Coronary artery bypass grafting versus drug-eluting stents in multivessel coronary disease. A meta-analysis on 24,268 patients☆

DOI: 10.1016/j.ejcts.2009.03.012 2009;36:611-615 Eur J Cardiothorac Surg

Roscitano, Brenno Fiorani, Gian Domenico Di Nucci and Riccardo Sinatra Umberto Benedetto, Giovanni Melina, Emiliano Angeloni, Simone Refice, Antonino

disease. A meta-analysis on 24,268 patientsCoronary artery bypass grafting versus drug-eluting stents in multivessel coronary

This information is current as of August 11, 2010

http://ejcts.ctsnetjournals.org/cgi/content/full/36/4/611located on the World Wide Web at:

The online version of this article, along with updated information and services, is

ISSN: 1010-7940. European Association for Cardio-Thoracic Surgery. Published by Elsevier. All rights reserved. Printfor Cardio-thoracic Surgery and the European Society of Thoracic Surgeons. Copyright © 2009 by The European Journal of Cardio-thoracic Surgery is the official Journal of the European Association

by on August 11, 2010 ejcts.ctsnetjournals.orgDownloaded from

Coronary artery bypass grafting versus drug-eluting stentsin multivessel coronary disease.

A meta-analysis on 24,268 patients§

Umberto Benedetto *, Giovanni Melina, Emiliano Angeloni, Simone Refice,Antonino Roscitano, Brenno Fiorani, Gian Domenico Di Nucci, Riccardo Sinatra

Cardiac Surgery Department, II School of Medicine, University of Rome ‘‘La Sapienza’’, Policlinico S. Andrea, Via di Grottarossa 1039, Rome, Italy

Received 3 September 2008; received in revised form 23 February 2009; accepted 6 March 2009; Available online 25 April 2009

Abstract

Objective: Coronary artery bypass grafting (CABG) has been shown to provide better results than percutaneous coronary intervention (PCI) inmultivessel coronary disease. Drug-eluting stents (DES) have significantly improved results of PCI in terms of restenosis but the advantages of sucha treatment compared to CABG remain uncertain. This meta-analysis summarizes available data from observational cohorts comparing DES-PCIversus CABG. Methods: We performed a systematic literature search for observational cohorts comparing CABG versus DES-PCI in patients withmultivessel coronary disease. Themixed model method was used to obtain the pooled hazard ratio (HR) for outcomes of interest. Results: A totalof nine observational nonrandomized studies were identified and analyzed including a total of 24,268 patients with multivessel coronary diseasewho underwent DES-PCI (n = 13,540) and CABG (n = 10,728). Mean follow-up time was 20 months. Pooled analysis showed that DES-PCI and CABGwere comparable in terms of composite occurrence of death, acute myocardial infarction and cerebrovascular accidents (HR = 0.94; 95%CI = 0.72—1.22; p = 0.66). However, there was a significantly higher risk of repeat revascularization in the DES-PCI group (HR = 4.06; 95%CI = 2.64—6.24; p < 0.001). Overall major adverse cardiac and cerebrovascular events rate in the DES-PCI was higher compared to the CABG group(HR = 1.86; 95% CI = 1.36—2.54; p < 0.001). Conclusions: In the ‘real world’ clinical practice, overall major adverse cardiac and cerebrovascularevents rate continues to be higher after DES-PCI due to an excess of redo revascularization compared with CABG.# 2009 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved.

Keywords: Coronary artery bypass grafting; Drug-eluting stents; Percutaneous coronary intervention

www.elsevier.com/locate/ejctsEuropean Journal of Cardio-thoracic Surgery 36 (2009) 611—615

1. Introduction

The two primary interventions for patients with multi-vessel coronary artery disease (MVD) are coronary arterybypass grafting (CABG) and percutaneous coronary interven-tion (PCI).

It has been shown that there are comparable overallsafety outcomes (death, acute myocardium infarction,cerebrovascular accident) in CABG and PCI patients, butrepeat revascularization procedures remain higher after PCI[1,2].

The recent introduction of drug-eluting stents (DES) hasfurther improved the results of PCI in term of restenosis thusincreasing the percentage of MVD patients treated with acatheter-based approach. However the advantages of such atreatment compared to CABG remain uncertain [5—15].

