A collaborative systematic review and meta-analysison 1278 patients undergoing percutaneousdrug-eluting stenting for unprotected leftmain coronary artery diseaseGiuseppe G.L. Biondi-Zoccai, MD,a,o Marzia Lotrionte, MD,b,o Claudio Moretti, MD,a Emanuele Meliga, MD,a

Pierfrancesco Agostoni, MD,c Marco Valgimigli, MD, PhD,d,e Angela Migliorini, MD,f David Antoniucci, MD,f

Didier Carrie, MD,g Giuseppe Sangiorgi, MD,h,i Alaide Chieffo, MD,h,i Antonio Colombo, MD,h,i

Matthew J. Price, MD,j Paul S. Teirstein, MD,j Evald H. Christiansen, MD,k Antonio Abbate, MD,l Luca Testa, MD,b

Julian P.G. Gunn, MD,m Francesco Burzotta, MD,b Antonio Laudito, MD,n Gian Paolo Trevi, MD,a andImad Sheiban, MDa Turin, Rome, Ferrara, Gussago, Florence, and Milan, Italy; Antwerp, Belgium; Toulouse,France; La Jolla, CA; Aarhus, Denmark; Richmond, VA; and Sheffield, United Kingdom

Background Cardiac surgery is the standard treatment for unprotected left main disease (ULM). Drug-eluting stent (DES)implantation has been recently reported in patients with ULM but with unclear results. We systematically reviewed outcomes ofpercutaneous DES implantation in ULM.

Methods Several databases were searched for clinical studies reporting on ≥20 patients and ≥6-month follow-up. Theprimary end point was major adverse cardiovascular events (MACEs; ie, death, myocardial infarction, or target vesselrevascularization [TVR]) at the longest follow-up. Incidence and adjusted risk estimates were pooled with generic inversevariance random-effect methods (95% CIs).

Results From 823 initial citations, 16 studies were included (1278 patients, median follow-up 10 months). Eight wereuncontrolled registries, 5 nonrandomized comparisons between DES and bare-metal stents and 3 nonrandomizedcomparisons between DES and CABG, with no properly randomized trial. Meta-analysis for DES-based PCI showed, at thelongest follow-up, rates of 16.5% (11.7%-21.3%) MACE, 5.5% (3.4%-7.7%) death, and 6.5% (3.7%-9.2%) TVR. Comparisonof DES versus bare-metal stent disclosed adjusted odds ratios for MACE of 0.34 (0.16-0.71), and DES versus CABG showedadjusted odds ratios for MACE plus stroke of 0.46 (0.24-0.90). Meta-regression showed that disease location predictedMACE (P = .001) and TVR (P = .020), whereas high-risk features predicted death (P = .027).

Conclusions Clinical studies report apparently favorable early and midterm results in selected patients with ULM.However, given their limitations in validity and the inherent risk for DES thrombosis, results from randomized trials are stillneeded to definitely establish the role of DES implantation instead of the reference treatment, surgery. (Am Heart J2008;155:274-83.)

Coronary artery bypass grafting (CABG) has beenconsidered the standard treatment for significant unpro-tected left main coronary disease (ULM) since the late

From the aInterventional Cardiology, Division of Cardiology, University of Turin, Turin, ItalybInstitute of Cardiology, Catholic University, Rome, Italy, cAntwerp Cardiovascular InstitutMiddelheim, AZ Middelheim, Antwerp, Belgium, dInstitute of Cardiology, University oFerrara, Ferrara, Italy, eCardiovascular Research Center, Salvatore Maugeri FoundationIRCCS, Gussago, Italy, fDivision of Cardiology, Careggi Hospital, Florence, ItalygCardiology Department, Rangueil Hospital, Toulouse, France, hEMO Centro CuorColumbus, Milan, Italy, iInterventional Cardiology Unit, St Raffaele Hospital, Milan, ItalyjDivision of Cardiovascular Disease, Scripps Clinic, La Jolla, CA, kDepartment of CardiologyAarhus University Hospital (Skejby), Aarhus, Denmark, lVirginia Commonwealth UniversitPauley Heart Center, Medical College of Virginia Campus, Virginia CommonwealtUniversity, Richmond, VA, mCardiovascular Research Unit, University of Sheffield, SheffieldUnited Kingdom, and nDivision of Cardiac Surgery, University of Turin, Turin, Italy.

oBoth authors equally contributed to this work.Dr Biondi-Zoccai consulted for Boston Scientific, Boston, MA; Cordis, Miami, FL; andMediolanum Cardio Research, Milan, Italy; and received lecture fees from Bristol-MyersSquibb, New York, NY. Dr Agostoni has received lecture fees from Cordis.Submitted April 13, 2007; accepted October 1, 2007.Reprint requests: Giuseppe G.L. Biondi-Zoccai, MD, Division of Cardiology, Universitof Turin, S. Giovanni Battista-Molinette Hospital, Corso Bramante 88-90, 10126 TurinItaly.E-mail: gbiondizoccai@gmail.com0002-8703/$ - see front matter© 2008, Mosby, Inc. All rights reserved.

