Long-term Clinical Outcomes Following Coronary Stenting FREE

In 2003, drug-eluting stents (DES) became available in the United States. That year, an estimated 664 000 percutaneous coronary intervention (PCI) procedures were performed on 652 000 US patients, with 84% including stent implantation.¹ Since that time, DES have been used in increasingly complex cases of coronary artery disease (CAD), resulting in a marked decrease in the number of coronary artery bypass graft (CABG) surgical procedures.²

The advent of DES was heralded as a major breakthrough for preventing restenosis after percutaneous coronary artery revascularization, with randomized trials reporting 70% to 80% relative reductions in repeated procedures.³ The US Food and Drug Administration (FDA)-approved indication for DES is generally limited to discrete, de novo lesions in native vessels with reference vessel diameters of 2.5 to 3.5 mm.⁴ There are currently 2 FDA-approved DES devices: a polymer-based sirolimus-eluting stent (SES) (Cypher; Cordis, Miami Lakes, Florida) and a polymer-based paclitaxel-eluting stent (PES) (Taxus; Boston Scientific, Natick, Massachusetts).⁵ However, there is evidence that patient populations examined in the initial DES randomized studies represent a minority of patients who receive these devices in community practice.⁶

Currently, little evidence exists regarding the long-term effectiveness of DES vs bare metal stents (BMS) in community practice settings, where patient risk levels are higher than in clinical trials. This study sought to provide an in-depth evaluation of 2-year outcomes of DES and BMS in a large academic practice setting and to estimate the clinical benefit of these devices in patients with single-vessel and multivessel CAD.

The Duke Databank for Cardiovascular Disease (DDCD) is a clinical data resource containing detailed clinical, angiographic, therapeutic, and outcome data on approximately 54 000 patients following their cardiac catheterization procedures performed at the Duke Heart Center in Durham, North Carolina.^7- 9 The DDCD patients who underwent revascularization with BMS (from January 1, 2000) or DES (from April 1, 2003) through July 31, 2005, were included in the study population. Follow-up on those patients was extended to September 7, 2006, giving all patients a minimum of 12 months of follow-up data. Exclusion criteria were prior CABG surgery or PCI procedure, congenital heart disease, moderate to severe valvular heart disease, or significant (≥75% stenosis) left main coronary artery disease.

To analyze differences between BMS and DES, only patients receiving 1, or both, of these 2 treatments during their initial revascularization procedures were included in the study population. Patients with multiple lesions often underwent revascularization in a single treatment session. Patients receiving both a DES and a BMS during the same intervention were categorized as DES patients because the DES procedure would determine their postprocedure antiplatelet treatment regimen.

Clinical trials comparing DES vs BMS frequently use either late-loss, target lesion revascularization (TLR), or target vessel revascularization (TVR) as primary end points.¹⁰ However, the determination of these end points requires the use of protocol-driven relook coronary angiography procedures regardless of patient symptoms, which are not routinely performed in US clinical practice. We have identified symptom-driven TVR as a practice-based surrogate outcome for angiographically driven TLR or TVR. It is generally accepted that the rate of clinically driven TVR is approximately one-half that of angiographic TVR. We analyzed follow-up information on 8 events: death, nonfatal myocardial infarction (MI), TVR, any revascularization, non-TVR revascularization, combination of death or MI, combination of death or TVR, and the combination of death or MI or TVR. We have defined the combination of death or MI and TVR for the overall study cohort as the 2 primary end points for these analyses.

All patients were contacted at 6 and 12 months after their initial procedure and annually thereafter. An independent mortality committee reviewed follow-up results to confirm deaths. Follow-up was considered complete if the mortality committee confirmed the patient's death or if the patient was successfully contacted at the scheduled follow-up interval. Ninety-seven percent of all scheduled follow-up contacts were completed. Patients with incomplete follow-up were censored at the time of last contact in all analyses. The Duke University Medical Center (DUMC) institutional review board gave approval with a waiver of informed consent on February 27, 2006 (Registry No. 8223-06-2R0ER).

