Study Design

Percutaneous Coronary Interventional

Strategies for the Treatment of Coronary In-

Stent Restenosis: Systematic Review and

Network Meta-analysis

George CM Siontis, MD1; Giulio G Stefanini, MD1; Dimitris Mavridis, PhD2,3;

Konstantinos C Siontis, MD4; Adnan Kastrati, MD5,6; Robert A Byrne, MB,

BCh, PhD 5,6; Fernando Alfonso, MD, PhD7; Bernhard Meier, MD1; Georgia

Salanti, PhD2; Peter Jüni, MD8; Stephan Windecker, MD1

1 Department of Cardiology, Bern University Hospital, Bern 3010, Switzerland. 2 Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina, Greece. 3 Department of Primary Education, University of Ioannina, Ioannina, Greece. 4 Department of Medicine, Mayo Clinic, Rochester, MN, USA. 5 Deutsches Herzzentrum, Technische Universität, Munich, Germany. 6 DZHK (German Centre for Cardiovascular Research), partner site Munich Heart Alliance, Munich, Germany. 7 Cardiac Department, Hospital Universitario de La Princesa Madrid, Madrid, Spain. 8 Institute of Social and Preventive Medicine and Clinical Trials Unit, Department of Clinical Research, University of Bern, Bern, Switzerland.

Contact address : Stephan Windecker, Professor and Chief-of-Cardiology,

Department of Cardiology, Bern University Hospital, 3010 Bern, Switzerland. E-mail:

stephan.windecker@insel.ch

2  

This is the protocol for a systematic review and there is no abstract. We aim to conduct a

systematic review and network meta-analysis of randomized controlled trials (RCTs) that

compare different percutaneous invasive strategies for the treatment of any type of coronary in-

stent restenosis (ISR). To create this protocol we used the proposed template for network meta-

analyses that is available on the following website mtm.uoi.gr.

3  

Background D e s c ri pti o n o f the c o ndi ti o n

Percutaneous coronary interventions (PCI) are routinely performed with the use of bare metal

(BMS) or drug eluting stents (DES) in a broad spectrum of clinical and anatomic settings. Stent

design has continuously improved through technological advances over the last two decades,

with the aim to optimize clinical and angiographic outcomes. Introduced in 2002, DES

represented a paradigm shift in PCI primarily by mitigating the risk of in-stent restenosis (ISR),

defined as a recurrent diameter stenosis >50% within the stented segment or its edges (5-mm

proximal or distal segments adjacent to the stent),(1,2) which has been the major limitation of

BMS.(3-5)

Early-generation DES have been largely replaced by new-generation DES, which have

further improved the safety and efficacy of PCI. (3,6-8) However, ISR requiring repeat

revascularization is still observed in 5-10% of patients undergoing PCI with DES (9) and in 15-

30% of patients undergoing PCI with BMS.(10-12) Although ISR reaches a peak during the first

year after stenting, late restenosis has been described for both DES and BMS.(13,14)

Accumulating evidence suggests important differences not only in ISR frequency but also pattern

of ISR between BMS and DES.(15-17) Of note, the risk of recurrent ISR remains high and

various treatments options including balloon angioplasty (BA), BMS, vascular brachytherapy

(VBT), DES, and drug-coated balloons (DCB) have been applied.(16,17) However, to date the

optimal treatment of ISR has not been determined and it remains unclear whether there are

differences in terms of treatment choice between BMS- and DES-ISR.

D e s c ri pti o n o f the interventions

Numerous treatment strategies have been investigated, including BA, rotablation, VBT, BMS,

DES with different antiproliferative drugs, and DCB. All of these percutaneous catheter-based

interventions aim to restore coronary blood flow by effectively treating in-stent neointimal

hyperplasia and preventing and recurrent restenosis.

BA refers to the intracoronary dilatation with a balloon within the diseased coronary

segment. Rotational atherectomy (rotablation) is a debulking technique whereby a diamond

tipped device spins at high revolutions to remove atheroma or tissue. VBT consists of the local

application of an intracoronary ionizing radiation source at the site of the diseased coronary

4  

segment. BMS are metallic prostheses which are implanted within the diseased coronary

segment. DES are metallic prostheses which – beyond providing scaffolding properties –locally

release antiproliferative agents over time. Finally, with DCB interventions, a balloon coated with

an antiproliferative agent dilates a coronary ISR lesion and the released drug is retained in the

treated tissue.

