|
|
| |||||||||||||||||||||||||||||||||||||||||
Background The placement of drug-eluting stents decreases the frequency of repeat revascularization procedures in patients undergoing percutaneous coronary intervention (PCI) in randomized clinical trials. However, there is uncertainty about the effectiveness of drug-eluting stents, and increasing concern about their safety, in routine clinical practice.
Methods From the Cardiac Care Network of Ontario's population-based clinical registry of all patients undergoing PCI in Ontario, Canada, we identified a well-balanced cohort of 3751 pairs of patients, matched on the basis of propensity score, who received either bare-metal stents alone or drug-eluting stents alone during an index PCI procedure between December 1, 2003, and March 31, 2005. The primary outcomes of the study were the rates of target-vessel revascularization, myocardial infarction, and death.
Results The 2-year rate of target-vessel revascularization was significantly lower among patients who received drug-eluting stents than among those who received bare-metal stents (7.4% vs. 10.7%, P<0.001). Drug-eluting stents were associated with significant reductions in the rate of target-vessel revascularization among patients with two or three risk factors for restenosis (i.e., presence of diabetes, small vessels [<3 mm in diameter], and long lesions [
Conclusions Drug-eluting stents are effective in reducing the need for target-vessel revascularization in patients at highest risk for restenosis, without a significantly increased rate of death or myocardial infarction.
20 mm]) but not among lower-risk patients. The 3-year mortality rate was significantly higher in the bare-metal–stent group than in the drug-eluting–stent group (7.8% vs. 5.5%, P<0.001), whereas the 2-year rate of myocardial infarction was similar in the two groups (5.2% and 5.7%, respectively; P=0.95).
In this study, we examined a database, initiated in 2003 in Ontario, Canada, in which outcome data from all patients undergoing a PCI and receiving a coronary stent are recorded. We compared the outcomes for the patients who received a drug-eluting stent and those for patients who received a bare-metal stent during the first 16 months of a new provincial program to monitor the introduction and use of drug-eluting stents in Ontario.15,16 During this period, Ontario had a policy of providing 1 year of clopidogrel therapy after PCI for patients receiving either type of stent who were 65 years of age or older and who were enrolled in the universal Ontario Drug Benefits plan.
We studied the effectiveness and safety of both types of stents, using a propensity-score–matched cohort of patients who received either a drug-eluting stent alone or a bare-metal stent alone.17 We also studied the use of three well-established risk factors for restenosis (presence of diabetes, small vessels, and long lesions) to predict which patients have the greatest benefit from implantation of a drug-eluting stent.18,19
Methods
Data Sources
The Cardiac Care Network (CCN) of Ontario maintains a prospective clinical registry of all patients undergoing invasive cardiac procedures in Ontario that are funded by the government. Clinical nurse research coordinators gather information on each patient from cardiac-procedure referral forms completed by the referring physician and from charts of patients. The registry includes information on demographic characteristics of patients and their cardiac history, cardiac procedures, and relevant coexisting conditions. All 12 PCI centers in Ontario were required to collect additional data on patients receiving a bare-metal stent or a drug-eluting stent in this registry, as a condition of receiving incremental funding for drug-eluting stents.
For the current study, we used the detailed clinical and stent information captured in the CCN PCI registry and supplemented it by linking that information to other population-based administrative databases in Ontario, using unique encrypted identifiers of patients. We identified an index cohort of all 18,314 patients who underwent a PCI with placement of either a single drug-eluting or bare-metal stent, or multiple stents (all the same type), in Ontario between December 1, 2003, and March 31, 2005. A total of 4961 patients were excluded from this cohort. Patients who had an invalid Ontario health card number were excluded because we could not ascertain their long-term outcomes. Patients who had missing information on potentially important prognostic factors such as socioeconomic status, type or location of lesion, or Canadian Cardiovascular Society angina classification were excluded because of our inability to use their propensity scores for matching. We also excluded from our index cohort any patient who had undergone a previous PCI in the past year. Patients undergoing PCI for left main coronary artery disease were excluded because of concern that there would not be a sufficient number of patients receiving bare-metal stents to allow for propensity-score matching. If a patient who had an index PCI had a second PCI within 1 year after the index (first) PCI, the second PCI was not included in the creation of the index PCI cohort.
