Stettler Windecker Meta Analysis 18k 10 20 07 V04

A Pooled Analysis of 38 Trials Comparing A Pooled Analysis of 38 Trials Comparing Drug-Eluting Stents and Bare-Metal StentsDrug-Eluting Stents and Bare-Metal Stents

Funding and Validation of Statistical Models The Swiss National Science Foundation

Stettler C., et al., Lancet 2007;370:937-48.

Outcomes Associated with Drug-Eluting and Bare-Metal Stents:

A Collaborative Network Meta-Analysis

IntroductionIntroduction• The long-term safety of drug-eluting stents (DES) has been

questioned by recent studies reporting rates of death, myocardial infarction (MI), or late stent thrombosis (ST) compared with bare-metal stents (BMS)

– Daemen J, et al., Lancet 2007; 369(9562):667-78.– Pfisterer M, et al., J Am Coll Cardiol 2006; 48(12):2584-91.– Nordmann AJ, et al.,. Eur Heart J 2006; 27(23):2784-814.– Camenzind E, et al., Circulation 2007; 115(11):1440-55.– Katritsis DG, Am J Cardiol 2005;95(5):640-3.

• However, these studies were hampered by low numbers of patients limited duration of follow-up, or observational studies

• A subsequent series of pooled analyses of randomized trials comparing DES with BMS in a single issue of NEJM in February 2007 found no evidence for death or MI rates with DES and inconsistent evidence for risk of LST with DES


Rationale for Network AnalysisRationale for Network Analysis• Novel analytical techniques called network meta-

analysis1,2 or mixed treatment comparison3,4,5 allow a unified, coherent analysis of the entire set of all randomized controlled trials comparing either of the two DES with BMS, or performed a head-to-head comparison of SES and PES, while fully respecting randomization1 Lumley T. Stat Med 2002;21(16):2313-24.2 Psaty BM, et al., JAMA 2003;289(19):2534-44.3 Lu G, Ades AE. Stat Med 2004;23(20):3105-24.4 Caldwell DM, et al., BMJ 2005;331(7521):897-900.5 and Cooper NJ, et al., Arch Intern Med 2006;166(12):1269-75.

• The authors established a collaborative group of investigators, who provided trial data based on standardized definitions of outcomes, and conducted a network meta-analysis


ObjectivesObjectives

• Analysis of all 38 randomized controlled trials, including 18,023 patients, was conducted by the Swiss National Science Foundation in an effort to increase the precision of estimates for effectiveness and safety of DES compared to BMS

• A subgroup analysis of mortality in 3,870 diabetic patients was conducted to address recent concerns about SES in this population


MethodsMethods

• Literature searches were performed in MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials for relevant studies in any language from inception to March 2007

• Websites dedicated to dissemination of results from cardiovascular trials (www.acc.org, www.tctmd.com, www.theheart.org, www.clinicaltrialresults.org) were also searched

• Other sources included conference abstracts, relevant book chapters, proceedings of relevant FDA advisory panels

• 2 investigators independently assessed reports for eligibility


http://www.acc.org/

http://www.tctmd.com/

http://www.theheart.org/

http://www.clinicaltrialresults.org/

Study Eligibility CriteriaStudy Eligibility Criteria

• Randomized controlled trials (RCT)

• Patients with signs of myocardial ischemia due to coronary artery disease

• Comparison of SES and PES

• Comparison of either DES with BMS

• Clinical follow-up duration of at least 6 months


Identification of EligibleIdentification of EligibleRandomized TrialsRandomized Trials

870 potentially eligible reportsidentified and screened for retrieval

786 reports excluded:424 with different interventions254 reviews or pooled analyses

98 observational studies10 case reports

84 reports retrieved fordetailed evaluation (41 trials)

6 reports excluded (3 trials):2 (1 trial) with different interventions

4 (2 trials) subgroup analyses

78 reports included(38 trials with 40 comparisons):

