Clopidogrel Use and Long-term Clinical Outcomes After Drug-Eluting Stent Implantation

Context Recent studies of drug-eluting intracoronary stents suggest that current antiplatelet regimens may not be sufficient to prevent late stent thrombosis.

Objective To assess the association between clopidogrel use and long-term clinical outcomes of patients receiving drug-eluting stents (DES) and bare-metal stents (BMS) for treatment of coronary artery disease.

Design, Setting, and Patients An observational study examining consecutive patients receiving intracoronary stents at Duke Heart Center, a tertiary care medical center in Durham, NC, between January 1, 2000, and July 31, 2005, with follow-up contact at 6, 12, and 24 months through September 7, 2006. Study population included 4666 patients undergoing initial percutaneous coronary intervention with BMS (n = 3165) or DES (n = 1501). Landmark analyses were performed among patients who were event-free (no death, myocardial infarction [MI], or revascularization) at 6- and 12-month follow-up. At these points, patients were divided into 4 groups based on stent type and self-reported clopidogrel use: DES with clopidogrel, DES without clopidogrel, BMS with clopidogrel, and BMS without clopidogrel.

Main Outcome Measures Death, nonfatal MI, and the composite of death or MI at 24-month follow-up.

Results Among patients with DES who were event-free at 6 months (637 with and 579 without clopidogrel), clopidogrel use was a significant predictor of lower adjusted rates of death (2.0% with vs 5.3% without; difference, −3.3%; 95% CI, −6.3% to −0.3%; P = .03) and death or MI (3.1% vs 7.2%; difference, −4.1%; 95% CI, −7.6% to −0.6%; P = .02) at 24 months. However, among patients with BMS (417 with and 1976 without clopidogrel), there were no differences in death (3.7% vs 4.5%; difference, −0.7%; 95% CI, −2.9% to 1.4%; P = .50) and death or MI (5.5% vs 6.0%; difference, −0.5%; 95% CI, −3.2% to 2.2%; P = .70). Among patients with DES who were event-free at 12 months (252 with and 276 without clopidogrel), clopidogrel use continued to predict lower rates of death (0% vs 3.5%; difference, −3.5%; 95% CI, −5.9% to −1.1%; P = .004) and death or MI (0% vs 4.5%; difference, −4.5%; 95% CI, −7.1% to −1.9%; P<.001) at 24 months. However, among patients with BMS (346 with and 1644 without clopidogrel), there continued to be no differences in death (3.3% vs 2.7%; difference, 0.6%; 95% CI, −1.5% to 2.8%; P = .57) and death or MI (4.7% vs 3.6%; difference, 1.0%; 95% CI, −1.6% to 3.6%; P = .44).

Conclusions The extended use of clopidogrel in patients with DES may be associated with a reduced risk for death and death or MI. However, the appropriate duration for clopidogrel administration can only be determined within the context of a large-scale randomized clinical trial.

Conclusions Published online December 5, 2006 (doi:10.1001/jama.297.2.joc60179).

The incidence of early vessel closure after coronary stent implantation was markedly reduced by the adoption of thienopyridine antiplatelet therapy.¹ The widespread adoption of dual antiplatelet therapy (aspirin and thienopyridines) has further reduced the risk of subacute thrombosis after bare-metal stent implantation to 0.5% to 1.9%.^1-3

Instructions for the use of drug-eluting stents commercially available in the United States specify treatment with clopidogrel for at least 3 months (for sirolimus-coated stents) or 6 months (for paclitaxel-coated stents) after implantation. Premature discontinuation of this minimum antiplatelet therapy has been associated with stent thrombosis.^4,5 However, studies of late thrombosis events among patients with a drug-eluting stent have cast doubt on whether the recommended regimens are sufficient.^6,7 An observational analysis from BASKET-LATE (Basel Stent Kosten-Effekivitats Trial-Late Thrombotic Events) examined the incidence of clinical events after cessation of clopidogrel therapy.⁸ This study identified 746 patients who were without major adverse events 6 months after drug-eluting or bare-metal stent placement. All patients had stopped taking clopidogrel and were followed up for an additional 12 months. At 18-month follow-up, there was no difference between patients with a drug-eluting or bare-metal stent in cumulative rates of death or myocardial infarction (MI). However, after clopidogrel discontinuation patients receiving drug-eluting vs bare-metal stents experienced higher rates of death and MI (4.9% vs 1.3%, respectively). These results have created uncertainty regarding the minimal necessary duration of antiplatelet therapy after drug-eluting stent implantation.

