drug-coated balloon versus plain balloon angioplasty for...

J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 9 , N O . 1 6 , 2 0 1 6

ª 2 0 1 6 B Y T H E AM E R I C A N C O L L E G E O F C A R D I O L O G Y F O UN DA T I O N I S S N 1 9 3 6 - 8 7 9 8 / $ 3 6 . 0 0

P U B L I S H E D B Y E L S E V I E R h t t p : / / d x . d o i . o r g / 1 0 . 1 0 1 6 / j . j c i n . 2 0 1 6 . 0 6 . 0 0 8

Drug-Coated Balloon Versus PlainBalloon Angioplasty for the Treatmentof Femoropopliteal Artery DiseaseAn Updated Systematic Review and Meta-Analysis ofRandomized Clinical Trials

Daniele Giacoppo, MD,a Salvatore Cassese, MD, PHD,a Yukinori Harada, MD,a Roisin Colleran, MBCHB,a

Jonathan Michel, MBBS,a Massimiliano Fusaro, MD,a Adnan Kastrati, MD,a,b Robert A. Byrne, MBCHB, PHDa

ABSTRACT

Fro

for

fro

pli

an

the

Ma

OBJECTIVES This study sought to assess the risk of target lesion revascularization (TLR) and all-cause death at

12 months and at the maximum available follow-up. Secondary objectives included the identification of factors which

could have influenced general findings.

BACKGROUND Recently several randomized trials comparing drug-coated balloon (DCB) with conventional plain

balloon (PB) for the treatment of femoropopliteal artery disease have been reported, but no updated meta-analyses are

available and questions remain surrounding the long-term antirestenotic effectiveness of the 2 therapies.

METHODS We searched main electronic databases for randomized trials comparing DCB and PB for femoropopliteal

artery disease. Random effects models were used to estimate the risk of TLR and all-cause death at 12 months, whereas

long-term TLR and death risk were assessed by mixed effects Poisson regression models and incident rates of each

outcome per patient-year. Main analyses were supplemented by sensitivity analyses, Bayesian estimates, and trial

sequential analysis.

RESULTS A total of 8 eligible trials were identified. DCB was associated with a marked 12-month TLR risk reduction as

compared with PB (risk ratio: 0.33; 95% confidence interval [CI]: 0.19 to 0.57). The risk of death was similar between

groups (risk ratio: 0.96; 95% CI: 0.47 to 1.95). Long-term outcomes assessment showed a reduced incidence of TLR with

DCB (0.35; 95% CI: 0.24 to 0.51) and a similar incidence of all-cause death (incidence rate ratio: 1.13; 95% CI: 0.60 to

2.15). Similar findings were observed in Bayesian analyses. Significant heterogeneity was present with evidence of dif-

ferential efficacy across devices. Trial sequential analysis indicated that available evidence is sufficient to prove superior

antirestenotic efficacy of DCB over PB.

CONCLUSIONS DCB significantly reduces the risk of TLR as compared with PB without any effect on all-cause

death. Evidence exists for differential efficacy according to the type of device used. Future trials investigating DCB

angioplasty should include potentially more effective comparator therapies. (J Am Coll Cardiol Intv 2016;9:1731–42)

© 2016 by the American College of Cardiology Foundation.

S ubstantial improvements in endovascular tech-niques and outcomes mean that percutaneoustransluminal angioplasty is now the first-line revascularization strategy for patients with

m the aDeutsches HerzzentrumMünchen, Technische Universität Münche

Cardiovascular Research), partner site Munich Heart Alliance, Munich, Ge

m the EAPCI (European Association Percutaneous Coronary Intervention

cations in relation to drug-eluting stent technology. Dr. Byrne has received

d Boston Scientific; and institutional research grants from Boston Scientific

y have no relationships relevant to the contents of this paper to disclose

nuscript received April 19, 2016; revised manuscript received May 26, 20

symptomatic peripheral arterial disease (1). Plainballoon (PB) angioplasty for femoropopliteal arterydisease has a high rate of procedural success and anacceptable safety profile, however, rates of restenosis

n, Munich, Germany and the bDZHK (German Centre

rmany. Dr. Giacoppo has been awarded with a grant

). Dr. Kastrati has reported submission of patent ap-

lecture fees from B. Braun Melsungen AG, Biotronik

and Heartflow. All other authors have reported that

.

16, accepted June 2, 2016.

http://crossmark.crossref.org/dialog/?doi=10.1016/j.jcin.2016.06.008&domain=pdfhttp://dx.doi.org/10.1016/j.jcin.2016.06.008

ABBR EV I A T I ON S

AND ACRONYMS

CI = confidence interval

DCB = drug-coated balloon

IRR = incidence rate ratio

PB = plain balloon

PRISMA = Preferred Reporting

Items for Systematic Reviews

and Meta-Analyses

RR = risk ratio

TLR = target lesion

revascularization

Giacoppo et al. J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 9 , N O . 1 6 , 2 0 1 6

DCB vs. PB for Femoropopliteal Artery Disease A U G U S T 2 2 , 2 0 1 6 : 1 7 3 1 – 4 21732

are considerable (2). For this reason, a num-ber of alternative percutaneous treatmentstrategies have been investigated (3–8).

Drug-coated balloons (DCBs) are standardballoon angioplasty catheters surface coatedwith a thin layer of antiproliferative drugcombined with an excipient or spacer sub-stance, which facilitates drug transfer to thevessel wall (9). The advantages of DCB ther-apy include drug delivery and inhibition ofneointimal proliferation without require-ment for a permanent metallic implant, moreuniform drug–tissue transfer, potential

amelioration of vessel healing due to the absence ofproinflammatory durable polymer surface coating,and preservation of arterial regulatory functions (9).

SEE PAGE 1743

Recently, several randomized clinical trialscomparing DCB with conventional PB angioplasty forthe treatment of femoropopliteal artery disease havebeen reported but no updated meta-analyses areavailable and data relating to the long-term assess-ment of the 2 therapies is scant (10–12). In addition,there continues to be ongoing discussion of the use-fulness of systematic use of DCB instead of PB for denovo lesions and results obtained with differentDCBs may be not uniform (13). Against this back-ground, we conducted a comprehensive meta-analysis of randomized clinical trials comparingDCB versus PB for the treatment of femoropoplitealartery disease with the primary objective to assessthe treatment effect for need for repeat targetlesion revascularization (TLR) and death at 12months and at the longest available follow-up.Secondary aims were the assessment of trial-levelfactors that could have influenced the anti-restenotic effectiveness of the 2 devices and in-troduced heterogeneity, the exploration of thepotential differential efficacy among available typesof DCB, and the definition of functional benefits ofa DCB-based revascularization.

METHODS

This meta-analysis was conducted in accordancewith the Preferred Reporting Items for SystematicReviews and Meta-Analyses (PRISMA) statement andCochrane’s Collaboration recommendations (14,15).The PRISMA checklist is reported in the OnlineAppendix. Data used were from intention-to-treatanalyses. Statistical analyses were performed usingR (version 3.2.3), WinBUGS (version 1.4.3), and TSA(version 0.9).

LITERATURE SEARCH AND STUDY SELECTION.

We searched PubMed, ScienceDirect, Scopus, Web ofKnowledge, and Cochrane Library electronic data-bases for randomized trials comparing DCB versus PBfor the treatment of femoropopliteal artery diseasefrom the date of inception to December 1, 2015. Nolanguage restrictions or specific clinical subsets wereimposed. The search algorithm applied for trialsidentification and the corresponding results are re-ported in the Online Table 1. Tangential explorationof relevant scientific websites (Online Table 1) as wellas bibliography screening of relevant reviews on thetopic was conducted to minimize the risk of missingreports.