Tables of baseline and angiographic characteristics were categorized by stent group (DES vs BMS) and within the DES population by stent type (SES vs PES). Baseline characteristics were summarized as counts and percentages for categorical variables and as medians with 25th and 75th percentiles for continuous variables. Comparisons of baseline variables were conducted using nonparametric tests.

Characteristics used in the adjustment models include patient cardiovascular history and physical examination (body mass index; systolic blood pressure; heart rate; history of MI, mild valvular heart disease, third heart sound, cerebrovascular disease, peripheral vascular disease, carotid bruits; history and severity of congestive heart failure), demographics (sex, race, age), CAD risk factors (smoking history, hypertension, diabetes mellitus), diagnostic catheterization findings (left ventricular ejection fraction, extent of CAD), stent characteristics (mean stent diameter and total length of stents), and comorbid conditions (Charlson Index, history of chronic obstructive pulmonary disease, metastatic cancer, solid tumor, connective tissue disease, renal disease, liver disease). Mean stent diameter and total length of stents were used as surrogate variables for vessel diameter and lesion length.

Propensity scores were the primary tool used to adjust for differences in the observed patient characteristics between treatment groups.¹¹ For this study's analyses, the estimated propensity scores were the model-based probabilities of a patient receiving a particular stent (DES or BMS for the overall analysis, and SES or PES for the DES analysis) conditional on the observed covariates. Because DES therapy was not available in the initial years of the study, we did not include year of revascularization in the logistic regression models used to estimate the propensity scores.¹² Variables in the propensity score models included patient demographics, CAD risk factors, cardiovascular history and physical examination findings, diagnostic catheterization results, comorbid conditions, and stent characteristics. Adjusted comparisons were based on inverse probability weighted (IPW) estimators using estimated propensity scores.¹³ The IPW adjustment was applied to the baseline variables to assess the adequacy of the estimated propensity scores and the comparability of patient groups after propensity score adjustment.

Adjusted and unadjusted cumulative incidence rates for the study outcomes were calculated using IPW estimators with partitioning of data into monthly intervals as described by Bang and Tsiatis¹⁴ and Anstrom and Tsiatis.¹⁵ The IPW-unadjusted estimates were based on functions of Kaplan-Meier estimates for the treatment-specific censoring distributions.^14- 15 The IPW-adjusted estimates were based on weights that are functions of estimated propensity scores and Cox proportional hazards estimates of the treatment-specific censoring distributions. Point estimates for event rates were obtained using generalized estimating equations with robust standard errors used to compute 95% confidence intervals (CIs) and P values.¹⁶ Comparisons between the treatment groups correspond with the standard DDCD follow-up intervals for these patients and were conducted at 6, 12, and 24 months following the index stent procedure.

Because censoring patients after death has been shown to be associated with overestimation of cumulative nonfatal incidence rates, we applied IPW estimators to estimate all nonfatal MI, TVR, and non-TVR revascularization, and all revascularization end points.¹⁷ For time-to-event end points (death, death or nonfatal MI, death or TVR, death or nonfatal MI or TVR), we used propensity score–weighted survival techniques to estimate adjusted hazard ratios (HRs) and 95% CIs.¹⁸

Clinical trials and observational studies have shown that the clinical effectiveness of CAD revascularization procedures is in part dependent on patient characteristics. In recent years, the American College of Cardiology/American Heart Association guidelines for CABG surgery concluded that the primary survival benefits of CABG, PCI, and medical therapy can be stratified by patient CAD severity.^19- 21 In the present study, we analyzed CAD severity using the number of major epicardial vessels that are significantly (≥75%) diseased and performed stratified analyses comparing BMS vs DES for patients with 1-, 2-, and 3-vessel disease. In addition, we compared BMS vs DES for the subset of patients with diabetes mellitus. For DES patients, we performed analyses comparing SES vs PES. Statistical significance was set at the P  = .05 level with no correction for multiple comparisons. All calculations were performed using SAS statistical software (version 8.2; SAS Institute Inc, Cary, North Carolina).