We assessed the association between clopidogrel use and long-term rates of death and death or MI following initial percutaneous coronary intervention (PCI) with drug-eluting or bare-metal stents.

The study population included consecutive patients from the Duke Heart Center who had an initial PCI with at least 1 bare-metal stent between January 1, 2000, and July 31, 2005; and with at least 1 drug-eluting stent between April 1, 2003, and July 31, 2005. The follow-up period extended through September 7, 2006, to ensure that all patients had an opportunity for at least 12 months of follow-up information. Exclusion criteria included congenital heart disease, moderate to severe valvular heart disease, prior coronary artery bypass graft surgery or PCI procedure, and significant (≥75% stenosis) left main coronary artery disease. Patients also were excluded if interventions other than stent placement occurred during their PCI procedure, or if they were not contacted for follow-up medication use for each analysis period. All study information was stored in the Duke Databank for Cardiovascular Disease. Duke University Medical Center Institutional Review Board approval was obtained on February 27, 2006, with a waiver of the requirement for written informed consent (Registry No. 8223-06-2R0ER).

Baseline Data. Baseline demographic, medical history, physical examination, and initial cardiac catheterization results information was collected prospectively, as previously described.^9-11 Initial demographic data (including patient race) were received from Duke University Medical Center's administrative systems, clinical data were collected by the Duke Heart Center, and ZIP code data were obtained from the 2000 US Census report. Patient race was collapsed into 2 categories (black [African American] and other race) for this analysis.

Follow-up Clinical Event and Medication Data. As part of the standard Duke Databank for Cardiovascular Disease follow-up protocol, all patients were contacted at 6 and 12 months after their initial procedure (bare-metal or drug-eluting stent implantation), and annually thereafter. We analyzed follow-up information on occurrence of 2 events (death and nonfatal MI) and use of 2 medications (clopidogrel and aspirin). An independent mortality committee reviewed follow-up results to confirm deaths. Follow-up MI was based on clinical diagnoses assigned by the patient's physician and was not centrally adjudicated. 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. Follow-up was 98% complete for all scheduled contacts as of September 7, 2006. Patients with incomplete follow-up were censored at the time of last contact.

During their follow-up contacts at 6, 12, and 24 months, patients were asked to provide information regarding their current medications. We considered patients to be using clopidogrel at the time of follow-up when this medication was specifically listed. Patient aspirin use was determined from the same medication list and from responses to a question asking whether patients were regular aspirin users. No attempt was made to verify patient-reported medication use.

Stent Type. Patients who received both bare-metal and drug-eluting stents during the same procedure were assigned to the drug-eluting stent group, because subsequent antiplatelet therapy requirements would be based on the presence of this device.

Landmark Analyses Based on Clopidogrel Use. Landmark analysis is a form of survival analysis that classifies patients based on some intermediate (nonoutcome) event that occurs during follow-up.¹² Prognosis is then evaluated from this landmark time point. In our analyses, we define landmark time and study outcomes in terms of their elapsed time from a patient's index procedure. Two landmarks were used in this study: 6-month clopidogrel use (yes or no) and 12-month clopidogrel use (yes or no) (Figure 1). Patients who were event-free (no death, MI, or revascularization) at 6 months and completed the 6-month follow-up contact, including questions regarding medication use, were assigned to 1 of 4 groups: drug-eluting stent with clopidogrel, drug-eluting stent without clopidogrel, bare-metal stent with clopidogrel, and bare-metal stent without clopidogrel. Outcomes for these groups were evaluated up to 24 months after the initial PCI procedure. Similarly, patients who were event-free at 12 months and completed the 12-month follow-up contact, including medication use, were assigned to a second landmark analysis of 4 groups (by stent type and clopidogrel use), and their 24-month outcomes were evaluated. When classifying groups, a window of 90 days before and after the follow-up points was allowed because of potential time lags in the follow-up process. For the 12-month landmark analysis, patients with PCI procedures occurring after July 31, 2004, were excluded because they did not have the opportunity for follow-up at 24 months.