,ef,,e,,yh,

1970s. Percutaneous coronary intervention (PCI) for ULMwas attempted as early as 1977 with balloon angioplasty,but the risk of failure was soon recognized.1 Bare-metalstents (BMSs) later provided important early and midterm

doi:10.1016/j.ahj.2007.10.009

-

y,

Figure 1

Reviewing process.

Table I. Included studies

Study design andfirst author/study name Country

Patientstreatedwith DES

Observational cohorts on DESAgostoni et al9 Netherlands 58de Lezo et al15 Spain 52Dudek et al16 Poland 28KOMATE18 Korea 54Lozano et al20 Spain 42Migliorini et al21 ⁎ Italy 156Price et al23 United States 50Wood et al25 ⁎ United States 100

Nonrandomized studies of DES versus BMSCarrié et al12 ⁎ France 120Chieffo et al10† Italy 85Christiansen et al14 ⁎ Denmark 46Han et al17 China 138Park et al11 Korea 102Sheiban et al24 ⁎ Italy 85

Nonrandomized studies of PCI vs CABGChieffo et al13† Italy 107Lee et al19 United States 50Palmerini et al22 Italy 94

KOMATE, KOrean Multicenter Angoplasty TEam.⁎Updated after direct communication with the original investigators.†Duplicate publication likely or acknowledged.

Biondi-Zoccai et al 275American Heart JournalVolume 155, Number 2

clinical benefits to the overall group of patients under-going PCI by reducing periprocedural risks (eg, the needfor emergency CABG), restenosis, and target lesionfailure; however, the impact of BMS on PCI for ULM hasbeen limited.2 Drug-eluting stents (DESs) have beenintroduced into clinical practice in the last few years andproved able to reduce the risk of restenosis.3 Thus, theyhold the promise of improving the outlook of PCI forULM. Several reports have indeed been published on theshort- and midterm results of DES-based PCI for ULM, butmost were limited by retrospective nature, small size,and/or single-center design. To thoroughly appraise therisk-benefit balance of percutaneous DES implantation forULM, in itself and in comparison with BMS-based PCI andCABG, we performed a systematic review and meta-analysis of pertinent studies.

MethodsThis review was reported according to established methods,4

minimizing duplication risks.5,6 BioMedCentral, clinicaltrials.gov, Google Scholar, and PubMed were searched (January 2000-September 2006),6 without language restrictions. Citations werescreened at title/abstract level and retrieved as full reports.Studies were included if the following criteria applied: (a) PCIwith stent implantation, (b) for ULM disease, (c) using DES, and(d) in ≥20 patients. Exclusion criteria were as follows: (a)duplicate publication, (b) ongoing/unpublished study, (c)publication only as an abstract or as conference proceedings, (d)inclusion of patients not undergoing PCI of the ULM, and (e) lackof ≥6-month follow-up.

The primary end point was major adverse cardiovascularevents (MACEs), defined as composite of death, nonfatalmyocardial infarction, or target vessel revascularization (TVR).The coprimary end point was major adverse cerebrocardiovas-cular events (MACCEs). Secondary end points were individualcomponents of MACE. Other details on patients, procedures, andstudy designs were abstracted,6 and internal validity of includedstudies was appraised.7

All principal investigators of included studies wererepeatedly contacted.Continuous variables are reported as mean (SD) or median

(range) and categorical variables as n (%). Poolingwas performedaccording to random-effectmodels with generic inverse varianceweighting with RevMan 4.2 (The Nordic Cochrane Center,Købehvn, Denmark) for computing incidence estimates (95%CIs). For nonrandomized controlled comparisons, adjusted riskestimates were pooled after logarithmic transformation accord-ing to random-effect models with generic inverse varianceweighting. Subgroup,6 sensitivity, and meta-regression analyses8

were performed to appraise robustness and explore moderators(SPSS 11.0; SPSS, Chicago, IL). Statistical heterogeneity wasappraised with I

2 and small study bias with funnel plots andregression analysis. Unadjusted P values are reported through-out, with hypothesis testing set at the 2-tailed .05 level.