How the interventions might work

The interventions work by initially restoring flow with mechanical means which can be followed

by interventions to prevent recurrent neointimal hyperplasia and recurrent restenosis. This can be

achieved with the use of antiproliferative drugs in stents or dilatation balloons, or local radiation

therapy. Restoration of coronary flow and prevention of restenosis may reduce symptoms and

incidence of adverse cardiovascular outcomes (such as myocardial infarction) and ultimately

prolong survival.

Why i t i s important to do this re v i e w

Although the incidence of ISR has declined with the systematic use of DES, ISR still occurs in

up to 10% of patients treated with new generation DES within 4 years of the intervention.(9)

Moreover, patients with ISR have a high risk of recurrent restenosis after different

interventions.(18,19) So far, randomized trials have directly compared several of the available

percutaneous catheter-based interventions for the treatment of ISR. However, the absence of

head-to-head trials for most of the treatments and the limited number of patients and trials leave

considerable uncertainty for decision-making and the optimal percutaneous treatment strategy

for patients presenting with ISR is still a matter of debate.(18)

OBJECTIVES

We aim to systematically review the literature and quantitatively synthesize currently available

evidence in order to compare directly and indirectly clinical and angiographic outcomes of

different percutaneous catheter-based coronary interventional strategies for the treatment of

patients with any type of coronary ISR after BMS or DES implantation. We aim to create a

clinically meaningful hierarchy of the effectiveness of available invasive strategies.

5  

Methods

Criteria for considering studies for this review Ty pes o f s tudie s

Randomized controlled trials (RCTs) comparing different invasive strategies for the treatment of

coronary ISR will be identified through a broad systematic literature search. We will include trials

that investigated any type of PCI strategy under different clinical settings for the treatment

of ISR. Trials with 2 or more arms of interventions irrespective of the type of ISR (BMS or DES

related) will be eligible. Trials with 2 or more arms for which a subset of interventions satisfy the

inclusion criteria will be kept in the analysis after having discarded those arms that do not satisfy

inclusion criteria. No language or year restrictions will be applied. From our systematic review,

we will exclude studies that allowed a mixture of interventions of interest in one study arm (i.e.,

balloon angioplasty, rotablation, or stenting); and studies that were not completed at the time of

our search (October 2014).

Ty pes o f patients inc luded

Eligible studies will include patients with any type of coronary ISR (following BMS or DES

implantation) that present with stable coronary artery disease (CAD) or an acute coronary

syndrome (ACS) and receive a percutaneous invasive strategy as treatment.

Ty pes o f interventions

We will consider trials that compare two or more of the following percutaneous invasive

strategies for the treatment of coronary ISR after BMS or DES implantation: BA, VBT, BMS, as

well as DES and DCB with different antiproliferative drugs. We will not consider studies that

examined cutting balloon or laser atherectomy as the only interventions in a study arm since

these interventions are exceedingly rare as single therapies. All possible comparisons are

illustrated in the network plots in Figure 1A and 1 B . Nodes refer to interventions and edges

between nodes refer to the fact that there are studies directly comparing the interventions defined

by the nodes. We will include any type of ISR (BMS- and DES-ISR) in the same analysis, while

analyses for BMS- and DES-ISR will be also performed separately. For the primary and any

additional analyses we will consider DES separately according to the type of antiproliferative

agent but we will also present analyses grouping the different DES types together. In case that

other than the pre-specified percutaneous coronary interventions will be identified through our

6  

search and deemed eligible, these will be also considered and included in our analysis. We are

mainly interested in the following two interventions of different types of DES versus DCB for

the treatment of ISR. We will also include supplementary interventions in the network (i.e. VBT,

rotablation) to increase the amount of available (indirect) information in the analysis.(20)

Ty pes o f outcome measures

As most of the clinical trials have used primary angiographic endpoints for the assessment of the

efficacy of ISR therapies, percent diameter stenosis was chosen as the primary endpoint in the

present study to provide sufficient precision and to arrive at a conclusive answer. It is known that

currently applied interventions for ISR have little effect on hard clinical outcomes, and most trials

have been underpowered to detect a difference for such outcomes. Percent diameter

stenosis is a most sensitive endpoint for the evaluation of angiographic effectiveness as

compared to binary restenosis, which is included as a secondary endpoint in our analysis.

