Drug-eluting stents (DES) have reduced the incidence of angiographic and clinical restenosis compared with bare-metal stents (BMS) in randomized clinical trials (RCT) of highly selected patients (1- 2). This benefit appears to persist for up to 4 years after stent implantation (3). This has led to the widespread use of DES, including in patients who would not have met the eligibility criteria for inclusion in the RCT of DES versus BMS. However, recent data suggest that DES may be associated with an increased rate of late (>1 year) stent thrombosis (ST), myocardial infarction (MI), and death compared with BMS (4- 6), particularly in patients not receiving clopidogrel (7). Because of concern of adverse late events with DES, the U.S. Food and Drug Administration (FDA) convened a panel to review available data from both pivotal RCT of DES versus BMS and post-RCT registry and single-center studies (8- 9). Based on review of these data, the panel concluded that when DES were used for their approved indications (“on-label”) the risk of late DES thrombosis did not outweigh the advantages over BMS in reducing rates of repeat revascularization (10). In contrast, the panel concluded, and the FDA concurred, that adverse late events occurred at a sufficient incidence to raise concern about the safety of “off-label” DES use (10).

Despite the panel’s conclusions, there are no RCT comparing outcomes of “on-label” and “off-label” stent treatments between DES and BMS (9). Moreover, it is uncertain that an adequately powered clinical trial to evaluate these comparisons could be performed or that it would be representative of the results under the conditions of routine practice (11). To test the hypothesis that late DES outcomes may be inferior to BMS when used in “off-label” stent treatments, we assessed the clinical outcomes in consecutive patients treated with DES when DES utilization was ≥90% and compared them with patients who received BMS before the availability of DES. The majority (>75%) of both stent groups were “off-label.”

Patients at our institution undergoing percutaneous coronary intervention (PCI) from April 2002 to April 2005 were included in the study. Of these, 1,164 consecutive patients, representing all patients who underwent coronary artery stenting between April 2002 and April 2003, before FDA approval of DES in the U.S., received BMS and served as the control group. The study group consisted of 1,285 consecutive patients who received DES after these stents were fully available (February 2004) and had replaced BMS as our routine stents of choice (≥90% utilization). Patients were excluded if they received both BMS and DES (n = 8) or were unavailable for follow up (BMS, n = 29; DES, n = 35). Patients were not excluded from the study for any other reason. Thus, 1,135 BMS and 1,242 DES patients composed the control and study groups, respectively. The study was approved by the Institutional Review Board of Wake Forest University Baptist Medical Center. We previously reported the 9-month follow-up of most of these patients (12).

Percutaneous coronary intervention was performed according to standard techniques. Because sirolimus-eluting stents were available much earlier than paclitaxel-eluting stents, they composed most of the DES used in the study: sirolimus-eluting stents, n = 971; paclitaxel-eluting stents, n = 259; both, n = 12. Anticoagulation during PCI was accomplished with unfractionated heparin or bivalirudin per standard protocol. Patients received glycoprotein IIb/IIIa receptor inhibition according to usual protocol with abciximab or eptifibatide at the discretion of the interventionalist (12). All patients were treated with aspirin (81 to 325 mg/day) before PCI and indefinitely thereafter. Patients also received clopidogrel (300 to 600 mg as a loading dose, given before or immediately after the procedure, followed by 75 mg/day). Clopidogrel was given for a minimum of 1 month in BMS-treated patients, for a minimum of 3 months for sirolimus-eluting stent–treated patients, and for a minimum of 6 months for paclitaxel-eluting stent–treated patients. Additional clopidogrel use was at the discretion of the physician responsible for clinical care of the patient.

Before hospital discharge, patient and procedural data and hospital outcomes were entered into the Wake Forest University Baptist Medical Center Cardiovascular Information Services Database. Collection of data and outcomes measures conformed to the American College of Cardiology National Cardiovascular Database Registry definitions for cardiovascular data (13). All patients reported in this study had equivalent follow-up duration. Clinical follow-up was obtained as follows: Independent chart review, including follow-up visit with a cardiologist, was available for 80% of patients; scripted phone interviews were obtained for 18% of patients, who had no clinical follow-up but no reported hospitalization since their index procedures; and review of the Social Security Death Index for 2% of patients, where the death records were the only available follow-up. Follow-up was censored at 2 years ± 30 days, with complete follow-up available in 95% of BMS and 90% of DES patients.