ResultsFrom 823 initial citations (Figure 1), we excluded 806

hits, leaving 17 studies,9-25 (Tables I and II). The 17included studies reported on a total of 16 nonduplicatecohorts, enrolling 1278 subjects undergoing PCI withDES for ULM disease. Eight studies were observationalreports on 540 patients treated with

Table II. Major excluded studies

Study Country Patients Design Reason for exclusion

Arampatzis et al (2003) Netherlands 31 Observational cohort Duplicate publicationArampatzis et al (2004) Netherlands 16 Observational cohort Duplicate publicationBerenguer et al (2005) Spain 7 Observational cohort Duplicate publicationCARDIA (2006) UK 600 RCT of PCI vs CABG OngoingCOMBAT Worldwide 1800 RCT of PCI vs CABG On holdeCYPHER (2006) Worldwide 171 Observational cohort UnpublishedErglis et al Europe 103 RCT of DES vs BMS UnpublishedEuropean Registry Europe 224 Observational cohort UnpublishedFrench Multicenter Taxus Study France 150 Observational cohort OngoingHerz et al (2005) Israel 4 Non-RCT of PCI vs CABG b20 Patients includedKim et al (2006) Korea 116 Observational cohort Duplicate publicationKorean Randomized Study Korea 124 RCT of PCI vs CABG OngoingLE MANS (2005) Poland 37 Observational cohort UnpublishedLefevre et al France 146 Observational cohort UnpublishedLeipzig Study Germany 200 RCT of PCI vs CABG OngoingLopez-Palop et al (2004) Spain 10 Observational cohort b20 Patients includedMunich Study Germany 340 RCT of CYPHER vs Taxus OngoingPalmerini et al (2005) Italy 42 Observational cohort Duplicate publicationPeszek-Przybyla et al (2006) Poland 62 Observational cohort b20 Patients includedPohl et al (2004) Germany 23 RCT of PCI vs CABG BMS use onlyREVASCULARIZE United States NA RCT of PCI vs CABG OngoingSECURE United States 20 Observational cohort UnpublishedSYNTAX Worldwide 1500 RCT of PCI vs CABG OngoingTeplitsky et al (2004) Israel 11 Observational cohort b20 Patients includedTRUE Europe 115 Observational cohort UnpublishedValgimigli et al (2005) Netherlands 80 Before-after study Duplicate publicationValgimigli et al (2006) Netherlands 110 Observational cohort Duplicate publication

References are available from the corresponding author upon request. NA, Not available or applicable; RCT, randomized clinical trial.

276 Biondi-Zoccai et alAmerican Heart Journal

February 2008

DES,9,15,16,18,20,21,23,25 6 studies were nonrandomizedcomparisons of PCI with DES (n = 576) versus BMS (n =509),10-12,14,17,24 and 3 were nonrandomized comparisonsof PCI with DES (n = 251) versus CABG (n = 419).13,22,23

Median follow-up was 10 months (range 6-19 months).The overall internal validity of all included studies wasonly moderate, given the presence of several potentialthreats to study quality. Details on included studies areavailable in Tables III and IV.All studies provided enough details on the primary end

point (ie, MACE or MACCE); however, incompletereporting of other events was common (Table V).Overall analysis (Figures 2 and 3) showed inhospital deathin 2.3% (1.1-3.4, I2 = 7.4%) and inhospital myocardialinfarction in 2.5% (1.2-3.8, I2 = 27%). Overall midtermfollow-up showed MACE in 16.5% (11.7-21.3, I2 = 84%),death in 5.5% (3.4-7.7, I2 = 64%), and TVR in 6.5%(3.7-9.2, I2 = 80%). Comparing MACE in studies limitingfollow-up to 6 months14,18,19,21,22 versus those extendingfollow-up to N6 months9-13,15-17,20,23-25 showed rates of,respectively, 14.2% (5.4-23.0, I2 = 85%) and 17.7%(11.7-23.8, I2 = 86%).Multivariable adjusted estimates for DES vs BMS were

reported only by a handful of studies,10,12,24 for a total of206 patients treated with DES versus 190 with BMS(Figure 4). Pooled analysis showed the superiority of DESboth in follow-up MACE and TVR (respectively, odds ratio

[OR] = 0.34 [0.16-0.71], P = .004, I2 = 45.3%, and OR =0.34 [0.12-0.94], P = .04, I2 = 0%).Adjusted estimates for the DES-based PCI versus