§ Presented at the 22nd Annual Meeting of the European Association forCardio-thoracic Surgery, Lisbon, Portugal, September 14—17, 2008.* Corresponding author. Tel.: +39 06 33775311; fax: +39 06 33775481.E-mail address: [email protected] (U. Benedetto).

1010-7940/$ — see front matter # 2009 European Association for Cardio-Thoracicdoi:10.1016/j.ejcts.2009.03.012

ejcts.ctsnetjournaDownloaded from

Preliminary results from a single ongoing randomizedcontrolled trial comparing DES-PCI versus CABG on a limitednumber of MVD patients have been recently presented [16].As generally occurs in randomized trials, the results of thisstudy might be subject to trial design bias by preferentiallyenrolling relatively low-risk patients.

Observational studies reflect the more heterogeneous ‘realworld’ clinical practice and they should be considered tosupport the results from randomized controlled trials andtherefore indetermining the standardof care forMVDpatients.

With the present investigation, we aimed to gain a betterinsight in the treatment of MVD from a meta-analysis ofavailable observational reports comparing CABG and DES-PCI.

2. Materials and methods

The meta-analysis of Observational Studies in Epidemiol-ogy guidelines were followed. To identify studies eligible forthis meta-analysis, a computerized search was performed inPubMed and Ovid databases for all English-only journals

Surgery. Published by Elsevier B.V. All rights reserved.

by on August 11, 2010 ls.org

U. Benedetto et al. / European Journal of Cardio-thoracic Surgery 36 (2009) 611—615612

Table 1Selected studies of the literature.

Study (publication date), ref. Location Operative era No. DES/ALL/CABG Mean follow-up (months)

Javaid et al. (2007) [5] USA nr 979/1680/701 12Briguori et al. (2007) [6] Italy 2002—2004 69/218/149 12Park et al. (2008) [7] Korea 2003—2005 1547/3042/1495 31Hannan et al. (2008) [8] USA 2003—2004 9963/17400/7437 19Yang et al. (2007) [9] China 2003—2004 235/466/231 25Lee et al. (2007) [10] USA 2003—nr 102/205/103 12Yang et al. (2008) [11] Korea 2003—2005 441/831/390 12Varani et al. (2007) [12] Italy 2003—2005 111/206/95 12Tarantini et al. (2009) [13] Italy 2004—2005 93/220/127 24

nr: not reported; DES: drug-eluting stents; CABG: coronary artery bypass grafting;.

published until December 2008. All peer-reviewed retro-spective observational studies that dealt with trials compar-ing outcomes after DES-PCI versus CABG for treatment ofmultivessel coronary disease were identified and reviewed.The text string employed (formatted for PubMed) was:(multivessel coronary artery disease OR multivessel coronarydisease) AND (coronary artery bypass grafting OR surgicalrevascularization OR percutaneous coronary intervention ORdrug-eluting stent OR coronary artery stenting).

Hierarchical study end points were: the compositeoccurrence, during follow-up, of death or acute myocardialinfarction (AMI) or cerebrovascular accident (CVA); repeatrevascularization; and overall major adverse cardiac andcerebrovascular events (MACCE) including death, acutemyocardial infarction, cerebrovascular accident and repeatrevascularization. The outcome definition used by theoriginal researchers was accepted.

To minimize temporal bias, studies comparing DES-PCIresults to historical CABG cohorts were excluded. Whenseveral articles reported on the same patient material, onlythe most recent article was included. Two reviewers (UB andEA) abstracted the data independently.