7 PES vs. BMS16 SES vs. BMS14 PES vs. SES

1 PES vs. SES vs. BMS


16 SES vs. BMS Trials Included16 SES vs. BMS Trials Included

RAVEL 238 Late Lumen Loss 18 2.5-3.5 19SIRIUS 1,058 CD*, AMI, TVR 15-30 2.5-3.5 53E-SIRIUS 352 Min Lumen Diam 15-32 2.5-3.0 35C-SIRIUS 100 Min Lumen Diam 15-32 2.5-3.0 8SES-SMART 257 Binary Restenosis 33 < 2.75 20DIABETES 160 Late Lumen Loss NR < 4.0 4Pache et al 500 Binary Restenosis NR NR 2PRISON II 200 Binary Restonsis NR NR 2SCANDSTENT 322 Min Lumen Diam 15* 2.25-4.5 4TYPHOON 715 VRD**, AMI, TVR 30 2.25-3.5 48SESAMI 320 Binary Restenosis NR NR 1DECODE 83 Late Lumen Loss NA NA NASCORPIUS 200 Late Lumen Loss 42 2.5-3.5 16RRISC 75 Late Lumen Loss 66 2.5-4.0 1MISSION 308 Late Lumen Loss NA NA 1Ortolani et al. 104 Late Lumen Loss 28 NA 1

* Cardiac death** Vessel-related death

SES vs. BMS Study N Primary Endpoint Lesion

Length (mm)RVD(mm)

# ofSites


7 PES vs. BMS Trials Included7 PES vs. BMS Trials Included

TAXUS I 61 Death, AMI, TVR, ST 12 3.0-3.5 3

TAXUS II 536 Neointimal Proliferation 12 3.0-3.5 38

TAXUS IV 1314 TVR 10-28 2.5-3.75 73TAXUS V 1172 TVR 10-46 2.25-4.0 66TAXUS VI 446 TVR 18-40 2.5-3.75 44

PASSION 619 Cardiac Death, AMI, TLR NR > 2.5 2

HAAMU-STENT 164 NA NA NA 1

PES vs. BMS Study N Primary Endpoint Lesion

Length (mm)RVD(mm)

# ofSites


15 SES vs. PES Trials Included15 SES vs. PES Trials Included

TAXi 202 Death, AMI, TLR, ST NR NR 1ISAR-DESIRE 200 Binary Restenosis NR NR 2ISAR-DIABETES 250 Late lumen Loss NR NR 2SIRTAX 1012 MACE NR 2.25-4.0 2CORPAL 652 Binary Restenosis < 20 < 2.5 2REALITY 1353 Binary Restenosis > 15 2.25-3.0 90ISAR-SMART-3 360 Late Lumen Loss NR < 2.8 2Zhang et al 449 Death, AMI, TVR NA 2.5-3.5 1LONG DES II 500 Binary Restenosis 25 2.5 5PROSIT 231 Late Lumen Loss NA NA NA

SORT OUT II 2098 Cardiac Death, AMI, TLR, TVR, TVF NA NA 5

Cervinka et al. 70 NIH > 20 < 2.5 1Petronio et al. 100 NIH 16 2.5-3.7 1Han et al. 416 NA NA NA NA

SES vs. PES Study N Primary Endpoint Lesion

Length (mm)RVD(mm)

# ofSites

BASKET 826 Cardiac Death, MI, TVR NR 4 1

SES vs. PES vs. BMS Study N Primary Endpoint Lesion

Length (mm)RVD(mm)

# ofSites


Key Baseline Predictors of Restenosis in Key Baseline Predictors of Restenosis in Complex DES vs. BMS Trials* Complex DES vs. BMS Trials*

SES-SMART1 SES RVD 2.75 257 25 13.01 ± 6.53 16.99 ± 5.71 2.22 ± 0.29PRISON II2 SES CTO 200 13 16.0 ± 9.3** 31.9 ± 15.3 2.53 ± 0.67RRISC3 SES SVG 75 15 NA 23.4 ± 7.0 2.38 ± 0.57DECODE4 SES