We identified the presence or absence of various coexisting conditions using the CCN PCI database and supplemented this information with data from the 24 secondary-diagnosis fields in the hospital-discharge administrative database of the Canadian Institute for Health Information.20 The socioeconomic status of patients was inferred from Statistics Canada census data on the median household income for each geographic area.21 Study patients were grouped into five quintiles, according to the income level of their neighborhoods, with similar numbers of patients in the five quintiles.
Clinical outcomes were identified through linkages to clinical and administrative databases, with the use of the most recent follow-up data that were available; a minimum of 1 year of follow-up data was required for inclusion in the study. Target-vessel revascularization was defined as a follow-up PCI performed in the same vessel as the index PCI, a follow-up PCI without implantation of a stent, or coronary-artery bypass grafting, as recorded in the CCN database through March 31, 2006. We collected data on admissions for myocardial infarction (codes I21 and I22 in the International Classification of Diseases, 10th Revision), before and after the index PCI, through March 31, 2006, using linkages to the Canadian Institute for Health Information database.22 Deaths from any cause were identified through November 30, 2006, by means of a linkage to the Ontario Registered Persons Database. The median duration of follow-up for death was similar in the drug-eluting–stent group (840 days) and the bare-metal–stent group (817 days).
Our study was approved by the Research Ethics Board at Sunnybrook Health Sciences Centre. The requirement for written informed consent from patients was waived, since participation in the CCN database is mandatory under Ontario's legislation regarding the privacy of health information.
Statistical Analysis
Before performing propensity-score matching, we conducted a univariate analysis of all potentially available explanatory factors for target-vessel revascularization, myocardial infarction, or death. A total of 21 factors were identified that were significantly associated with one or more of these three primary outcomes (P<0.05). We compared the prevalences of these risk factors in the bare-metal–stent group and the drug-eluting–stent group before the propensity-score matching, using a t-test for means and chi-square tests for percentages.
A propensity score for the predicted probability of receipt of a drug-eluting stent in each patient was estimated with the use of a logistic-regression model fit with the 21 factors. We created a propensity-score–matched cohort by attempting to match each patient who received a drug-eluting stent with one who received a bare-metal stent (a 1:1 match). A nearest-neighbor–matching algorithm with a "greedy" heuristic (one that always implements the best immediate, or local, solution) was used to match patients on the basis of their diabetes status (presence or absence of diabetes) and the logit of their propensity score, with matching occurring if the difference in the logits of the propensity scores was less than 0.2 times the standard deviation of the scores (the caliper width).23 In greedy nearest-neighbor matching, a patient with a drug-eluting stent was randomly selected and matching was attempted with the "nearest" patient with a bare-metal stent. This process was repeated until matches had been attempted for all patients with drug-eluting stents. Each matched pair was unique, and data for unmatched patients in either group were not used in subsequent analyses.
We also assessed the degree of balance in measured covariates between the drug-eluting–stent group and the bare-metal–stent group. The distributions of categorical variables and continuous variables between the two groups in the matched cohort were compared with the use of McNemar's test and paired t-tests, respectively. We computed the standardized difference between the two groups for each type of distribution, with differences of less than 0.1 taken to indicate good balance in the matched cohort.17 Kaplan–Meier survival curves were estimated in the matched cohort for patients who received a drug-eluting stent and those who received a bare-metal stent. The survival curves were compared according to methods appropriate for matched data.24
In prespecified subgroup analyses of target-vessel revascularization, patients in the drug-eluting–stent group and those in the bare-metal–stent group were matched on the basis of lesion length (long [20 mm] vs. short [<20 mm]), vessel diameter (small [<3 mm] vs. large [3 mm]), diabetes status (presence vs. absence of diabetes), and the logit of the propensity score. The total stent length was used as a proxy for lesion length, and the narrowest diameter of the inserted stents was used as a proxy for vessel diameter. Within each of the eight subgroups, the reduction in the risk of the outcome was compared between the drug-eluting–stent group and the bare-metal–stent group with the use of a Cox regression model, with stent type as the sole predictor variable. Robust standard errors that accounted for the clustering of pairs in the matched cohort were obtained. The number of patients who would need to be treated to prevent one target-vessel revascularization with the use of a drug-eluting stent was also calculated for each of the eight subgroups.25
All reported P values are two-sided, and P values of less than 0.05 were considered to indicate statistical significance. SAS software, version 9.1 (SAS Institute), and the R programming language were used for statistical analyses.