CABG were reported only by Chieffo et al13 and Leeet al,19 for a total of 157 patients treated with DES and265 with CABG (Figure 4). Meta-analysis for theoccurrence of MACCE at follow-up suggested the super-iority of DES versus CABG (OR = 0.46 [0.24-0.90],P = .02, I2 = 0%), even if this estimate should be viewedwith caution, given the small number of patientsincluded, the width of the ICs, and the caveats on theinternal validity of included studies.Subgroup analyses were performed for (a) DES

implantation for nonbifurcational ULM, (b) DESimplantation for ULM in low-risk patients, and(c) DES implantation for ULM in high-risk subjects.Given the limited details provided by the individualstudies on these subgroups despite repeated efforts tocontact the primary authors, computations performedon the available data are hereby reported as exploratoryfindings only.Including a recent update from a multicenter registry

and excluding duplicate sources, a total of 285 patientswith nonbifurcational ULM and treated with DES haddetailed follow-up data.9,11,14,15,21,24-26 Pooled analysisshowed inhospital death in 0.9% (0%-2.1%, I2 = 0%) andmyocardial infarction in 3.2% (0%-5.6%, I2 = 0%),

Table III. Characteristics of included studies

StudyAge(y)

Male(%)

Diabetic(%)

ACS(%)

NonbifurcationalULM (%)

Surgicalhigh-risk

features(%) ⁎

LVEF(%)

COPD(%)

RF(%)

Angiographicfollow-up

(%)

Oralantiplateletregimen †

SES/PESuse(%)

Agostoniet al9

63 ± 13 69 33 32 45 NA 48 ± 12 NA NA NA A + C,6 m

NA

Carriéet al12

66 ± 12 75 29 63 0 NA 59 ± 11 NA NA 100 A + C,6m

0/100

Chieffoet al10

63 ± 12 84 21 31 19 45 51 ± 11 NA NA 85 A + C,≥6 m

NA

Chieffoet al13

64 ± 10 NA 19 32 19 32 52 ± 10 NA 6 85 NA 51/49

Christiansenet al14

NA NA NA NA 46 43 NA NA NA NA A + C,≥3 m

NA

de Lezoet al15

63 ± 11 42 35 83 58 NA 57 ± 13 NA NA 67 A + C,12 m

100/0

Dudeket al16

NA NA NA NA NA NA NA NA NA NA A + C,6-12 m

46/54

Hanet al17‡

62 ± 11 NA 29 45 29 NA NA NA NA 36 A + C,6-9 m

NA

KOMATE18 59 ± 9 68 27 65 67 NA 60 ± 18 NA 4 44 A + C,6 m

65/35

Lee et al19 72 ± 15 50 36 66 40 64 51 ± 15 NA 16 42 A + C,6 m

84/16

Lozanoet al20

70 ± 11 60 33 17 31 100 37‡ NA NA 57 A + C,3-6 m

71/19

Miglioriniet al21

70 ± 10 80 32 69 15 61 27 § NA 27 84 A + C,6 m

26/74

Palmeriniet al22

73 ± 11 70 26 63 20 64 52 ± 14 16 20 66 NA 68/32

Park et al11 60 ± 11 75 28 60 29 NA 60 ± 8 NA NA 84 A + C,6 m ‖

100/0

Price et al23 69 ± 13 64 26 34 6 58 24 ¶ NA 16 98 A + C,indefinitely

100/0

Sheiban24 68 ± 10 77 22 67 40 46 55 ± 10 NA 4 61 A + C orA + T,6 m

100/0

Woodet al25

68 ± 13 64 30 NA 31 NA 47 ± 13 NA NA NA A + C,≥6 m

NA

ACS, Acute coronary syndrome; LVEF, left ventricular ejection fraction; COPD, chronic pulmonary obstructive disease; RF, renal failure; SES, sirolimus-eluting stent (of all DESs); PES,paclitaxel-eluting stent (of all DESs).⁎Proportion with LVEF ≤40%.†A stands for aspirin, C for clopidogrel, and T for ticlopidine, followed by duration in months of dual antiplatelet therapy.‡ Patients with ejection fraction (LVEF) ≤45%.§ Proportion with LVEF ≤35%.‖Cilostazol was also administered for 1 month after the procedure.¶Defined as EuroSCORE ≥6 or Parsonnet score ≥15.