DM83 100 15.06 ± 6.34 20.9 ± 8.45 2.51 ± 0.35

SCORPIUS5 SES 200 100 11.2 ± 4.6 19.6 ± 9.0 2.54 ± 0.45DIABETES6 SES 160 100 14.6 ± 8 NA 2.33 ± 0.6SCANDSTENT7 SES

High-Risk†

322 18 18.8 ± 13.0 26 2.86 ± 0.53TAXUS V8 PES 1172 32 17.3 ± 9.0 28.7 ± 13.2 2.68 ± 0.58TAXUS VI9 PES 446 20 20.94 ± 7.21 33.7 ± 10.79 2.81 ± 0.48TYPHOON10 SES

AMI

715 16 NA 22.1 ± 8.6 2.78 ± 0.50SESAMI11 SES 320 21 NA 16.9 ± 4.1 NAMISSION12 SES 308 NA NA 26.5 ± 12.8 2.75 ± 0.54PASSION13 PES 619 11 NA 19.0 ± 5.6 3.13 ± 0.43HAAMU-STENT14 PES 164 15 Not reported

StudyLes. Length

(mm)RVD(mm)

DM(%)

* Excludes RCTs enrolling simple/moderate patients/lesions: RAVEL, SIRIUS trials, Pache, and Ortolani, TAXUS II, and TAXUS IV** Occlusion length† SCANDSTENT enrolled patients with occluded, bifurcational, ostial, or angulated lesions; TAXUS V focused enrollment to include patients with longer lesions and patients requiring either 2.25 or 4.0 mm stents. TAXUS VI focused on enrolling patients with longer lesions.

TypeStent Length

(mm)nPt / Vessel

Type

Citations: 1. JAMA 2004;292:2727-34 2. Circulation 2006;114:921-28 3. JACC 2006;48:2423-31 4. Am J Cardiol 2005;96(7A):31H 5. Annual Transcatheter Cardiovascular Therapeutics meeting Oct 22-27, 2006 Abstract 288 6. Circulation 2005;112:2175-83 7. JACC 2006;47:449-55 8. JAMA 2005;294:1215-239. Circulation 2005;112:3306-13 10. NEJM 2006;355:1093-104 11. JACC 2007;49:1924-30 12. Annual AHA meeting Nov 12-15, 2006 13. NEJM 2006;355:1105-13 14. Annual Transcatheter Cardiovascular Therapeutics meeting Oct 22-27, 2006 Abstract 178

Pre-Specified EndpointsPre-Specified Endpoints• Overall mortality

• Cardiac death, defined as any death due to cardiac cause, procedure-related deaths and those related to concomitant treatment and death of unknown cause

• MI, including fatal and non-fatal non-Q-wave or Q-wave MI

• Composite of death or MI

• Definite stent thrombosis (confirmed by angiography or post-mortem examination) in accordance with the Academic Research Consortium (ARC)

– Early (0 to 30 days) and Late Events (> 30 days) were also evaluated separately

• Target Lesion Revascularization– TVR was used as a proxy for 3 studies

(BASKET, HAAMU-STENT, and Zhang, et al.,)Stettler C., et al., Lancet 2007;370:937-48.

Additional Analyses Related to Additional Analyses Related to Stent Thrombosis (ST)Stent Thrombosis (ST)• Post hoc analyses:

– ST occurring > 30 days to 1 year – ST occurring > 1 year to 4 years – Analyses using per protocol definitions of RCTs

• Sensitivity analyses:– Restricted the network to trials with adequate

concealment of allocation, blind adjudication of clinical outcomes, and intention-to-treat analysis and high quality trials satisfying all 3 criteria

– Since strut thickness or stent platform may influence clinical outcomes, the authors performed sensitivity analyses adjusted for strut thickness and adjusted for type of stent platform

• Pooled estimates were derived from standard random-effects meta-analyses of direct within-trial comparisons


