budget. Technology Assessments produced in the UK (Hill 2004; Hill 2007) using both trial and audit data from a UK context, concluded that only around 5% of cases would warrant the use of DES. This finding was based on the then current UK price premiums for DES over BMS; prices for DES could change and thereby influence their cost-effectiveness. It is worthy of note that even in healthcare economies such as the USA, DES may not be commercially attractive as hospitals lose turnover from reduced readmissions for repeat revascularisation and reduced income from reimbursement, since each repeat procedure incurs a further reimbursement payment in insurance-based healthcare economies. Also, profit from coronary artery bypass graft surgery, which may be displaced by new stent technologies, is considerably higher than that for DES implantation procedures. Why it is important to do this review In 2006, an estimated 1,313,000 PCI procedures were performed in the US; more than 70% of these were performed with DES as opposed to BMS (AHA Statistics Committee 2009). In Europe, approximately 769,766 stenting procedures were reported for 2004, with DES being used in 26% of these (Cook 2007). The number of PCIs carried out in the UK in 2008 was 80,331; 92% involved stents, 57% of which were DES (BCIS 2008). Manufacturers, professional bodies and patients have reported considerable enthusiasm for the use of DES, but these are currently only available at a considerable increment in price compared to non-des. The additional cost may be justified by reduced requirement for subsequent re-intervention at the target site; however, the additional cost of this new technology may limit the widespread adoption of drug-eluting stents to particular patient subgroups. In England and Wales, current guidance by the National Institute for Health and Clinical Excellence (NICE) recommends that DES are only used in the treatment of coronary artery disease if the artery to be treated is less than 3 mm in diameter or the affected section of the artery is longer than 15 mm, and the additional cost of the DES over BMS is 300 or less. NICE 2008 This review examines the available evidence related to the clinical effectiveness of DES in relation to the reduction in cardiac events. It is a composite of two stages, an initial (unpublished) review submitted in 2004 and the present review, which is a substantial update. The methods for each stage are reported separately; however, the results are presented in aggregate. O B J E C T I V E S To assess the clinical effectiveness of routine stenting with drugeluting compared with non-eluting coronary artery stents in adults with stable angina or acute coronary syndrome (including acute myocardial infarction and unstable angina). M E T H O D S Criteria for considering studies for this review Types of studies 4

8 Randomised controlled trials (RCTs), published or unpublished. Types of participants Adults with stable angina or acute coronary syndrome (including acute myocardial infarction (ST segment elevation and depression, Q wave and non-q wave) and unstable angina). Types of interventions Stent (non-drug eluting) versus drug-eluting stent. Drug-eluting stents included those releasing any type of drug. Types of outcome measures Primary outcomes: death (both cardiac and non-cardiac); combined event rate e.g. major adverse cardiac events (MACE) including acute myocardial infarction (AMI), target or lesion vessel stenosis or need for revascularisation or stent thrombosis, major adverse cardiac and cerebrovascular events (MACCE), or other composites of the events: Secondary: AMI; target vessel revascularisation (TVR); target lesion revascularisation (TLR); repeat treatment (PTCA, stent or coronary artery bypass graft - CABG); thrombosis. Search methods for identification of studies The search incorporated a number of methods to identify completed or ongoing RCTs. We applied no language limitations. New England Journal of Medicine; Transcatheter Cardiovascular Therapeutics; as well as the journals of the American College of Cardiology; American Heart Association; British Cardiac Society; European Society of Cardiology. We also took out subscriptions to -based information newsletters and regular examination of web pages (including those supported by stent manufacturers) relevant to the review topic; carried out searching of bibliographies of identified sources; and made use of publicly available submissions to NICE, London, UK. We also searched internet-based resources including: Stent manufacturer webpages. Data collection and analysis Electronic searches The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2008, Issue 4), MEDLINE (1990 to April 2009) and EMBASE (1980 to January 2009) were searched (see Appendix 1 for details of search strategies). We also undertook electronic handsearching of journals and conferences in relevant fields (January 2008); these included: American Heart Journal; American Journal of Cardiology; BMJ; Cardiovascular Revascularization Therapy; Catheterization and Cardiovascular Interventions; Circulation; European Heart Journal; Heart; International Journal of Cardiology; Journal of the American College of Cardiology; JAMA; Journal of Thoracic and Cardiovascular Surgery; Lancet; Study selection and quality assessment Two review authors independently applied the inclusion criteria. In the first instance reviewers independently scanned the titles and abstracts of identified references. We obtained full details of selected trials and two review authors independently applied inclusion criteria. Two authors independently assessed methodological quality of included trials by two reviewers, in duplicate, using a scheme based on Schultz 1995, methods proposed by the Heart Collaborative Review Group (Heart CRG) and grading similar to that used in Villanueva 2003: Adequacy of the randomisation process A - Adequate sequence generation is reported (such as computer generated random numbers and random number tables, whilst inadequate approaches will include the use of alternation, case record numbers, birth dates or days of the week). B - Did not specify one of the adequate reported methods in (A) but mentioned randomisation method. C - Other methods of allocation that appear to be unbiased. 5