Biondi-Zoccai et al 277American Heart JournalVolume 155, Number 2

whereas after a median follow-up of 10 months, MACEoccurred in 14.7% (6.2%-23.2%, I2 = 75%), death in 4.1%(1.7%-6.6%, I2 = 0%), and TVR in 6.7% (0.9%-12.4%,I2 = 64%). Available studies reported on a total of 260patients with surgical low-risk features (European Systemfor Cardiac Operative Risk Evaluation b6 or Parsonnetscore b15).10,14,21,24 Meta-analysis showed inhospitaldeath in 3.0% (0%-6.1%, I2 = 0%) and myocardialinfarction in 3.0% (0.1%-6.0%, I2 = 0%), whereas after 8months, MACE occurred in 15.7% (7.7%-23.7%, I2 = 74%),death in 4.8% (1.6%-8.0%, I2 = 0%), and TVR in 8.5% (1.9%-15.1%, I2 = 77%). Conversely, detailed event rates were

available for 312 patients with surgical high-risk features(European System for Cardiac Operative Risk Evaluation≥6 or Parsonnet score ≥15).10,14,20,21,24 Random-effectquantitative pooling yielded inhospital death in 6.6%(3.6%-9.7%, I2 = 0%) and myocardial infarction in 1.3%(0%-3.0%, I2 = 0%), whereas after 8 months, MACEoccurred in 20.6% (11.9%-29.2%, I2 = 75%), death in12.0% (7.5%-16.6%, I2 = 26%), and TVR in 6.4%(0.7%-12.1%, I2 = 77%).Weighted least squares meta-regression disclosed a

highly statistically significant interaction between therate of nonbifurcational ULM and the risk of MACE

Table IV. Internal validity of included studies

StudyProspective

designMulticenterenrolment

Selectionbias

Performancebias

Attritionbias

Detectionbias

Multivariable adjustmentfor potential confounders

Agostoni et al9 No No B B D B Probably adequateCarrié et al12 Yes No B B A B Probably adequateChieffo et al10 No Yes B B A B Probably adequateChieffo et al13 No No B B A B Probably adequateChristiansen et al14 Yes No C B D B None reportedde Lezo et al15 No Yes B B A B None reportedDudek et al16 Unclear No B B D B None reportedHan et al17 No No C B C B None reportedKOMATE18 Unclear Yes B B C C None reportedLee et al19 No No B B D B Probably adequateLozano et al20 Yes Yes B B B C None reportedMigliorini et al21 No No B B B B Probably adequatePalmerini et al22 Yes No B B C C None reportedPark et al11 No No B B A C None reportedPrice et al23 No No B B A B Probably adequateSheiban et al24 No No B B A B Probably adequateWood et al25 No No B B A B None reported

Risk of bias is expressed as A (low), B (moderate), C (high), or D (incomplete reporting).

Table V. Unadjusted clinical outcomes

Study NInhospitaldeath (%)

InhospitalMI (%)

Follow-up(m)

Follow-upcompletion

(%)MACE(%)

MACCE(%)

Death(%)

MI(%)

Stroke(%)

TVR(%)

ST(%)

Agostoni et al9 58 2 3 14 100 16 NA 5 3 NA 7 NACarrié et al12–BMS arm ⁎ 57 0 5 10 100 66 NA 7 5 NA 26 0Carrié et al12–DES arm ⁎ 49 4 3 10 100 13 13 10 3 0 2 1Chieffo et al10–BMS arm 64 0 8 12 100 42 NA 14 NA NA 30 0Chieffo et al10–DES arm 85 0 6 12 100 25 NA 4 NA NA 19 1Chieffo et al13–CABG arm 142 2 26 12 100 42 44 8 27 2 6 NAChieffo et al13–DES arm 107 0 9 12 100 33 34 3 10 1 20 1Christiansen et al14–BMS arm ⁎ 39 31 3 6 100 44 44 31 5 0 8 NAChristiansen et al14–DES arm ⁎ 42 2 0 6 100 7 7 5 2 0 5 2de Lezo et al15 52 0 4 12 100 6 NA 0 4 NA 2 0Dudek et al16 36 NA NA 9 NA 14 NA NA NA NA 0 NAHan et al17–BMS arm 34 NA NA 12 NA 27 NA 9 3 NA 15 NAHan et al17–DES arm 138 NA NA 12 NA 11 NA 7 1 NA 10 NAKOMATE18 54 2 0 6 81 5 NA 2 0 NA 2 0Lee et al19–CABG arm 123 5 2 6 NA 14 17 11 2 8 1 NALee et al19–DES arm 50 2 0 6 NA 11 11 4 4 0 7 0Lozano et al20 42 10 0 10 100 26 NA 20 4 NA 2 1Migliorini et al21 ⁎ 156 7 1 6 100 24 24 11 1 0 12 0Palmerini et al22–CABG arm 154 NA NA 6 NA 14 NA 11 13 NA NA NAPalmerini et al22–DES arm 94 NA NA 6 NA 22 NA 9 13 NA NA NAPark et al11–BMS arm 121 0 8 12 100 25 NA 0 8 NA 17 0Park et al11–DES arm 102 0 7 12 100 9 NA 0 7 NA 2 0Price et al23 50 0 8 8 100 54 NA 10 10 NA 44 4Sheiban et al24–BMS arm ⁎ 69 10 3 10 99 36 36 20 3 0 13 0Sheiban et al24–DES arm ⁎ 77 3 3 10 100 9 10 3 3 1 4 0Wood et al25 ⁎ 100 2 3 19 100 19 NA 6 3 NA 7 1