Baseline characteristics and event rates were summarized for patient groups as number (percentage) for categorical variables and as median (interquartile range) for continuous variables. Tables of baseline and angiographic characteristics and follow-up aspirin and clopidogrel use were categorized by treatment modality. Binary variables were compared across interventions using the Pearson χ² test. Continuous and ordinal categorical variables were compared using the Wilcoxon rank sum test. Statistical significance was determined at the 2-sided α=.05 level (P≤.05).

Unadjusted and adjusted cumulative incidence rates were calculated using inverse probability weighted estimators.^13-15 The inverse weighted estimators were based on partitioning the data into monthly intervals.¹³ Unadjusted estimates were based on weights that are a function of Kaplan-Meier estimates for the treatment-specific censoring distributions. Inverse probability weighted adjusted estimates were based on estimated propensity scores and Cox proportional hazards regression model estimates of the treatment-drug group specific censoring distributions.¹⁶ SAS version 8.2 (SAS Institute Inc, Cary, NC) with robust SEs was used to estimate treatment effects, 95% confidence intervals (CIs), and P values.¹⁷ Weighted Cox proportional hazards regression models and adjusted cumulative incidence curves were constructed using inverse probability weights.¹⁸

Four treatment-drug group propensity scores were estimated using logistic regression models. The following variables were used in our propensity score and Cox proportional hazard regression models: patient demographics (race, age, sex), coronary artery disease risk factors (smoking history, hypertension, diabetes mellitus), cardiovascular history and physical examination (body mass index [calculated as weight in kilograms divided by height in meters squared], systolic blood pressure, carotid bruits, heart rate, history and severity of congestive heart failure, history of MI, mild valvular heart disease, third heart sound, history of cerebrovascular disease, history of peripheral vascular disease), diagnostic catheterization findings (left ventricular ejection fraction, extent of coronary artery disease), comorbid conditions (Charlson Index, history of chronic obstructive pulmonary disease, connective tissue disease, renal disease, liver disease, metastatic cancer, solid tumor), stent characteristics (average stent diameter and total length of stents), socioeconomic status (ZIP code level median income per household and average house value), and patient-reported aspirin use.

Between January 1, 2000, and July 31, 2005, 4927 patients received an initial PCI procedure at the Duke Heart Center. We excluded 261 patients, 156 with balloon angioplasty without a stent device and 105 receiving a nonstent device (eg, atherectomy, excimer laser, brachytherapy). Of the 4666 patients remaining, 3165 received a bare-metal stent and 1501 received a drug-eluting stent.

Baseline Characteristics. Our population included 3609 patients who were without major adverse events 6 months after their initial stent procedure. All 4 groups were similar with regard to age, race, and sex (Table 1). However, fewer patients in the bare-metal stent without clopidogrel group had a history of diabetes, while more patients in the drug-eluting stent without clopidogrel group had a history of congestive heart failure. Although both drug-eluting stent groups had fewer patients with a history of MI than the bare-metal stent groups, they also had more patients with multivessel disease, and their patients resided in ZIP codes with greater household incomes and house values. There was also significant variation across the groups in self-reported regular aspirin use. By 24 months, clopidogrel use had diminished among patients who reported using it at 6 months (Table 1). In the same period, clopidogrel use increased among patients who did not report taking it at 6 months. Thus, by 24-month follow-up, there was a 40.7% difference in clopidogrel use between the drug-eluting stent with clopidogrel vs drug-eluting stent without clopidogrel groups; and a 54.1% difference between the bare-metal stent with clopidogrel vs bare-metal stent without clopidogrel groups.