Stent thrombosis was defined following the recommendations of the Academic Research Consortium (ARC) for definite and probable ST as presentation with acute coronary syndrome and definite angiographic or pathologic evidence of ST, unexplained death within 30 days of stent placement, or target vessel infarction in the absence of angiography (14). “On-label” stent use was defined by the study criteria for the initial randomized DES studies (1- 2) as follows: >18 years old, single de novo native coronary artery lesions <30 mm in length without thrombus, left ventricular ejection fraction ≥25%, no MI within 7 days of the procedure, and no evidence of renal failure (serum creatinine ≤2.0 mg/dl). Stent use in all other patients was defined as “off-label.” This definition of “on-label” use is similar to the indications for both Cypher (Cordis Corporation, Miami, Florida) and Taxus (Boston Scientific, Billerica, Massachusetts), with the exception that renal failure was not specifically listed as a contraindication for DES use in the indications. Nonfatal MI was defined as ischemic symptoms and an elevation of creatine kinase-MB >2× the upper limit of normal, with or without ST-segment elevation or development of Q waves.

Descriptive statistics (means and standard deviation of continuous factors, frequency counts, and relative frequencies of categoric factors) were calculated and compared using the Wilcoxon rank sum test for continuous factors and chi-square testing for categoric factors. Hazard ratios (HR) are presented along with their 95% confidence intervals (CI). Kaplan-Meier plots of cumulative incidence were constructed from index procedure to 2 years of follow-up. The log rank test was used to test for differences between DES and BMS incidence curves. Cox proportional hazards modeling was used to assess independent predictors of outcomes at 2 years to account for follow-up data censored before 2 years. The proportional hazards assumption was tested for all variables by examining log-log survival curves. No variables in the final models violated the proportional hazards assumption. The SAS version 9.1 statistical software package (SAS Institute, Cary, North Carolina) was used for all statistical analyses.

The baseline clinical characteristics of the BMS and DES stent groups were similar (Table 1). Acute coronary syndromes were present overall in 71% of patients, (p = 0.23; BMS vs. DES). Procedural characteristics were similar with the exception that average length stented was longer in DES-treated patients, 25 ± 8 mm, compared with BMS-treated patients, 20 ± 10 mm; p < 0.001. By study design, the baseline characteristics of “on-label” and “off-label” patients differed, with more recent MIs, comorbid illnesses such as renal failure, and multilesion and vessel procedures in the “off-label” stent treatments (Table 1). However, within the “on-label” and “off-label” treatment groups, patient and procedural characteristics were similar for BMS and DES. Medication use during the follow-up period was available for some of the patients in the study (Table 2). Aspirin was used in 92% of the BMS patients and 91% of the DES patients; p = 0.86 (at 2 years). Clopidogrel use differed between BMS and DES patients, but only at 1 year of follow-up. At 1 year, clopidogrel use was 40% for BMS and 51% for DES (p < 0.001), and at 2 years clopidogrel use was 38% for BMS and 42% for DES (p = 0.23). Statin use at 2 years was 83% for BMS and 85% for DES; p = 0.32.

Kaplan-Meier plots of the cumulative incidence of selected outcomes at 2 years are shown for the entire BMS and DES groups in (Figure 1). The hazard of TVR was lower with DES compared with BMS: HR 0.62 (95% CI 0.48 to 0.80). The mortality curves of the 2 stent groups overlap until approximately 90 days and then diverge, with a lower hazard of death in the DES group compared with the BMS group that persisted out to 2 years: HR 0.71 (95% CI 0.54 to 0.92). Mortality or nonfatal MI was also lower in the DES patients compared with the BMS patients: HR 0.77 (95% CI 0.62 to 0.95). Stent thrombosis, however, was similar, with HR 0.97 (95% CI 0.49 to 1.91) for DES compared with BMS.

A total of 530 patients (22%) in this study met the criteria for “on-label” stent use (281 BMS, 249 DES), and the remaining 1,847 (78%) were “off-label” stent procedures (854 BMS, 993 DES). Cumulative Kaplan-Meier plots of incidence of selected outcomes for all 4 groups are shown in (Figure 2). Overall, the hazard of nonfatal MI or death was higher for “off-label” compared with “on-label” stent treatments: HR 2.53 (95% CI 1.80 to 3.57). However, within both “on-label” and “off-label” stent treatment groups, all-cause mortality and nonfatal MI or death were lower for DES compared with BMS. For “off-label” stent treatments, the hazard of nonfatal MI or death was 0.78 (95% CI 0.62 to 0.98) and all-cause mortality 0.72 (95% CI 0.54 to 0.94) for DES compared with BMS. There were no stent thromboses in “on-label” stent treatments during the 2-year study period (Figure 3). For “off-label” stent treatments, there were a total of 16 stent thromboses in BMS and 17 in DES patients; p = 0.93. The hazard of ST was 0.91 (95% CI 0.46 to 1.80) for DES compared with BMS. Late stent thromboses (>1 year) occurred only in “off-label” DES patients (0.8%); Fisher exact test versus “off-label” BMS: p = 0.016. Three of these patients reported taking clopidogrel at the time of their DES thrombosis.