MI, Myocardial infarction; ST, stent thrombosis.⁎Updated after communication with investigators.

278 Biondi-Zoccai et alAmerican Heart Journal

February 2008

(β = −0.015 [−0.008 to −0.022], P = .001) (Figure 5).Intriguingly, location of ULM disease could explain asmuch as 58% of the observed variability (R2 = 0.61,

adjusted R2 = 0.58). The prevalence of high-risk

features proved to be a statistically significant predictorof midterm follow-up case fatality (β = 0.017 [0.003-

Figure 2

Overall risk of inhospital death (A) and myocardial infarction (B) in patients treated with DESs for ULM.

Biondi-Zoccai et al 279American Heart JournalVolume 155, Number 2

0.031], P = .027, R2 = 0.58, adjusted R2 = 0.51) but

not of midterm MACE (P = .55). On the other hand,TVR was significantly associated with the rate ofnonbifurcational ULM (β = −0.016 [−0.003 to −0.030],P = .020). Angiographic follow-up rate, despite being acommon cause of event inflation in coronaryintervention studies,27 was not significantly associatedwith MACE or TVR.Sensitivity analyses conducted with fixed-effect, and

DerSimonian-Laird methods confirmed the findings of the

original analyses. Testing for small study bias showed nostatistical significance (P = .63).

DiscussionThe present meta-analysis, reporting on DES implanta-

tion for ULM, has the following implications: publishedreports on DES use now include N1200 subjects,confirming that this treatment option has been alreadyadopted into clinical practice despite the lack of a sound

Figure 3

Overall risk of MACEs (A), death (B), and TVR (C) at a median follow-up of 10 months (range 6-19 months) in patients treated with DESs for ULM.

280 Biondi-Zoccai et alAmerican Heart Journal

February 2008

evidence base; although overall results suggest anapparently favorable risk of MACE in selected patients,most studies are fraught with validity threats; moreover,

clinical follow-up is to date still limited at the midtermthreshold; thus, results from randomized trials comparingpercutaneous versus surgical revascularization are still

Figure 3 (continued)

Figure 4

Forest plots for the adjusted comparisons between DESs versus BMSs (A) or DES versus bypass surgery (CABG) in patients with ULM (B).

Biondi-Zoccai et al 281American Heart JournalVolume 155, Number 2

Figure 5

L'Abbé plots for the association between prevalence of nonbifurcational ULM and rate of midterm follow-up MACEs) (A); prevalence ohigh-risk features and midterm follow-up death (B); and prevalenceof high-risk features and midterm TVR (C).

282 Biondi-Zoccai et alAmerican Heart Journal

February 2008

-f

needed to establish the role of percutaneous DESimplantation in the evidence-based medicine hierarchy incomparison with the reference treatment, surgery,especially given the recent concerns raised on late stentthrombosis.28

Unprotected left main disease occurs in 5% of patientsundergoing coronary angiography29 and has been, fordecades, a strict indication for CABG. Since DESapproval,3 interventionists have begun to use them notonly in patients with ULM at very high surgical risk butalso in those with more favorable characteristics.30 Mostreports were, however, small single-center experiences,with obvious limitations in external validity. The availableevidence on patient selection and risk stratification inpatients undergoing PCI with DES for ULM has alreadyestablished that nonbifurcational ULM is associated with amore favorable prognosis than distal (ie, bifurcation)ULM, but whether surgery should be the first-linetreatment even for nonbifurcational ULM remainsdebated.26 Recent data have lead to an overall reappraisalof the long-term safety of DES,31 given the likely increasein stent thrombosis, and this has had even greaterimplications on DES-based PCI for ULM.Notwithstanding the limitations of included studies,

