Author + information
- Received October 25, 2007
- Revision received January 4, 2008
- Accepted January 28, 2008
- Published online April 1, 2008.
- Edward L. Hannan, PhD⁎,⁎ (, )
- Michael Racz, PhD†,
- Gary Walford, MD‡,
- David R. Holmes, MD§,
- Robert H. Jones, MD∥,
- Samin Sharma, MD¶,
- Stanley Katz, MD# and
- Spencer B. King III, MD, MACC⁎⁎
- ↵⁎Reprint requests and correspondence:
Dr. Edward L. Hannan, School of Public Health, State University of New York, Department of Health Policy, Management, and Behavior, SUNY University at Albany, One University Place, Rensselaer, New York 12144-3456.
Objectives The purpose of this study was to compare outcomes for drug-eluting stents (DES) and bare-metal stents (BMS) for patients with ST-segment elevation myocardial infarction (STEMI).
Background Despite some controversy related to late stent thromboses, DES are approved for use in many patients undergoing stenting. However, there are several types of patients in whom implanting a DES would be regarded as off-label use, and this study compares DES and BMS for one of these groups.
Methods New York’s percutaneous coronary intervention registry was used to identify 772 patients undergoing percutaneous coronary intervention (PCI) for STEMI who received BMS and 1,154 STEMI patients who received DES between October 1, 2003, and December 31, 2004. These patients were tracked through December 31, 2005. Mortality, target vessel PCI, and subsequent coronary artery bypass graft (CABG) surgery within 2 years of undergoing the procedure were captured. Adverse outcomes were adjusted using proportional hazards methods to account for baseline differences in patients’ severity of illness.
Results The BMS patients had significantly higher mortality (adjusted hazard ratio [HR] vs. DES = 2.01, 95% confidence interval [CI] 1.21 to 3.34, risk-adjusted mortality = 8.6% vs. 5.0%, p = 0.007) and significantly higher subsequent CABG surgery (adjusted HR vs. DES = 2.33, 95% CI 1.31 to 4.16, risk-adjusted rate = 6.4% vs. 3.0%, p = 0.004) rates. There was no difference by type of DES (adjusted HR for paclitaxel-eluting stent versus sirolimus-eluting stent; mortality 0.72, 95% CI 0.30 to 1.72), subsequent CABG surgery (adjusted HR = 0.60, 95% CI 0.26 to 1.40), and target vessel PCI (adjusted HR = 0.74, 95% CI 0.35 to 1.58).
Conclusions In this observational study, DES were associated with lower mortality and subsequent CABG surgery when used for STEMI patients.
Despite the fact that stenting has largely replaced conventional balloon angioplasty as the preferred mode of percutaneous coronary intervention (PCI) in the last several years, there exists considerable controversy as to the relative merits of bare-metal stents (BMS) and drug-eluting stents (DES) or for what types of patients each type of stent is superior (1–19).
Current Food and Drug Administration approval has restricted DESs to short de novo lesions in native coronary arteries between 2.5 and 3.5 mm in length for sirolimus-eluting stents (SES) (20) and between 2.5 and 3.75 mm for paclitaxel-eluting stents (PES) (21). Patients with long or complicated lesions, chronic total occlusions, vein graft disease, multivessel coronary artery disease, or urgent coronary syndromes (including ST-segment elevation myocardial infarction [STEMI]) are not included.
Few studies have compared outcomes of DES and BMS for off-label uses despite the fact that off-label use of DES constitutes a high percentage of all stent cases. Although some randomized controlled trials (RCTs) have compared BMS and DES for STEMI patients, those studies were quite small and somewhat inconclusive (22–25). The purpose of this study is to examine 1 particular, important type of off-label use, for STEMI patients, and to document the use of DES for these patients in a population-based study. Also, adverse outcomes for STEMI patients with BMS are compared with adverse outcomes with DES after adjusting for baseline differences in patient risk factors.