3. Statistical analysis

The most appropriate effect index, to evaluate timerelated effects of CABG and PCI was the hazard ratio (HR) andits 95% confidence limits. We assumed CABG as the goldstandard and DES-PCI as the challenging strategy (HR <1,favors DES-PCI; HR >1, favors CABG). When only the graphedsurvival curves and the baseline sample sizes in thecomparison groups were provided, the method of Parmaret al. [3] was used to determine the desired hazard ratioestimate and its variance. Themethod ofWilliamson et al. [4]was used to obtain the hazard ratio estimates and variances ifthe number of persons at risk at each of several time pointsfor the graphed curves was available. Estimated ratios andtheir variances were used to construct summary 95%confidence intervals for parameters of interest for individualstudies as well as for the summary measure. Publication biaswas evaluated using Begg and Mazumdar rank correlationtest. The mixed model method was used to account for therandom variability between studies. Heterogeneity of thehazard ratios among the included studies was assessed by I2

and a value >50% was considered as indicative of hetero-geneity. To assess sources of heterogeneity, subgroupanalyses were performed (1): including only diabetic

ejcts.ctsnetjournDownloaded from

patients; (2) and comparing studies with shorter (12 months)versus longer follow-up (>12 months). Comprehensive Meta-Analysis, Version 2 (Borenstein and Rothstein) was used toanalyze abstracted data and to generate Forrest plots fordisplaying the results.

4. Results

A total of nine observational nonrandomized studies [5—13] were identified and analyzed including a total of 24,268patients with multivessel coronary disease who underwentDES-PCI (n = 13,540) and CABG (n = 10,728). They aresummarized in Table 1 according to time of publicationand geographical location of the study.

ERACI III [14] and ARTS II [15] studies were excluded fromthe analysis since the DES-PCI results were compared tohistorical CABG arms of ERACI II and ARTS I, respectively.

The total follow-up summed 479,524 months patients.Mean follow-up time was 20 months. The studies of Javaidet al. [5], Briguori et al. [6], Lee et al. [10], Yang et al. [11]and Varani et al. [12] have the shortest follow-up (12months), while those of Yang et al. [9] and Park et al. [7] havethe longest (25—31 months). All studies included madepossible the computation of the hazard ratio for theoutcomes of interest (Table 2).

Pooled analysis showed that DES-PCI and CABG werecomparable in terms of composite occurrence of death,acute myocardial infarction and cerebrovascular accidents(HR = 0.94; 95% CI = 0.72—1.22; p = 0.66; Fig. 1). However,there was a significantly higher risk of repeat revasculariza-tion in the DES-PCI (HR = 4.06; 95% CI = 2.64—6.24;p < 0.001; Fig. 2). Overall MACCE in the DES-PCI group washigher (HR = 1.86; 95% CI = 1.36—2.54; p < 0.001) due to anexcess of redo revascularization compared with CABG.

There was significant heterogeneity among studies fordeath/AMI/CVA (I2 = 69%), repeat revascularization (I2 = 91%)and MACCE (I2 = 91%). When analysis was restricted to diabeticpatients, the results did not significantly change from theoverall population with regard to the composite occurrence ofdeath/AMI/CVA (HR:1.23; 95%CI: 0.79—1.91;p = 0.34); repeatrevascularization (HR: 4.33; 95% CI: 2.68—6.98; p < 0.001) andoverall MACCE (HR: 2.03; 95% CI: 1.35—3.05; p = 0.001).

Pooling data from studies with shorter (12 months) versuslonger (>12 months) follow-up, the estimated HRs did notsignificantly differ from the overall population with regard tothe composite occurrence of death/AMI/CVA (HR: 1.21;p = 0.45 and HR: 0.82; p = 0.09, respectively), repeat

by on August 11, 2010 als.org

U. Benedetto et al. / European Journal of Cardio-thoracic Surgery 36 (2009) 611—615 613

Table 2Hazard ratio of DES-PCI versus CABG.

HR LCL UCL p value

Death/AMI/CVAPark et al. 0.70 0.53 0.91 0.009Hannan et al. 0.99 0.89 1.098 0.87Briguori et al. 1.03 0.53 1.99 0.92Yang et al. 0.70 0.33 1.49 0.36Lee et al. 1.29 0.68 2.46 0.42Yang et al. 0.75 0.43 1.31 0.32Javaid et al. 1.93 1.37 2.73 <0.001Varani et al. 0.47 0.16 1.37 0.169Tarantini et al. 0.56 0.20 1.56 0.27