During our study period, 3165 patients received BMS, and 1501 received DES (Table 1). Patients receiving DES vs BMS were more likely to have a history of diabetes mellitus but less likely to have a history of smoking or MI. The DES patients also had less single- and more multivessel CAD, resulting in a greater total length of stents but smaller mean stent diameters. Among DES patients, 1132 received SES, 283 received PES, 76 received both, and 10 patients received DES not classified as either. The DES patients receiving both SES and PES differed notably in many characteristics from those patients receiving only 1 type of DES, including being more likely to have multivessel CAD with greater total length of stents.

After adjustment by propensity scores, patients receiving DES and BMS had similar baseline characteristics (Table 2). The P values for all propensity score model variables comparing BMS and DES patient and stent characteristics were larger than P  = .40. After excluding DES patients who received both SES and PES, propensity score adjusted baseline characteristics for the SES and PES groups were balanced (P >.25 for all comparisons).

Across all outcomes, event rates for BMS and DES patients in the second year of follow-up were lower than those in the initial year (Table 3). At 2 years, there was no notable difference between BMS and DES patients with regard to mortality. There was a small but statistically significant difference in nonfatal MI in favor of DES, and a larger difference in TVR, but no difference in non-TVR revascularization. Of note, for the death or MI composite outcome, there was a statistically significant advantage for DES vs BMS at 6 months, which was not significant at 12 and 24 months (see Table 3 for P values).

After adjusting for differences in baseline characteristics, the use of DES vs BMS was not associated with a mortality difference at 24 months; however, there were statistically significant reductions in nonfatal MI and TVR, and in all composites except for death or MI (see Table 4 for P values, and the Figure). The propensity score–weighted survival models reinforced these results. The estimated DES vs BMS HR for mortality was 1.05 (95% CI, 0.85-1.29), whereas the HR for death or MI was 0.96 (95% CI, 0.80-1.15; P = .66). A large benefit for DES vs BMS was observed for the time to death or MI or TVR end point with an estimated HR of 0.66 (95% CI, 0.57-0.77; P < .001).

Comparisons of adjusted outcomes by number of diseased vessels served to amplify the results observed in the entire population (see Table 5 for P values). There was some suggestion of a benefit in death or nonfatal MI associated with the use of DES vs BMS in patients with multivessel CAD. There were significant differences in the rates of TVR, such that the benefit associated with DES vs BMS use increased with the number of diseased vessels. There also was a significant difference at 24 months in revascularizations of other than the target vessel favoring DES vs BMS. Of note, the differential rates of TVR-related composite outcomes (death or TVR, and death or MI or TVR) at 24 months for patients with 2-vessel CAD are twice those for patients with 1-vessel CAD, and the rates for those with 3-vessel CAD are 3 times those for patients with 1-vessel CAD.

There were no statistically significant treatment differences in the rates of death or MI for the subset of 1206 patients with diabetes mellitus (24-month rates: DES, 15.4%; BMS, 16.2%; difference, −0.8%; 95% CI, −6.1% to 4.5%; P = .77). As with the overall study cohort, the TVR benefit for DES vs BMS was large and highly statistically significant (24-month rates: DES, 7.5%; BMS, 19.3%; difference, −11.8%; 95% CI, −16.1% to −7.6%; P < .001). Among patients with diabetes mellitus and multivessel CAD (n = 537), there was no significant difference for death or MI (24-month rates: DES, 20.5%; BMS, 20.9%; difference, −0.4%; 95% CI, −9.1% to 8.2%; P = .93) but a large benefit for DES for the TVR end point (24-month rates: DES, 8.2%; BMS, 22.4%; difference, −14.2%; 95% CI, −21.4% to −7.0%; P < .001).

Because of the small number of PES patients with 24 months of follow-up, results by DES type are given only at 6 and 12 months (see Table 6 for P values). Although there are no statistically significant differences between SES and PES for death, nonfatal MI, or TVR, there were trends favoring SES. These resulted in a statistically significant difference in the composite of death or MI or TVR at 6-months' follow-up but no significant difference at 12 months.