9 Adequacy of the allocation concealment process A - Adequate measures to conceal allocations. We deemed concealment adequate where randomisation was centralised or pharmacycontrolled, or where the following were used: serially numbered containers, on-site computer-based systems where assignment is unreadable until after allocation, other methods with robust methods to prevent foreknowledge of the allocation sequence to clinicians and patients. B - Unclearly concealed trials, in which the authors either did not report an allocation concealment approach at all, or reported an approach that did not fall into one of the categories in (A). C - Inadequately concealed trials, in which method of allocation was not concealed. Inadequate approaches included: use of alternation, case record numbers, days of the week, open random number lists and serially numbered envelopes even if opaque. Potential for selection bias after allocation A - Studies where an intention to treat analysis is possible and few exclusions (with adequate reporting of these exclusions). B - Studies which reported exclusions as reported in (A), but exclusions were less than 10%. C - No reporting of exclusions; exclusions of 10% or more or wide differences in exclusions between groups. Data synthesis We conducted meta-analyses for event rates, mortality, myocardial infarction, thrombosis, target lesion revascularisation, target vessel revascularisation. We analysed data in the form of odds ratios (OR) and 95% confidence intervals (95% CI) using the Mantel- Haenszel method, fixed-effect model provided by RevMan Analyses 1.0. We tested for heterogeneity between trial results using a multistep process whereby we completed the following tasks. We examined forest plots and noted the presence or absence of overlap in the CIs (lack of overlap of confidence intervals may indicate heterogeneity). We performed the Chi 2 test for heterogeneity. We obtained the I 2 statistic to describe the proportion of the variability due to heterogeneity. Where quantitative heterogeneity between trials was indicated, we have presented analysis using a random-effects model in addition to the fixed-effect estimate. For convenience, we have grouped trials in the meta-analysis plots according to type of drug eluted from the stent. Adequacy of masking A - Double (or triple) blind. B - Single blind. C - Non-blind. D - Unclear. Where disagreement arose on the suitability of a trial for inclusion in the review or on its quality, the reviewers reached a consensus by discussion and/or consulting a third authors. R E S U L T S Description of studies See: Characteristics of included studies; Characteristics of excluded studies; Characteristics of studies awaiting classification; Characteristics of ongoing studies. Data extraction Three authors extracted data using pre-tested data extraction forms; a second author cross-checked all data by a second reviewer. The extraction included the outcome measures (detailed above), as well as information on trial design, participants (including baseline characteristics and co-morbidity in terms of diabetes and previous heart disease). For binary outcome measures, we sought data on the number of patients with each outcome event, by allocated treatment group, irrespective of compliance and whether or not the patient was later thought to be ineligible or otherwise excluded from treatment or follow up to allow an intention-to-treat analysis. Results of the search The results of our searches yielded 1867 non-duplicate records from our electronic search of CENTRAL, MEDLINE and EM- BASE. Of these, we selected 119 records for closer examination and application of the inclusion/exclusion criteria. We identified a further 15 additional records (relating to ACTION; DECODE; DESSERT; DEVINE; EAGLE; ENDEAVOR II; FUTURE I; FUTURE II; HAAMU-STENT; Li 2007; MISSION; Pasceri 2003; PATENCY; STEALTH; TAXUS VI) through handsearching and selected them for inclusion in the review. The authors agreed on all records characterised for inclusion/exclusion. We did not find any RCTs relating to quality of life. 6

10 Included studies Forty-seven trials met the inclusion criteria: ACTION; APPLAUSE; ASPECT; BASKET-SES BASKET-PES; C-SIRIUS; DECODE; DELIVER; DESSERT ; DIABETES; Diaz 2007; EAGLE; ELUTES; ENDEAVOR II; E-SIRIUS; ETHOS I 1/SR; ETHOS I 2/MR; FUTURE I; FUTURE II; HAAMU-STENT; Han 2006; Han 2007 ; JUPITER II; König 2007; Li 2004; MISSION; Ortolani 2007; Pache 2005; PASSION; PATENCY; PRISON II; RAVEL; SCANDSTENT; SCORE; SCORPIUS; SELECTION; SES-AMI; SES-SMART; SIRIUS; SPIRIT FIRST; STEALTH; STRATEGY; Strozzi 2007; TAXUS I; TAXUS II (SR); TAXUS II (MR); TAXUS IV; TAXUS V; TAXUS VI; TYPHOON. ETHOS I 1/SR and TAXUS II (MR) compared two treatment groups, slow and moderate release DES; BASKET-PES and BASKET-SES are one trial that assessed two types of stents. We have counted each of these as single trials, but have discussed and analysed their data separately. We have identified four studies that are ongoing studies and two that are awaiting assessment. We anticipate including data from these studies in a review update. Preliminary examination of the results of the studies awaiting assessment indicates that their inclusion would not alter the conclusions of the review. Trials began recruiting from 1999 onwards. The majority were conducted in Europe; four were conducted internationally, four were US-based, four were based in Asia. The most frequently investigated drug was sirolimus, utilised in 21 trials (BASKET-SES; C-SIRIUS; DECODE; DESSERT; DIABETES; Diaz 2007; E-SIRIUS; Li 2004; MISSION; Ortolani 2007; Pache 2005; PRISON II; RAVEL; SCANDSTENT; SCORPIUS; SES-AMI; SES-SMART; SIRIUS; STRATEGY; Strozzi 2007; TYPHOON). Paclitaxel eluting compounds were investigated in 15 trials (APPLAUSE; ASPECT; BASKET-PES; DELIVER; EAGLE; ELUTES; HAAMU-STENT; PASSION; PATENCY; SELECTION; TAXUS I; TAXUS II (SR); TAXUS II (MR); TAXUS IV; TAXUS V; TAXUS VI). Everolimus-eluting stents were the subject of three trials (FUTURE I; FUTURE II; SPIRIT FIRST), dexamethasone 2 trials (Han 2006; König 2007) and tacrolimus 2 trials (Han 2007; JUPITER II). There was a single trial for each of the following: actinomycin (ACTION), biolimus A9 (STEALTH) QP2/7 hexanoylitaxel (SCORE), 17 beta- Estradiol (ETHOS I 1/SR; ETHOS I 2/MR) and zotarolimus/ ABT 578 (ENDEAVOR II). We have provided information on the trials and patient populations in the Characteristics of included studies table. Thirty-four trials stated that patients received treatment on a single lesion/ vessel. Male patients predominated in all trials with the exception of DESSERT, EAGLE and STEALTH. The mean age of patients ranged from 57 years (ASPECT) to 70 years (DESSERT). The proportion of participants with diabetes mellitus varied from 9% (ACTION) to 48% (König 2007). Four trials exclusively included patients with diabetes mellitus (DECODE; DESSERT; DIABETES; SCORPIUS) whilst the FUTURE I trial excluded these patients. A proportion of trials focused on subgroups, namely: acute coronary syndrome (König 2007; Strozzi 2007); ST elevated myocardial infarction (Diaz 2007; HAAMU-STENT; MISSION; PASSION; SELECTION; SES-AMI; STRATEGY; TYPHOON); total coronary occlusions (PRISON II); small vessels (Li 2004; SES-SMART); small vessels and long lesions (C-SIRIUS; E-SIRIUS); complex lesions (SCANDSTENT) and long lesions (TAXUS VI). Three trials were exclusively populated by unselected patients - that is there were no specific inclusion or exclusion criteria etc (BASKET-SES; BASKET-PES; Han 2007). A total of participants were included in the 47 trials, 6146 evaluating sirolimus, 5753 paclitaxel, 1197 zotarolimus, 532 tacrolimus, 367 actinomycin, 266 QP2/7-Hexanoyltaxol, 212 dexamethasone, 