Data were obtained from New York State’s Percutaneous Coronary Interventions Reporting System (PCIRS), a mandatory registry in New York that was initially developed in 1991 to collect detailed information for each patient in the state who was undergoing PCI. The PCIRS contains information regarding demographics; pre-procedural risk factors; periprocedural complications; types of devices used; extent of disease; vessels and lesions treated; dates of admission, discharge, and procedure; discharge disposition and destination; and hospital and operator identifiers. These data are matched to New York administrative data and are audited by the New York State Department of Health’s utilization review agent to ensure completeness and accuracy.
The PCIRS data were matched to New York’s vital statistics data so that these patients could be followed after discharge for evidence of subsequent death. Also, data from New York’s Cardiac Surgery Reporting System were used along with PCIRS data to identify repeat target vessel PCIs and subsequent coronary artery bypass grafts (CABGs) after discharge through December 31, 2005.
Patients and end points
All patients who had experienced a STEMI within 12 h prior to undergoing stenting with either BMS or DES between October 1, 2003, and December 31, 2004, were included in the study except patients who had been previously revascularized (264 patients), had left main disease (50 patients), had prior thrombolytic therapy within 7 days (503 patients), were from out of state (58 patients), or had combinations of different stent types (166 patients). All other primary angioplasty patients undergoing stenting between October 1, 2003, and December 31, 2004 (772 BMS patients and 1,154 DES patients: total 1,926 patients) were followed through December 31, 2005. Of the 1,154 DES patients, 706 received SES and 448 received PES. End points were mortality, target vessel PCI, and subsequent CABG surgery. The PCI registry enabled us to identify the target vessels in the index admission because it contained information on attempted vessels. However, the CABG registry did not contain attempted vessels, so subsequent target vessel revascularization was not available, and subsequent CABG surgery was used as the end point.
Differences between primary angioplasty patients undergoing stenting with BMS and with DES in the prevalence of various patient risk factors (demographics, comorbidities, left ventricular function, hemodynamic state, vessels diseased, time since onset of symptoms) as well as differences in in-hospital mortality and need for CABG surgery in the same admission were tested using Fisher exact test (for binary risk factors) and the chi-square test (for other risk factors).
To test for risk-adjusted differences in longer-term outcomes (mortality, target vessel PCI, and subsequent CABG surgery), stepwise Cox proportional hazard models with a robust covariance matrix that accounts for correlation of survival times for individuals within a hospital or operator cluster (26) were developed for each adverse outcome measure after having confirmed that the proportional hazards assumption was justified (27). Candidate independent variables included the patient risk factors available in the registry. Also, individual hospitals were controlled for in the model in case outcomes were related to individual hospitals as well the type of stent used. In an additional model, hospital volumes were controlled for in lieu of individual hospitals. Type of stent (BMS, DES) was used in each model as the study independent variable, and BMS/DES adjusted hazard ratios (HRs) were obtained by exponentiation of the coefficient of that variable. Adjusted survival curves were constructed for BMSs and DESs for each of the 2 outcomes using the Cox proportional hazards models and methods for calculating adjusted survival (28).
Each type of DES (SES, PES) was then compared against BMS for each adverse outcome by creating similar proportional hazards models with each of the DES types as a binary indicator variable and BMS as the reference. Again, HRs and their confidence intervals were calculated. A SES was compared directly with a PES for each outcome by restricting the database to DES patients and using PES as the reference and SES as an indicator variable.
To test for selection bias, a propensity model was developed (29,30). The risk factors in Table 1 were used as independent variables in a logistic regression model with a binary dependent variable representing the use of BMS. The propensity score was subdivided into quintiles and HRs for DES/BMS mortality, CABG surgery, and target vessel PCI were compared for each quintile. Hazard ratios were compared across quintiles for each outcome to determine if there was any tendency for them to be affected by the tendency to use a BMS versus a DES. All tests were 2-sided and conducted at the 0.05 level, and all analyses were conducted in SAS 9.1 (SAS Institute Inc., Cary, North Carolina).