Pre-specified inclusion criteria were: 1) random-ized trials of patients receiving DCB versus PB; 2)single-treatment strategy, either DCB or PB, withbailout stenting in case of unsuccessful angioplastywith balloon; 3) treatment of femoropopliteal lesionswith critical stenosis ($70%); and 4) original resultspublished in a peer-reviewed medical journal.Exclusion criteria included: 1) observational studies;2) treatments other than DCB or PB; 3) use of othertreatments in combination with DCB or PB; 4) appli-cation of DCB or PB only after stenting (post-dilation);and 5) lesion location in nonfemoropopliteal arterialsegment (below-the-knee arteries disease, iliacartery, and so on). Trials including both de novo andrestenotic lesions were allowed. Additional informa-tion about search and selection methods is reportedin the Online Appendix. The risk of bias in each trialwas qualitatively assessed as recommended by theCochrane Collaboration (15).

PRIMARY AND SECONDARY OUTCOMES. The pri-mary objective of this meta-analysis was the evalua-tion of the risk of TLR at 12 months and at long-termfollow-up. Secondary outcomes of interest were12-month and long-term all-cause death.

STATISTICAL ANALYSES. The analyses of 12-monthTLR and 12-month all-cause death were performedby using DerSimonian–Laird random effects models(16,17). Effect size was estimated as risk ratio (RR) and95% confidence intervals (CIs).

The analyses of long-term TLR and long-term all-cause death were performed by using mixed–effectsPoisson regression models with random study effects(18). The analyses used the incident rate of theoutcome per patient-years to obtain the pooled inci-dence rate ratio (IRR) with 95% CI of DCB versus PB(18). IRR was considered the most appropriateoutcome for this analysis because it allowed incor-porating the different follow-up durations of theincluded trials. A Bayesian analysis was also

http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008

TABLE 1 Design of the Included Trials

Study (Ref.#)Randomization

(DCB:PB) Center (n) Region Identification Number* Enrolment TimePrimaryEndpoint

MaximumAvailable Follow-Up

(months)

Biolux P-I (25) 1:1 5 Austria, Germany NCT01221610 Oct 2010/Aug 2011 6-month LLL 12

FAIR (8) 1:1 5 Germany NCT01305070 Jan 2010/Nov 2012 6-month BR* 12

FemPac (26) 1:1 2 Germany NCT00472472 July 2004/Jan 2006 6-month LLL 24

IN.PACT SFA(27,30)

2:1 57 Austria, Belgium,Germany, Italy,Switzerland, USA

NCT01175850 Sep 2010/Apr 2011(Phase I)

Apr 2012/Jan 2013(Phase II)

12-month PP† 12

LEVANT I (28) 1:1 9 Belgium, Germany, USA NCT00930813 Jun 2009/Dec 2009 6-month LLL 24

LEVANT 2 (6) 2:1 54 Austria, Germany, USA NCT01412541 Jul 2011/Jul 2012 12-month PP† 12

PACIFIER (29) 1:1 3 Germany NCT01083030 Mar 2010/Aug 2011 6-month LLL 24

THUNDER (7,31) 1:1:1‡ 3 Germany NCT00156624 Jun 2004/Jun 2005 6-month LLL 60

*Assessed by Duplex ultrasonography: peak systolic velocity ratio $2.4 w $50% luminal reduction. †The third group of the THUNDER trial (paclitaxel diluted in the contrastmedia) was not included in the study. ‡PP was the composite of freedom from clinically driven target lesion revascularization and restenosis as determined by a Duplexultrasonography–derived peak systolic velocity ratio of #2.4.

Biolux P-I ¼ A Prospective, Multi-centre, Randomized Controlled, First in Man Study to Assess the Safety and Performance of the Passeo-18 Lux Paclitaxel Releasing PTABalloon Catheter vs. the Uncoated Passeo 18 Balloon Catheter in Patients With Stenosis and Occlusion of the Femoropopliteal Arteries; DCB ¼ drug-coated balloon; FemPac ¼Paclitaxel Coated Balloon Catheter for Inhibition of Restenosis in Femoropopliteal Arteries; FAIR¼ Randomized Femoral Artery In–Stent Restenosis; IN.PACT SFA¼ The IN.PACTSFA Clinical Study for the Treatment of Atherosclerotic Lesions in the Superficial Femoral Artery and/or Proximal Popliteal Artery Using the IN.PACT Admiral� Drug-ElutingBalloon in a Chinese Patient Population; LEVANT I ¼ A Prospective, Multicenter, Single Blind, Randomized, Controlled Trial Comparing the Lutonix Catheter vs. StandardBalloon Angioplasty for Treatment of Femoropopliteal Arteries With and Without Stenting; LEVANT 2 ¼ A Prospective, Multicenter, Single Blind, Randomized, Controlled TrialComparing the Moxy Drug Coated Balloon vs. Standard Balloon Angioplasty for Treatment of Femoropopliteal Arteries; LLL ¼ late lumen loss; PACIFIER ¼ Paclitaxel-coatedBalloons in Femoral Indication to Defeat Restenosis; PB ¼ plain balloon; PP ¼ primary patency; THUNDER ¼ Local Taxan With Short Time Contact for Reduction of Restenosis inDistal Arteries.

J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 9 , N O . 1 6 , 2 0 1 6 Giacoppo et al.A U G U S T 2 2 , 2 0 1 6 : 1 7 3 1 – 4 2 DCB vs. PB for Femoropopliteal Artery Disease

1733

performed. Hierarchical models (binomial or Poissonlikelihood and logit or log-link function, respec-tively) with random effects were computed by Mar-kov Chain Monte Carlo method with Gibbs samplingand posterior inference was based on 100,000 sim-ulations following discard of a “burn-in” of 50,000simulations (19,20). Noninformative priors wereused (21) and convergence was graphically appraisedaccording to Gelman–Brooks (22). Posterior inferencewas expressed as RR or IRR, as appropriate, and theaccompanying 95% credible intervals came fromthe 2.5th and 97.5th centiles of the posteriordistribution.

HETEROGENEITY AND PUBLICATION BIAS/SMALL

STUDY EFFECT. Heterogeneity and publication bias/small study effect assessment are described in theOnline Appendix.

SENSITIVITY AND SUBGROUP ANALYSES. Rationaland specifications of sensitivity and subgroup anal-yses are described in the Online Appendix.

TRIAL SEQUENTIAL ANALYSIS. We performed a trialsequential analysis to assess whether cumulativeevidence deriving from randomized trials was suffi-ciently large to declare the superiority of one treat-ment over the other (23,24). Considering the realdistribution of the events in the 2 groups of patients,we anticipated a 25% relative risk reduction (a ¼ 0.05;1–b ¼ 0.80) in the risk of 12-month TLR by the

O’Brien–Fleming a-spending function which allowsgenerating monitoring boundaries accounting forrepeated statistical testing. Accordingly, we calcu-lated the required diversity-adjusted information sizeas number of patients (24).