Our results suggest that patients who receive DES vs BMS have similar overall long-term rates of death or MI but substantially lower rates of target vessel revascularization. Although the current FDA-approved indication for DES is limited to discrete, de novo lesions in native vessels with reference vessel diameters of 2.5 to 3.5 mm,⁴ our results indicate that the target vessel revascularization benefits associated with DES vs BMS occur in patients with 1-, 2-, and 3-vessel CAD and are greater in patients with multivessel disease vs those with single-vessel disease. We found no notable differences in 12-month outcomes between patients receiving SES and those receiving PES.

Our results for death or MI agree with those from a pooled analysis of 9 randomized studies (n = 5261) that found no statistically significant differences between DES and BMS for death or nonfatal MI over a 4-year follow-up period.⁵ These results also agree with findings from the 12 395 patients in the Western Denmark Heart Registry²² who underwent percutaneous coronary intervention from 2002 to 2005 and who were followed for 15 months. A total of 11 730 coronary lesions were treated with BMS, and 5422 lesions were treated with DES. Both BMS and DES patients received clopidogrel bisulfate for 12 months following stent implantation. There were no significant differences detected in death (unadjusted rates: BMS, 6.2%; DES, 4.4%; adjusted HR, 0.90; P = .29) or nonfatal MI (unadjusted cumulative incidence: BMS, 3.0%; DES, 3.2%; P = .65).

However, our study differs from the Swedish Coronary Angiography and Angioplasty Registry (SCAAR)²³ of 19 771 patients, which estimated an adjusted mortality HR of 1.18 (95% CI, 1.04-1.35), indicating increased risk for DES patients over 3 years of follow-up. We believe that these results may in part be explained by differences in the use of antiplatelet therapy following DES implantation. In the SCAAR study,²³ most of the DES patients stopped taking clopidogrel at 6 months, and these researchers observed a reversal in the mortality HR at that time. Previous work²⁴ from our group has demonstrated the importance of long-term dual antiplatelet therapy in DES patients. Thus, the death and MI results in our present study should be interpreted within the context of a community practice with self-reported clopidogrel usage rates among DES patients of 52.4%, 46.8%, and 35.9% at 6, 12, and 24 months, respectively.²⁴

The advantage of DES over BMS for initial revascularization seems to lie principally in the reduced need for repeated revascularization of the same coronary vessel. At the 2-year follow-up, patients receiving DES vs BMS in our study showed a 9.7% absolute reduction in TVR, which supports prior clinical trial results.⁵ Interestingly, our results from a community-based setting suggest that the TVR benefit for DES vs BMS seems to be even larger in patients with multivessel CAD vs those with single-vessel CAD. This result is not surprising, given that patients with multiple stents would have more opportunities to undergo TVR procedures. However, the early DES clinical trials were restricted to stenting of a single-vessel segment per patient and therefore have not provided evidence to support the use of these devices in multivessel CAD populations.

Our findings suggest that event rates in clinical practice differ from those in clinical trials. The use of coronary stents in clinical practice is associated with smaller reductions in TVR rates than in clinical trials and with greater risk for death. In our population, the risk for 1-year mortality (6.4% for DES and 5.4% for BMS) was greater than that observed in pooled clinical trials. In the SES clinical trials, the 1-year rates of death were 1.2% for DES patients and 0.8% for BMS patients. Similarly, in the PES clinical trials, the 1-year rates of death were 1.6% for DES patients and 1.8% for BMS patients. In contrast, the 1-year mortality rate in the SCAAR population²³ was approximately 4%. Thus, our rates are closer to those observed in the Swedish and Danish registries than those reported in randomized trials.