166 everolimus, 120 biolimus A9 and beta- Estradiol. Numbers randomised to DES versus BMS were not equal since three trials (ACTION; ELUTES; ASPECT) assessed various concentrations of drug-elution, but used single control groups; the BASKET trial included both a sirolimus and paclitaxel group with a single BMS control; ETHOS I explored a slow release and moderate release elution profile in two separate groups compared with a single (BMS) control group. The TAXUS II trial also explored two different DES elution profiles (one slow (SR), one moderate (MR) release) in two separate cohorts, but each had its own (BMS) control group. Trial sizes varied from 30 (APPLAUSE) to 1314 (TAXUS IV). Of the 47 trials, 34 were reported to have been conducted at more than one centre. Of the 47 trials, 65% acknowledged industry funding, at least in part; 14% specifically stated that the funding source was from a non-industry source (ie a Foundation). The remaining trials did not state the source of the funding. Excluded studies The Characteristics of excluded studies table shows that we excluded 14 trials. The majority (eight) were not RCTs, one amalgamated the data for two different DES, one had pre-treatment with cutting balloon, one re-assigned patients to the DES arm partway through the trial, one did not present separate data for individual patients (vessels randomised), one combined two types of stents and one compared titanium oxide stents with stainless steel stents. Risk of bias in included studies We constructed funnel plots (Figure 1; Figure 2) to examine the risk of publication bias for the main outcome of MACE for trials reporting outcomes at six months and 12 months. We found no evidence to suggest publication bias. 7

11 Figure 1. All trials reporting outcomes at 6 months 8

12 Figure 2. All trials reporting outcomes at 12 months The majority of trials were published as peer-reviewed publications. The limited information on trial methodology in conference abstracts and presentations affected our ability to assess the quality of four trials (EAGLE; HAAMU-STENT; PATENCY; SPIRIT FIRST). The results of the quality assessment are presented in Table 3. 9

13 Overall, five trials achieved As across all elements of the quality assessment (ELUTES; ENDEAVOR II; SIRIUS; TAXUS IV; TAXUS V) With regard to the adequacy of the randomisation process, a rating of A was given to 12 of the 22 trials assessing sirolimus eluting stents, four of 15 trials assessing paclitaxel eluting stents, one of the two trials assessing tacrolimus eluting stents and the single trials assessing zotarolimus and actinomycin eluting stents. The remaining trials were all rated as B. Concerning the adequacy of the allocation concealment process, a rating of A was given to 11 of the 23 trials assessing sirolimus eluting stents, four of the 16 trials assessing paclitaxel eluting stents, one of the two trials assessing tacrolimus eluting stents and the single trials assessing zotarolimus and actinomycin eluting stents. The remaining trials were all rated as B. The reporting of exclusions and use of intention-to-treat analysis was very good across all trials: all but two (FUTURE II and EAGLE) were rated as A. Masking adequacy was more variable across the trials. Of the 23 trials assessing sirolimus eluting stents, five were double or triple blinded, five were single blinded and three were non-blinded. Blinding levels were unclear in the remaining 10 trials. In the 16 trials assessing paclitaxel eluting stents, eight were double or triple blinded, three were single blinded and blinding levels were unclear in five. Two of the three everolimus trials were single blinded, with the third trial unclear as to blinding levels. One of the three dexamethasone trials was unblinded, whilst blinding was unclear in the remainder. One of the tacrolimus trials was double or triple blinded whilst blinding was unclear in the other. Both the Biolimus A9 and 17 beta-estradiol trials were double or triple blinded, whilst the actinomycin and QP2 trials were single and unblinded respectively. It is unlikely that there has been any selective reporting of outcomes in these studies, as the outcomes for such studies are internationally standardised. There has been discussion however regarding the appropriateness (or not) of using composite endpoints. It is beyond the scope of this review to enter that debate. Effects of interventions We collated the stated primary and composite outcome for each trial and the composite event rates (generally major adverse cardiac events - MACE, but also target vessel failure - TVF). The definition of the composite event rate varied across trials (see Characteristics of included studies table) although all included a hierarchy of death, AMI, and some measure of coronary revascularisation. Rates of revascularisation were reported as either target vessel revascularisation and/or target lesion revascularisation. These syntheses report odds ratios using fixed-effect comparisons in the main, although in cases where high levels of heterogeneity (I 2 = 50%or greater) were indicated, random-effects comparisons were used and reported alongside. In the analyses, we have grouped the trials according to the eluted agent. The two cohorts in TAXUS II which received stents with differing doses of paclitaxel are reported separately in the meta-analysis (as TAXUS II (SR) and TAXUS II (MR) ). The two groups in the ETHOS I trial (ETHOS I 1/SR AND ETHOS I 2/MR) are also separately reported and the single control group results duplicated. Initially, we have adopted an inclusive, comprehensive approach as set out in the protocol for this review, and included all RCTs we identified. However, to serve the needs of the health decision maker, it is appropriate to consider evidence from trials of the DES (currently) with CE Marking (and/or FDA approved) and that are marketed, thus rendering them available. We have presented a second set of analyses in which only those DES which are available are included. A series of analyses present comparisons for subgroups of the included patient population: patients with more complex vessel involvements (eg small and narrow), unselected patients and those with diabetes mellitus. In the literature, the diabetic population tends to be considered a subgroup; however, it is our view (see Hill 2007) that vessel architecture is the critical factor and that patients with diabetes tend to also be those with small vessels and/or long lesions. The analyses include data from six months, twelve months and annually thereafter. Note that not all trials report outcomes at every time point included in the analysis. Trials do not necessarily report on all outcomes, therefore not all drugs can be included in all outcomes. Table 4 shows the length of available data per eluted drug. With regard to the event rate outcome, details of the events included in this measure for each trial can be seen in the Characteristics of included studies table. Two studies (BASKET-PES; BASKET-SES) also