RESULTS

A total of 8 randomized clinical trials were identified(6–8,25–31) (Online Appendix). Online Figure 1 illus-trates the selection process in detail (PRISMA flowdiagram). Trial design and methodology are shown inTable 1. Trial inclusion and exclusion criteria aresummarized in the Online Table 2. More than one-halfof the included trials were not powered to detectdifferences in TLR and had an angiographic endpoint(late lumen loss) as primary endpoint. The IN.PACTSFA (IN.PACT SFA Clinical Study for the Treatment ofAtherosclerotic Lesions in the Superficial FemoralArtery and/or Proximal Popliteal Artery Using theIN.PACT Admiral� Drug-Eluting Balloon in a ChinesePatient Population) and LEVANT 2 (A Prospective,Multicenter, Single Blind, Randomized, ControlledTrial Comparing the Moxy Drug Coated Balloon vs.Standard Balloon Angioplasty for Treatment ofFemoropopliteal Arteries), the 2 largest trials, werepowered for a composite endpoint of major adverseevents, including Duplex ultrasonography measure-ments (Table 1). Within-trial clinical characteristicswere comparable and described patients with high

http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008https://clinicaltrials.gov/ct2/show/NCT01221610?term=NCT01221610%26rank=1https://clinicaltrials.gov/ct2/show/NCT01305070?term=NCT01305070%26rank=1https://clinicaltrials.gov/ct2/show/NCT00472472?term=NCT00472472%26rank=1https://clinicaltrials.gov/ct2/show/NCT01175850?term=NCT01175850%26rank=1https://clinicaltrials.gov/ct2/show/NCT00930813?term=NCT00930813%26rank=1https://clinicaltrials.gov/ct2/show/NCT01412541?term=NCT01412541%26rank=1https://clinicaltrials.gov/ct2/show/NCT01083030?term=NCT01083030%26rank=1https://clinicaltrials.gov/ct2/show/NCT00156624?term=NCT00156624%26rank=1

TABLE 2 Main Trial-Level Clinical Characteristics

Study (Ref. #)DCB:PB

n:n (Total) Age (yrs)* Male DM Smokers HTN DLM CKD CADCarotid Artery

Disease

Biolux P-I (25) 30:30 (60) 71 56.7 (34) 33.3 (20) 68.3 (41) 73.3 (44) 61.6 (37) NR NR NR

FAIR (8) 62:57 (119) 68 68.9 (82) 37.8 (45) 31.9 (38) 88.2 (105) 78.1 (93) 15.1 (18) 40.3 (48) 20.2 (24)

FemPac (26) 45:42 (87) 69 59.8 (52) 47.1 (41) 41.4 (36) 79.3 (69) 58.1 (50) NR NR NR

IN.PACT SFA (27,30) 220:111 (311) 68 70.1 (218) 46.0 (143) 37.8 (125) 90.3 (299) 83.7 (277) NR 56.3 (182) 33.9 (105)

LEVANT I (28) 49:52 (101) 69 63.4 (64) 47.5 (48) 34.7 (35) 91.1 (92) 64.4 (65) NR 41.6 (42) NR

LEVANT 2 (6) 316:160 (476) 68 63.0 (300) 42.9 (204) 34.7 (165) 88.7 (422) 88.4 (421) 3.8 (18) 49.2 (234) NR

PACIFIER (29) 41:44 (85) 71 61.5 (56) 35.2 (32) 53.8 (49) 65.9 (60) 48.4 (44) NR 31.9 (29) NR

THUNDER (7,31) 48:54 (102) 68 65.7 (67) 49.0 (50) 22.5 (23) 81.4 (83) 65.7 (67) NR NR NR

Values are % (n) unless otherwise indicated. Original reports did not show within-trial differences for all these variables. *Pooled mean of arm-level rounded mean values.

CAD ¼ coronary artery disease; CKD ¼ chronic kidney disease; DLM ¼ dyslipidemia; DM ¼ diabetes mellitus; HTN ¼ hypertension; NR ¼ not reported; other abbreviations as in Table 1.

TABLE 3

Study

Biolux P-I

FAIR (8)

FemPac (2

IN.PACT S

LEVANT I

LEVANT 2

PACIFIER

THUNDER

Values are %THUNDER,difference i

BR ¼ bina



cardiovascular risk (Table 2). Superficial femoralartery disease was predominant and variable rates ofde novo target lesions, ranging from 0% to 94.9%,were observed (Table 3, Online Table 3). In 3 trials(6,7,25), bailout stenting was significantly morefrequent in the PB group and in the 5 remaining trials(8,26–29) was numerically higher (Table 2). Meanlesion length across trials was

TABLE 4 Main Characteristics of DCBs

Study (Ref. #) DCB Type Drug

Drug DoseDensity

(mg/mm2) Excipient Manufacturer

Biolux P-I (25) Passeo–18 Lux Paclitaxel 3.0 Butyryl-tri-n-hexylcitrate (BTHC)

Biotronik

FAIR (8) IN.PACTAdmiral

Paclitaxel 3.5 Urea Medtronic

FemPac (26) Paccocath* Paclitaxel 3.0 Iopromide Bavaria MedizinTechnologie*

IN.PACT SFA(27,30)

IN.PACTAdmiral

Paclitaxel 3.5 Urea Medtronic

LEVANT I (28) Lutonix†‡ Paclitaxel 2.0 Polysorbate andSorbitol

Lutonix†

LEVANT 2 (6) Lutonix†‡ Paclitaxel 2.0 Polysorbate andSorbitol

Lutonix†

PACIFIER (29) IN.PACT Pacific Paclitaxel 3.0 Urea Medtronic

THUNDER (7,31) Paccocath* Paclitaxel 3.0 Iopromide Bavaria MedizinTechnologie*

*Current DCB version (manufacturer) is SeQuent Please (Braun). †Current DCB version (manufacturer) is Lutonix(Bard). ‡Although in the original manuscripts of the LEVANT I and LEVANT 2 trials the DCB name was “Lutonix”,in some reports and in the protocol of the LEVANT 2 trial the device was designated as “Moxy.”

Abbreviations as in Table 1.


1735

the RR ranged from 0.28 (without LEVANT I orLEVANT 2, the trials in which DCB showed the lowestbenefit as compared with PB) to 0.38 (without theIN.PACT SFA, the trial in which DCB showedthe greatest benefit as compared with PB) and thesummary effect in all cases remained highly signifi-cant. Visual inspection of the contour-enhancedfunnel plot for 12-month TLR revealed an asym-metric distribution, quantified by “trim and fill” with4 missing trials to the right of the pooled effects andresulted significant (p ¼ 0.01) at Egger’s linearregression (Online Figure 5).

To analyze the individual impact on heterogeneity,a single trial was removed one at a time, and the in-dividual influence on I2 was estimated (Figure 2).Using this method, we identified 3 trials that ampli-fied the I2: the LEVANT I, IN.PACT SFA, and LEVANT 2trials (6,27,28). This was also graphically appraisableby Baujat plot (Figure 2, left panel). Subsequently,because the global I2 remained high, we investigatedthe possible combinations of trials, which resulted inan I2 value below the threshold of low heterogeneity(71.3%)(Figure 4). Results in the 2 groups remained consis-tent with the main analysis. Finally, although theFAIR (Randomized Femoral Artery In–Stent Reste-nosis) trial significantly differs from the others,enrolling only patients with in-stent restenoticlesions, its removal did not change the superiority ofDCB over PB (RR: 0.35; 95% CI: 0.20 to 0.63) and thepoint estimate after exclusion was similar to mainanalysis pooled value.

The impact of prevalence of target lesion total oc-clusion at baseline was explored by grouping trialsaccording to rate #27.6% or >27.6% (Online Figure 7).The cutoff rate was extracted by the median rate oftarget lesion total occlusion across the included trials.The subgroup analysis confirmed the results of mainanalysis.

The additional primary endpoint of long-term TLR(Figure 5, left panel) was assessed by using themaximum trial-level available follow-up (meanfollow-up time: 1.9 years; range: 1-5) for a total of1,843 patient-years. The meta-analysis of long-termTLR confirmed that the superior effectiveness ofDCB over PB was stable over time (IRR 0.35; 95%CI: 0.24 to 0.51). Heterogeneity was moderate(I2 ¼ 44.2%) and also in this case mainly due to theLEVANT I and LEVANT 2 trials (6,28), which showed anonsignificant effect moderately favoring DCB. Afterexcluding these 2 trials, heterogeneity was no longerdetected and the summary estimate favoring DCBseemed to be magnified compared with main analysis(Online Figure 8). The analyses were repeated using aBayesian framework with concordant results.