Repeat revascularizationPark et al. 2.56 1.96 3.4 <0.001Hannan et al. 5.88 5.31 6.51 <0.001Briguori et al. 4.01 1.67 9.60 0.002Yang et al. 13.26 4.15 42.34 <0.001Lee et al. 6.73 2.06 21.95 0.002Yang et al. 4.26 1.78 10.15 0.001Javaid et al. 2.43 1.73 3.41 <0.001Varani et al. 5.91 1.39 25.6 0.016Tarantini et al. 1.91 0.62 5.83 0.25

MACCEPark et al. 1.37 1.16 1.63 <0.001Hannan et al. 2.89 2.72 3.08 <0.001Briguori et al. 1.48 0.91 2.43 0.11Yang et al. 3.09 1.80 5.30 <0.001Lee et al. 2.25 1.36 3.72 0.001Yang et al. 1.41 0.93 2.13 0.10Javaid et al. 2.44 1.87 3.19 <0.001Varani et al. 1.52 0.70 3.28 0.28Tarantini et al. 0.96 0.48 1.92 0.91

HR: hazard ratio; DES: drug-eluting stents; PCI: percutaneous coronary inter-vention; CABG: coronary artery bypass grafting; LCL: low 95% confidencelimits; ULC: upper 95% confidence limits; AMI: acute myocardial infarction;CVA: cerebrovascular accident; MACCE: major adverse cardiac and cerebro-vascular event.

Fig. 1. Pooled hazard ratio with related 95% confidence limits for the compo-site occurrence of death, acute myocardial infarction (AMI) and cerebrovas-cular accident (CVA). Squares indicate individual trial and lozenges indicatepooled summary effect estimate. The hazard ratios and its 95% confidenceintervals (CI) are displayed on a logarithmic scale.

Fig. 2. Pooled hazard ratio with related 95% confidence limits for repeatrevascularization. Squares indicate individual trial and lozenges indicatepooled summary effect estimate. The hazard ratios and its 95% confidenceintervals (CI) are displayed on a logarithmic scale.

ejcts.ctsnetjournaDownloaded from

revascularization (HR: 3.4; p < 0.001 and HR: 4.2; p < 0.001,respectively) and overall MACCE (HR: 2.1; p < 0.001 and HR:2.6; p < 0.001, respectively, Fig. 3).

There was no evidence of a significant publication bias fordeath/AMI/CVA ( p = 0.10); repeat revascularization( p = 0.20) and overall MACCE ( p = 0.50).

Fig. 3. Pooled hazard ratio with related 95% confidence limits for majoradverse cardiac and cerebrovascular events stratified for studies with shorter(12 months) versus longer (>12 months) follow-up (FU). Squares indicateindividual trial, black lozenges indicate pooled summary effect estimatesstratified for follow-up duration and white lozenges indicate pooled summaryeffect estimate. The hazard ratios and its 95% confidence intervals (CI) aredisplayed on a logarithmic scale.

by on August 11, 2010 ls.org

U. Benedetto et al. / European Journal of Cardio-thoracic Surgery 36 (2009) 611—615614

5. Discussion

The present meta-analysis showed that in the ‘real world’clinical practice DES-PCI continues to be associated with afour-fold increased risk of repeat revascularization whencompared to CABG in the treatment of MVD. Overall MACCErate was higher in the DES-PCI group due to the excess of redorevascularization compared with CABG.

A recent meta-analysis of randomized controlled trials [2]showed that PCI with stenting was associated with astatistically significant increase in subsequent revasculariza-tion (PCI or CABG) and major adverse cardiovascular eventsrelative to CABG.

However, these results have not been translated to recentclinical practice because of the use of bare metal stents inthe PCI arms. The advent of antiproliferative drug-elutingstents has dramatically reduced restenosis and repeatrevascularization thus improving the results of PCI. Whetherit is time to consider PCI with DES as the preferred strategyfor the treatment of MVD patients remains lack of definitiveclinical evidence.

Recent comparisons between DES-PCI results and histor-ical CABG cohorts [14,15] showed similar MACCE rates in DES-PCI and CABG treated patients, with a non-statisticallysignificant trend towards decreased overall safety outcomes(death, AMI and CVA) in the DES-PCI group even if CABGcontinued to be associated with a lower rate of repeatrevascularization. However these comparisons are affectedby temporal bias that may reduce the strength of theirconclusions.