Outcomes Testing for Interaction with Outcomes Testing for Interaction with Presence of DiabetesPresence of Diabetes• In addition to the primary network meta-analyses

in all patients, the authors stratified analyses of mortality and the composite of death or MI according to the presence or absence of diabetes and performed a test for interaction between estimated hazard ratios and diabetes


Statistical Analysis Used by Swiss Statistical Analysis Used by Swiss National Science FoundationNational Science Foundation

• The study employed an extension of multivariable Bayesian hierarchical random effects models for mixed treatment comparisons using Markov chain Monte Carlo simulation methods with vague priors

• If performed appropriately, such an analysis will avoid confounding bias and increase the precision of the estimates of effectiveness and safety


Cumulative Incidence of TLRCumulative Incidence of TLR

TVR was used as a proxy for 3 studiesStettler C., et al., Lancet 2007;370:937-48.

Cumulative Incidence of All Death Cumulative Incidence of All Death


Cumulative Incidence of Cardiac DeathCumulative Incidence of Cardiac Death


Cumulative Incidence of Cumulative Incidence of Myocardial InfarctionMyocardial Infarction


Cumulative Incidence of Death or Cumulative Incidence of Death or Myocardial InfarctionMyocardial Infarction


Cumulative Incidence of ARC Definite Cumulative Incidence of ARC Definite Stent ThrombosisStent Thrombosis


Definite ARC Stent ThrombosisDefinite ARC Stent Thrombosis

0 Days- 4 Years BMS PES SESn = 50 n = 72 n = 66

1.38 (0.96-2.24) 0.14 1.00 (0.68-

1.63) 1.00 0.71 (0.48-1.13) 0.21

0 – 30 DaysBMS PES SES

n = 28 n = 30 n = 360.95 (0.38-

2.53) 0.90 1.02 (0.46-2.67) 0.96 1.05 (0.46-

3.17) 0.90

>30 Days – 4 Years BMS PES SES

n = 22 n = 42 n = 302.11 (1.19-

4.23) 0.02 1.14 (0.62-2.26) 0.71 0.54 (0.26-

0.98) 0.04

>30 Days – 1 YearBMS PES SES

n = 14 n = 16 n = 161.61 (0.65-

4.04) 0.23 1.14 (0.45-2.88) 0.78 0.68 (0.26-

1.64) 0.43

>1 - 4 YearsBMS PES SES

n = 8 n = 26 n = 143.57 (0.86-

16.85) 0.07 1.43 (0.27-6.24) 0.64 0.39 (0.09-

1.32) 0.10

PES vs. BMS HR 95% CI p

SES vs. BMS HR 95% CI p

SES vs. PES HR 95% CI p


0 Days- 4 Years BMS PES SES

n = 57 n = 96 n = 851.56 (0.84-

2.58) 0.13 1.03 (0.59-1.67) 0.92 0.65 (0.41-

1.06) 0.08

0 – 30 DaysBMS PES SES

n = 35 n = 35 n = 381.01 (0.48-

2.07) 0.98 0.86 (0.47-1.70) 0.59 0.84 (0.42-

1.89) 0.63

>30 Days – 4 Years BMS PES SES

n = 22 n = 47 n = 352.36 (1.23-

7.00) 0.01 1.13 (0.66-2.81) 0.57 0.45 (0.25-

0.79) 0.01

>30 Days – 1 YearBMS PES SES

n = 20 n = 20 n = 231.32 (0.66-

3.07) 0.62 0.92 (0.37-1.69) 0.80 0.74 (0.32-

1.35) 0.32

>1 - 4 YearsBMS PES SES

n = 2 n = 27 n = 1220.0 (3.92-

221.7) 0.001 5.82 (0.88-76.89) 0.07 0.30 (0.05-

0.98) 0.046

Protocol Stent ThrombosisProtocol Stent ThrombosisPES vs. BMS

HR 95% CI pSES vs. BMS

HR 95% CI pSES vs. PES

HR 95% CI p


Cumulative Incidence of Death Cumulative Incidence of Death Overall: DM vs. Non-DMOverall: DM vs. Non-DM