Results of the meta-analysis for the secondaryendpoints of 12-month all-cause death are illustratedin the Online Figure 9: the pooled risk of all-causedeath at 12 months was similar between the 2 treat-ments (RR: 0.96; 95% CI: 0.47 to 1.95). No significantasymmetry was visualized in the contour-enhancedfunnel plot for 12-month all-cause death and Egger’stest was nonsignificant (Online Figure 10). After per-forming a meta-analysis for the long-term all-causedeath outcome (Figure 5, right panel), summary IRR

http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008http://dx.doi.org/10.1016/j.jcin.2016.06.008

FIGURE 1 Risk of Target Lesion Revascularization. Risk at 12 Months Comparing DCB With PB

The forest plot illustrates the results of the main analysis: DCB compared with PB produced a 67% RR reduction in 12-month target lesion revascularization. Bayesian

estimate (lower summary effect, RR: 0.30; 95% credible intervals: 0.14 to 0.58) was consistent. CI ¼ confidence interval; DCB ¼ drug-coated balloon; PB ¼ plainballoon; RR ¼ risk ratio. *Credible intervals for the Bayesian estimate.



did not favor one treatment over the other (IRR: 1.13;95% CI: 0.60 to 2.15). An excess of mortality with DCBwas observed in 1 study, the IN.PACT SFA trial.Bayesian analysis was consistent with frequentistestimate. Other major adverse events were overallextremely rare and comparable in the 2 groups,although functional benefits at follow-up were lessevident than TLR reduction (Online Table 6).

Trial sequential analysis showed that the numberof available trials is likely sufficient to demonstrateoverall superior 12-month antirestenotic efficacy ofDCB over PB (Figure 6). Indeed, after sequentialaddition of trials according to a chronological order(Z-score), very early the cumulative evidence reachednot only the conventional boundary (standard esti-mate of required evidence), but also the a-spendingfunction monitoring boundary (adjusted estimate ofrequired evidence). The analysis predicted that a totalof 421 patients was required to gain sufficient statis-tical power and adjusting the CI of the main analysisfor repeated statistical testing the summary effectremained highly significant (adjusted 95% CI: 0.15to 0.69).

DISCUSSION

In this meta-analysis, we observed 5 key findings: 1)DCB is significantly superior to PB in reducing the riskof TLR at 12 months in patients with femoropopliteal

artery disease and this benefit appears to persist overtime with reduced rates of TLR at long-term follow-up; 2) the antirestenotic benefits of DCB are consis-tent across subsets of either de novo or restenoticlesions; 3) there was some evidence of differentialefficacy of available paclitaxel DCBs; 4) there was nodifference in terms of mortality between treatmentwith DCB or PB; and 5) additional randomized clinicaltrials comparing currently available DCB with PB in ageneral clinical and angiographic subset do not seemto be required.

Our meta-analysis differs from prior meta-analyses(10–12) in several aspects: 1) it is updated to includerecently published randomized trials; 2) we focus onclinical outcome measures, such as TLR and all-causedeath, both at 12 months and in the long term; 3) itsystematically assesses and explores reasons forobserved heterogeneity in trial-level results; and 4) itprovides novel insights regarding the availableevidence on DCB versus PB using trial sequentialanalysis.

This meta-analysis provides evidence of clear su-periority of DCB over PB for the treatment of femo-ropopliteal artery disease, in terms both of 12-monthand long-term TLR. Indeed, the durability of DCBtherapy superiority is particularly noteworthy. Con-cerns had been raised that the lower late lumen lossand binary restenosis observed at 6 months in patientswith lower limb disease treated with DCB enrolled in

http://dx.doi.org/10.1016/j.jcin.2016.06.008

FIGURE 2 Inspection of Individual Contribution to Main Analysis Heterogeneity

(Left) Baujat plot inspecting overall heterogeneity. (Right) Removing each trial at a time from the others, the individual impact on I2 was estimated. Trials contributing to

heterogeneity increase are disposed to the right of main analysis I2 value of 69.4% (red line), whereas trials reducing heterogeneity are disposed to the left. The

circle size is proportional to inverse of variance and I2 estimates reported into each circle correspond to the I2 value without the influence of that trial. The removal of

the LEVANT I, the IN.PACT SFA, and the LEVANT 2 trials reduced I2 below the main analysis value and therefore they increased heterogeneity. The extent of the I2

percentage variation is described on the horizontal axis. However, individual removal of trials did not lead I2 below 25% threshold (“low” heterogeneity) and p values

of significance testing remained significant. Removing 2 trials each time to the right of main analysis I2 value, heterogeneity was no longer detectable without the

LEVANT I and LEVANT 2 trials. After removal of other combinations of trials, heterogeneity remained high. In conclusion, the LEVANT I and LEVANT 2 trials majorly

contributed to the high heterogeneity observed in the main analysis.


1737

trials with planned angiographic surveillance may notpersist over the longer term. Our analysis, however,shows that DCB therapy continues to be associatedwith a reduced risk of TLR at a mean follow-up of 1.9years. This finding is also in agreement with therecently available long-term follow-up of trialscomparing DCB with PB in the treatment of coronaryin-stent restenosis (32,33). The cumulative incidenceof TLR in the PB group was slightly higher thanrecently reported (4,34). This may have exaggeratedthe magnitude of the results favoring DCB but themargin of significance of summary effect was clear.

The observed heterogeneity between trials seemedto be explained by the inclusion of the 2 trials usingthe Lutonix DCB (6,28). Indeed, heterogeneity wasnot detected by excluding the trials using the LutonixDCB and using an influence analysis we showed thatthe LEVANT I and LEVANT 2 trials majorly contrib-uted to I2 increase. The different effectiveness of theDCB treatment in the 2 trials using the Lutonix DCBhas 2 possible explanations: on the one hand, thisresult may suggest a lower efficacy of this type of DCBcompared with the others; in contrast, the findings

may reflect trial design characteristics specific toLEVANT I and LEVANT 2 trials. This interpretationof the heterogeneity is graphically expressed bythe strong asymmetry of the funnel plot, with theLEVANT I and LEVANT 2 trials falling in the non-significance area. However, the different results ofthese 2 trials may have introduced only heterogeneitywithout implying a publication bias (“true” hetero-geneity) and the 4 missing trials to the right of meaneffect required to make symmetric the funnel plotwere corresponding to the high-significance area (p <0.01), which is not associated generally with thepresence of a small study effect (35).

Another important finding of this meta-analysis isthe possible differential effectiveness observedamong DCBs. This finding is in agreement with theheterogeneity analyses, providing a reasonable clin-ical explanation. Indeed, the attenuated anti-restenotic effects associated with the trials using theLutonix DCB could be explained by the lower pacli-taxel dose density compared with the other DCBs(2 vs. $3 mg/mm2). In the animal model, DCB effec-tiveness appeared at the dose of 1 mg/mm2 with an

FIGURE 3 Subgroup Analysis According to DCB Catheter Type

The forest plots illustrate the results of the subgroup analysis according to DCB type. The 2 trials using the Lutonix DCB showed a mild

nonsignificant RR reduction, whereas antirestenotic efficacy of the other DCBs remained consistent with main analysis. Formal testing for

differences among groups was highly significant. Abbreviations as in Figure 1.



incremental antirestenotic effect up to 3 mg/mm2, butbeyond this value neointimal area remained compa-rable (36). However, paclitaxel dose density is onlyone of the factors influencing DCB efficacy. Indeed,excipients are key components of the balloon coatingand regulate paclitaxel elution (37). The Lutonix DCBcoating drug carrier consists of polysorbate and sor-bitol, which have not been extensively explored inanimal models and may have different effectivenesscompared with excipients of the other DCBs(6,10,28,38).