Results
Characteristics of Patients
Before Propensity-Score Matching
A total of 13,353 patients met the study inclusion criteria and met none of the exclusion criteria. The characteristics of patients chosen to receive a drug-eluting stent alone as compared with a bare-metal stent alone, according to Ontario's selective funding policy for drug-eluting stents, are shown in Table 1. Overall, 38% of the stents implanted in patients in Ontario during the study period were drug-eluting stents, with the percentage of patients receiving drug-eluting stents remaining roughly constant (see the Supplementary Appendix, available with the full text of this article at www.nejm.org). Patients chosen by their cardiologists to receive a bare-metal stent were more likely than those chosen to receive a drug-eluting stent to have an admission for acute myocardial infarction on the same day their PCI and stent implantation were performed (i.e., were more likely to require primary PCI or rescue PCI). The mean stent diameter was smaller, and the mean stent length was longer, in the drug-eluting–stent group than in the bare-metal–stent group, showing that cardiologists did selectively implant drug-eluting stents in patients at increased risk for restenosis.
|
After propensity-score matching was completed, there were 3751 matched pairs of patients (Table 2). There were no significant differences between the drug-eluting–stent group and the bare-metal–stent group in any of the 21 characteristics (P>0.20 for all comparisons). The standardized difference between the two groups was less than 0.03 for all variables. In the matched cohort, 82.9% of patients in the drug-eluting–stent group received a paclitaxel stent, and 17.1% received a sirolimus stent.
|
Overall, 10.7% of the patients receiving a bare-metal stent and 7.4% of those receiving a drug-eluting stent required target-vessel revascularization by year 2 of follow-up (P<0.001) (Figure 1A and Table 3). The maximal reduction in the need for repeat revascularization from the use of drug-eluting stents was seen at 6 months. The effectiveness of drug-eluting stents in reducing the need for repeat target-vessel revascularization varied widely, with the four subgroups of patients with two or three risk factors for restenosis (i.e., presence of diabetes, small vessels, and long lesions) showing a significant benefit (Table 4). None of the other four subgroups had significant reductions in the rates of target-vessel revascularization. The number needed to treat to prevent one target-vessel revascularization with the use of drug-eluting stents ranged from 10 to 167 among the eight subgroups.
|
|
|
The rate of myocardial infarction after PCI during the 2-year follow-up period did not differ significantly between the two groups (P=0.95) (Figure 1B and Table 3). However, the rate was 0.4 percentage point lower in the drug-eluting–stent group than in the bare-metal–stent group during the first 6 months of follow-up, whereas the rates of myocardial infarction crossed over at approximately 15.6 months of follow-up; the rate was 0.5 percentage point higher in the drug-eluting–stent group by year 2 of follow-up.
The rate of death after PCI in the matched cohort at year 3 of follow-up was 7.8% among patients receiving a bare-metal stent and 5.5% among those receiving a drug-eluting stent (P<0.001) (Table 3), with the survival curves gradually diverging over time in favor of drug-eluting stents (Figure 1C). The composite outcome of death or myocardial infarction also favored drug-eluting stents (Figure 1D). To investigate the potential contributions of target-vessel revascularization and myocardial infarction to the absolute differences in the survival rate, we compared the 30-day mortality among patients who had each outcome. The 30-day mortality among patients who underwent target-vessel revascularization was 2.5% in the bare-metal–stent group and 2.4% in the drug-eluting–stent group. In contrast, the 30-day mortality among patients who had a myocardial infarction was 6.9% in both groups.