Unadjusted Results. Among patients who were without major adverse events at 6 months, unadjusted 2-year differences between the 4 groups revealed disparities in event rates. At 24 months, the drug-eluting stent with clopidogrel group vs the drug-eluting stent without clopidogrel group had significantly lower rates of death, nonfatal MI, and death or MI (Table 2). However, there were no statistically significant differences between patients in the bare-metal stent with clopidogrel and bare-metal stent without clopidogrel groups with regard to these events. Patients in the drug-eluting stent with clopidogrel group had significantly lower rates of death and death or MI than did patients in the bare-metal stent with clopidogrel or bare-metal stent without clopidogrel groups, but no difference was observed in nonfatal MI.

Adjusted Results. Patients in the drug-eluting stent with clopidogrel group had significantly lower rates of death and death or MI than did patients in the drug-eluting stent without clopidogrel group, but no difference was observed in nonfatal MI (Table 2 and Figure 2). In the weighted Cox proportional hazard regression model, the adjusted hazard ratio (HR) for death in the drug-eluting stent without clopidogrel group compared with the drug-eluting stent with clopidogrel group was 2.43 (95% CI, 1.12-5.26; P = .03). The adjusted HR for death or MI in the drug-eluting stent without clopidogrel group compared with the drug-eluting stent with clopidogrel group was 1.93 (95% CI, 1.05-3.56; P = .04). There were no differences between the bare-metal stent with clopidogrel and bare-metal stent without clopidogrel groups with regard to death, nonfatal MI, and death or MI. In this analysis, differences between the drug-eluting stent with clopidogrel and bare-metal stent with clopidogrel groups were not statistically significant; however, differences between the drug-eluting stent with clopidogrel and bare-metal stent without clopidogrel groups were statistically significant for death and death or MI.

Baseline Characteristics. Our population included 2518 patients who were without major adverse events at 12 months. Patients in all 4 groups were similar with regard to age, race, sex, history of congestive heart failure, and socioeconomic status (Table 3); however, fewer patients in the bare-metal stent without clopidogrel group had a history of diabetes. Both drug-eluting stent groups had fewer patients with a history of MI than the bare-metal stent groups, and the drug-eluting stent with clopidogrel group had the highest percentage of patients with multivessel disease. Most patients in all 4 groups were receiving aspirin at 6, 12, and 24 months; however, there appeared to be some crossover in patient-reported clopidogrel use between 6 and 12 months. By the 24-month follow-up, there was a 61.7% difference in clopidogrel use between the drug-eluting stent with clopidogrel group vs the drug-eluting stent without clopidogrel group, and a 64.2% difference between the bare-metal stent with clopidogrel group and the bare-metal stent without clopidogrel group.

Unadjusted Results. Patients in the drug-eluting stent with clopidogrel group vs the drug-eluting stent without clopidogrel group had significantly lower rates of death, nonfatal MI, and death or MI; whereas, there were no significant differences in the bare-metal stent with clopidogrel vs bare-metal stent without clopidogrel groups for these events (Table 4). Compared with patients in the bare-metal stent group (bare-metal stent with and without clopidogrel), patients in the drug-eluting stent with clopidogrel group had significant reductions in the outcomes of death, nonfatal MI, and death or MI.

Adjusted Results. Patients in the drug-eluting stent with clopidogrel group vs without clopidogrel group had lower rates of death, nonfatal MI, and death or MI (Table 4 and Figure 3). Hazard ratios were not reported as there were no events for patients in the drug-eluting stent with clopidogrel group. Again, there were no differences between patients in the bare-metal stent with clopidogrel group vs the bare-metal stent without clopidogrel group for these events. With regard to drug-eluting stent with clopidogrel vs bare-metal stent with clopidogrel, the drug-eluting stent with clopidogrel group had significantly lower rates of death and death or MI, but no statistically significant difference was observed in nonfatal MI. Patients receiving a drug-eluting stent with clopidogrel vs a bare-metal stent without clopidogrel had significantly lower rates of death, nonfatal MI, and death or MI.