Despite the clear reduction in TLR associated withDCB, there was no difference in terms of mortalitybetween DCB treatment and PB. This findingremained unchanged after comparing incidence rateof all-cause death in the 2 groups at longest availablefollow-up.

We did not meta-analyze other secondary clinicaloutcomes because the qualitative review of datashowed incomplete reporting (i.e., primary patency)and extremely rare occurrence (i.e., major amputation,thrombosis, and myocardial infarction). Nevertheless,

data review indicated some interesting findings.Indeed, with the exception of the FAIR trial, thenumber of target lesion thrombosis in patients treatedwith DCB was in all trials equal to or lower thanPB group, suggesting that concerns about a potentialDCB thrombogenic tendency compared with PB couldnot be supported. Moreover, periprocedural dissec-tions were comparable between both treatmentgroups though large differences among trials wereobserved.

Finally, we also performed a trial sequential anal-ysis with the aim to assess the requirement for furtherstudies investigating the comparative efficacy of DCBversus PB in the treatment of femoropopliteal arterydisease. Our findings indicate that evidence ofsuperiority of currently available DCBs is clear andsuggests that future investigations should be orientedto comparisons between DCB and underexploredpromising devices for femoropopliteal artery diseasetreatment, such as drug-eluting stents, or specificcommon high-risk clinical and angiographic subsets,such as diabetes, long lesion, in-stent restenosis, and

FIGURE 4 Subgroup Analysis According to Trial-Level Prevalence of De Novo Target Lesion

Considering the median prevalence of de novo lesions, included trials were divided in 2 groups according to their individual rate, either #71.3%

or >71.3%. Regardless more or less de novo target lesions, main analysis conclusions remained unchanged. Abbreviations as in Figure 1.


1739

total occlusion. However, although general pro-prieties of the IN.PACT, Paccocath and Lutonix DCBsseem to be well-evaluated, further evidence withPasseo–18 Lux is required.

STUDY LIMITATIONS. As with any meta-analysis, ourreport shares the limitations of the original trials andpotential sources of heterogeneity in clinical and

FIGURE 5 Long-Term Target Lesion Revascularization and All-Cause

The meta-analysis of IRR (mean follow-up time of 1.9 years for 1,843 pa

revascularization (left) and the secondary endpoint of all-cause death (r

toward the left side of the forest plots favor DCB, while effects toward

Bayesian estimate. IRR ¼ incidence rate ratio; other abbreviations as in

procedural characteristics cannot be fully exploredwithout individual patient data. Metaregression canonly partially overcome the absence of individualpatient data and given the strong dependence on thenumbers of trials was not performed (39).

More specifically, the results of our meta-analysisshould be interpreted taking the following limita-tions into account. First, 2 of the included trials

Death in DCB Versus PB

tient-years) for the other primary endpoint of long-term target lesion

ight) confirmed main analysis findings (12-month follow-up). Effects

the right side of the forest plots favor PB. *Credible intervals for the

Figure 1.

FIGURE 6 Trial Sequential Analysis for 12-Month Target Lesion Revascularization

Trial sequential analysis showed that available evidence seems sufficiently large to prove a superior antirestenotic effect associated with DCB as compared with PB.

Indeed, the anticipated adjusted number of patients required to have enough statistical power for a RRR of 25% was 421 and the Z-score (blue line) crossed

very early not only the conventional boundary (green line; 1.96 cumulative Z-score), but also the monitoring boundary (upper red line) generated by the

O’Brien–Fleming a-spending function accounting for repeated testing. The green zone illustrates the extent of the cumulative Z-score between conventional

boundary and monitoring boundary (conventional significance area), while the light blue zone describes the extent of the statistical significance reached by pooled

evidence (Z-score line, blue line) over the monitoring boundary. The lower red line (futility boundary) was not crossed, which means that available evidence

did not fall in the zone indicating uselessness of adding new trials (orange area). RRR ¼ relative risk reduction; other abbreviations as in Figure 1.



(Biolux P–I and FemPac [Paclitaxel Coated BalloonCatheter for Inhibition of Restenosis in Femo-ropopliteal Arteries]) included minimal rates (n ¼ 12)of target lesion involving below-the-knee arteries(25,26). Additionally, the analyses of 12-month TLRand all-cause death include 6-month data for theFemPac trial, because events at 12 months were notprovided (26). Moreover, the third arm of the THUN-DER (Local Taxan With Short Time Contact forReduction of Restenosis in Distal Arteries) trial(paclitaxel in contrast media) was not pooled,implying a remote possibility that benefits ofrandomization could have been lost (7). Second, inthe LEVANT I trial a 1:1 randomization was donefollowing stratification according to flow limiting

dissection or $70% resistant stenosis after initialtreatment and, although DCB or PB assignment wasrandom, 25% of patients (namely “stent group”)received provisional stenting (28). Third, we detectedsignificant differences between trials using theLutonix DCB and those using the other catheters.However, no randomized clinical trials directlycompared the different devices and the differenceobserved could be due to confounding factors.Fourth, only 50% of trials reported a clinical follow-up between 24 and 60 months and the mean follow-up in this meta-analysis was of 1.9 months.Although the methodology used attempted to ac-count for these issues, it cannot replace time-to-eventanalyses with individual patient data and uniform

PERSPECTIVES

WHAT IS KNOWN? In randomized clinical trials, DCBs are generally

associated with superior antirestenotic efficacy compared with plain

balloon. However, these trials are powered only for surrogate endpoints or

composite endpoints, including clinical and surrogate parameters. The

clinical impact of femoropopliteal artery revascularization with DCB is var-

iable across reports and not explored in an adequately large number of

patients. Moreover, the influence of target lesion type and the long-term

durability of DCB effects are poorly defined. Finally, no randomized trials


1741

long-term follow-up. Fifth, the significant funnel plotasymmetry in relation to TLR was considered likely tobe representative of “true” heterogeneity derivingfrom the different catheter types within DCB grouprather than a small study effect. However, due to thelimited number of trials, this question cannot becompletely answered. Finally, the paucity of majoradverse events and the significant variations inreporting functional variations of lower limb revas-cularization with DCB and PB across the trials did notallow the meta-analysis of these outcomes. Func-tional benefits at follow-up of femoropoplitealrevascularization with DCB over PB were less evidentthan antirestenotic effects. Future trials with DCBshould specifically address very late clinical andfunctional improvements.

directly comparing the different available DCB devices have been

conducted.

WHAT IS NEW? In patients undergoing femoropopliteal artery inter-

vention, DCB therapy should be preferred over conventional plain balloon

angioplasty due to superior antirestenotic effectiveness in terms of TLR.

At long-term follow-up, the lower incidence of TLR associated with DCB

seems to be durable. Although DCB performance seems not influenced by

target lesion type, differential effectiveness across currently available

devices was detected. Survival is not influenced by the revascularization

strategy.

WHAT IS NEXT? Differences between the 2 treatments in terms of

symptoms and functional improvement are not delineated sufficiently in

randomized clinical trials and the specific subgroups of patients which might

receive greatest benefit from revascularization with DCB is not defined.

Future trials on DCB should potentially include more effective comparators

such as drug-eluting stents.

CONCLUSIONS

The treatment of femoropopliteal artery disease withDCB significantly reduces the risk of 12-month TLRcompared with PB without any effect on all-causemortality. The observed treatment effect persists atlong-term follow-up. Results were consistent acrosssubsets of included lesions, but trials using theLutonix DCB introduced significant heterogeneity,suggesting an attenuated antirestenotic effect.Additional trials to confirm the superior anti-restenotic efficacy of currently available DCBs versusPB angioplasty in a general clinical and angiographicsubset of patients with femoropopliteal artery dis-ease are not required. Future trials should focuson the comparison between DCB and potentiallymore effective comparators such as drug-elutingstents.