Multivariate analysis of ST and nonfatal MI or death for the entire study population over the 2-year study period is shown in (Table 3). History of renal failure (HR 2.11 [95% CI 1.52 to 2.93]) and diabetes mellitus (HR 1.73 [95% CI 1.38 to 2.17]) were 2 of the strongest independent predictors of increased nonfatal MI or death, although DES use (HR 0.75 [95% CI 0.60 to 0.94]) was the strongest independent predictor of lower nonfatal MI or death. The HR of nonfatal MI or death for DES versus BMS were also compared across covariate strata (Figure 4). For almost all of the clinical and lesion variables assessed, the point estimate of the HR for nonfatal MI or death favored DES, although the upper boundaries of the 95% CI crossed the line of equivalency. The overall HR for nonfatal MI or death comparing DES with BMS was 0.77 (95% CI 0.62 to 0.95), favoring DES.

In this large contemporary experience comprising mainly coronary stent procedures classified as “off-label” (9- 10), we observed a significantly lower hazard of revascularization, nonfatal MI or death, and all-cause mortality at 2 years after DES compared with BMS. The overall incidence of ST using the ARC definition of definite and probable ST (14) was similar for DES and BMS at 2 years, although the only late stent thromboses were observed in “off-label” DES patients. “On-label” stent treatments were associated with fewer adverse events at 2 years than “off-label” stent treatments independent of stent type. This observation is consistent with the higher-risk baseline and procedural characteristics of patients receiving “off-label” treatments. However, “off-label” DES was associated with lower cumulative event rates at 2 years compared with BMS with equivalent safety (nonfatal MI or death) in spite of a 0.8% incidence of late DES thrombosis. Thus, the strategy of switching to routine use of DES from BMS as evaluated in this study was associated with lower rates of both repeat revascularization and hard cardiac end points (death and death or nonfatal MI).

As discussed previously, the FDA recently convened a panel amidst growing concerns about DES safety (8). After review of available data, however, the panel concluded that when DES were used for their approved indications (i.e., “on-label”), the risk of late DES thrombosis did not outweigh the advantages over BMS in reducing the rates of repeat revascularization (10,15). In contrast, the panel voiced concern of DES safety in patients treated with high-risk characteristics (i.e., “off-label” DES use) (9- 10) after reviewing studies examining outcomes conducted outside of RCT (8- 9,16). Other recent studies also have observed that high-risk patient and lesion factors contribute to the risk of late ST and nonfatal MI or death after DES. However, patients in these studies were drawn from study populations consisting exclusively of DES patients (17- 18). Whether the increased relative risk of DES use in patients with these high-risk features would be observed compared with a group of patients with similar features treated with BMS has not been previously evaluated (9). Our observations confirm that high-risk baseline patient and lesion characteristics are associated with higher rates of late adverse outcomes, regardless of the type of coronary stent used. However, our analysis of the relative risk/benefit of DES compared with BMS in patients with these high-risk features indicated that use of DES was as safe, if not better, than BMS with respect to cumulative outcomes in patients in whom these high-risk features were present. Longer follow-up of these patients will be necessary to determine if continued late DES thrombosis occurs and negates the benefits of DES observed at 2 years.

Stent thrombosis and associated nonfatal MI occur more than 1 year after stent implantation with DES, but not BMS, after both “on-label” and “off-label” stent procedures (15). However, the reported absolute incidence of these events, and their effect on overall outcomes, differs among studies. In the BASKET-LATE (Basel Stent Cost-Effectiveness Trial: Late Clinical Events Related to Late Stent Thrombosis After Stopping Clopidogrel) study, there was a significantly higher incidence of late (>6 months) ST and nonfatal MI with DES than with BMS, although the cumulative rates of nonfatal MI and death were similar for BMS and DES (4). In the SCAARS (Swedish Coronary Angiography and Angiolasty Registry Study), mortality at 3 years was higher with DES than with BMS (6). In contrast, Tu et al. (19) and the 3-year report of the T-SEARCH (Taxus Stent Evaluated at Rotterdam Cardiology Hospital) registry (20) both noted lower cumulative event rates with DES than with BMS. The source of variability of these study observations will likely be the subject of intense scrutiny over the next several years as the safety of DES is evaluated further.