More specifically, we will estimate the relative ranking of the competing interventions according

to the following outcomes:

Primary outcome:

• Percent diameter stenosis (% DS) at follow-up. Secondary outcomes:

• Binary restenosis (diameter stenosis of 50% or greater at follow-up angiography on the basis of the “in-segment” analysis)

• Target lesion revascularization (TLR)

• All-cause mortality

• Myocardial infarction (MI) Target-vessel MI and clinically-driven TLR will be extracted whenever available. Finally, when

TLR is not reported, we will extract target vessel revascularization (TVR) and consider it

equivalent to TLR.

Search methods for identification of studies Trials comparing any eligible invasive strategy for the treatment of coronary ISR will be

identified through a systematic literature search of the following databases: PubMed, EMBASE,

and Cochrane Library Central Register of Controlled Trials (CENTRAL); while the

7  

ClinicalTrials.gov and Current Controlled Trials (controlled-trials.com) registries will be

additionally scrutinized to ensure identification of all published trials. A modified search

algorithm will be developed and adapted for each database with a combination of relevant text

terms and key words (as provided in Appendix ). Potentially eligible trials that have been only

presented at conferences and of which a full manuscript is not yet available at the time of our

search, will be considered eligible and the principal investigators will be contacted to participate

in this meta- analysis by providing the required data. We will scrutinize for additional eligible

studies the reference lists of the retrieved publications, and relevant meta-analyses in the field.

No language, year of publication, or sample size restrictions will be applied.

Data collection and analysis S e l e c ti o n o f s tudie s

The aforementioned search strategy will be performed for each database separately and search

results will be entered in EROS (Early Review Organizing Software) (www.eros-systematic-

review.org), a web-based software designed specifically for the first stages of a systematic

review. Using this software, 2 investigators will scrutinize all the entries for eligibility in title

and abstract level.

Data ex trac tion and management

Two investigators will perform study identification, data abstraction, and risk of bias

assessment independently. Any discrepancies will be resolved by consensus and arbitration by a

third investigator.

We will review the main report, subsequent publications of each trial and any

supplementary material of the included trials and extract the following data:

• study characteristics: first author, study acronym, year of publication, interventions of

interest, study arms, number of patients in each arm of randomisation, and the sponsor of

the study.

• patient data: age, gender, comorbidities, and clinical indication for coronary intervention (stable angina or ACS).

• lesion characteristics: target vessel, bifurcations, ostial lesions, total occlusion, and index stent type of restenosis (BMS- or DES-ISR).

• quantitative angiographic measurements (in-segment and in-stent): before procedure [lesion length (mm), reference vessel diameter (RVD), minimal luminal diameter (MLD),

8  

and % of diameter stenosis (%DS)]; post-procedure [MLD and %DS]; and at available

angiographic follow-up [late luminal loss (LLL), MLD, %DS, and binary restenosis]. The

pre-specified time of angiographic and clinical follow-up will be recorded for each trial,

while for the clinical endpoints of interest we will consider the longest available follow-

up period.

• clinical outcomes: the clinical outcomes of interest will be extracted. Outcome data

From each eligible trial we will extract the number of individuals randomized to each arm. We

will also extract any summary metric (mean (standard deviation) or median (interquartile range))

that is reported for each arm of intervention for the main angiographic outcome of interest at

follow-up (percent diameter stenosis). For the secondary outcomes of interest (overall mortality,

myocardial infarction, target lesion revascularisation, binary restenosis), the respective number

of participants in each arm of interventions and the number of participants with one of the events

of interest, will be also recorded.

