Comparative Clinical Outcomes of Paclitaxel- and Sirolimus-Eluting Stents
Charles A. Simonton, MD, FACC, FSCAI; Bruce Brodie, MD; Barrett Cheek, MD; Fred Krainin, MD; Chris Metzger, MD; James Hermiller, MD; Stanley Juk, MD; Peter Duffy, MD; Angela Humphrey, MS; Marcy Nussbaum, MS; Sherry Laurent, PhD for the STENT Group J Am Coll Cardiol. 2007;50(13):1214-1222. ?2007 Elsevier Science, Inc.
Abstract and Introduction
Objectives: The purpose of this study was
to compare the 9-month clinical outcomes of patients treated with
paclitaxel-eluting stents (PES) or sirolimus-eluting stents (SES) for
coronary artery stenosis.
Since approval of the sirolimus-eluting stent (SES) (Cypher, Cordis Corp., Miami Lakes, Florida) in April 2003 and the paclitaxel-eluting stent (PES) (Taxus, Boston Scientific Corp., Natick, Massachusetts) in March 2004, these stents have gained widespread acceptance by interventional cardiologists for treatment of patients with obstructive coronary artery disease (CAD). Limited data are available from multicenter studies in the U.S. comparing clinical outcomes between these 2 stents. Previous comparisons of PES and SES in randomized multicenter trials have produced varying results and have been performed exclusively outside the U.S..[2-8]
The STENT (Strategic Transcatheter Evaluation of New Therapies) study is the first multicenter prospective registry to collect clinical outcomes with these 2 stents from a "real-world" patient population in the U.S. The registry was created to evaluate coronary stents in broad clinical settings after approval by the U.S. Food and Drug Administration (FDA) in May 2003. The primary purpose of the present study was to compare 9-month clinical outcomes in patients receiving a PES or SES to determine differences in mortality, clinically diagnosed myocardial infarction (MI), target vessel revascularization (TVR), or stent thrombosis (ST).
The STENT study is a multicenter multidevice percutaneous coronary intervention (PCI) registry that began enrollment in May 2003. For the present study, all patients enrolled from May 1, 2003, through September 30, 2005, were included for analysis. Centers were selected to represent a mix of large tertiary referral and smaller community-based interventional cardiology programs. All centers obtained institutional review board (IRB) approval to prospectively consent and enroll patients. Because there were no exclusion criteria, the intent was to enroll every patient undergoing PCI. Centers were expected to maintain a minimum prospective enrollment of 80% of all patients undergoing PCI and a 9-month clinical follow-up of 90%. For patients having more than 1 procedure during the study period, the first procedure was used for 9-month follow-up. Patients having lesions treated only with PES or SES were included. Lesions could be pretreated with balloon angioplasty or other devices and could receive postdilatation.
The 3 primary end points were death, clinically diagnosed MI, and TVR. A TVR was defined as a repeat procedure in the target vessel (including all of its branches) specifically, coronary artery bypass graft (CABG), or repeat PCI of the target vessel. The TVR was used instead of target lesion revascularization, because determinations were visually assessed without a core angiographic laboratory. Secondary end points consisted of major adverse cardiac events (MACE) (a composite of death, MI, or TVR), ST, and stroke. Stent thrombosis was defined as: 1) angiographic evidence of thrombus within the stent or a reduction in Thrombolysis In Myocardial Infarction grade to 0, 1, or 2 in the previously treated target vessel in association with an acute clinical cardiac presentation of unstable angina or MI (acute coronary syndrome); 2) sudden cardiac death; or 3) ST-segment elevation myocardial infarction (STEMI) in the target vessel distribution. Additional end points included proportional hazards-adjusted time to TVR and time to MACE events. Any questionable events, all TVR and ST events, and all cardiac deaths were adjudicated by the principal investigator at each site.
Data Collection and Management
Each of the 8 participating centers (Online Appendix A) had at least 1 coordinator for consenting, collecting, and entering data into a web-based system. Frequent conference calls and on-site meetings were conducted to promote homogeneity of data interpretation and data entry. Clinical and analytical integrity of the STENT registry were monitored by the Executive Steering Committee, which included a principal investigator from each participating institution. All data were reviewed by this committee on a regular basis through electronic communication, teleconferencing, and in-person meetings.