we aimed to provide a comprehensive synthesis of theevidence on DES for ULM. We identified several studies,including registries or nonrandomized comparison ofDES versus BMS or CABG, but no randomized trials.Pooled analyses of registry data from the present work,including N1200 patients, support the early and midtermsafety of DES implantation, especially in carefullyselected patients and under the premise that PCI isperformed by experienced operators in centers with on-site surgical backup. The beneficial impact of DES overBMS in these lesions is clearly and unanimously shownin the available controlled studies. Indeed, in-stentrestenosis has been called into question as a potentialcause of acute coronary syndromes in unspecifiedlesions, but in ULM restenosis, it is well known as apotential cause of sudden cardiac death or myocardialinfarction. Thus, the antiproliferative action of DES is ofparamount importance in ULM lesions, and to date, DESshould likely be recommended whenever PCI for ULMis envisioned. Nonetheless, our findings on DES ingeneral, and on the comparison with CABG inparticular, should be viewed in light of the need to waitfor the long-term results of ad hoc randomized trials (eg,COMBAT, SYNTAX, and REVASCULARIZE studies) beforedrawing any definitive conclusion.Among the limitations of this work, drawbacks of

systematic reviews of nonrandomized studies are wellknown, and this type of meta-analysis is obviously inferiorto randomized controlled trials in the evidencehierarchy.4 More specific limits of this work include theextent of statistical inconsistency, the differences inpatients and techniques among different centers, the lack

Biondi-Zoccai et al 283American Heart JournalVolume 155, Number 2

of data on MACE for the PCI versus CABG comparison,and the frequent limitations in included studies, whichshould be strongly borne in mind and prompt the designand conduction of more rigorous observational as well asrandomized studies.

We thank original authors for their help. This study is

part of a training project of the Meta-analysis and

Evidence-based Training in Cardiology Center, Turin,

Italy (http://www.metcadio.org).

References1. Meier B. The first coronary angioplasties in Zurich. In: Bertrand M,

editor. The evolution of cardiac catheterization and interventionalcardiology. St Albans, UK: Iatric Press; 2006. p. 61-74.

2. Takagi T, Stankovic G, Finci L, et al. Results and long-termpredictors of adverse clinical events after elective percutaneousinterventions on unprotected left main coronary artery. Circulation2002;106:698-702.

3. Hill RA, Dundar Y, Bakhai A, et al. Drug-eluting stents: an earlysystematic review to inform policy. Eur Heart J 2004;25:902-19.

4. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis ofobservational studies in epidemiology: a proposal for reporting.Meta-analysis Of Observational Studies in Epidemiology (MOOSE)group. JAMA 2000;283:2008-12.

5. Biondi-Zoccai GG, Lotrionte M, Abbate A, et al. Compliancewith QUOROM and quality of reporting of overlapping meta-analyses on the role of acetylcysteine in the prevention of contrastassociated nephropathy: case study. BMJ 2006;332:202-9.

6. Meta-analysis and Evidence-based Training in Cardiology Centerprotocol #1-2006. Available at: http://www.metcardio.org/protocols.html [last accessed on July 10, 2007].

7. The Cochrane Collaboration handbook for systematic reviews ofinterventions. Available at: http://www.cochrane.org/resources/handbook/ [last accessed on July 10, 2007].

8. Biondi-Zoccai GG, Abbate A, Agostoni P, et al. Long-term benefits ofan early invasive management in acute coronary syndromes sig-nificantly depend on intracoronary stenting and aggressive antiplatelettreatment: a metaregression. Am Heart J 2005;149:504-11.

9. Agostoni P, Valgimigli M, Van Mieghem CA, et al. Comparison ofearly outcome of percutaneous coronary intervention for unprotectedleft main coronary artery disease in the drug-eluting stent era withversus without intravascular ultrasonic guidance. Am J Cardiol2005;95:644-7.

10. Chieffo A, Stankovic G, Bonizzoni E, et al. Early and mid-term resultsof drug eluting stent implantation in unprotected left main. Circulation2005;111:791-5.

11. Park SJ, Kim YH, Lee BK, et al. Sirolimus-eluting stent implantation forunprotected left main coronary artery stenosis: comparison with baremetal stent implantation. J Am Coll Cardiol 2005;45:351-6.