Data on potential e f f e c t mo di f i e rs

We will report separate effect estimates for the primary outcome of interest only and the main

network (including any type of ISR and different types of DES separately (Figure 1B ))

with respect to the following potential effect modifiers:

1. Year of study publication. 2. Sample size of the trial. 3. Clinical indication for coronary intervention (stable angina or ACS) 4. Index stent type of restenosis (BMS- or DES-ISR)

Assessment of risk of bias in included studies We will investigate study limitations using the Cochrane risk of bias tool to evaluate the internal

validity and conduct of included studies. Two authors will independently assess risk of bias for

each included study using the risk of bias assessment tool.(21) We will resolve any

disagreements with discussion and involvement of a third author. Each item will be described as

being at low, high or unclear risk of bias. The areas that will be evaluated are:

random sequence generation: Was there adequate sequence generation (selection bias)?

allocation concealment: Was allocation adequately concealed (selection bias)?

9  

blinding: Was knowledge of the allocated intervention adequately prevented during the

study (detection bias)?

o participants and personnel

o outcome assessors incomplete outcome data: Were incomplete outcome data adequately addressed (attrition

bias)?

selective outcome reporting: Are reports of the study free of possible selective outcome

reporting (reporting bias)?

We will evaluate the aforementioned items not only in each study but also in each pairwise

comparison.(22) We will classify each piece of direct evidence in the network as low, moderate,

or high risk of bias. We will illustrate these assessments in the network plot for the primary

outcome with colored edges according to the risk of bias. We will also produce the contribution

matrix which gives the percentage contribution of each direct estimate to the network meta-

analysis estimates.(23) This will help to delineate the contribution of direct and indirect evidence

to each network meta-analysis estimate.

Measures of treatment effect Re l ati v e treatment e f f e c ts

First, we will conduct a pair-wise meta-analysis by synthesizing at least two studies that compare

the same interventions using a random-effects model (24) in STATA 12. A random-effects model

assumes that different studies assessed different yet related treatment effects. We will

estimate the pairwise relative treatment effects of the competing interventions using standardized

mean differences for the primary outcome (percent diameter stenosis). We will estimate odds

ratios (OR) for dichotomous outcomes. We will produce summary results for all outcomes and

give 95% confidence intervals. We will use restricted maximum likelihood to estimate

heterogeneity. If we have skewed data, we will use the methods presented in to pool results.(25)

Re l ati v e treatment ranking

We will estimate the ranking probabilities for all treatments of being at each possible rank for

each intervention using the mvmeta command in STATA.(26) We will obtain a hierarchy of the

competing interventions using rankograms.(27) We will obtain a treatment hierarchy using the

10  

surface under the cumulative ranking curve (SUCRA) and mean ranks.(28) We will produce the

relevant plots using the suite of STATA commands by Chaimani et al.(28)

U ni t o f anal y s i s i s s ue s

We expect that some studies will not report mean values and standard deviations (SD) but would

report certain quantiles instead. If a study reports the median, minimum, and maximum values,

we will use the methodology from (29) to estimate the respective mean and SD of the study

population. We will include also those studies that report the median and the interquartile range

(IQR) assuming that data are normally distributed and the standard deviation would be

s=IQR/1.35 while the mean would be equal to the median. However, use of median and IQR is

usually an indicator that data are not normal (Cochrane handbook chapter 7, available from

www.cochrane-handbook.org).(30) We will repeat the analysis excluding these studies as a

sensitivity analysis for the main network.

Studie s w i th mul tipl e treatment groups

We will take into account the correlations between effect sizes measured within a single trial. Assessment of reporting biases For each pairwise comparison that includes at least 10 trials we will draw funnel plots and

compute Egger’s test to test visually and statistically for small-study effects. For these

comparisons we will draw contour-enhanced funnel plots to disentangle small study effects from

publication bias.(31) We will also draw a comparison-adjusted funnel plot to explore for small

study effects assuming that small study effects favour the novel treatment.(28,32) In case

publication bias is suspected (by funnel plot asymmetry) and enough studies are included in the

synthesis, we will also apply a selection model to explore if probability of publication is

associated with magnitude of effect.(33)

Dealing with missing data Missing data and dropouts will be assessed in all included studies. Details and characteristics of

dropouts will be investigated and reported. We will explore if reasons for missing data are related

to the actual outcome and if missing data are balanced in the intervention arms. We will

extract if authors accounted for missing data and which method they used to do so. If the authors

have used naïve imputation methods (best/worst case scenarios, mean imputation), we will

11  

consider the study to be at high risk of bias whereas if they have used methods that take into

account the fact that data have been imputed using statistical methods that take uncertainty in the

imputed values into account (e.g. multiple imputation) we will consider the study to be at low

risk of bias.