Prospective observational methodology was used with no intent to influence the practices of the participating hospitals or physicians. All sites were trained on study variable definitions using American College of Cardiology (ACC) definitions (version 2.0). For variables that were not ACC defined, a definition was provided by Steering Committee consensus. Patient baseline clinical data at hospital admission and time of procedure were collected by staff at each institution. Intraprocedural data were collected and approved by the operating interventional cardiologist.
Reference vessel diameters, lesion lengths, and percentage of pre- and postprocedure stenosis in the target lesions were estimated visually by the operating physician. Stent length per lesion and total stent length per procedure (sum of the individual stent lengths deployed) consisted of the actual package lengths of the stents deployed. Post-discharge clinical follow-up at 9 months was conducted at each institution by telephone interview. At the time of follow-up, complete hospital records were obtained and reviewed for every patient reporting a cardiac event.
An independent data coordinating and statistical center (R. Stuart Dickson Institute for Health Studies, Charlotte, North Carolina) was responsible for the development and maintenance of the web-based data collection tools, extraction of the data, performance of scheduled quality checks, and all of the analytics. The center also employs a full-time research coordinator to assist sites with IRB compliance, data monitoring, and quality control and a full-time auditor for site audits. Regular audits of randomly selected procedures and data fields were performed against source documents on 10% of the first 4,000 procedures and 5% thereafter.
Demographic characteristics, pre-existing risk factors, procedure-related variables, and 9-month outcomes were summarized using mean and standard deviation for continuous variables and frequency and percentage for categoric variables. Baseline demographic, clinical, and procedural characteristics were compared between the 2 groups to determine any significant differences. Significance testing was conducted using t tests or the Kruskal-Wallis test for continuous variables and the chi-square test or Fisher exact test for categoric variables depending upon the distribution. A p value of <0.05 was considered to be statistically significant. All analyses were performed using SAS version 9.1 (SAS Institute, Cary, North Carolina).
Unadjusted time-to-event rates in the SES and PES groups were calculated and compared using Cox proportional hazards models to estimate the hazard ratios with 95% confidence intervals (CIs). Unadjusted Kaplan-Meier curves were constructed with corresponding log rank tests for TVR and MACE. Propensity scores were used to adjust for nonrandomization of stent selection. A logistic regression model was used where the type of stent (PES or SES) was the outcome variable and baseline characteristics with p < 0.01 from the bivariate analysis were the covariates. Propensity scores were generated and included as a regression (covariance) adjustment in each of the Cox proportional hazard regression models. The ability of the propensity score to effectively balance stent groups at baseline was confirmed using separate logistic models with PES/SES group type as the dependent variable and the covariate and propensity score as the independent variables (Online Appendix B). Because the propensity scores could not balance for site, it was removed from the propensity model and adjusted for directly in the outcome models. Owing to the possibility of collinearity between site and stent type, hazard ratios were calculated with and without site in the model.
During the study period, there were 18,379 PCI procedures performed at the 8 institutions (Fig. 1). Of these, informed consent was obtained in 16,442 (89.5%). A PES or SES only was performed in 10,707 procedures, and, of these, 9,835 patients underwent a first procedure in the registry receiving only a PES or SES device. A total of 9,226 patients (93.8% of eligible) were followed up at 9 months and constitute the study population.
The PES stent group was significantly older, had a higher percentage of whites, and had a higher incidence of MI on admission ( Table 1 ). The SES group had a significantly higher incidence of cigarette use, hypercholesterolemia, prior PCI, cardiogenic shock, cardiac arrests, and more frequent family history of CAD. All of these differences, although statistically significant, were numerically small. There were no other significant differences in risk factors, such as diabetes, or clinical presentation between the 2 groups.