Preliminary results from a single randomized controlledtrial [16] comparing DES-PCI versus CABG on a limited numberof MVD patients have been recently presented. No differ-ences were found in the composite occurrence of death, AMIand CVA in CABG between the two groups at 12 months butoverall MACCE in the DES-PCI group was significantly higherdue to an excess of redo revascularization compared withCABG. As frequently occurs in randomized controlled trials,this study may result in some degree of selectability as onlyabout 58% of patients screened for inclusion have actuallybeen enrolled. Therefore conclusions from ongoing rando-mized controlled studies must be confirmed and supported byresults coming from the more heterogeneous ‘real world’clinical practice.

The present meta-analysis, summarizing results fromobservational studies including a large ‘real world’ MVDpopulation, supports the conclusion that DES-PCI continuesto be associated with a significantly higher risk of overallMACCE due to an excess of redo revascularization comparedwith CABG repeat revascularization.

Our findings need to be confirmed in ongoing, largerandomized clinical trials. Finally, the present results mustbe interpreted with some caution since the design of studiesincluded lack random allocation to DES-PCI or CABG.

References

[1] Bravata DM, Gienger AL, McDonald KM, Sundaram V, Perez MV, Varghese R,Kapoor JR, Ardehali R, Owens DK, Hlatky MA. Systematic review: thecomparative effectiveness of percutaneous coronary interventions andcoronary artery bypass graft surgery. Ann Intern Med 2007;147:703—16.

ejcts.ctsnetjournDownloaded from

[2] Takagi H, Kawai N, Umemoto T. Meta-analysis of four randomized con-trolled trials on long-term outcomes of coronary artery bypass graftingversus percutaneous coronary intervention with stenting for multivesselcoronary artery disease. Am J Cardiol 2008;101:1259—62.

[3] Parmar MKB, Torri V, Stewart L. Extracting summary statistics to performmeta-analyses of the published literature for survival endpoints. StatistMed 1998;17:2815—34.

[4] Williamson PR, Smith CT, Hutton JL, Marson AG. Aggregate data meta-analysis with time-to-event outcomes. Statist Med 2002;21:3337—51.

[5] Javaid A, Steinberg DH, Buch AN, Corso PJ, Boyce SW, Pinto Slottow TL,Roy PK, Hill P, Okabe T, Torguson R, Smith KA, Xue Z, Gevorkian N, SuddathWO, Kent KM, Satler LF, Pichard AD, Waksman R. Outcomes of coronaryartery bypass grafting versus percutaneous coronary intervention withdrug-eluting stents for patients with multivessel coronary artery disease.Circulation 2007;116:I200—6.

[6] Briguori C, Condorelli G, Airoldi F, Focaccio A, D’Andrea D, Cannavale M,Abarghouei AA, Giordano S, De Vivo F, Ricciardelli B, Colombo A. Com-parison of coronary drug-eluting stents versus coronary artery bypassgrafting in patients with diabetes mellitus. Am J Cardiol 2007;99:779—84.

[7] Park DW, Yun SC, Lee SW, Kim YH, Lee CW, Hong MK, Kim JJ, Choo SJ, SongH, Chung CH, Lee JW, Park SW, Park SJ. Long-term mortality afterpercutaneous coronary intervention with drug-eluting stent implantationversus coronary artery bypass surgery for the treatment of multivesselcoronary artery disease. Circulation 2008;117:2079—86.

[8] Hannan EL, Wu C, Walford G, Culliford AT, Gold JP, Smith CR, Higgins RS,Carlson RE, Jones RH. Drug-eluting stents vs. coronary-artery bypassgrafting in multivessel coronary disease. N Engl J Med 2008;358:331—41.

[9] Yang ZK, Shen WF, Zhang RY, Kong Y, Zhang JS, Hu J, Zhang Q, Ding FH.Coronary artery bypass surgery versus percutaneous coronary interven-tion with drug-eluting stent implantation in patients with multivesselcoronary disease. J Interv Cardiol 2007;20:10—6.