REPRINT REQUESTS AND CORRESPONDENCE: Dr.Robert Byrne, Deutsches Herzzentrum München, Laza-rettstrasse 36, 80636,Munich, Germany. E-mail: [email protected].

RE F E RENCE S

1. Tendera M, Aboyans V, Bartekink ML, et al.ESC Guidelines on the diagnosis and treatmentof peripheral artery diseases: document coveringatherosclerotic disease of extracranial carotid andvertebral, mesenteric, renal, upper and lower ex-tremity arteries. Eur Heart J 2011;32:2851–906.

2. Norgren L, Hiatt WR, Dormandy JA, et al. Inter–Society Consensus for the Management ofPeripheral Arterial Disease (TASC II). Eur J VascEndovasc Surgery 2007;33 Suppl 1:S1–75.

3. Schillinger M, Sabeti S, Loewe C, et al. Balloonangioplasty versus implantation of nitinol stents inthe superficial femoral artery. N Engl J Med 2006;354:1879–88.

4. Krankenberg H, Schluter M, Steinkamp HJ, et al.Nitinol stent implantation versus percutaneoustransluminal angioplasty in superficial femoralartery lesions up to 10 cm in length: the femoral

artery stenting trial (FAST). Circulation 2007;116:285–92.

5. Dake MD, Ansel GM, Jaff MR, et al. Sustainedsafety and effectiveness of paclitaxel–elutingstents for femoropopliteal lesions: 2–year follow–up from the Zilver PTX randomized and single–armclinical studies. J Am Coll Cardiol 2013;61:2417–27.

6. Rosenfield K, Jaff MR, White CJ, et al. Trial of apaclitaxel–coated balloon for femoropoplitealartery disease. N Engl J Med 2015;373:145–53.

7. Tepe G, Zeller T, Albrecht T, et al. Local deliveryof paclitaxel to inhibit restenosis during angio-plasty of the leg. N Engl J Med 2008;358:689–99.

8. Krankenberg H, Tubler T, Ingwersen M, et al.Drug–coated balloon versus standard balloon forsuperficial femoral artery in–stent restenosis: theRandomized Femoral Artery In–Stent Restenosis(FAIR) Trial. Circulation 2015;132:2230–6.

9. Byrne RA, Joner M, Alfonso F, et al. Drug–coated balloon therapy in coronary and peripheralartery disease. Nat Rev Cardiol 2014;11:13–23.

10. Cassese S, Byrne RA, Ott I, et al. Paclitaxel–coated versus uncoated balloon angioplasty re-duces target lesion revascularization in patientswith femoropopliteal arterial disease: a meta–analysis of randomized trials. Circ CardiovascInterv 2012;5:582–9.

11. Fusaro M, Cassese S, Ndrepepa G, et al. Pacli-taxel–coated balloon or primary bare nitinol stentfor revascularization of femoropopliteal artery: ameta–analysis of randomized trials versus un-coated balloon and an adjusted indirect compari-son. Int J Cardiol 2013;168:4002–9.

12. Katsanos K, Spiliopoulos S, Karunanithy N,Krokidis M, Sabharwal T, Taylor P. Bayesiannetwork meta–analysis of nitinol stents, covered

mailto:[email protected]:[email protected]://refhub.elsevier.com/S1936-8798(16)30830-5/sref1http://refhub.elsevier.com/S1936-8798(16)30830-5/sref1http://refhub.elsevier.com/S1936-8798(16)30830-5/sref1http://refhub.elsevier.com/S1936-8798(16)30830-5/sref1http://refhub.elsevier.com/S1936-8798(16)30830-5/sref1http://refhub.elsevier.com/S1936-8798(16)30830-5/sref1http://refhub.elsevier.com/S1936-8798(16)30830-5/sref2http://refhub.elsevier.com/S1936-8798(16)30830-5/sref2http://refhub.elsevier.com/S1936-8798(16)30830-5/sref2http://refhub.elsevier.com/S1936-8798(16)30830-5/sref2http://refhub.elsevier.com/S1936-8798(16)30830-5/sref3http://refhub.elsevier.com/S1936-8798(16)30830-5/sref3http://refhub.elsevier.com/S1936-8798(16)30830-5/sref3http://refhub.elsevier.com/S1936-8798(16)30830-5/sref3http://refhub.elsevier.com/S1936-8798(16)30830-5/sref4http://refhub.elsevier.com/S1936-8798(16)30830-5/sref4http://refhub.elsevier.com/S1936-8798(16)30830-5/sref4http://refhub.elsevier.com/S1936-8798(16)30830-5/sref4http://refhub.elsevier.com/S1936-8798(16)30830-5/sref4http://refhub.elsevier.com/S1936-8798(16)30830-5/sref4http://refhub.elsevier.com/S1936-8798(16)30830-5/sref5http://refhub.elsevier.com/S1936-8798(16)30830-5/sref5http://refhub.elsevier.com/S1936-8798(16)30830-5/sref5http://refhub.elsevier.com/S1936-8798(16)30830-5/sref5http://refhub.elsevier.com/S1936-8798(16)30830-5/sref5http://refhub.elsevier.com/S1936-8798(16)30830-5/sref6http://refhub.elsevier.com/S1936-8798(16)30830-5/sref6http://refhub.elsevier.com/S1936-8798(16)30830-5/sref6http://refhub.elsevier.com/S1936-8798(16)30830-5/sref7http://refhub.elsevier.com/S1936-8798(16)30830-5/sref7http://refhub.elsevier.com/S1936-8798(16)30830-5/sref7http://refhub.elsevier.com/S1936-8798(16)30830-5/sref8http://refhub.elsevier.com/S1936-8798(16)30830-5/sref8http://refhub.elsevier.com/S1936-8798(16)30830-5/sref8http://refhub.elsevier.com/S1936-8798(16)30830-5/sref8http://refhub.elsevier.com/S1936-8798(16)30830-5/sref8http://refhub.elsevier.com/S1936-8798(16)30830-5/sref9http://refhub.elsevier.com/S1936-8798(16)30830-5/sref9http://refhub.elsevier.com/S1936-8798(16)30830-5/sref9http://refhub.elsevier.com/S1936-8798(16)30830-5/sref10http://refhub.elsevier.com/S1936-8798(16)30830-5/sref10http://refhub.elsevier.com/S1936-8798(16)30830-5/sref10http://refhub.elsevier.com/S1936-8798(16)30830-5/sref10http://refhub.elsevier.com/S1936-8798(16)30830-5/sref10http://refhub.elsevier.com/S1936-8798(16)30830-5/sref10http://refhub.elsevier.com/S1936-8798(16)30830-5/sref11http://refhub.elsevier.com/S1936-8798(16)30830-5/sref11http://refhub.elsevier.com/S1936-8798(16)30830-5/sref11http://refhub.elsevier.com/S1936-8798(16)30830-5/sref11http://refhub.elsevier.com/S1936-8798(16)30830-5/sref11http://refhub.elsevier.com/S1936-8798(16)30830-5/sref11http://refhub.elsevier.com/S1936-8798(16)30830-5/sref12http://refhub.elsevier.com/S1936-8798(16)30830-5/sref12http://refhub.elsevier.com/S1936-8798(16)30830-5/sref12



stents, drug–eluting stents, and drug–coated bal-loons in the femoropopliteal artery. J Vasc Surg2014;59:1123–33.e8.

13. Cortese B, Granada JF, Scheller B, et al. Drug-coated balloon treatment for lower extremityvascular disease intervention: an internationalpositioning document. Eur Heart J 2016;37:1096–103.

