Outcomes Following Coronary Stenting: A National Study of Long Term, Real-World Outcomes of Bare-Metal and Drug-Eluting Stents Pamela S. Douglas, J. Matthew Brennan, Kevin J. Anstrom, Eric L. Eisenstein, David Dai, Ghazala Haque, David F. Kong, Ralph Brindis, Art Sedrakyan, David Matchar, Eric D. Peterson Duke Clinical Research Institute Duke University Medical Center

Funding and Disclosures Disclosures: ( See for full information) n Pamela S. Douglas: None relevant n J. Matthew Brennan: None n Kevin Anstrom: Research Support from AstraZeneca, Bristol Myers Squibb, Eli Lilly and Medtronic; consultant for Johnson & Johnson and Pfizer. n Eric L. Eisenstein: Research Support from Medtronic Vascular and Eli Lilly n David Dai: None n Ghazala Haque: None n David F. Kong: None relevant n Ralph Brindis: None n Art Sedrakyan: None n David Matchar: None relevant n Eric D. Peterson: Research Support from BMS/Sanofi and Merck/Schering Sponsor and funding: AHRQ CV Research Consortium Additional support: ACC-NCDR

Background n Clinical trials demonstrate reduced restenosis with drug eluting coronary stents (DES) compared to bare metal stents (BMS) n However, some trials and registries have reported late stent thrombosis and higher mortality with DES n Questions remain regarding the effectiveness and safety of DES in the real world and among understudied patient populations

Goal and Population Goal To examine comparative effectiveness and safety of DES vs BMS in a national PCI cohort Study population All PCI pts > 65 yo in NCDR CathPCI 1/04-12/06 Follow up obtained through linkage to CMS inpatient claims data using indirect identifiers; 76% matched Final cohort 262,700 pts 83% DES; 46% Cypher, 55% Taxus

Analysis n 30 month outcomes l Death, MI, Stroke, Revascularization, Major bleeding Overall and in important subgroups Outcomes adjustments l Inverse propensity weighted model (102 covariates) l Cox proportional hazards model (60 covariates) Sensitivity analyses Results in RCT-like population Non-CV cause of death

Patient Characteristics DES (217,675) vs BMS (45,025) UnadjustedIPW Adjusted DESBMSDESBMS Age * Female43%40%*43% Caucasian90%91%*90% Diabetes 32% Renal Failure 6%8%* 7% Hypertension80%80%*80%81% Prior PCI28%26%*28% Prior CABG22%28%*23% Urgent Status38%36%*37%38% STEMI10%16%*11% *p< DES vs BMS

DES and BMS Event Rates: 30-month Unadjusted Events Requiring Rehospitalization Rate / 100 patients

DES and BMS Event Rates: 30-month Adjusted HR = 0.91 (0.85,0.98) HR = 0.96 (0.88,1.04) HR = 0.75 (0.73,0.77) HR = 0.76 (0.72,0.80) HR = 0.91 (0.89,0.94) Rate / 100 patients

Landmark Display: Mortality

Landmark Analysis: MI STEMI HR = 0.78 NSTEMI HR = 0.73

Subgroup Analyses DeathMIRevasc Favors DES

Sensitivity Analysis: Patient Selection n RCT - like population n N = 49,355 (19%) n Inclusion criteria l Elective PCI, < 2 stents l Native vessel, de novo l Class A or B lesions l Lesion length, diameter l ASA, clopidogrel OK l No CKD Favors DES

Sensitivity Analysis: Device Selection Cause of Death in DES v BMS n Using 1º hosp dx, cause extracted in 90% deaths n HR 0.80 favoring DES for CHF/MI death n HR 0.74 favoring DES Non CV death n Sicker patients may preferentially receive BMS Unadjusted rates/ 100

Potential Limitations n Medicare, NCDR data sources l All patients were > 65 yo, inpatients, NCDR sites (n=650) l No clinical data available for bleeding, revascularization, stent thrombosis l No information on follow-up medications n Observational data: potential for bias, unmeasured confounders

Conclusions n Linkage of clinically rich NCDR data to claims data is feasible; Data analysis allows a robust, longitudinal assessment of clinical effectiveness n Comparing outcomes of DES to BMS at 30 mo: l No major DES safety concerns l Lower death and MI rates in DES patients l Slightly lower revascularization, bleeding rates l Similar stroke rates n Results consistent among all patient subgroups n Caveat: The apparent benefit of DES may be affected by selection bias and unmeasured confounders present in this real world cohort

THANK YOU

Study Population All PCI admissions for patients who had some stent implantation 450,242 PCI admissions, 662 sites, 390,973 patients CMS Matched 290,438 PCI admissions, 650 sites, 290,438 patients Exclude patients w index stent application w not during Fee- For-Service (FFS) enrollment 14,225 PCI admissions, 597 sites, 14,225 pts Patients w index stent application during FFS enrollment 276,213 PCI admissions, 650 sites, 276,213 pts Exclude admissions with both DES and BMS 12,822 PCI admissions, 583 sites, 12,822 pts Admissions with either DES only or BMS only 263,391 PCI admissions, 650 sites, 263,391 pts Exclude records with missing candidate variables and time-to-censor out of range [ days] Final study population 262,700 PCI admissions, 650 sites, 262,700 patients

CMS Matched vs Not Matched Patient Clinical Characteristics Total (387,849) CMS-Matched (290,438) CMS-Not- Matched (97,411) Age74.3 ± ± ± 6.3 Female41.3%42.4%38.2% Caucasian88.8%90.3%84.5% Diabetes – Non-insulin23.1%22.9%23.7% – Insulin9.5%9.6%9.1% Prior Renal Failure – Non-dialysis4.9%5.0%4.4% – Dialysis1.6% 1.4% Hypertension * 80.3% 80.4% Prior PCI28.1%27.9%28.4% Prior CABG22.8%23.2%21.8%

CMS Match vs Not Matched Patient Clinical Characteristics Total (387,849) CMS Matched (290,438) CMS Not-Matched (97,411) Status – Urgent15.2%15.6%14.0% – STEMI11.2%11.6%10.0% Number of Diseased Vessels – One41.7%41.3%43.0% – Two31.4%31.5%31.1% – Three DES 22.9%23.2%22.1% – Some Cypher46.1%46.2%46.0% – Some Taxus55.5%55.4%55.8% – Some Off-label70.6%70.2%71.9%

Unmeasured Confounder: Post-PCI Clopidogrel Use n Used published Duke data for clopidogrel use and survival benefit to correct HR for death: l Prevalence DES use = 50%, BMS use = 12 mo l Clopidogrel benefit: 50% 12 mo in DES, BMS n HR death corrects from 0.75 to 0.90 W/o clopidogrel use HR = 0.75 With 50%, 20% cHR = 0.90 BMS family of curves: 0 to 50% use JAMA :159 Biometrics :948

Comparability with Other Registry Data – Mortality Favors DES Favors BMS