[10] Lee MS, Jamal F, Kedia G, Chang G, Kapoor N, Forrester J, Czer L, ZimmerR, DeRobertis M, Trento A, Makkar RR. Comparison of bypass surgery withdrug-eluting stents for diabetic patients with multivessel disease. Int JCardiol 2007;123:34—42.

[11] Yang JH, Gwon HC, Cho SJ, Hahn JY, Choi JH, Choi SH, Lee YT, Lee SH,Hong KP, Park JE. Comparison of coronary artery bypass grafting withdrug-eluting stent implantation for the treatment of multivessel coronaryartery disease. Ann Thorac Surg 2008;85:65—70.

[12] Varani E, Balducelli M, Vecchi G, Aquilina M, Maresta A. Comparison ofmultiple drug-eluting stent percutaneous coronary intervention andsurgical revascularization in patients with multivessel coronary arterydisease: one-year clinical results and total treatment costs. J InvasiveCardiol 2007;19:469—75.

[13] Tarantini G, Ramondo A, Napodano M, Favaretto E, Gardin A, Bilato C,Nesseris G, Tarzia V, Cademartiri F, Gerosa G, Iliceto S. PCI versus CABGfor multivessel coronary disease in diabetics. Catheter Cardiovasc Interv2009;73:50—8.

[14] Rodriguez AE, Maree AO, Mieres J, Berrocal D, Grinfeld L, Fernandez-Pereira C, Curotto V, Rodriguez-Granillo A, O’Neill W, Palacios IF. Late lossof early benefit from drug-eluting stents when compared with bare-metalstents and coronary artery bypass surgery: 3 years follow-up of the ERACIIII registry. Eur Heart J 2007;28:2118—25.

[15] Daemen J, Kuck KH, Macaya C, LeGrand V, Vrolix M, Carrie D, Sheiban I,Suttorp MJ, Vranckx P, Rademaker T, Goedhart D, Schuijer M, Wittebols K,Macours N, Stoll HP, Serruys PW, ARTS-II Investigators. Multivessel cor-onary revascularization in patients with and without diabetes mellitus: 3-year follow-up of the ARTS-II (Arterial Revascularization Therapies Study-Part II) trial. J Am Coll Cardiol 2008;52:1957—67.

[16] Ong AT, Serruys PW, Mohr FW, Morice MC, Kappetein AP, Holmes Jr DR,Mack MJ, van den Brand M, Morel MA, van Es GA, Kleijne J, Koglin J,Russell ME. The SYNergy between percutaneous coronary interventionwith TAXus and cardiac surgery (SYNTAX) study: design, rationale, andrun-in phase. Am Heart J 2006;151:1194—204.

Appendix A. Conference discussion

Dr J. Takkenberg (Rotterdam, The Netherlands): I have a few remarksabout your methods and I have two questions.

A systematic review with a meta-analysis is not an easy task. I can tell itfrom my own experience. It requires careful planning, precise selection ofevidence, and thorough advanced statistical analysis. It is sometimes calledthe work of a monk, referring to the Middle Ages when monks would spend

by on August 11, 2010 als.org

U. Benedetto et al. / European Journal of Cardio-thoracic Surgery 36 (2009) 611—615 615

years rewriting one book. I think your book is not yet completely rewritten. Inparticular, your hypothesis should be clearly stated, as well as your inclusioncriteria. For example, there seem to be two studies in your systematic reviewwith only diabetes patients. So there is some selection bias, for sure. Astraightforward plan for data analysis needs to be added, and your searchneeds to be more thorough and include databases like EMBASE and theCochrane database, and, of course, screening of recent meeting abstracts.Your results should include the reasons for excluding papers that wereretrieved in the initial data search. Anybody who attempts to replicate yourstudy should be able to come up with the same results. And that is not easy, Irealize that, but I am sure you are going to manage.

A systematic review like yours may or may not contain a meta-analysis. It isactually quite debatable. A meta-analysis is the systematic review’s statisticalanalysis of the results from independent studies, and it generally aims toproduce a single estimate of a treatment effect, but this is usually done usingrandomized controlled trials. In recent years, systematic reviews with meta-analysis of observational data like yours have becomemore common. However,observational studies are much more likely to be biased, and performing ameta-analysis on this type of data should not be primarily done to obtain apooled point estimate, but, more importantly, to assess carefully the possiblesources of heterogeneity by means of subgroup analysis and meta-regression.Well, there obviously is a lot of heterogeneity in your meta-analysis and thatrequires further exploration, and that actually leads to my first question.