Assessment of clinical and methodological heterogeneity within treatment

comparisons There are three different types of heterogeneity, namely clinical, methodological and statistical

heterogeneity (Cochrane handbook chapter 9).(34) To evaluate the presence of clinical

heterogeneity we will generate descriptive statistics for trial and study population characteristics

across all eligible trials that compare each pair of interventions. We will assess the presence of

clinical heterogeneity within each pairwise comparison by comparing these characteristics. We

will assess methodological heterogeneity by evaluating the design of the studies. Statistical

heterogeneity refers to differences in true effect sizes.

Assessment of transitivity across treatment comparisons Although participants are randomized within a study, treatment strategy comparisons are not

randomized across studies. We assume that an intervention is missing from a trial for reasons

not associated with its relative effectiveness and any patient that meets the inclusion criteria is, in

principle, equally likely to be randomized to any of the eligible interventions.(35,36) This is a key

assumption in network meta-analysis called transitivity. It states that we can genuinely learn

about the relative effectiveness between two treatments via an indirect route. If for example the

treatments “rotablation” and “BMS” are both directly compared to “BA”, then we can assess

“rotablation” vs. “BMS” indirectly through “BA”. This assumption entails that “BA” is similar

when it appears in “BA” vs. “rotablation” and “BA” vs. “BMS” trials and also that the

distribution of effect modifiers is similar in “BA” vs. “rotablation” and “BA” vs. “BMS” trials.

This will be extended in all treatment comparisons. We will assess the assumption of transitivity

by comparing the distribution of the potential effect modifiers across the different pairwise

comparisons. We suspect that year of study publication, sample size of the trial, clinical

indication for coronary intervention (stable angina or ACS), and index stent type of restenosis

(BMS- or DES-ISR) can be effect modifiers and we will explore if the distribution of these

differs across treatment comparisons. We assume that the most common treatment (BA) that

would be probably used for indirect comparisons is similar when it appears in different trials.

12  

The assessment of statistical heterogeneity in the entire network will be based on the

magnitude of the heterogeneity variance parameter (τ2) estimated from the network meta-analysis models. We will also estimate a total I-squared value for heterogeneity in the network as described elsewhere.(37)

Data synthesis Methods f o r direc t treatment compari sons

We will conduct pairwise meta-analyses in Stata by assuming a random effects model

(DerSimonian1986) for every treatment comparison with at least two studies.

Methods f o r indi rec t and mi x e d compari sons

We will use network meta-analysis to compare different percutaneous coronary interventions for

coronary in-stent restenosis. Network meta-analysis synthesizes both direct and indirect

evidence, estimates the relative effectiveness between pair of interventions even if these

interventions have never been compared directly in RCTs and provides a ranking of

interventions.(38-41) For a comparison, for example BA vs. rotablation, direct evidence is

provided by trials directly comparing these two interventions whereas indirect evidence is

provided if there is an indirect path linking these two treatments (e.g., if both BA and rotablation

are compared to BMS).(42) By combining direct and indirect evidence we obtain estimates with

increased precision. We will perform network meta-analysis in Stata using the mvmeta command

(37,43) and self-programmed Stata routines available at http://www.mtm.uoi.gr/index.php/stata-

routines-for-network-meta-analysis.(28) We will use the restricted maximum likelihood method

to estimate heterogeneity assuming a common estimate for the heterogeneity variance across the

different comparisons.

Assessment of statistical heterogeneity A s s umpti o ns w he n e s timating he te ro g e ne i ty

In standard pairwise meta-analyses, we assume different heterogeneity estimates for different

comparisons. In network meta-analysis we assume that heterogeneity is the same for all

treatment comparisons. We estimate heterogeneity using restricted maximum likelihood both in

pairwise and network meta-analysis.

Measures and te s ts f o r he te ro g e ne i ty

13  

We will assess statistical heterogeneity visually by inspecting the forest plot for each pairwise

comparison. We will also compute the I! and its 95% confidence interval that shows the

percentage of variation that is not attributed to random error.