Diabetics Non-Diabetics


Cumulative Incidence of OverallCumulative Incidence of OverallDeath or MI: DM vs. Non-DMDeath or MI: DM vs. Non-DM

Diabetics Non-Diabetics


BMS Strut-ThicknessBMS Strut-Thickness

140

50

130

0

50

100

150

200

Bx Velocity ACS RX Multi-Link Express

Stru

t Thi

ckne

ss (u

m)

*

ISAR-STEREO: RCT enrolling 628 patients with CAD in > 2.8 mm native vessels comparing thin (50 m ACS RX Multi-Link) vs. thick (140 m; Bx Velocity) strut stents:- Angiographic Restenosis at 6 months: 15.0% vs. 25.8%; P=0.003- Clinical restenosis at 1 year: 8.6% vs. 13.8%; P=0.03

Kastrati A., et al., Circulation 2001;103;2816-2821

This concept is still a debated issue since several analyses of BMS studies have found that strut thickness is not a predictor of TLR: www.TCTMD.com Cutlip D., Strut Thickness: Strut Thickness Has a Minor Impact on Restenosis-Most Bare Metal Stents Are More Alike Than Different!; Monday, December 08, 2003).

http://www.tctmd.com/

Analyses Adjusted for Type of Stent Analyses Adjusted for Type of Stent PlatformPlatform

• “Thirty-seven trials (17,859 patients) contributed to the analyses adjusted for type of stent platform and adjusted for strut thickness.”

• “Results were generally robust to the different analytical approaches used in the sensitivity analyses.”

• “When adjusting for the type of stent platform, however, differences in ST between PES and the other two stent types tended to become more pronounced.”



SES vs. PES (Adjusted for strut thickness):0.85 (0.78-0.99)

Influence of Strut Thickness:Influence of Strut Thickness:Death or MIDeath or MI

TVR was used as a proxy for 3 studiesStettler C., et al., Lancet 2007;370:937-48.

SES vs. PES (Adjusted for strut thickness):0.64 (0.52-0.78)

Influence of Strut Thickness: TLRInfluence of Strut Thickness: TLR

Summary of Key Network Analysis FindingsSummary of Key Network Analysis FindingsPES vs. BMSHR (95% CI)

SES vs. BMSHR (95% CI)

SES vs. PESHR (95% CI)

Death 1.03 (0.84-1.22)p = 0.75

1.00 (0.82-1.25)p = 0.89

0.96 (0.83-1.24)p = 0.80

MI 1.00 (0.81-1.23) p = 0.99

0.81 (0.66-0.97) p = 0.030

0.83 (0.71-1.00)p = 0.045

TLR 0.42 (0.33-0.53)p<0.0001

0.30 (0.24-0.37)p < 0.0001

0.70 (0.56-0.84)p = 0.0021

ARC-Definite ST

Late (>30 days-1yr) 1.61 (0.65-4.04)p = 0.23

1.14 (0.45-2.88) p = 0.78

0.68 (0.26-1.64) p = 0.43

Very Late (>1yr-4yrs) 3.57 (0.86-16.85) p = 0.071

1.43 (0.27-6.24)p = 0.64

0.39 (0.09-1.32)p = 0.10

Overall (0 days-4yrs) 1.38 (0.96-2.24) p = 0.14

1.00 (0.68-1.63) p = 1.00

0.71 (0.48-1.13) p = 0.21

Protocol-Defined ST

Late (>30 days-1yr) 1.32 (0.66-3.07)p = 0.62

0.92 (0.37-1.69) p = 0.80

0.74 (0.32-1.35)p = 0.32

Very Late (>1yr-4yrs) 20.02 (3.92-221.7)p = 0.001

5.82 (0.88-76.89) p = 0.07

0.30 (0.05-0.98)p = 0.046

Overall (0 days-4yrs) 1.56 (0.84-2.58) p = 0.13

1.03 (0.59-1.67) p = 0.92

0.65 (0.41-1.06) p = 0.08


Statistical HeterogeneityStatistical Heterogeneity

• Estimates of statistical heterogeneity between trials were low and criteria for an adequate fit of the model were all satisfied for all outcomes, except TLR