To assess whether aspirin use confounded the clopidogrel results, we analyzed the subset of patients who reported aspirin use at 6 months (Figure 4). The adjusted cumulative mortality and death or MI data in the aspirin cohort mimicked the results for adjusted outcomes in the 6- and 12-month landmark analyses, except that the event rates were lower. Compared with patients in the drug-eluting stent without clopidogrel group, patients in the drug-eluting stent with clopidogrel group tended to have lower mortality and lower rates of death or MI. Thus, these results reinforce the results observed in our overall analyses.

Our observational results suggest that patients who received long-term clopidogrel therapy following a PCI using at least 1 drug-eluting stent had a significantly improved prognosis compared with similar patients not receiving this therapy. Current US Food and Drug Administration–approved indications for clopidogrel use following drug-eluting stent implantation call for 3 to 6 months of therapy,^19-21 depending on the specific device used. Such regimens were shown to be safe and effective in the clinical trials for the Cypher and TAXUS stents when judged by 1 year outcomes.^22-24 Recently, several lines of evidence, including unpublished long-term follow-up from these same trials,²⁵ have suggested that drug-eluting stent use is associated with a late increased risk of catastrophic stent thrombosis at a rate significantly higher than with bare-metal stents. These data have led to speculation that drug-eluting stents may require protracted and possibly indefinite clopidogrel therapy. There are, however, no clinical trials to our knowledge that currently address the effectiveness of such a strategy or its required duration.

Therefore, our study provides new evidence that continued clopidogrel therapy conveys an important prognostic benefit after drug-eluting stent implantation. The absence of a similar benefit for patients with a bare-metal stent and taking clopidogrel provides important reassurance that the differences observed in our study were not simply the effect of clopidogrel regimens extended in response to some unmeasured prognostic factors unrelated to type of stent. With 600 000 US patients hospitalized and receiving stent devices each year,²⁶ the need for definitive evidence on this issue has major public health importance.

Our study results foster interesting hypotheses for future investigations. By simultaneously comparing patients in 4 treatment groups defined by stent type and clopidogrel use, we found that patients with a drug-eluting stent receiving clopidogrel 6 and 12 months after their initial procedure have significantly lower rates of death and death or MI compared with patients with a drug-eluting stent not receiving this medication. These results complement those from the PREMIER Registry⁵ and together with the Clopidogrel for the Reduction of Events During Observation (CREDO) trial²⁷ suggest that all patients with drug-eluting stents should continue to take clopidogrel for at least 12 months after PCI, and possibly indefinitely, while the bare-metal stent may be a more appropriate stent choice for patients unable to take clopidogrel for an extended length of time. These possible benefits of clopidogrel appear to be maintained for at least 24 months; however, further research is required to determine the optimal duration of clopidogrel use in patients who are more clinically and angiographically complex than were enrolled in the pivotal drug-eluting stent clinical trials.

There are 2 important caveats to our analysis. First, clopidogrel use was not randomly assigned. Thus, the decision to continue the drug beyond the periods recommended by the relevant clinical trials may have been correlated with unmeasured prognostic factors. However, for such confounding to create the appearance of better prognosis for the patients in the drug-eluting stent with clopidogrel group, the bias in treatment selection would have to be toward use in lower-risk patients, which is counterintuitive. Obvious biases, such as treatment of younger patients with less severe coronary artery disease or less major comorbidity were controlled for in this analysis.

The second important caveat is that clopidogrel use in our analysis was identified by patient report at 2 discrete points (6- and 12-month follow-up). Therefore, these data are subject to recall bias. Furthermore, the indications and rationale for long-term clopidogrel regimens and for its discontinuation were not collected. Because follow-up contacts did not necessarily occur at exactly 6 and 12 months, we used a 90-day window around the anniversary date to determine follow-up contact and medication use. Narrowing this window to 30 days produced outcomes with similar results.