You explored the obvious potential sources of heterogeneity such aspatient age, left main disease, and diabetes. But also factors like, for example,the proportion of off-pump CABG and the post-PCI clopidogrel may be sourcesof clinical heterogeneity, and my question would be, did you investigate these?

My second question concerns the final sentence in your conclusion. Yourefer to the randomized trials that are underway on this subject. Do you reallyneed to confirm your findings with these trials? I wonder. Although randomizedtrials are considered to represent the highest quality of research, they areusually done in highly selected patients who may not represent your averagecoronary patient population. For example, last month Michael Mack and

ejcts.ctsnetjournaDownloaded from

colleagues published an excellent prospective registry of patients whounderwent CABG and PCI. You may want to add that to your systematic review.They found that, compared to randomized controlled trial results, theiroutcomes were inferior. They were obviously dealing with a differentpopulation. So do you think randomized controlled trials are necessary toconfirm your findings?

Dr Benedetto: For the first question, yes, we found a high variabilityamong studies included, and we tried to address this issue performing a meta-regression analysis, including study population characteristics such as age,female sex, diabetes, left main disease. However, we could not perform ameta-regression including the off-pump procedure because only four studiesreported the number of off-pump procedures performed, and we know that toperform ameta-regression with a low number of cases is not regular. However, Ithink that clinical practice showed us that off-pump techniques do not reducethe incidence of death, of a hard clinical end point. So I think that off-pumpprocedures may not change the results in comparison of CABG and PCI withDESs.

A different thinking is for clopidogrel discontinuation since clinicalpractice shows us that a longer period of antiplatelet therapy following DESimplantation has better results. However, in this meta-analysis only threepapers reported the policy for clopidogrel discontinuation. So we did not showthe meta-regression analysis for this issue because it did not reach statisticalpower.

For the last question, I think that randomized controlled trials have low tomoderate power to detect significant differences in a hard end point such asmortality. In addition, as you said, they had selected patients, but weextended the results to the population as a whole, and this may be a limit. Ofnote, 70% of this usage is now off-label without any evidence from randomizedcontrolled trials. And so I think to gain additional information comparing CABGwith PCI with DES or bare-metal stents, we need randomized controlled trialsbut also registry, observational registry, which often may reach a differentconclusion, and the tools of meta-analysis for an observational registry may beuseful to summarize data from registries.

by on August 11, 2010 ls.org

DOI: 10.1016/j.ejcts.2009.03.012 2009;36:611-615 Eur J Cardiothorac Surg

Roscitano, Brenno Fiorani, Gian Domenico Di Nucci and Riccardo Sinatra Umberto Benedetto, Giovanni Melina, Emiliano Angeloni, Simone Refice, Antonino

disease. A meta-analysis on 24,268 patientsCoronary artery bypass grafting versus drug-eluting stents in multivessel coronary

This information is current as of August 11, 2010

& ServicesUpdated Information

http://ejcts.ctsnetjournals.org/cgi/content/full/36/4/611including high-resolution figures, can be found at:

References http://ejcts.ctsnetjournals.org/cgi/content/full/36/4/611#BIBL

This article cites 16 articles, 6 of which you can access for free at:

Citations

shttp://ejcts.ctsnetjournals.org/cgi/content/full/36/4/611#otherarticleThis article has been cited by 1 HighWire-hosted articles:

Subspecialty Collections

http://ejcts.ctsnetjournals.org/cgi/collection/coronary_disease Coronary disease

following collection(s): This article, along with others on similar topics, appears in the

Permissions & Licensing

http://ejcts.ctsnetjournals.org/misc/Permissions.shtmlor in its entirety can be found online at: Information about reproducing this article in parts (figures, tables)

Reprints http://ejcts.ctsnetjournals.org/misc/reprints.shtml

Information about ordering reprints can be found online:

by on August 11, 2010 ejcts.ctsnetjournals.orgDownloaded from