Similar to most recent studies (4- 7), we observed late ST with DES- but not BMS-treated patients. However, in contrast to some of those studies, the rates of late ST and nonfatal MI in the present study were relatively low and the cumulative incidence of nonfatal MI or death after 2 years were lower for DES than for BMS patients. The reason for these apparent differences is not clear. However, several factors may be important. First, the 2-year cumulative incidence of DES thrombosis of 1.5% in the present study is not statistically different than the 2-year DES thrombosis rate of 1.2% reported by Mauri et al. (14); p = 0.41, using the same ARC definition. Second, the absolute number of late adverse events attributable to DES in both the BASKET-LATE study (4) and the Duke Registry (7) were small. Therefore, the power to generalize their findings may be limited. Third, the outcomes of the present study represent the strategy of using DES for all coronary stent treatments at a time when they were fully available and considerable experience in their implantation had been obtained. In this manner, we minimized the potential selection bias that likely influences the nonrandomized studies comparing outcomes of patients who were chosen to receive BMS with those who received DES. In contrast, the impact of the varied use and experience with DES on study outcomes that was present in other studies, particularly the SCAARS, where use of DES ranged from 1% to 60% in the participating study centers, is uncertain (6). Finally, we compared DES to a historical, not contemporaneous, control group receiving BMS. It is possible that advances in medical treatment, such as more aggressive statin use and more prolonged use of clopidogrel, may have affected the outcomes of this study (7,21- 22). However, we observed that use of these medications (aspirin, clopidogrel, and statins) was similar for both BMS- and DES-treated patients at 2 years. Further study of the effect of medical treatment on late outcomes merits attention.

The potential mechanism of a lower 2-year cumulative incidence of nonfatal MI or death with DES was not examined in the present study. Evaluation of the benefit of DES has focused on the ability of this therapy to reduce restenosis. Although restenosis is traditionally considered to be a benign process, recent data suggest that restenosis may present as an MI in up to 9.5% of patients (23- 24). Moreover, and independent of the clinical presentation, those receiving treatment for restenosis have a small incidence of nonfatal MI or death in the year following this repeat procedure (25). Our observations support this possibility, because most of the apparent benefit occurred in the first 9 months after stent implantation, coincident with the period of restenosis vulnerability with BMS. Additional mechanisms of benefit may be relevant, such as passivation of potentially vulnerable plaque owing to stenting of longer lengths of the coronary artery with DES compared with BMS. However, further studies will need to be performed to address these potential mechanisms.

Observational studies such as this one may be subject to event bias due to unequal follow-up. However, we obtained nearly complete follow-up, so that ≥90% of the patients had follow up available at 2 years. The present study may also be confounded by selection bias. However, the DES and BMS patients had very similar baseline clinical and lesion characteristics. Moreover, use of a recent historical control group avoids the potential selection bias of stent therapies when both BMS and DES are available. Randomized clinical trials would provide the fairest evaluation of DES efficacy and safety, but RCT usually exclude the very type of high-risk patients that are of interest (11). Although there were more than 1,000 patients in each treatment group, the study was underpowered to evaluate differences in the incidence of ST. Finally, our study did not examine outcomes beyond 2 years. Hopefully, longer-term follow-up of cohorts such as this will provide valuable information concerning the relative incidence of late adverse events after drug-eluting coronary stent treatment.

The observations of lower cumulative rates of nonfatal MI and all-cause mortality at 2 years in DES-treated patients compared with BMS-treated patients in spite of a low incidence of late DES thrombosis is reassuring that use of DES in routine practice is safe. Moreover, the observations that “off-label” DES in patients with high-risk baseline and lesion characteristics were associated with better outcomes than with similar patients treated with BMS provide clinical evidence supporting broader, not more restricted, use of DES. In spite of the benefits of DES observed in this study, recent studies raise important questions about long-term DES safety in a broader population of patients than was studied in the initial RCT. Longer-term follow-up safety and effectiveness studies of DES use will need to be obtained to determine if late DES-related events will affect the balance of safety and effectiveness more than 2 years after DES implantation.

The authors gratefully acknowledge Tammy Davis for manuscript preparation and Angelina Pack, Aruna Joel, Sabrina Smith, and Robin Taylor for data collection and database entry.