reported costeffectiveness analysis. These cost-effectiveness results have been combined within a comprehensive review of DES and no attempt is made to replicate this here (Hill 2004; Hill 2007). Suffice it to say that health policy makers have taken into consideration these results and have in some cases limited the use of DES if their acquisition costs exceed set limits (NICE 2008). In costeffectiveness analyses, the costs per QALY were high ( 183, ,000) and outside of the normal range of acceptability. Analysis 1: All DES Main findings The included studies assessed 10 different drugs: sirolimus, paclitaxel, everolimus, zotarolimus, tacrolimus, dexamethasone, biolimus A9, actinomycin, QP2 and 17 beta-estradiol. Of these, data were available for up to five years (sirolimus, paclitaxel), three years (everolimus, zotarolimus), 12 months (dexamethasone, tacrolimus, QP2, actinomycin), and six months (17 beta-estradiol, biolimus A9). On the composite measure, five of the 10 drug groups (sirolimus, paclitaxel, zotarolimus, dexamethasone and everolimus) were observed to have a statistically significant 10

14 benefit compared to BMS; three had little (tacrolimus) or no benefit (biolimus A9, 17-beta Estradiol) and one (actinomycin) was noted to have significantly more adverse events than the comparator BMS. The majority of benefit was seen in significantly reduced revascularisation rates as no statistically significant differences between nine drug groups and BMS were found on rates of mortality, MI or thrombosis. The MI rate for QP2 was found to be greater in the DES group than for patients receiving BMS. Mortality Analysis 2.1, Analysis 2.2, Analysis 2.3, Analysis 2.4, Analysis 2.5 and Analysis 2.6 show mortality rates at six months, 12 months, two years, three years, four years and five years respectively. We found no difference in mortality between DES and BMS in any analysis. Data beyond 12 months were available for the sirolimus group (two trials, n = 1296) out to five years, the paclitaxel group out to five years (single trial, n = 440) everolimus (single trial, n = 54) and zotarolimus (single trial, n = 1183) out to three years. Event rate Analysis 1.1, Analysis 1.2, Analysis 1.3, Analysis 1.4, Analysis 1.5 and Analysis 1.6 show the findings for event rates at six months, 12 months, two years, three years, four years and five years respectively. Data for the sirolimus group indicate that the significantly reduced event rates in the DES arm compared to BMS were consistent from six months (OR 0.28; 95% CI 0.23 to 0.35) out to five years (OR 0.53; 95% CI 0.41 to 0.67). Data from 12 trials (n = 3437) were included at six months and from two trials (n = 1296) at five years. The paclitaxel group likewise demonstrated significantly reduced event rates in the DES arm compared to BMS. The findings were consistent from six months (OR 0.59; 95% CI 0.50 to 0.69) to four years (OR 0.62; 95% CI 0.49 to 0.80). Data from 14 trials (n = 5717) informed the six-month outcome, whilst two trials (n = 1375) informed the outcome at four years. At five years, the single trial (n = 440) failed to show any difference between DES and BMS; however, the assessed stent was experimental only. The single trial of zotarolimus (n = 1183) reported a significant benefit of DES over BMS at six months (OR 0.47; 95% CI 0.32 to 0.69) and at three years (OR 0.52; 95% CI 0.38 to 0.72). Dexamethosone significantly reduced event rates in the DES arm compared to BMS at 12 months (only time point available). Data were taken from two small (n = 212) trials (OR 0.42; 95% CI 0.22 to 0.82). Data for the everolimus group were available at six months (three trials, n = 189) and three years (single trial, n = 58). At six months the everolimus group had significantly reduced event rates compared to BMS (OR 0.32; 95% CI 0.11 to 0.90); this difference was not evident at three years. In the tacrolimus group, event rates were calculated based on the results of two trials (n = 524). Event rates were significantly reduced in the DES arm compared to BMS at six months (OR 0.59; 95% CI 0.35 to 0.99) although it is worth noting that the CI is wide and close to one; however at 12 months (single trial, n = 314), no difference remains. Of the remaining four drug groups, biolimus A9, 17 beta Estradiol and QP2 event rates were found to be no different to BMS. In the case of actinomycin, the use of DES was less effective than the comparator BMS. Myocardial infarction Analysis 3.1, Analysis 3.2, Analysis 3.3, Analysis 3.4, Analysis 3.5 and Analysis 3.6 show MI rates at six months, 12 months, two years, three years, four years and five years respectively.there was no difference in incidence of MI between and DES and BMS up to 12 months, with the exception of the QP2 (SCORE) trial which reported statistically significantly higher rate of MI at six months and 12 months in the DES arm (OR 10.75; 95% CI 3.15 to 36.66). This latter finding resulted in the early termination of the trial (as mandated in the trial protocol) and accelerated follow up of patients. Data were available up to five years for the sirolimus group (two trials, n = 1296), five years for paclitaxel (single trial, n = 440) and three years for everolimus (single trial, n = 54) and zotarolimus (single trial, n = 1156). We noted no difference in rates of MI compared to BMS. TLR Analysis 4.1, Analysis 4.2, Analysis 4.3, Analysis 4.4, Analysis 4.5 and Analysis 4.6 show TLR rates at six months, 12 months, two years, three years, four years and five years respectively. The trial assessing actinomycin did not report TLR rates, reducing the groups in this analysis to nine. Benefits of sirolimus DES compared to BMS were found consistently from six months (OR 0.18; 95% CI 0.14 to 0.25) out to five years (OR 0.21; 95% CI 0.15 to 0.29). The six-month data were derived from 10 trials (n = 2692); data at five years from two trials (n = 1296). We noted some heterogeneity within the sirolimus trials at one, two and three years; however analysis using the random-effects model did not affect the overall outcomes. Similarly, data for paclitaxel compared to BMS showed significant and consistent benefits of DES from six months (OR 0.35; 95% CI 0.28 to 0.45) to four years (OR 0.33; 95% CI 0.24 to 0.45). At five years, no significant difference was observed. The six-month data were derived from 10 trials (n = 3884); the four year data from three trials (n = 1850). The single zotarolimus trial (n = 1183) reported significant outcomes for DES compared with BMS at both six months (OR 0.36; 95% CI 0.22 to 0.56) and three years (OR 0.46; 95% CI 0.31 to 0.67). 11