Mean follow-up times were 619 days for BMS (SD = 183) and 522 days for DES (SD = 142). As demonstrated in Table 1, STEMI patients undergoing BMS and DES implantation had very similar characteristics. The only significant differences were with regard to Hispanic ethnicity (9.6% for BMS vs. 6.7% for DES, p = 0.024), and intravenous (IV) GP IIb/IIIa platelet inhibitors given prior to the operation (65.4% for BMS vs. 56.8% for DES, p < 0.001). Although BMS patients had higher observed and expected mortality and same-stay CABG rates, these differences were not statistically significant (Table 2).
With regard to long-term outcomes, Table 3 indicates that the DES patients experienced lower unadjusted mortality rates (unadjusted BMS/DES HR = 1.53, 95% confidence interval [CI] 1.02 to 2.29) and subsequent CABG surgery rates (unadjusted HR = 1.74, 95% CI 1.12 to 2.71).
After adjusting for the individual hospital in which the PCI occurred, IV GP IIb/IIIa platelet inhibitors given prior to the operation, the number of vessels diseased, region of disease (left anterior descending involvement or proximal left anterior descending involvement), age, female gender, ejection fraction, peripheral vascular disease, cerebrovascular disease, hemodynamic instability, shock, diabetes and renal failure, DES patients continued to exhibit lower adverse outcome rates (Table 3). For mortality, the adjusted HR was 2.01, 95% CI 1.21 to 3.34. For subsequent CABG surgery and target vessel PCI, the respective adjusted HRs were 2.33, 95% CI 1.31 to 4.16 and 1.15, 95% CI 0.74 to 1.78. Adjusting for hospital volumes in lieu of individual hospitals yielded similar results.
As indicated in Figure 1, the risk-adjusted mortality rates at 2 years were 8.6% for BMS and 5.0% for DES, and this difference was significant (p = 0.007). Figure 2 demonstrates that the risk-adjusted rates for subsequent CABG surgery were 6.4% for BMS and 3.0% for DES, and this difference was also significant (p = 0.004). The rates were 7.6% and 6.7% for target vessel PCI (Fig. 3), and this difference was not significant.
When we subdivided the DES patients into 706 SES patients and 448 PES patients and repeated the analyses, we found that both SES and PES patients had significantly lower mortality than BMS patients (for SES, adjusted HR = 0.55, 95% CI 0.31 to 0.98, and for PES, adjusted HR = 0.40, 95% CI 0.18 to 0.85; data not in tables). Both SES and PES were associated with lower subsequent CABG rates than BMS (for SES, adjusted HR = 0.52, 95% CI 0.27 to 0.98, and for PES, adjusted HR = 0.33, 95% CI 0.14 to 0.76). With respect to target vessel PCI, there were no significant differences (for SES/BMS, adjusted HR = 1.11, 95% CI 0.69 to 1.79, and for PES/BMS adjusted HR = 0.51, 95% CI 0.25 to 1.03).
After confining the analyses to patients who received either all SES or all PES, we found that there was no significant difference between the 2 types of DES for mortality (for PES/SES, adjusted HR = 0.72, 95% CI 0.30 to 1.72; data not in tables). Also, there was no significant difference for subsequent CABG surgery (adjusted HR = 0.60, 95% CI 0.26 to 1.40) or for target vessel PCI (adjusted HR = 0.74, 95% CI 0.35 to 1.58).
With regard to the potential for selection bias, it should first be noted that the prevalence of important risk factors was very similar for patients receiving DESs and BMSs for primary angioplasty. The only significant differences were with regard to Hispanic ethnicity and IV GP IIb/IIIa platelet inhibitors given prior to the operation. Nevertheless, a propensity analysis was undertaken to test for selection bias.