One-year adjusted TVR rates in our community practice (4.4% for DES vs 13.2% for BMS) are lower than the TVR rates reported in clinical trials (mean rates, 6.2% for DES vs 16.6% for BMS). At the 2-year follow-up, our adjusted TVR rates (6.6% for DES vs 16.3% for BMS) were less than those reported for the RAVEL trial²⁵ (7.7% for DES vs 30.6% for BMS), but comparable with those from the RESEARCH registry²⁶ (8.2% for DES vs 14.8% for BMS). Prior clinical trial protocols have mandated repeated angiography that has been shown to increase the frequency of repeated TVR procedures even in patients without clinical symptoms. In contrast, restenosis in clinical practice is typically detected when symptoms prompt repeated angiography. The general rule of thumb is that clinically-driven revascularization rates are approximately half those of angiographically driven revascularization rates. Our data support the hypothesis that TVR rates are lower when they are clinically driven, but the rates in our study are higher than the “rule-of-thumb” proportion that is often quoted. As with the results from prior clinical trials, the 2-year TVR rate differences in our clinical practice for DES vs BMS were largely responsible for observed differences in the composite of death or MI or TVR (16.2% for DES vs 25.6% for BMS). However, these differences should not be trivialized because a substantial proportion of in-stent restenosis cases present as MI or unstable angina and require hospitalization.^27- 28 Thus, the downstream risks of restenosis may be clinically important.

Our study did not observe statistically significant differences in 12-month event rates between patients receiving SES vs PES, and we lacked sufficient data on patients with PES to assess 24-month results (see Table 5 for P values). A pooled analysis⁵ of 9 double-blind trials comparing SES and PES with BMS suggested a slightly better HR for the SES than for the TVR end point (HR of 0.38 for SES vs BMS; HR of 0.62 for PES vs BMS). However, an observational study²⁹ of 6509 patients receiving PES or SES found no significant differences for death or MI over a 1-year follow-up period. This registry found no statistically significant differences between PES and SES for TVR (1-year rates, PES 5.5% vs SES 6.3%; P = .20). Results from our clinical practice and the pooled DES vs BMS clinical trials do not suggest any clear advantage of SES or PES for the death or nonfatal MI end points. Nonetheless, larger studies with longer follow-up will be required to address the long-term effectiveness of SES vs PES in actual practice settings.²⁹

Our study was observational and conducted during a period of rapid technology diffusion that had important implications for the DUMC CAD practice.²⁴ In addition, clinical practice at the DUMC may not represent a typical community practice. Although we believe that we have adjusted for measured differences in our analyses, we recommend that readers consider our practice-based results along with those from clinical trials and other registries when making their assessments of the relative efficacy of DES vs BMS. In addition, although we have demonstrated improved outcomes for DES vs BMS in patients with multivessel CAD, our study is limited to the long-term comparison of DES vs BMS and did not consider CABG surgery, which may be a more appropriate comparator treatment strategy for these patients.^30- 31

Because our analyses relied on DDCD data, detailed TVR information was available only for procedures occurring at DUMC. Therefore, a few TVR end points occurring at non-DUMC hospitals may not have been captured by our follow-up methods. However, for patients initially treated with stents who subsequently underwent revascularization, 99% of follow-up PCIs and 95% of follow-up CABG surgical procedures were performed at DUMC. Thus, we believe that our results give a fair representation of what one would expect to observe in actual community practice.

In conclusion, in our clinical practice there was a significant reduction in TVR for DES vs BMS, but no reduction in the composite end point of death or MI. Our results from a community practice setting serve to support, although at a diminished level, the TVR benefits reported for clinical trials of DES vs BMS but also suggest a higher risk of death for both types of stents than found in clinical trials. Our 2-year follow-up period extends well beyond the time horizon of most clinical trials, and our results demonstrate the sustainability of the reduction in TVR. In addition, our results do suggest that the TVR benefits extend to patients with multivessel disease.

Correspondence: Kevin J. Anstrom, PhD, Duke Clinical Research Institute, PO Box 17969, Durham, NC 27715 (Kevin.Anstrom@duke.edu).

Accepted for Publication: January 27, 2008.

Author Contributions:Study concept and design: Anstrom, Kong, Califf, Kramer, Matchar, and Eisenstein. Acquisition of data: Anstrom, Shaw, Califf, and Matchar. Analysis and interpretation of data: Kong, Shaw, Kramer, Rao, Matchar, Mark, Harrington, and Eisenstein. Drafting of the manuscript: Anstrom, Kong, Matchar, and Eisenstein. Critical revision of the manuscript for important intellectual content: Kong, Shaw, Califf, Kramer, Rao, Matchar, Mark, Harrington, and Eisenstein. Statistical analysis: Anstrom and Shaw. Obtained funding: Kong, Califf, and Matchar. Administrative, technical, and material support: Califf, Matchar, Mark, Harrington, and Eisenstein. Study supervision: Kong, Rao, Matchar, and Eisenstein.