12. Carrié D, Lhermusier T, Hmem M, et al. Clinical andangiographic outcome of paclitaxel-eluting stent implantation forunprotected left main coronary artery bifurcation narrowing.EuroIntervention 2006;1:396-402.

13. Chieffo A, Morici N, Malsano F, et al. Percutaneous treatment withdrug-eluting stent implantation versus bypass surgery forunprotected left main stenosis. A single-center experience. Circulation2006;113:2542-7.

14. Christiansen EH, Lassen JF, Andersen HR, et al. Outcome ofunprotected left main percutaneous coronary intervention in surgical

low-risk, surgical high-risk, and acute myocardial infarction patients.EuroIntervention 2006;1:403-8.

15. de Lezo JS, Medina A, Pan M, et al. Rapamycin-eluting stents for thetreatment of unprotected left main coronary disease. Am Heart J2004;148:481-5.

16. Dudek D, Heba G, Giszterowicz D, et al. Stenting ofunprotected left main coronary artery in patients with lowpreoperative risk of coronary artery bypass grafting. Kardiol Pol2006;64:929-36.

17. Han YL, Wang SL, Jin QM, et al. Efficacy of stenting for unprotectedleft main coronary artery disease in 297 patients. Chin Med J (Engl)2006;119:544-50.

18. Lee SH, Ko YG, Jang Y, et al, for the Korean Multicenter AngioplastyTeam (KOMATE) Investigators. Sirolimus- versus paclitaxel-elutingstent implantation for unprotected left main coronary artery stenosis.Cardiology 2005;104:181-5.

19. Lee MS, Kapoor N, Jamal F, et al. Comparison of coronary arterybypass surgery with percutaneous coronary interventionwith drug-eluting stents for unprotected left main coronary arterydisease. J Am Coll Cardiol 2006;47:864-70.

20. Lozano I, Herrera C, Moris C, et al. Drug-eluting stents in patients withleft main coronary lesions who are not candidates for surgicalrevascularization. Rev Esp Cardiol 2005;58:145-52.

21. Migliorini A, Moschi G, Giurlani L, et al. Drug-eluting stentsupported percutaneous coronary intervention for unprotected leftmain disease. Catheter Cardiovasc Interv 2006;68:225-30.

22. Palmerini T, Marzocchi A, Marrozzini C, et al. Comparison betweencoronary angioplasty and coronary artery bypass surgery for thetreatment of unprotected left main coronary artery stenosis (theBologna Registry). Am J Cardiol 2006;98:54-9.

23. Price MJ, Cristea E, Sawhney N, et al. Serial angiographic follow-upof sirolimus-eluting stents for unprotected left main coronary arteryrevascularization. J Am Coll Cardiol 2006;47:871-7.

24. Sheiban I, Meliga E, Moretti C, et al. Sirolimus-eluting stents vs baremetal stents for the treatment of unprotected left main coronaryartery stenosis. EuroIntervention 2006;2:356-62.

25. Wood F, Bazemore E, Schneider JE, et al. Technique of left mainstenting is dependent on lesion location and distal branch protection.Catheter Cardiovasc Interv 2005;65:499-503.

26. Chieffo A, Park SJ, Valgimigli M, et al. Favorable long-termoutcome after drug-eluting stent implantation in nonbifurcationlesions that involve unprotected left main coronary artery. Amulticenter registry. Circulation 2007;116:158-62.

27. Ruygrok PN, Melkert R, Morel MA, et al. Does angiography sixmonths after coronary intervention influence management andoutcome? BENESTENT II Investigators. J Am Coll Cardiol 1999;34:1507-11.

28. Bavry AA, Kumbhani DJ, Helton TJ, et al. Late thrombosis ofdrug-eluting stents: a meta-analysis of randomized clinical trials. AmJ Med 2006;119:1056-61.

29. Giannoglou GD, Antoniadis AP, Chatzizisis YS, et al. Prevalence ofnarrowing N or = 50% of the left main coronary artery among 17,300patients having coronary angiography. Am J Cardiol 2006;98:1202-5.

30. Huang HW, Brent BN, Shaw RE. Trends in percutaneous versussurgical revascularization of unprotected left main coronarystenosis in the drug-eluting stent era—a report from the AmericanCollege of Cardiology–National Cardiovascular data registry(ACC-NCDR). Catheter Cardiovasc Interv 2006;68:867-72.

31. Biondi-Zoccai GGL, Agostoni P, Moretti C, et al. Making sense of therecent meta-analytical confusion concerning the safety of drug-eluting stents. EuroIntervention 2007 [in press].