14. Liberati A, Altman DG, Tetzlaff J, et al. ThePRISMA statement for reporting systematicreviews and meta–analyses of studies that eval-uate healthcare interventions: explanation andelaboration. BMJ 2009;339:b2700.

15. Higgins JPT, Green S. Cochrane handbook forsystematic reviews of interventions 5.1.0 [Upda-ted March 2011]. Available at: www.cochrane–handbook.org. 2011. Accessed December 10,2015.

16. DerSimonian R, Laird N. Meta–analysis in clin-ical trials. Control Clin Trials 1986;7:177–88.

17. Borenstein M, Hedges LV, Higgins JPT,Rothstein HR. Introduction to meta–analysis. WestSussex (UK): John Wiley & Sons, 2009.

18. Bagos PG, Nikopoulos GK. Mixed–effectsPoisson regression models for meta–analysis offollow–up studies with constant or varying dura-tions. Int J Biostat 2009;5:1–33.

19. Spiegelhalter D, Abrams K, Myles J. Bayesianapproaches to clinical trials and health–care eval-uation. Statistics in practice. New York: John Wileyand Sons, 2004.

20. Dias S, Sutton AJ, Welton NJ, et al. Evidencesynthesis for decision making 6: embedding evi-dence synthesis in probabilistic cost–effectivenessanalysis. Med Decis Making 2013;33:671–8.

21. Lambert PC, Sutton AJ, Burton PR, et al. Howvague is vague? A simulation study of the impactof the use of vague prior distributions in MCMCusing WinBUGS. Stat Med 2005;24:2401–28.

22. Brooks SP, Gelman A. General methods formonitoring convergence of iterative simulations.J Comput Graph Stat 1998;7:434–55.

23. Wetterslev J, Thorlund K, Brok J, Gluud C. Trialsequential analysis may establish when firm

evidence is reached in cumulative meta–analysis.J Clin Epidemiol 2008;61:64–75.

24. Wetterslev J, Thorlund K, Brok J, et al. Esti-mating required information size by quantifyingdiversity in random–effects model meta–analyses.BMC Med Res Methodol 2009;9:86.

25. Scheinert D, Schulte KL, Zeller T, et al. Pacli-taxel–releasing balloon in femoropopliteal lesionsusing a BTHC excipient: twelve–month resultsfrom the BIOLUX P–I randomized trial. J EndovascTher 2015;22:14–21.

26. Werk M, Langner S, Reinkensmeier B, et al.Inhibition of restenosis in femoropopliteal arteries:paclitaxel–coated versus uncoated balloon:femoral paclitaxel randomized pilot trial. Circula-tion 2008;118:1358–65.

27. Tepe G, Laird J, Schneider P, et al. Drug–coated balloon versus standard percutaneoustransluminal angioplasty for the treatment of su-perficial femoral and popliteal peripheral arterydisease: 12–month results from the IN.PACT SFArandomized trial. Circulation 2015;131:495–502.

28. Scheinert D, Duda S, Zeller T, et al. TheLEVANT I (Lutonix paclitaxel–coated balloon forthe prevention of femoropopliteal restenosis) trialfor femoropopliteal revascularization: first–in–human randomized trial of low–dose drug–coatedballoon versus uncoated balloon angioplasty.J Am Coll Cardiol Intv 2014;7:10–9.

29. Werk M, Albrecht T, Meyer DR, et al. Pacli-taxel–coated balloons reduce restenosis afterfemoro–popliteal angioplasty: evidence from therandomized PACIFIER trial. Circ Cardiovasc Interv2012;5:831–40.

30. Laird JR, Schneider PA, Tepe G, et al. Durabilityof treatment effect using a drug–coated balloon forfemoropopliteal lesions: 24–month results of IN.PACT SFA. J Am Coll Cardiol 2015;66:2329–38.

31. Tepe G, Schnorr B, Albrecht T, et al.Angioplasty of femoral–popliteal arteries withdrug–coated balloons: 5–year follow–up of theTHUNDER trial. J Am Coll Cardiol Intv 2015;8:102–8.

32. Xu B, Qian J, Ge J, et al. Two-year results andsubgroup analyses of the PEPCAD China in-stent

restenosis trial: a prospective, multicenter, ran-domized trial for the treatment of drug-elutingstent in-stent restenosis. Catheter CardiovascInterv 2016;87:624–9.

33. Kufner S, Cassese S, Valeskini M, et al. Long–Term efficacy and safety of paclitaxel–elutingballoon for the treatment of drug–eluting stentrestenosis: 3–year results of a randomized controlledtrial. J Am Coll Cardiol Intv 2015;8:877–84.

34. Dake MD, Ansel GM, Jaff MR, et al. Paclitaxel–eluting stents show superiority to balloon angio-plasty and bare metal stents in femoropoplitealdisease: twelve–monthZilver PTX randomized studyresults. Circ Cardiovasc Interv 2011;4:495–504.

35. Sterne JA, Sutton AJ, Ioannidis JP, et al. Rec-ommendations for examining and interpretingfunnel plot asymmetry in meta–analyses of rand-omised controlled trials. BMJ 2011;343:d4002.

36. Kelsch B, Scheller B, Biedermann M, et al.Dose response to Paclitaxel-coated balloon cath-eters in the porcine coronary overstretch and stentimplantation model. Invest Radiol 2011;46:255–63.

37. Radke PW, Joner M, Joost A, et al. Vasculareffects of paclitaxel following drug-elutingballoon angioplasty in a porcine coronary model:the importance of excipients. EuroIntervention2011;7:730–7.

38. Yazdani SK, Pacheco E, Nakano M, et al.Vascular, downstream, and pharmacokinetic re-sponses to treatment with a low dose drug-coatedballoon in a swine femoral artery model. CatheterCardiovasc Interv 2014;83:132–40.

39. Thompson SG, Higgins JP. How shouldmeta–regression analyses be undertaken andinterpreted? Stat Med 2002;21:1559–73.

KEY WORDS balloon angioplasty,drug-coated balloon, meta-analysis,peripheral arterial disease

APPENDIX For supplemental methods aswell as tables and figures, please see the onlineversion of this article.