In the propensity analyses, the propensity model demonstrated low ability to identify predictors of DES use (C = 0.58), which was not surprising given that Table 1 shows that the prevalence of patient risk factors were very similar for DES and BMS patients. Despite this fact, the HR favored DES (was >1) for all but 1 of the 15 quintile comparisons across the 3 adverse outcome measures (1 for each measure), and there was no trend toward better outcomes for BMS for the quintiles in which BMS were used more frequently.
Despite the fact that DES have proven to have lower restenosis and repeat revascularization rates than BMSs in numerous studies (1–15), some recent studies (16–19) have reported higher late stent thrombosis rates in DES patients, and this has caused considerable concern about their safety.
A Food and Drug Administration panel that was convened to examine the relative safety of DES and BMS concluded that DES are associated with a small increase in stent thrombosis rates compared to BMS, but that this problem does not outweigh the benefits of DES (31).
It was also noted by the Food and Drug Administration that patient outcomes for off-label use may be worse than those observed for on-label use (31). In fact, 2 recent studies confirmed that when DES are used off-label, they are associated with higher adverse event rates than when they are used on-label (32,33). However, this is not surprising because the risk factors that constitute off-label use have been demonstrated to be related to higher long-term adverse outcome rates. What is most important is how DES patients with risk factors associated with off-label use fare as compared with BMS patients with the same risk factors.
A recent meta-analysis of 7 RCTs has examined this issue for patients with acute myocardial infarction. Combining trials with a total of 1,177 DES patients and 1,180 BMS patients with a follow-up of 8 to 12 months, Pasceri et al. (34) found that the incidence of death or myocardial infarction was not significantly different (5.8% for DES and 6.9% for BMS, risk ratio = 0.84, 95% CI 0.62 to 1.15). Target lesion revascularization rates were lower for DES patients (4.8% vs. 12.0%, risk ratio = 0.40, 95% CI 0.30 to 0.54).
Our study found that prior to adjustment, DES were associated with significantly better outcomes (for mortality, BMS/DES HR = 1.53, 95% CI 1.02 to 2.29, and for subsequent CABG surgery, HR = 1.74, 95% CI 1.12 to 2.71). After adjusting for patient factors related to adverse outcomes as well as for the hospital in which stenting was performed, DES continued to have significantly better outcomes (for mortality, adjusted HR= 2.01, 95% CI 1.21 to 3.34, and for subsequent CABG surgery, adjusted HR = 2.33, 95% CI 1.31 to 4.16), although repeat PCI was not significant.
Thus, our target vessel revascularization findings were quite similar to those of large RCTs or meta-analyses of RCTs, in which DES were usually found to have significantly lower target vessel revascularization rates. However, our study is the first one to find a significant mortality advantage for DES, and the difference does not appear to be merely a function of statistical power because our sample sizes are similar to those in some of the other studies.
It is also notable that there were no significant differences in adjusted outcomes between SES and PES. Also, both types of DES demonstrated a mortality and subsequent CABG advantage compared with BMS.
There are several limitations to the study. First, it is an observational study in which patients were not randomized to DES and BMS, and therefore is subject to concerns related to selection bias, whereby 1 treatment may appear to have superior outcomes because the treatment was associated with patients who had lower prevalences of risk factors for adverse outcomes. We investigated this concern by comparing risk factor prevalences for DES and BMS patients and only found 2 variables (Hispanic ethnicity and IV GP IIb/IIIa platelet inhibitors given prior to the operation) for which there was a significant difference in prevalences. This was confirmed when we developed a propensity model that had very low ability to predict which patients underwent DES versus BMS based on the numerous risk factors available in the registry. These results lead us to conclude that because the advantage of DES was not limited to groups of patients with very low probabilities of undergoing BMS, there is no evidence of selection bias with respect to variables contained in the registry.