The 24-month event rates of 0% for death, nonfatal MI, and death or MI for the drug-eluting stent with clopidogrel group in the 12-month landmark analysis underestimate the true event rates. However, 14 patients receiving drug-eluting stents died or had a nonfatal MI, and none of these 14 patients was in the drug-eluting stent with clopidogrel group. We believe that our results along with those from BASKET-LATE serve to identify key parameters for subsequent research in this area.²⁸ Extended clopidogrel therapy has its own risks and our analysis does not evaluate the long-term nonfatal implications of its use. Other studies are required to assess relationships between long-term clopidogrel use and the risk for major bleeding events, the role for devices used to facilitate stent deployment in patients with drug-eluting stents receiving long-term clopidogrel therapy, as well as relationships between target-vessel and nontarget-vessel stenosis in patients with drug-eluting and bare-metal stents. If the use of drug-eluting stents is, “committing millions of patients to lifelong potent antithrombotic therapy,”²⁸ our society must consider how this therapy will be delivered to patients without adequate financial resources.

In a large consecutive cohort of contemporary patients receiving PCI, the long-term risk for death and major cardiac events was significantly increased among patients in the drug-eluting stent group who had discontinued clopidogrel therapy at 6 or 12 months. Extended-duration clopidogrel therapy following drug-eluting stent implantation was associated with a lower incidence of death or MI, a finding that has immediate implications for clinical practice. We propose a 3-group clinical trial to further investigate these results. Patients in 2 groups would be randomized to discontinue clopidogrel therapy at 12 and 24 months after drug-eluting stent implantation; whereas, patients in the third group would continue clopidogrel through 3 years of follow-up. We estimate that a sample size of approximately 10 000 patients would be required to detect a 25% reduction in death or MI at 3 years.

Corresponding Author: Eric L. Eisenstein, DBA, Duke Clinical Research Institute, Duke University Medical Center, Box 3865, 2400 Pratt St, Room 0311, Durham, NC 27710 (eric.eisenstein@duke.edu).

Published Online: December 5, 2006 (doi:10.1001/jama.297.2.joc60179).

Author Contributions: Dr Eisenstein had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Eisenstein and Anstrom contributed equally to this work and share the lead authorship of this article.

Study concept and design: Eisenstein, Anstrom, Kong, Shaw, Califf.

Acquisition of data: Anstrom, Shaw, Califf.

Analysis and interpretation of data: Eisenstein, Anstrom, Kong, Shaw, Tuttle, Mark, Kramer, Harrington, Matchar, Kandzari, Peterson, Schulman, Califf.

Drafting of the manuscript: Eisenstein, Anstrom, Kong, Shaw, Mark, Peterson.

Critical revision of the manuscript for important intellectual content: Eisenstein, Anstrom, Kong, Shaw, Tuttle, Kramer, Harrington, Matchar, Kandzari, Peterson, Schulman, Califf.

Statistical analysis: Eisenstein, Anstrom, Shaw, Tuttle.

Obtained funding: Kong, Matchar, Schulman, Califf.

Administrative, technical, or material support: Eisenstein, Kong, Tuttle, Kramer, Harrington, Matchar, Peterson, Califf.

Study supervision: Eisenstein, Kong, Mark, Matchar, Kandzari.

Funding/Support: This study was funded under contract 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.

Role of the Sponsors: The sponsors did not participate in the design or conduct of this study, in the collection, management, analysis, or interpretation of data, in the writing of the manuscript, or in the preparation, review, approval, or decision to submit this manuscript for publication.

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

Acknowledgment: We thank the faculty and staff of the cardiac catheterization laboratories at Duke Hospital for collecting the data. We thank Judith A. Stafford, MS, Duke Clinical Research Institute, for her programming in support of this project; Charles B. McCants, Jr, BS, Duke Clinical Research Institute, for updating the Duke Databank for Cardiovascular Disease follow-up files so that the results from these analyses would be current; and Maqui Ortiz, Duke Clinical Research Institute, for the editing and review of these materials and for her administrative support. All were compensated by the Duke Clinical Research Institute for their work. We also thank Elise Berliner, PhD, Agency for Healthcare Research and Quality, for her consistent work as an advocate for and supporter of this project, and her invaluable feedback during the course of the Treatment of In-Stent Restenosis project. Ms Berliner was compensated by the AHRQ.