The assessment of statistical heterogeneity in the entire network will be based on the

magnitude of the heterogeneity variance parameter (𝜏!) estimated from the network meta-

analysis models. For dichotomous outcomes magnitude of heterogeneity variance will be

compared with the empirical distribution as derived by Turner. We will also estimate a total 𝛪! 

value for heterogeneity as described elsewhere.

Assessment of statistical inconsistency Lo cal approaches f o r evaluating i nc o ns i s te nc y

To evaluate the presence of inconsistency locally we will use the loop-specific approach.(42)

This method evaluates the consistency assumption in each closed loop of the network separately

as the difference between direct and indirect estimates for a specific comparison in the loop

(inconsistency factor). Then, the magnitude of the inconsistency factors and their 95% CIs can be

used to infer about the presence of inconsistency in each loop. We will assume a common

heterogeneity estimate within each loop.(44) We will present the results of this approach

graphically in a forest plot using the ifplot command in Stata.(28) Dias et al. suggested a method

which separates evidence on a particular comparison into ‘direct’ and ‘indirect’.(45) This ‘node-

splitting’ method excludes one direct comparison at a time and estimates the indirect treatment

effect for the excluded comparison. The direct and indirect evidence is subsequently compared to

see if direct and indirect evidence is in agreement.

Global approaches f o r evaluating i nc o ns i s te nc y

To check the assumption of consistency in the entire network we will use the ‘design-by-

treatment’ model as described by Higgins and colleagues.(46) This method accounts for different

source of inconsistency that can occur when studies with different designs (two-arm trials vs.

three-arm trials) give different results as well as disagreement between direct and indirect

evidence. Using this approach we will infer about the presence of inconsistency from any source

in the entire network based on a chi-square test. The design-by-treatment model will be

performed in Stata using the mvmeta command.

Inconsistency and heterogeneity are interweaved; to distinguish between these two sources

of variability we will employ the I-squared for inconsistency (47) that measures the percentage

14  

of variability that cannot be attributed to random error or heterogeneity (within comparison

variability).

Inv e s ti g ati o n o f he te ro g e ne i ty and i nc o ns i s te nc y

If we find important heterogeneity or/and inconsistency, we will explore the possible sources. If

sufficient studies are available, we will perform meta-regression or subgroup analyses for the

primary outcome by using the following effect modifiers as possible sources of inconsistency

and or heterogeneity: (i) year of study publication; (ii) sample size of the trial; (iii) clinical

indication for coronary intervention (stable angina or ACS); (iv) index stent type of restenosis

(BMS- or DES-ISR).

Sensitivity analyses For the primary angiographic outcome, we will repeat the analysis including those studies that

report only the median and the IQR assuming that data are normally distributed. For the primary

outcome and the main network we will repeat the analysis excluding those studies that are at

high or unclear risk of bias.

15  

Figures Figure 1 : Network plots illustrating all possible comparisons of the available interventions

for the treatment of any type of ISR. In panel A, any type of DES has been included in the same

group, while in panel B different types of DES have been included separately.

16  

Appendices Appendix 1 : Search algori thms .

Database Search algorithm

PUBMED #1 randomized controlled trial [pt]

#2 controlled clinical trial [pt]

#3 randomized [tiab]

#4 clinical trials as topic [mesh: noexp]

#5 randomly [tiab]

#6 trial [ti]

#7 restenosis [tiab]

#8 re-stenosis [tiab]

#9 in-stent restenosis [tiab]

#10 #1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9

#11 ((#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9) AND

coronar* [tiab]

CENTRAL #1 coronar*: ti,ab,kw

#2 re*stenos*: ti,ab,kw

#3 random*: ti,ab,kw

#4 *stent*: ti,ab,kw

#5 #1 AND #2

#6 #3 AND #4 AND #5

EMBASE #1 in*stent re*stenos*s/mj

#2 coronar*:ti,ab

#3 random*:ti,ab

#4 (#1 near/5 #2):ti,ab

#5 #3 AND #4

17  

References

1. Mehran R, Dangas G, Abizaid AS et al. Angiographic patterns of in-stent restenosis: classification and implications for long-term outcome. Circulation 1999;100:1872-8.