15 Two small trials (n = 212) reported results of dexamethasone at 12 months only; these indicated benefits of DES compared to BMS (OR 0.35; 95% CI 0.17 to 0.71). Of the five remaining drugs (everolimus, tacrolimus, biolimus A9, 17 beta-estradiol, QP2) no benefit of DES was observed. TVR Analysis 5.1, Analysis 5.2, Analysis 5.3, Analysis 5.4, Analysis 5.5 and Analysis 5.6 show TVR rates at six months, 12 months, two years, three years, four years and five years respectively. Findings noted significant and consistent benefits of DES over BMS for the sirolimus group at six months (OR 0.31; 95% CI 0.22 to 0.42) through to five years (OR 0.45; 95% CI 0.34 to 0.61). These findings were based on four trials (n = 1978) at six months and a single trial (n = 1058) at five years. Paclitaxel was also consistently significantly better than BMS from six months (OR 0.49; 95% CI 0.40 to 0.60) out to three years (OR 0.55; 95% CI 0.43 to 0.70). These results were based on eight trials (n = 4111) at six months and three trials (n = 1785) at three years. Beyond three years, the TVR rate in the paclitaxel group was not significantly different to BMS. The single trial (n = 1183) in the zotarolimus group reported significant benefits of DES over BMS at six months (OR 0.41; 95% CI 0.27 to 0.63) and three years. A single, small trial (n = 58) reported data at three years for everolimus which showed DES to be superior to BMS, although the wide confidence intervals should be noted here (OR 0.18; 95% CI 0.03 to 0.92). Of the remaining groups (tacrolimus, dexamethasone, 17 beta estradiol, QP2, actinomycin) TVR rates in the DES arms were not statistically different to those in the BMS group. Thrombosis Analysis 6.1, Analysis 6.2, Analysis 6.3, Analysis 6.4, Analysis 6.5 and Analysis 6.6, show rates of thrombosis at six months, 12 months, two years, three years, four years and five years respectively. The trials assessing actinomycin and dexamethasone did not report thrombosis rates, reducing the numbers of drugs in this analysis to eight. For seven of these, we found no difference between DES and BMS on rates of thrombosis. However, the trial assessing QP2 (SCORE) reported significantly more cases of thrombosis in the DES group than in the BMS group (OR 10.69; 95% CI 1.34 to 85.63; OR 15.99; 95% CI 2.06 to ). Data beyond 12 months were available for sirolimus (five years), paclitaxel (five years), everolimus and zotarolimus (three years). Analysis 2: Available DES The DES included in this analysis are limited to those currently (2009) available for use, ie they are marketed and have CE marking (and or FDA approval). The number of drugs is thereby reduced to seven (sirolimus, paclitaxel, everolimus, tacrolimus, zotarolimus, biolimus A9 and dexamethasone). This analysis included 39 trials (APPLAUSE; BASKET-SES; BASKET-PES; C-SIRIUS; DECODE; DELIVER; DESSERT; DIABETES; Diaz 2007; E-SIRIUS; EAGLE; ENDEAVOR II; HAAMU-STENT; Han 2006; Han 2007; JUPITER II; König 2007; MISSION; Ortolani 2007; Pache 2005; PASSION; PATENCY; PRISON II; RAVEL; SCANDSTENT; SCORPIUS; SELECTION; SES-AMI; SES- SMART; SIRIUS; SPIRIT FIRST; STEALTH; STRATEGY; TAXUS I; TAXUS II (MR); TAXUS II (SR); TAXUS IV; TAXUS V; TYPHOON. Main findings On the composite event rate measure, we observed four groups of DES to have a significant benefit compared to BMS (sirolimus, paclitaxel, zotarolimus, dexamethasone), four reported little (tacrolimus) or no benefit and one reported significantly more adverse events. The majority of benefit was seen in significantly reduced revascularisation rates. We found no significant differences between DES and BMS for rates of mortality, MI or thrombosis. Event rate Analysis 7.1, Analysis 7.2, Analysis 7.3, Analysis 7.4, Analysis 7.5 and Analysis 7.6 show event rates at six months, 12 months, two years, three years, four years and five years respectively. In the sirolimus group, we noted significantly reduced event rates in the DES arm compared to BMS at six months (OR 0.29; 95% CI 0.23 to 0.36) out to five years (OR 0.53; 95% CI 0.41 to 0.67). The data were derived from 11 trials (n = 3358) at six months and two trials (n = 1296) at five years. We observed similar findings in the paclitaxel group, wherein significantly fewer event rates were recorded in the DES arm compared to BMS at six months (OR 0.55; 95% CI 0.45 to 0.67) out to four years (OR 0.62; 95% CI 0.49 to 0.80). The data were derived from eight trials (n = 3566) at six months and two trials (n = 1375) at five years. The single trial (n = 1183) reporting outcomes for zotarolimus noted significantly fewer events in the DES compared to BMS groups at six months (OR 0.47; 95% CI 0.32 to 0.69) and at three years (OR 0.52; 95% CI 0.38 to 0.72). The two small (n = 212) trials within the dexamethasone group indicated a significant difference in event rates at 12 months (only time point available) compared to BMS (OR 0.42; 95% CI 0.22 to 0.82). For tacrolimus, the significant difference in events rates in favour of DES at six months was no longer in evidence at 12 months. It is noteworthy that the confidence intervals reported at six months are wide and the upper limit is close to one (OR 0.59; 95% CI 0.35 to 0.99). The event rates in the DES arm of the everolimus and biolimus A9 groups were not significantly different to those found in BMS. 12

16 Mortality We found no significant difference in mortality rates between DES and BMS in any analysis. Data beyond 12 months were available for sirolimus (two trials, n = 1296) out to five years, paclitaxel out to four years (three trials, n = 1640) everolimus (single trial, n = 56) and zotarolimus (single trial, n = 1156) out to three years. Myocardial infarction We found no significant difference in MI rates between DES and BMS in any analysis. Data beyond 12 months were available for sirolimus (two trials, n = 1296) out to five years, paclitaxel out to four years (three trials, n = 1640), everolimus (one trial, n = 54) and zotarolimus (one trial, n = 1156) out to three years. TLR In the sirolimus group, significantly fewer TLRs were noted in the DES arm compared to BMS at six months (OR 0.18; 95% CI 0.14 to 0.25) out to five years (OR 0.30; 95% CI 0.22 to 0.42). The data were derived from nine trials (n = 2613) at six months and two trials (n = 1296) at five years. We noted similar findings in the paclitaxel group, wherein significantly fewer TLRs were noted in the DES arm compared to BMS at six months (OR 0.37; 95% CI 0.28 to 0.48) out to four years (OR 0.35; 95% CI 0.25 to 0.48). The data were derived from seven trials (n = 3003) at six months and two trials (n = 1580) at four years. It should be noted that results at three years showed these DES to be non-significantly different to BMS; however, the findings were based on a single, small (n = 61) trial. The single trial (n = 1183) reporting TLRs for zotarolimus found significantly fewer TLRs in the DES compared to BMS groups at six months (OR 0.36; 95% CI 0.22 to 0.56) and three years (OR 0.46; 95% CI 0.31 to 0.67). The two small (n = 212) trials within the dexamethasone group indicated a significant difference in rates of TLR at 12 months (only time point available) compared to BMS (OR 0.35; 95% CI 0.17 to 0.71). The TLR rates in the DES arm of the everolimus, tacrolimus, biolimus A9 and QP2 groups were not significantly different to those noted in the BMS arm at any time point. TVR In