There were few significant differences between the 2 groups at the time of the procedure ( Table 2 ). On presentation to the catheterization laboratory, the SES group had more urgent procedures compared with the PES group, which had more emergent procedures mostly because of a slightly higher frequency of acute evolving MI (PES 9.9% vs. SES 8.3%; p = 0.01). Other procedural factors that could impact clinical outcome, such as left ventricular ejection fraction, congestive heart failure, and degree of CAD, were not significantly different.
The PES was used significantly more frequently in high-risk lesions (calcification and bifurcations) and in smaller vessels (<3.0 mm) ( Table 3 ). In comparison, the SES stent was used significantly more in saphenous vein graphs (SVG), chronic total occlusions (CTO), and restenosis procedures. Other lesion factors, such as lesion length, lesion location, and total stent length per procedure, were not statistically different. At 98%, acute procedural success was identical for both types of stents.
Nine-month unadjusted clinical outcomes ( Table 4 ) show no statistically significant differences between PES and SES groups. Of note, TVR occurred in 4.1% of PES procedures and 4.3% of SES (p = 0.76) and MACE in 7.5% of PES and 8.0% of SES (p = 0.37). At 0.7%, ST was identical for both groups, with the majority (66.7% of PES and 80.7% of SES) occurring within the first 30 days after the procedure. Stroke at 9 months was infrequent for both groups (PES 0.3%, SES 0.5%).
Unadjusted and adjusted hazard ratios for death, MI, TVR, and MACE are shown in Table 5 . After adjustment, using PES as the reference group, the association between stent type and the outcome components of MACE changed very little, remaining nonsignificant for death (hazard ratio [HR] 1.25, 95% confidence interval [CI] 0.91 to 1.72; p = 0.18), MI (HR 0.97, 95% CI 0.71 to 1.32; p = 0.85), TVR (HR 0.88, 95% CI 0.70 to 1.10; p = 0.26), and MACE (HR 0.95, 95% CI 0.81 to 1.12; p = 0.56). When site was removed from the models, the point estimates changed but the conclusions of the tests remained the same (all were still nonsignificant). The unadjusted Kaplan-Meier curves generated for time to TVR and MACE through the first 9 months after procedure illustrate the time-dependent differences in clinical events (Fig. 2). This analysis shows no statistically significant differences between the 2 stent products for time to event for either TVR or MACE up to 9 months.
To determine whether the similar outcomes for PES and SES observed in the overall population were consistent across complex patient subgroups, unadjusted hazard ratios for MACE were calculated for a number of subgroups (Fig. 3). A similar risk of MACE was observed across all but 3 subgroups: men, patients with prior CABG, and patients with SVG lesions. However, after propensity score adjustment for baseline differences in these 3 groups, only the SVG subgroup remained significant, but with wide CIs because of the small sample size.
Hazard ratios (HRs) of stent type for MACE in the overall patient population and specified subgroups. *These 3 subgroups show statistically significant unadjusted HRs. However, the male and prior coronary artery bypass graft (CABG) subgroups become nonsignificant when adjusted by propensity score. Saphenous vein graph (SVG) lesions remain significant (HR 2.38; 95% CI 1.21 to 4.69; p = 0.012). LAD = left anterior descending artery; STEMI = ST-segment elevation myocardial infarction; other abbreviations as in Figures 1 and 2.
The present study represents the largest multicenter prospective comparison of PES and SES procedures in patients receiving these devices in the U.S. These data describe clinical outcomes from a "real-world" population of patients having more complex clinical, lesion, and procedural characteristics than patients included in the randomized trials (TAXUS [Treatment of De Novo Coronary Disease Using a Single Paclitaxel-Eluting Stent]-I, -II, -IV, and -V, SIRIUS [Sirolimus-Coated Stents in De Novo Coronary Lesions], E [European]-SIRIUS, and C [Canadian]-SIRIUS).[12-18] This is indicated by the percentage of patients in the present study with acute evolving MI (9.1%), cardiogenic shock (0.8%), multilesion procedures (22.8%), bifurcating lesions (7.6%), SVG lesions (5.5%), thrombotic lesions (0.6%), CTO lesions (1.7%), in-stent restenosis lesions (4.5%), and many other lesions excluded in the randomized trials.