2. Kuntz RE, Baim DS. Defining coronary restenosis. Newer clinical and angiographic paradigms. Circulation 1993;88:1310-23.

3. Bangalore S, Kumar S, Fusaro M et al. Short- and long-term outcomes with drug-eluting and bare-metal coronary stents: a mixed-treatment comparison analysis of 117 762 patient-years of follow-up from randomized trials. Circulation 2012;125:2873-91.

4. Stettler C, Allemann S, Wandel S et al. Drug eluting and bare metal stents in people with and without diabetes: collaborative network meta-analysis. BMJ 2008;337:a1331.

5. Stettler C, Wandel S, Allemann S et al. Outcomes associated with drug-eluting and bare-metal stents: a collaborative network meta-analysis. Lancet 2007;370:937-48.

6. Stefanini GG, Baber U, Windecker S et al. Safety and efficacy of drug-eluting stents in women: a patient-level pooled analysis of randomised trials. Lancet 2013;382:1879-88.

7. Raber L, Magro M, Stefanini GG et al. Very late coronary stent thrombosis of a newer-generation everolimus-eluting stent compared with early-generation drug-eluting stents: a prospective cohort study. Circulation 2012;125:1110-21.

8. Stefanini GG, Holmes DR, Jr. Drug-eluting coronary-artery stents. N Engl J Med 2013;368:254-65.

9. Taniwaki M, Stefanini GG, Silber S et al. 4-year clinical outcomes and predictors of repeat revascularization in patients treated with new-generation drug-eluting stents: a report from the RESOLUTE All-Comers trial (A Randomized Comparison of a Zotarolimus-Eluting Stent With an Everolimus- Eluting Stent for Percutaneous Coronary Intervention). J Am Coll Cardiol 2014;63:1617-25.

10. Go AS, Mozaffarian D, Roger VL et al. Heart disease and stroke statistics-- 2014 update: a report from the American Heart Association. Circulation 2014;129:e28-e292.

11. Morice MC, Urban P, Greene S, Schuler G, Chevalier B. Why are we still using coronary bare-metal stents? J Am Coll Cardiol 2013;61:1122-3.

12. Dehmer GJ, Weaver D, Roe MT et al. A contemporary view of diagnostic cardiac catheterization and percutaneous coronary intervention in the United States: a report from the CathPCI Registry of the National Cardiovascular Data Registry, 2010 through June 2011. J Am Coll Cardiol 2012;60:2017-31.

13. Yamaji K, Inoue K, Nakahashi T et al. Bare metal stent thrombosis and in- stent neoatherosclerosis. Circ Cardiovasc Interv 2012;5:47-54.

18  

14. Byrne RA, Iijima R, Mehilli J et al. Durability of antirestenotic efficacy in drug-eluting stents with and without permanent polymer. JACC Cardiovasc Interv 2009;2:291-9.

15. Byrne RA, Cassese S, Windisch T et al. Differential relative efficacy between drug-eluting stents in patients with bare metal and drug-eluting stent restenosis; evidence in support of drug resistance: insights from the ISAR- DESIRE and ISAR-DESIRE 2 trials. EuroIntervention 2013;9:797-802.

16. Byrne RA, Joner M, Tada T, Kastrati A. Restenosis in bare metal and drug- eluting stents: distinct mechanistic insights from histopathology and optical intravascular imaging. Minerva Cardioangiol 2012;60:473-89.

17. Alfonso F. Treatment of drug-eluting stent restenosis the new pilgrimage: quo vadis? J Am Coll Cardiol 2010;55:2717-20.

18. Alfonso F, Byrne RA, Rivero F, Kastrati A. Current Treatment of In-Stent Restenosis. J Am Coll Cardiol 2014;63:2659-2673.

19. Latib A, Mussardo M, Ielasi A et al. Long-term outcomes after the percutaneous treatment of drug-eluting stent restenosis. JACC Cardiovasc Interv 2011;4:155-64.

20. Ades AE, Caldwell DM, Reken S, Welton NJ, Sutton AJ, Dias S. Evidence synthesis for decision making 7: a reviewer's checklist. Med Decis Making 2013;33:679-91.