However, it should be noted that propensity matching cannot control for bias related to an imbalance in unknown or unmeasured risk factors between the 2 interventions. Factors that could influence PCI outcomes in general that were not contained in the registry include vessel size, lesion length, clot burden, contraindication to antiplatelet medications, compliance with dual antiplatelet therapy, hyperlipidemia, limited life expectancy, active bleeding/bleeding risk, and the use of statins and angiotensin-converting enzyme inhibitors. Also, ejection fraction was treated as a categorical variable.
Another limitation is that myocardial infarction and stent thrombosis were not included as end points. This is potentially important considering the concern about stent thrombosis in off-label use of DES, especially in patients with STEMI.
Furthermore, the cause of death was unknown, and to the extent that the cause may have related to stenting more among DES patients, this would bias the study against BMS. Also, because the study was not a RCT, there were no central blinded clinical event committees nor core laboratories.
Another caveat is that because we used the New York State vital statistics data to track mortality after discharge and New York State registries to track subsequent CABG and target vessel PCI, we restricted the study to New York State patients. However, if patients moved outside of the state after discharge and then died or underwent revascularization out of state, these events would have been missed by our study. Also, the study would miss New York State patients who underwent a repeat revascularization out of state. There is no reason why there would be a bias in favor of either type of stent with respect to missed patients, and an earlier study (35) demonstrated that there was not a bias in this regard.
In conclusion, the mortality advantage in favor of DES is contrary to the results of RCTs, and as such is hypothesis generating, in particular because the possible mechanism underlying this finding is unknown. In particular, our finding could be at least partially related to selection bias, and it is recommended that a large, adequately powered clinical trial be conducted to confirm or disprove the finding.
Dr. Sharma is the recipient of a research grant from Boston Scientific and has Speakers’ Bureau appointments with Boston Scientific, Abbott Vascular, and Eli Lilly & Co. Dr. Katz serves as an Expert Witness for Medical Liability Mutual Insurance Company and as an advisory board member of General Electric, Boston Scientific, and Medtronic. Dr. King receives royalties from Cordis and serves as a consultant/advisory board member of Medtronic. Cindy L. Grines, MD, acted as Guest Editor for this paper.
- Abbreviations and Acronyms
- bare-metal stent(s)
- coronary artery bypass graft
- drug-eluting stent(s)
- hazard ratio
- percutaneous coronary intervention
- Percutaneous Coronary Interventions Reporting System
- paclitaxel-eluting stent(s)
- randomized controlled trial
- sirolimus-eluting stent(s)
- ST-segment elevation myocardial infarction
- Received October 25, 2007.
- Revision received January 4, 2008.
- Accepted January 28, 2008.
- American College of Cardiology Foundation
- Holmes D.R. Jr..,
- Leon M.B.,
- Moses J.W.,
- et al.
- Stone G.W.,
- Ellis S.G.,
- Cox D.A.,
- et al.
- Colombo A.,
- Drzewiecki J.,
- Banning A.,
- et al.
- Hermiller J.B.,
- Raizner A.,
- Cannon L.,
- et al.
- Grube E.,
- Silber S.,
- Hauptmann K.E.,
- et al.
- Ong A.T.,
- Aoki J.,
- van Mieghem C.A.,
- et al.
- Bavry A.A.,
- Kumbhani D.J.,
- Helton T.J.,
- Bhatt D.L.
- Regar E.,
- Serruys P.W.,
- Bode C.,
- et al.
- Tanabe K.,
- Serruys P.W.,
- Grube E.,
- et al.
- U.S. Food and Drug Administraciton
- U.S. Food and Drug Administration
- Menichelli M.,
- Parma A.,
- Pucci E.,
- et al.
- Anderson H.V.,
- Smalling R.W.,
- Henry T.D.
- Allison P.
- Rosenbaum P.R.,
- Rubin D.B.
- U.S. Food and Drug Administration
- Hannan E.L.,
- Racz M.J.,
- McCallister B.D.,
- et al.