Although a number of baseline characteristics differed between PES and SES procedures (increased age, more STEMI, and smaller vessels in the PES group and more SVG, CTO, and in-stent restenosis lesions in the SES group), these differences were numerically small. Importantly, most of the baseline characteristics in the 2 groups were similar, including the prevalence of diabetes, lesion length, and total length of deployed stent. Both unadjusted and adjusted clinical outcomes at 9 months in this large comparison indicate that, in "real world" practice, selection of either of these 2 drug-eluting stent devices resulted in almost identical rates of adverse events. The clinical restenosis rate (TVR) observed here compares favorably with that seen in simpler lesions in previous randomized controlled trials in the U.S..[14,19] The TVR rate for SES in SIRIUS at 9 months was 6.2% and for PES in TAXUS-IV was 4.7%, compared with 4.3% for SES and 4.1% for PES in the present study. Although the patients in the STENT registry are more complex than earlier randomized trial patients, late event rates are similar and infrequent, most likely because of the lack of angiographically driven TVR seen in randomized trials.
The results of the present study are consistent with previous reports from outside of the U.S., such as the RESEARCH (Rapamycin-Eluting Stent Evaluated At Rotterdam Cardiology Hospital) and T (Taxus)-SEARCH registries from the Netherlands, the Milan registry, and the Latin American Society for Interventional Cardiology registry.[22,23] In the RESEARCH and T-SEARCH registries, PES and SES from nonconcurrent time periods were compared and showed 1-year clinically driven TVR rates of 5.4% and 3.7%, respectively (p = 0.30) and overall MACE rates of 13.9% and 10.5%, respectively (p = 0.74). The highest TVR and MACE rates reported come from the Milan experience in complex lesions with a median follow-up of 10.6 months, showing TVR rates of 12.5% for PES and 16.5% for SES (p = ns), MACE rates of 18.1% for PES and 21.0% for SES (p = ns), and a propensity analysis confirming the similarity between devices (HR 0.87, 95% CI 0.62 to 1.25). The SES outcomes in the present study are also similar to those reported from the e-CYPHER registry in both Europe and the U.S., with 1-year TVR and MACE rates in e-CYPHER of 3.1% and 5.8%, respectively. The PES event rates in the present study are also similar to those reported from other PES registries such as MILESTONE-II (Scimed Boston Scientific Online Stents European Registry)[25,26] and ARRIVE I (TAXUS Peri-Approval Registry: A Multi-Center Safety Surveillance Program) at 1 year, with TVR and MACE rates in the MILESTONE-II registry of 5.5% and 7.5%, respectively, and for the ARRIVE I registry of 5.6% and 7.2%, respectively.
When comparing the results of this study with earlier randomized trials in Europe, the similar clinical outcomes with PES and SES found in the present study reinforce the results of the REALITY (Prospective Randomized Multicenter Head-to-Head Comparison of the Sirolimus-Eluting Stent [Cypher] and the Paclitaxel-Eluting Stent [Taxus]) study, but contrast with the meta-analysis studies of PES versus SES by Kastrati et al.[7,28,29] and by Windecker et al.. In the REALITY study, which included routine angiographic follow-up, the TVR rates at 1 year for PES and SES were 7.9% and 8.0%, respectively (p = 0.99), and the MACE rates were 11.4% vs. 10.7%, respectively (p = 0.73). The 2 published meta-analyses of Kastrati et al. and Windecker et al. suggested significantly lower clinical event rates for SES compared with PES patients. However, these pooled analyses are limited, in that the majority of studies included are from one or two center studies as opposed to multicenter studies. In addition, they pool trials of broader populations, such as the REALITY and SIRTAX (Sirolimus-Eluting Stent Compared With Paclitaxel-Eluting Stent for Coronary Revascularization) studies, with other narrowly selected populations, as in the ISAR-DIABETES (Paclitaxel-Eluting Stent Versus Sirolimus-Eluting Stent for the Prevention of Restenosis in Diabetic Patients With Coronary Artery Disease) and ISAR-DESIRE (Drug-Eluting Stents for In-Stent Restenosis) studies. The limitation of patient selection and effects of routine angiographic follow-up may reduce the ability to generalize from these randomized trials to a broad spectrum of patients.