21. Higgins JP, Altman DG, Gotzsche PC et al. The Cochrane Collaboration's tool for assessing risk of bias in randomised trials. BMJ 2011;343:d5928.

22. Balshem H, Helfand M, Schunemann HJ et al. GRADE guidelines: 3. Rating the quality of evidence. J Clin Epidemiol 2011;64:401-6.

23. Krahn U, Binder H, Konig J. A graphical tool for locating inconsistency in network meta-analyses. BMC Med Res Methodol 2013;13:35.

24. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177-88.

25. Higgins JP, White IR, Anzures-Cabrera J. Meta-analysis of skewed data: combining results reported on log-transformed or raw scales. Stat Med 2008;27:6072-92.

26. White IR. Multivariate random-effects meta-analysis. Stata Journal 2009;9:40- 56.

27. Salanti G, Ades AE, Ioannidis JP. Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial. J Clin Epidemiol 2011;64:163-71.

28. Chaimani A, Higgins JP, Mavridis D, Spyridonos P, Salanti G. Graphical tools for network meta-analysis in STATA. PLoS One 2013;8:e76654.

29. Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Med Res Methodol 2005;5:13.

30. Higgins JPT DJeCSsacdIHJ, Green S (editors), Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011). The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org.

19  

31. Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Contour-enhanced meta-analysis funnel plots help distinguish publication bias from other causes of asymmetry. J Clin Epidemiol 2008;61:991-6.

32. Chaimani A, Salanti G. Using network meta-analysis to evaluate the existence of small-study effects in a network of interventions. . Res Synth Method 2012;3:161-176.

33. Mavridis D, Sutton A, Cipriani A, Salanti G. A fully Bayesian application of the Copas selection model for publication bias extended to network meta- analysis. . Statist Med 2013;32:51-66.

34. Deeks JJ, Higgins JPT, Altman DG (editors). Chapter 9: Analysing data and undertaking meta-analyses. In: Higgins JPT, Green S (editors). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011). The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org.

35. Salanti G, Kavvoura FK, Ioannidis JP. Exploring the geometry of treatment networks. Ann Intern Med 2008;148:544-53.

36. Salanti G. Indirect and mixed-treatment comparison, network, or multiple- treatments meta-analysis: many names, many benefits, many concerns for the next generation evidence synthesis tool. Research Synthesis Methods 2012;3:17.

37. White IR. Consistency and inconsistency in network meta-analysis : model estimation using multivariate meta-regression. Research Synthesis Methods 2012;3:111-125.

38. Cipriani A, Higgins JP, Geddes JR, Salanti G. Conceptual and technical challenges in network meta-analysis. Ann Intern Med 2013;159:130-7.

39. Salanti G, Higgins JP, Ades AE, Ioannidis JP. Evaluation of networks of randomized trials. Stat Methods Med Res 2008;17:279-301.

40. Caldwell DM, Ades AE, Higgins JP. Simultaneous comparison of multiple treatments: combining direct and indirect evidence. BMJ 2005;331:897-900.

41. Lu G, Ades AE. Combination of direct and indirect evidence in mixed treatment comparisons. Stat Med 2004;23:3105-24.

42. Bucher HC, Guyatt GH, Griffith LE, Walter SD. The results of direct and indirect treatment comparisons in meta-analysis of randomized controlled trials. J Clin Epidemiol 1997;50:683-91.

43. White IR. Multivariate random-effects meta-regression: Updates to mvmeta. The Stata Journal 2011;11:255-270.

44. Veroniki AA, Vasiliadis HS, Higgins JP, Salanti G. Evaluation of inconsistency in networks of interventions. Int J Epidemiol 2013;42:332-45.

45. Dias S, Welton NJ, Caldwell DM, Ades AE. Checking consistency in mixed treatment comparison meta-analysis. Stat Med 2010;29:932-44.

46. Higgins JP JD, Barrett JK, Lu G, Ades AE, White R. Consistency and inconsistency in network meta-analysis: concepts and models for multi-arm studies. Research Synthesis Methods 2012;3:98-110.

20  

47. Jackson D, Barrett JK, Rice S, White IR, Higgins JP. A design-by-treatment interaction model for network meta-analysis with random inconsistency effects. Stat Med 2014.