http://refhub.elsevier.com/S1936-8798(16)30830-5/sref12http://refhub.elsevier.com/S1936-8798(16)30830-5/sref12http://refhub.elsevier.com/S1936-8798(16)30830-5/sref12http://refhub.elsevier.com/S1936-8798(16)30830-5/sref13http://refhub.elsevier.com/S1936-8798(16)30830-5/sref13http://refhub.elsevier.com/S1936-8798(16)30830-5/sref13http://refhub.elsevier.com/S1936-8798(16)30830-5/sref13http://refhub.elsevier.com/S1936-8798(16)30830-5/sref13http://refhub.elsevier.com/S1936-8798(16)30830-5/sref14http://refhub.elsevier.com/S1936-8798(16)30830-5/sref14http://refhub.elsevier.com/S1936-8798(16)30830-5/sref14http://refhub.elsevier.com/S1936-8798(16)30830-5/sref14http://refhub.elsevier.com/S1936-8798(16)30830-5/sref14http://www.cochrane%96handbook.orghttp://www.cochrane%96handbook.orghttp://refhub.elsevier.com/S1936-8798(16)30830-5/sref16http://refhub.elsevier.com/S1936-8798(16)30830-5/sref16http://refhub.elsevier.com/S1936-8798(16)30830-5/sref17http://refhub.elsevier.com/S1936-8798(16)30830-5/sref17http://refhub.elsevier.com/S1936-8798(16)30830-5/sref17http://refhub.elsevier.com/S1936-8798(16)30830-5/sref18http://refhub.elsevier.com/S1936-8798(16)30830-5/sref18http://refhub.elsevier.com/S1936-8798(16)30830-5/sref18http://refhub.elsevier.com/S1936-8798(16)30830-5/sref18http://refhub.elsevier.com/S1936-8798(16)30830-5/sref19http://refhub.elsevier.com/S1936-8798(16)30830-5/sref19http://refhub.elsevier.com/S1936-8798(16)30830-5/sref19http://refhub.elsevier.com/S1936-8798(16)30830-5/sref19http://refhub.elsevier.com/S1936-8798(16)30830-5/sref20http://refhub.elsevier.com/S1936-8798(16)30830-5/sref20http://refhub.elsevier.com/S1936-8798(16)30830-5/sref20http://refhub.elsevier.com/S1936-8798(16)30830-5/sref20http://refhub.elsevier.com/S1936-8798(16)30830-5/sref21http://refhub.elsevier.com/S1936-8798(16)30830-5/sref21http://refhub.elsevier.com/S1936-8798(16)30830-5/sref21http://refhub.elsevier.com/S1936-8798(16)30830-5/sref21http://refhub.elsevier.com/S1936-8798(16)30830-5/sref22http://refhub.elsevier.com/S1936-8798(16)30830-5/sref22http://refhub.elsevier.com/S1936-8798(16)30830-5/sref22http://refhub.elsevier.com/S1936-8798(16)30830-5/sref23http://refhub.elsevier.com/S1936-8798(16)30830-5/sref23http://refhub.elsevier.com/S1936-8798(16)30830-5/sref23http://refhub.elsevier.com/S1936-8798(16)30830-5/sref23http://refhub.elsevier.com/S1936-8798(16)30830-5/sref24http://refhub.elsevier.com/S1936-8798(16)30830-5/sref24http://refhub.elsevier.com/S1936-8798(16)30830-5/sref24http://refhub.elsevier.com/S1936-8798(16)30830-5/sref24http://refhub.elsevier.com/S1936-8798(16)30830-5/sref25http://refhub.elsevier.com/S1936-8798(16)30830-5/sref25http://refhub.elsevier.com/S1936-8798(16)30830-5/sref25http://refhub.elsevier.com/S1936-8798(16)30830-5/sref25http://refhub.elsevier.com/S1936-8798(16)30830-5/sref25http://refhub.elsevier.com/S1936-8798(16)30830-5/sref26http://refhub.elsevier.com/S1936-8798(16)30830-5/sref26http://refhub.elsevier.com/S1936-8798(16)30830-5/sref26http://refhub.elsevier.com/S1936-8798(16)30830-5/sref26http://refhub.elsevier.com/S1936-8798(16)30830-5/sref26http://refhub.elsevier.com/S1936-8798(16)30830-5/sref27http://refhub.elsevier.com/S1936-8798(16)30830-5/sref27http://refhub.elsevier.com/S1936-8798(16)30830-5/sref27http://refhub.elsevier.com/S1936-8798(16)30830-5/sref27http://refhub.elsevier.com/S1936-8798(16)30830-5/sref27http://refhub.elsevier.com/S1936-8798(16)30830-5/sref27http://refhub.elsevier.com/S1936-8798(16)30830-5/sref28http://refhub.elsevier.com/S1936-8798(16)30830-5/sref28http://refhub.elsevier.com/S1936-8798(16)30830-5/sref28http://refhub.elsevier.com/S1936-8798(16)30830-5/sref28http://refhub.elsevier.com/S1936-8798(16)30830-5/sref28http://refhub.elsevier.com/S1936-8798(16)30830-5/sref28http://refhub.elsevier.com/S1936-8798(16)30830-5/sref28http://refhub.elsevier.com/S1936-8798(16)30830-5/sref29http://refhub.elsevier.com/S1936-8798(16)30830-5/sref29http://refhub.elsevier.com/S1936-8798(16)30830-5/sref29http://refhub.elsevier.com/S1936-8798(16)30830-5/sref29http://refhub.elsevier.com/S1936-8798(16)30830-5/sref29http://refhub.elsevier.com/S1936-8798(16)30830-5/sref30http://refhub.elsevier.com/S1936-8798(16)30830-5/sref30http://refhub.elsevier.com/S1936-8798(16)30830-5/sref30http://refhub.elsevier.com/S1936-8798(16)30830-5/sref30http://refhub.elsevier.com/S1936-8798(16)30830-5/sref31http://refhub.elsevier.com/S1936-8798(16)30830-5/sref31http://refhub.elsevier.com/S1936-8798(16)30830-5/sref31http://refhub.elsevier.com/S1936-8798(16)30830-5/sref31http://refhub.elsevier.com/S1936-8798(16)30830-5/sref31http://refhub.elsevier.com/S1936-8798(16)30830-5/sref32http://refhub.elsevier.com/S1936-8798(16)30830-5/sref32http://refhub.elsevier.com/S1936-8798(16)30830-5/sref32http://refhub.elsevier.com/S1936-8798(16)30830-5/sref32http://refhub.elsevier.com/S1936-8798(16)30830-5/sref32http://refhub.elsevier.com/S1936-8798(16)30830-5/sref32http://refhub.elsevier.com/S1936-8798(16)30830-5/sref33http://refhub.elsevier.com/S1936-8798(16)30830-5/sref33http://refhub.elsevier.com/S1936-8798(16)30830-5/sref33http://refhub.elsevier.com/S1936-8798(16)30830-5/sref33http://refhub.elsevier.com/S1936-8798(16)30830-5/sref33http://refhub.elsevier.com/S1936-8798(16)30830-5/sref34http://refhub.elsevier.com/S1936-8798(16)30830-5/sref34http://refhub.elsevier.com/S1936-8798(16)30830-5/sref34http://refhub.elsevier.com/S1936-8798(16)30830-5/sref34http://refhub.elsevier.com/S1936-8798(16)30830-5/sref34http://refhub.elsevier.com/S1936-8798(16)30830-5/sref35http://refhub.elsevier.com/S1936-8798(16)30830-5/sref35http://refhub.elsevier.com/S1936-8798(16)30830-5/sref35http://refhub.elsevier.com/S1936-8798(16)30830-5/sref35http://refhub.elsevier.com/S1936-8798(16)30830-5/sref36http://refhub.elsevier.com/S1936-8798(16)30830-5/sref36http://refhub.elsevier.com/S1936-8798(16)30830-5/sref36http://refhub.elsevier.com/S1936-8798(16)30830-5/sref36http://refhub.elsevier.com/S1936-8798(16)30830-5/sref36http://refhub.elsevier.com/S1936-8798(16)30830-5/sref37http://refhub.elsevier.com/S1936-8798(16)30830-5/sref37http://refhub.elsevier.com/S1936-8798(16)30830-5/sref37http://refhub.elsevier.com/S1936-8798(16)30830-5/sref37http://refhub.elsevier.com/S1936-8798(16)30830-5/sref37http://refhub.elsevier.com/S1936-8798(16)30830-5/sref38http://refhub.elsevier.com/S1936-8798(16)30830-5/sref38http://refhub.elsevier.com/S1936-8798(16)30830-5/sref38http://refhub.elsevier.com/S1936-8798(16)30830-5/sref38http://refhub.elsevier.com/S1936-8798(16)30830-5/sref38http://refhub.elsevier.com/S1936-8798(16)30830-5/sref39http://refhub.elsevier.com/S1936-8798(16)30830-5/sref39http://refhub.elsevier.com/S1936-8798(16)30830-5/sref39

Drug-Coated Balloon Versus Plain Balloon Angioplasty for the Treatment of Femoropopliteal Artery DiseaseMethodsLiterature search and study selectionPrimary and secondary outcomesStatistical analysesHeterogeneity and publication bias/small study effectSensitivity and subgroup analysesTrial sequential analysis

ResultsDiscussionStudy Limitations

ConclusionsReferences

drug-coated balloon versus plain balloon angioplasty for...

Documents