the sirolimus group, significantly fewer TVRs were noted in the DES arm compared to BMS at six months (OR 0.31; 95% CI 0.22 to 0.42) out to five years (OR 0.45; 95% CI 0.34 to 0.61). The data were derived from four trials (n = 2613) at six months and a single trial (n = 1058) at five years. Similar findings were noted in the paclitaxel group, wherein significantly fewer TVRs were noted in the DES arm compared to BMS at six months (OR 0.52; 95% CI 0.42 to 0.66) out to three years (OR 0.35; 95% CI 0.25 to 0.48). The data were derived from seven trials (n = 3003) at six months and two trials (n = 1355) at three years. Although results at four years showed these DES to be non-significantly different to BMS, as noted above, the findings were based on a single, small (n = 61) trial. The single trial (n = 1183) reporting TVRs for zotarolimus noted significantly less in the DES compared to BMS groups at six months (OR 0.41; 95% CI 0.27 to 0.63) and three years (OR 0.49; 95% CI 0.35 to 0.70). The TVR rates in the DES arm of the everolimus, tacrolimus and dexamethasone groups were not significantly different to those noted in the BMS arm at any time point. Thrombosis No differences were found in any of the drug groups assessed (sirolimus, paclitaxel, tacrolimus, zotarolimus, biolimus A9, dexamethasone or everolimus at any time point. Analysis 3: diabetes mellitus This is an analysis of trials that exclusively included patients with diabetes mellitus. Sirolimus was the only drug investigated. The four included trials were DECODE, DESSERT, DIABETES, SCORPIUS. Data were available to two years. Main findings On the composite measure, DES fared significantly better than BMS at six months, 12 months, and two years. As with all other analyses, the majority of benefit was seen in significantly reduced revascularisation rates with no significant differences noted between DES and BMS for rates of mortality, acute myocardial infarction or thrombosis. Event rate A significant benefit of DES over BMS was in evidence at six months (OR 0.27; 95% CI 0.17 to 0.45). This continued at 12 months (OR 0.32; 95% CI 0.21 to 0.51) and at two years (OR 0.21; 95% CI 0.09 to 0.47) Mortality There was no benefit of sirolimus over BMS at any time point. Myocardial infarction There was no benefit of sirolimus over BMS shown for any analysis. 13

17 TLR The TLR rates were significantly lower for the patients in the DES group compared to BMS at all time points: six months (OR 0.15; 95% CI 0.08 to 0.30), 12 months (OR 0.11; 95% CI 0.05 to 0.26) and two years (OR 0.07; 95% CI 0.02 to 0.18). TVR Only two trials (n = 233) reported TVR, both at 12 months. Rates of TVR were significantly lower in the DES compared to the BMS group (OR 0.22; CI 0.09 to 0.55). Thrombosis There was no benefit of sirolimus over BMS at any time point. Analysis 6: unselected patients These trials had no specific inclusion or exclusion criteria for patients. They were designed to reflect general practice. Data related to three drugs are reported (sirolimus, paclitaxel and tacrolimus) and are available at six months only. The three included trials were BASKET-PES, BASKET-SES and Han Event rate A significant effect was noted for both sirolimus (OR 0.44; 95% CI 0.23 to 0.82) and tacrolimus (OR 0.36; 95% CI 0.13 to 0.97) but not for paclitaxel. Mortality No significant difference was found between any DES and BMS. Myocardial infarction No significant difference was found between any DES and BMS. TLR No significant benefit for DES compared to BMS was noted; only data for tacrolimus (n = 200) were available. Analysis 4: Long lesions, small vessels, complex lesions Main findings On the composite measure, the majority of benefit was seen in significantly reduced revascularisation rates. No significant differences between DES and BMS were found for rates of mortality, acute myocardial infarction or thrombosis. Two drugs (sirolimus and paclitaxel) were investigated in five trials: C-SIRIUS, E-SIRIUS, SIRIUS, SES-SMART, SCANDSTENT, TAXUS VI. The latter was the only trial to assess paclitaxel. Event rate The DES in the sirolimus group consistently reported significantly lower event rates than BMS. Data were available from six months (OR 0.29; 95% CI 0.22 to 0.40) up to four years (OR 0.44; 95% CI 0.33 to 0.60). The single paclitaxel trial (n = 440) did not produce any evidence of a benefit on this measure at any time point. Mortality Neither sirolimus nor paclitaxel were significantly different to BMS on this measure. Myocardial infarction With the exception of the sirolimus group at three years, no benefit of DES compared to BMS was noted. It should be noted that the CI for sirolimus at three years is wide and is close to one at the upper end. TLR In the DES groups, both sirolimus and paclitaxel were shown to be superior to BMS in all analyses from six months (OR 0.21; 95% CI 0.15 to 0.30; OR 0.31; 95% CI 0.17 to 0.59) up to three years in the case of paclitaxel (OR 0.49; 95% CI 0.29 to 0.84) out to four years (OR 0.27; 95% CI 0.19 to 0.40) for sirolimus. TVR Significantly lower TVR rates were noted for sirolimus (OR 0.37; 95% CI 0.16 to 0.84) but not for paclitaxel. There were no data for tacrolimus. Thrombosis There were no data for this outcome. TVR Forest plots to show that there were benefits of sirolimus from six months (OR 0.30; 95% CI 0.20 to 0.45) to three years (OR 0.35; 95% CI 0.25 to 0.49). The benefits of paclitaxel were evidenced at six months (OR 0.42; 95% CI 0.24 to 0.74) to two years (OR 0.57; 95% CI 0.34 to 0.94) but not at three or five years. 14

18 Thrombosis Neither drug was significantly different to BMS on any analysis. D I S C U S S I O N Key findings Overall, there were no differences in rates of mortality, AMI, or thrombosis between any DES and BMS in any of our analyses. (The noted exception is QP2 which showed higher rates in the DES arm in all three outcomes). There were advantages for DES on the outcomes of event rate, TLR and TVR. Those DES eluting sirolimus, paclitaxel, everolimus, tacrolimus, zotarolimus and dexamethasone had reduced rates of composite cardiac events (underpinned by reductions in repeat revascularisations) compared to BMS. The evidence shows that the benefits of zotarolimus continue at least to three years, whilst those for sirolimus and paclitaxel may last up to four years. The advantages of tacrolimus were in evidence only at six months and there were data for dexamethasone for one year only. In terms of TLR, those DES which eluted sirolimus, paclitaxel, dexamethasone or zotarolimus had benefits over BMS, both at six months and one year (dexamethasone), three years (zotarolimus), four years (paclitaxel), and five years (sirolimus). For TVR, DES eluting sirolimus, paclitaxel, zotarolimus and everolimus showed benefits over BMS at six months up to three years (zotarolimus, paclitaxel, and everolimus) and four years (sirolimus). Actinomycin and QP2 appeared to be harmful. Four trials were halted early, three (ACTION, DECODE, SCORE) on the advice of the