• In conventional meta-analyses, all estimates of statistical heterogeneity between trials were low, except for comparisons of SES versus BMS and PES versus BMS on TLR


Consistency of the NetworkConsistency of the Network•Criteria for consistency were satisfied for all outcomes, except ST and TLR:

– Stent Thrombosis:• The goodness of fit of the model was excellent and estimates of

between-trial heterogeneity were low.• A p-value for inconsistency of 0.69 suggested that the observed

inconsistency may have been due to chance alone.– Target Lesion Revascularization:

• Goodness of fit of the model was not optimal and there was some evidence for heterogeneity between trials and inconsistency of the network.

• The magnitude of effects, the fully concordant results of conventional random-effects meta-analyses of direct within-trial comparisons and the lack of heterogeneity between trials for the comparison of SES versus PES in network and conventional meta-analysis on this outcome indicate, however, that observed differences in TLR between stent types are real.

• Results were also robust in sensitivity analyses restricted to trials of high methodological quality and after adjusting for the strut thickness or the type of stent platform used

•The authors stated that the estimates are reliable also for ST and TLR.Stettler C., et al., Lancet 2007;370:937-48.


Consistency of Efficacy FindingsConsistency of Efficacy FindingsPatient-level Direct Comparison

HR 0.72Network Analysis

HR 0.70

Kastrati A., et al., J Am Coll Cardiol 2007; 50: e-publication (August 21, 2007).

SES

PES

SES

PES


Death or MI Through Latest Follow-up0.5 1.0 1.5

Favors SES Favors PES

0.86 (95 CI: 0.72-1.01)

P = 0.07

Direct ComparisonHR 0.86

Network AnalysisHR 0.92

Consistency of Safety Findings:Consistency of Safety Findings:Death or MIDeath or MI

Kastrati A., et al., J Am Coll Cardiol 2007; 50: e-publication (August 21, 2007).

Conclusions from Publication Conclusions from Publication (Efficacy)(Efficacy)• “A secondary analysis showed a marked reduction in

TLR with both DES, which was more pronounced for SES than for PES.”

• “About six patients will have to receive a SES rather than a BMS to prevent one TLR over 4 years; 35 would need to receive a SES rather than a PES to prevent one such event.”


Summary from PublicationSummary from Publication

• “This collaborative network meta-analysis of randomized controlled trials indicates that overall and cardiac mortality associated with DES and BMS are similar. Relevant harms associated with SES compared with BMS are unlikely, while rates of TLR and MI are lower with SES than with PES and BMS. We conclude, therefore, that SES seem to be clinically better than BMS and PES.”


Conclusions from Publication (Safety)Conclusions from Publication (Safety)• “Our collaborative network meta-analysis indicates that drug-

eluting stents and bare-metal stents are associated with similar rates of overall and cardiac mortality, and that the use of sirolimus-eluting stents is associated with a reduction in the risk of myocardial infarction compared with the use of bare-metal and paclitaxel-eluting stents.”

• “About 100 patients will have to receive SES, rather than BMS or PES, to prevent one MI over 4 years .”

• “Although there was little evidence of an overall increase in definite stent thrombosis associated with DES, we found PES to be associated with an increased incidence of late stent thrombosis (>30 days-4 years) compared with BMS and SES.”

– “Wide credibility intervals precluded definite conclusions about a potential increase of late stent thrombosis with SES compared with BMS.”

• “Lastly, we found little evidence of an increased risk of mortality associated with either DES in diabetic patients, but wide credibility intervals precluded definite conclusions.”


Health & Medicine

Stettler Windecker Meta Analysis 18k 10 20 07 V04