The results of the present study indicate that for patients receiving PES and SES described in this all-inclusive registry, the clinical outcomes at 9 months are indistinguishable between the 2 stent types. Based on these results, the decision of which stent type to use could be made by the physician based on comfort with the mechanical properties of a particular stent (e.g., deliverability, feel, balloon compliance) rather than expected differences in clinical outcomes. These findings show that when the operators in this group of centers selected a drug-eluting stent without forced assignment (randomization) and based on their own judgment of patient and lesion characteristics, outcomes were similar between PES and SES patients. The possibility that specific subgroups of patients may have had different outcomes cannot be determined definitively, although it is of interest that similar outcomes were seen for the 2 stents across most of the complex subgroups.
Because this registry was observational, study results may be confounded by the nonrandomized assignment of each treatment. In addition, registry findings can be limited by low rates of enrollment, lack of an independent clinical events committee, and under-reporting of events. However, this appears less likely in the present study, because of the close and comprehensive structure of this registry resulting from a limited number of study sites, homogeneity of methodology, and frequent auditing.
Conclusions and Clinical Implications
This study from a large multicenter registry shows that PES and SES were associated with low and indistinguishable overall MACE rates, including similar rates of ST, at 9 months' follow-up. The implications of these findings are that selection of a particular drug-eluting stent may be determined more by the operator's comfort with the mechanical properties of the stent for a given lesion rather than by any apparent differences in clinical outcomes.
Table 1. Baseline Patient Characteristics by Procedure Group (PES and SES)
Table 4. Unadjusted Adverse Outcomes at 9 Months Comparing SES and PES Groups
Table 5. Unadjusted and Adjusted Hazard Ratios for Clinical Outcomes Comparing SES and PES Groups at 9 Months
For a list of the STENT registry study centers and a table showing the propensity adjustment (p values for the statistically significant predictors of stent type before and after propensity adjustment), please see the online version of this article.
Special appreciation is extended to the STENT registry central coordinator, Susan Christopher, and executive assistant, Connie Poole, for their tremendous commitment to this study. In addition, each of the individual site coordinators, data coordinators, and physician operators are thanked for their diligence and success with patient consents, data collection, and follow-up.
Supported by unrestricted research grants from Cordis Corporation, Boston Scientific Corporation, Possis Medical, and The Medicines Company.
MACE = major adverse cardiac events; MI = myocardial infarction; PCI = percutaneous coronary intervention; PES = paclitaxel-eluting stent(s); SAT = subacute stent thrombosis; SES = sirolimus-eluting stent(s); ST = stent thrombosis; TVR = target vessel revascularization
Dr. Charles A. Simonton, Sanger Clinic, 1001 Blythe Boulevard, Suite 300, Charlotte, North Carolina 28203. (Email: Charles.Simonton@carolinashealthcare.org )
Charles A. Simonton, MD, FACC, FSCAI,* Bruce Brodie, MD,? Barrett Cheek, MD,? Fred Krainin, MD,? Chris Metzger, MD,|| James Hermiller, MD,# Stanley Juk, MD,** Peter Duffy, MD,?? Angela Humphrey, MS,?? Marcy Nussbaum, MS,?? Sherry Laurent, PhD,?? for the STENT Group
*Carolinas Heart Institute, Carolinas Medical Center, Charlotte, North Carolina
?LeBauer Cardiovascular Research Foundation, Greensboro, North Carolina
?High Point Regional Health System, High Point, North Carolina
?McLeod Regional Medical Center, Florence, South Carolina
||Holston Valley Medical Center, Kingsport, Tennessee
#Indiana Heart Institute, Indianapolis, Indiana
**Sisters of Charity Providence Hospitals, Columbia, South Carolina
??Moore Regional Hospital, Pinehurst, North Carolina
??R. Stuart Dickson Institute for Health Studies, Carolinas Medical Center, Charlotte, North Carolina.