DSMB in response to unacceptable adverse event rates in the DES arm; one trial (PATENCY) was terminated for business reasons (Serruys and Gershlick). It is unlikely that the results from these trials would have a major impact on our overall conclusions since our second analysis assessed DES that are currently available; this eliminated ACTION and SCORE, and the numbers of patients in the DECODE and PATENCY trials was relatively small. These findings reflect those of two previous reviews produced by our group, the Hill 2004 version of this review, a monograph completed in 2005 and recently published by the NCCHTA ( Hill 2007). However, the present review represents a considerable update to all of the foregoing as it includes supplementary followup data, the addition of a substantial number of new trials and an analysis of thrombosis rates. Overall conclusions The DES included in this analysis performed no better than BMS on the outcomes of mortality, AMI or thrombosis. Four drugs (sirolimus, paclitaxel, zotarolimus and dexamethasone) were superior to BMS on the primary outcome of event rate (which often includes TLR or TVR) some with evidence of effect at up to five years (sirolimus), four years (paclitaxel), three years (zotarolimus) and one year (dexamethasone). TLR and TVR rates for these drugs (where recorded) were lower than in the BMS groups, again there was evidence of longer-term effectiveness. Other drugs are either not effective or, in the cases of actinomycin and QP2, cause harm. These findings were largely mirrored in the analyses of data from the identified subgroups. The early and rapid uptake in the use of DES was based on early trial results. As noted earlier, subsequent analysis of cost effectiveness and possible stent thrombosis with DES has limited their use in clinical practice. Limitations We acknowledge that inclusion of unpublished data may be controversial, as full quality assurance of trials may not be possible and data may vary between unpublished and peer-reviewed reports (Dundar 2004). Noting the key findings of this review, it is important to appreciate several issues which limit appraisal of DES and correspondingly our review. Current clinical guidelines (AHA PCI guidelines; ESC PCI guidelines) recommend post-procedural use of clopidogrel for at least 12 months to prevent occurrences in late stent thrombosis. The duration of post-procedural clopidogrel in the trials included this review varied; many of the patients in the earlier trials were treated with clopidogrel for considerably less than 12 months. Since the world revenue of stents is estimated to be up to $6 billion annually (Bischof 2009), the competition between companies is far greater than collaboration. This lack of collaboration to have standardised trial designs across companies leads to numerous differences in protocol design, trial conduct and outcome definitions between trials, which challenges easy comparison between different drug-eluting devices. For example, having all patients undergoing protocol mandated follow-up angiography will lead to a higher repeat revascularisation rate compared to patients for whom the protocol only requires clinical follow up. This is related to the fact that cardiologists who identify renarrowing at the follow-up angiogram are likely to treat the lesion even when the patient has no symptoms or signs of ischaemia. This phenomena is known as the oculostenotic reflex. Patients undergoing clinical follow up only are unlikely to have a further angiogram unless prompted by further symptoms or signs of ischaemia and therefore such trials will have lower rates of repeat procedures compared to those with angiographic follow up. Furthermore, there is likely to be more renarrowing seen in patients with angiography scheduled at nine to 12 months than if scheduled at six to seven months, since restenosis occurs to a lesser but significant degree even after 15

19 six months (Cutlip 2002). Moreover, since angiographic analysis is a speciality skill performed at only a few angiographic core labs in the world, any differences between these labs may skew the comparison between different trials (although not between groups in a single trial). A third issue relates to the use of MACE as an outcome. This composite endpoint has been widely and differently interpreted across trials, for example death may be reported as all death, cardiac death, or not specified. The use of MACE may also obscure real and important difference in outcomes. For example, repeat revascularisations are reported as events in the same way and with the same weight as a clinical MI or death. In practice, given the rarity of coronary death or MI, the majority of MACE events are made up of revascularisation procedures. We were unable to report outcomes uniformly across all timepoints and for all outcomes for all drugs. Not all of the trials reported outcomes at every timepoint we included in our analyses (ie some trials reported outcomes at one year, whilst others reported outcomes at six months). Few trials reported outcomes beyond one or two years. In addition, not all trials included all outcomes selected for this review. Revascularisation may be reported as target lesion revascularisation (TLR), target vessel revascularisation (TVR), revascularisation by a particular technique (CABG or PCI) or it may not be specified. There are also limited data on total revascularisation; for example, a patient may have another procedure carried out in a vessel other than the one originally treated. This reporting is appropriate for the assessment of the efficacy of a particular stent, but data related to any revascularisation are needed when assessing the practical effectiveness and costs of patient treatment. As noted above, the definition of revascularisation is affected by the trial protocol. Two trials (BASKET-PES; ETHOS I 1/SR) use the same control group for the two intervention arms of the trial. To use this data in the meta-analysis required the double counting of the patients in the control arms of these trials. the follow up on the SIRIUS trials to eight years and for 10 other RCTs to five years. Beyond clinical outcomes Although the main focus has been on reducing restenosis, it is also important to appreciate patient preferences. Given that the disutility (aversion to negative consequences) of ischaemia recurrence may be contained in only a few days in pain until a further angiographic procedure and intervention can be performed, in terms of events, any significant differences in mortality rates with BMS are likely to outweigh the restenosis rate reduction benefits seen. One estimate, using the terminology of health economics, was that one premature death averted would be worth 125 times the quality adjusted utility benefit of preventing restenosis (Cohen 2001). The conclusion of our economic analysis (Hill 2007) was that all patients considered together, the calculated cost per QALY ratios are high ( 183, ,000) and outside the normal range of