Financial Disclosure: Dr Anstrom has received research and salary support from Alexion, AstraZeneca, Bristol Myers Squibb, Lilly, Eyetech, Innocoll Pharmaceuticals, Medtronic, Medtronic Vascular, Pacific Therapeutics, Pfizer, and Proctor & Gamble. Dr Kong has received research and salary support from the Agency for Healthcare Research and Quality, IBM, Novartis, Proctor & Gamble, and Terumo Corp, and personal income for consulting services from Allmed Healthcare Management. Dr Califf has received research and salary support from Novartis Pharmaceutical and Schering-Plough; has reported activities that generate income for Duke from Heart.org/Concepts, Kowa Research Institute, Merck, Novartis Pharmaceutical, Sanofi-Aventis, Schering-Plough, Scios Pharma, and Vertex; has reported activities that generate personal income from Avalere Health, AstraZeneca, Biogen, Bayer Corp, Brandeis University, Bristol Myers Squibb, Sanofi-Aventis, Heart.org/Concepts, Five Prime, Kowa Research Institute, Merck, NITROX, Novartis Pharmaceutical, Schering-Plough, Scios Pharma, and Vertex; and has equity in NITROX. Dr Kramer has received research and salary support from Pfizer and personal income for consulting services from Icagen and Lilly. Dr Peterson has received research and salary support from Bristol Myers Squibb, Millennium Pharmaceutical, Sanofi-Aventis, Schering-Plough, Pfizer, and Scios; and has received personal income for consulting services from Amgen, Bayer Corp, and CV Therapeutics. Dr Rao has received research funding from Cordis Corp and Momenta Pharmaceuticals and has acted as a consultant and engaged in the Speakers' Bureau for Sanofi-Aventis and The Medicines Company. Dr Mark has received research, salary, and miscellaneous support from Aventis, AstraZeneca, Medtronic, Novartis, National Institutes of Health (NIH)/National Heart, Lung, and Blood Institute, NIH/Agency for Healthcare Research and Quality, Proctor & Gamble, Pfizer, Medtronic, Alexion Pharmaceuticals, Medicure, and Mosby. Dr Harrington has reported educational activities that generate revenue for Duke with Schering-Plough; consulting services that generate personal income with AstraZeneca, Baxter, Bayer AG, Bristol Myers Squibb, Indigo Pharmaceuticals, KAI Pharmaceuticals, Medicure, Merck Group, Millennium Pharmaceutical, NicOx, OLG Research, Sanofi-Aventis, Sanofi-Synthelabo, Schering-Plough, Seredigm, The Medicines Company, and WebMD. Dr Eisenstein has received research and salary support from Medtronic Vascular Inc.

Funding/Support: This study was funded under contract No. 290-05-0032 from the Agency for Healthcare Research and Quality (AHRQ), US Department of Health and Human Services as part of the Developing Evidence to Inform Decisions about Effectiveness (DEcIDE) program.

Disclaimer: The authors of this article are responsible for its content. Statements in the article should not be construed as endorsement by the AHRQ or the US Department of Health and Human Services. This article is based on the AHRQ DEcIDE report titled “Treatment of In-Stent Restenosis” (AHRQ publication, under review).

Additional Contributions: The staff and faculty at DUMC assisted in collecting the data. Allyn Meredith, MA, and Maqui Ortiz, at Duke Clinical Research Institute (DCRI), assisted in the editing and review of these materials; Judith A. Stafford, MS, DCRI, assisted with the programming; Charles B. McCants Jr, BS, DCRI, updated the DDCD follow-up files. (All were compensated by the DCRI for their work.) Elise Berliner, PhD, of the AHRQ, provided support and invaluable feedback during the course of the “Treatment of In-Stent Restenosis” project. (Ms Berliner was compensated by the AHRQ.)