acceptability. The use of DES would be best targeted at the subgroups of patients with the highest risks of requiring re-intervention, and could be considered cost-effective in only a small percentage of such patients. This conclusion reflects that of the BASKET trial authors (BASKET-PES; BASKET-SES). Continuing development of DES This is a rapidly evolving technology and new types of DES are under investigation. For example, bioabsorbable stents (eg BVS), stents which have biodegradable polymer coating (eg Infinnium, Conor), stents designed to promote endothelialisation (Genous R stent), multi-layer drug stents (eg SymBio), a stent which elutes bevacizumab, as well as stents designed to improve deliverability and application to a more diverse range of vessel sizes and architecture. Future directions Long-term performance of DES A key question is whether benefits of DES are maintained over five years; such data have been requested by the FDA from some of the landmark trials (SIRIUS; TAXUS IV). In this review, data for sirolimus from the RAVEL and SIRIUS trials were available at five years and show that the benefits appear to remain at five years. The one trial of paclitaxel reporting data at five years (TAXUS VI, assessing an experimental stent) provided no evidence of benefit at this time point. Data at three and four years for zotarolimus and paclitaxel respectively do provide evidence of their longer-term benefit. Johnson and Johnson has recently committed to extend A U T H O R S C O N C L U S I O N S Implications for practice Stents eluting sirolimus, paclitaxel and zotarolimus represent better short-term outcomes in terms of preventing restenosis compared with BMS. However, they are expensive and long-term efficacy and safety data are yet to be confirmed. In light of this, a number of regulatory bodies have investigated the cost-effectiveness of the use of DES with a variety of outcomes. Some limit their use to specific patient populations (eg patients with small or long vessels), while others limit the number of DES that can be utilised. The National Institute for Health and Clinical Excellence have advised their use only if the price premium for DES can be reduced to

20 Implications for research Secondary research should involve continued systematic reviewing (to incorporate longer-term follow up, data on newer BMS and DES (although these may be less forthcoming in the future since the trend for evaluating new DES is likely to be in head-to-head trials). Individual patient data level reviews to stratify risk groups and determine which patients are most likely to benefit from the use of DES. As longer term follow up and expanded outcomes (such as utility data) become available there will be a continuing need to update this review. In addition, it has yet to be established how long anti-platelet therapy should continue after DES implantation. Primary research should move towards standardisation of trial protocols, utilise clinically driven outcomes and measures relevant to patients such as the impact on quality of life of revascularisation. To this end, the ARC group recommendations for the type and definition of clinical endpoints in coronary stent trials, including the reporting of two composite endpoints, one that is device oriented and one that is patient oriented. The report further recommends the discontinuation of MACE (and other acronyms) given that historically MACE has been interpreted in a variety of different ways. Finally, ARC advise on the optimal basis for DES evaluation: overall cardiovascular outcomes from the patient s perspective, including all death, MI, and repeat revascularization procedures. These outcomes reflect the complex interplay between device performance, revascularization strategy, and secondary prevention and key patient descriptors. Both the time course and the composite selected should characterize patient wellbeing related to the pathophysiology of the implanted DES device and its impact on underlying coronary artery disease outcome. For example, whether a device improves functional capacity and quality of life, but does not affect MI rates or mortality--as is the case for percutaneous intervention in elective cases--should be clear so that regulatory authorities, clinicians and reimbursement agencies can carefully weigh the net benefit against possible safety concerns. ARC 2007 A C K N O W L E D G E M E N T S We are grateful to Ruaraidh Hill for producing the original review that formed the basis of this document and Dr. Ameet Bakhai for his input into the first review. R E F E R E N C E S References to studies included in this review ACTION {published and unpublished data} Linnemeier T. The ACTION Study. (Slide presentation, Cardiovascular Research Foundation [CRF] Drug- Eluting Stent Symposium. Transcatheter Cardiovascular Therapeutics [TCTMD] online database) [accessed 31/10/ 2002]. ACTION.pdf Serruys P. Final ACTION results (ActinomycinD) ACTinomycineluting stent improves outcomes by reducing neointimal hyperplasia -ACTION TCT 2002, Washington DC, September [accessed 21/01/2003]. TCT (Transcatheter Cardiovascular Therapeutics). 2002, issue 21/01/2003. Serruys P, Ormiston F, Sousa E, Grube E, den Heijer P, Feyter P, et al.actinomycin-eluting stent for coronary revascularization: a randomized feasibility and safety study: the ACTION trial. Journal of the American College of Cardiology 2004;44: Serruys PW. Actinomycin eluting stent improves outcomes by reducing neointimal hyperplasia (ACTION). European Society of Cardiology XXIV Annual Meeting. Berlin, August 31-September 4, Available from: centre/meeting int/detail.asp? acr trial=action [accessed 31/10/2002]. APPLAUSE {published data only (unpublished sought but not used)} Grube E, Mueller R, Lim V, Schmidt T, Gerckens U, Buellesfeld L. Evaluation of a new polymer-coated paclitaxel-eluting stent for treatment of de novo lesions: Six-month clinical and angiographic follow-up results of the APPLAUSE trial. Journal of Invasive Cardiology 2006;18 (5): ASPECT {unpublished data only} Heldman A. Six-month results of ASPECT. TCT (Transcatheter Cardiovascular Therapeutics) 2001 Conference (Presented on behalf of Park SJ). Available from: news item id=2405 [accessed 31/10/2002] Hong MK, Mintz GS, Lee CW, Song JM, Han KH, Kang DH, et al.paclitaxel coating reduces in-stent intimal hyperplasia in human coronary arteries: a serial volumetric intravascular ultrasound analysis from the ASian Paclitaxel- Eluting Stent Clinical Trial (ASPECT). Circulation 2003; 107(4): Liakishev AA. [Clinical trial of paclitaxel-eluting stents. Results of ASPECT]. [Russian]. Kardiologiia 2003;43(6): 72. Park SJ, Shim WH, Ho DS, Raizner AE, Park SW, Hong MK, et al.a paclitaxel-eluting stent for the prevention of coronary restenosis. New England Journal of Medicine 2003; 348(16): BASKET-PES {published data only (unpublished sought but not used)} Kaiser C, Brunner-La Rocca H, Buser P, Bonetti P, Osswald S, Linka A, et al.incremental cost-effectiveness of drug- 17