Discussion
Our study provides information from a large population-based database that could be used to help clinicians determine whether a bare-metal stent or drug-eluting stent is better for a particular patient. We found that drug-eluting stents were most effective in reducing the need for target-vessel revascularization in patients at the highest risk for restenosis (i.e., those who had two or three risk factors — presence of diabetes, small vessels, and long lesions). In contrast, we found small reductions, which were not significant, in the rates of target-vessel revascularization among patients at low or intermediate risk for restenosis (e.g., those without diabetes and with large vessels or short lesions). The relatively low rate of clinically driven target-vessel revascularization in most of the subgroups of patients with bare-metal stents in our study is in contrast to the rate of 15 to 20% or higher, in association with protocol-driven angiography and other factors, reported in clinical trials of drug-eluting stents.1,3,4
Although the benefit of drug-eluting stents in reducing the need for target-vessel revascularization in the real world has been confirmed in a few other, recent registry reports,8,26 there has been great concern expressed recently about whether this benefit is outweighed by the risk of late stent thrombosis.27,28,29,30 Our results suggest that the risk–benefit equation for overall mortality favors drug-eluting stents, since the overall rate of death at year 3 of follow-up was lower in the drug-eluting–stent group than in the bare-metal–stent group. A mortality benefit of drug-eluting stents has not been found in clinical trials conducted to date, but this could reflect the relatively small numbers of patients, limited long-term follow-up, or a lower risk in the populations in many of these studies.1,2,3,4
Some have argued that the risk of restenosis, which may be manifested as angina or myocardial infarction, may outweigh the slightly increased rate of late stent thrombosis found in recent studies.31 We report a 30-day mortality of 2.5% among patients who received a bare-metal stent and underwent repeat procedures, showing that the need for target-vessel revascularization is a clinically significant adverse event. However, the increased incidence of repeat procedures among our patients who received a bare-metal stent is insufficient to explain fully the small absolute difference in the survival rate in favor of drug-eluting stents. Alternatively, our mortality data could reflect unmeasured confounding factors, and we suggest caution in concluding, on the basis of our study alone, that drug-eluting stents save lives.
Because we collected follow-up data from linkages to the administrative databases, it was not possible for us to ascertain the exact rate of late stent thrombosis. However, we found that the rate of myocardial infarction was slightly lower in the drug-eluting–stent group than in the bare-metal–stent group during the first 6 months of follow-up, whereas this benefit was lost during the second year, with a slightly higher rate of myocardial infarction in the drug-eluting–stent group at 2 years. This crossover of rates of myocardial infarction in year 2 has also been reported for patients in the Swedish PCI stent registry and is consistent with findings from other studies that have shown an increased frequency of late thrombotic events, such as nonfatal myocardial infarction, among patients receiving a drug-eluting stent within a few months after clopidogrel therapy was stopped.8,12,14 If confirmed in additional studies, these results would suggest that dual antiplatelet therapy with aspirin and clopidogrel may need to be continued beyond the 1-year period currently recommended in American and European guidelines.10,32,33
Our study has many strengths but also important limitations. The strengths include the availability of a large population-based cohort of patients who underwent PCI, which permitted a comparison of outcomes among patients receiving drug-eluting stents and concurrent controls receiving bare-metal stents with the use of a rigorous cohort analysis involving propensity-score matching. The opportunity to create our data set using multiple linked clinical and administrative databases of Canada's single-payer universal health care system allowed us to achieve virtually 100% follow-up of outcome rates for both groups of patients. Our study was entirely funded from public sources, without any involvement from industry.
Important limitations of our study include its observational nature and the fact that our findings may not be generalizable to countries without universal health insurance and drug-benefit plans. The benefits of drug-eluting stents are in part related to the prolonged use of dual antiplatelet therapy, which was available for 1 year at a minimal cost to most patients in our study. Although we were able to establish a large and well-balanced cohort of patients receiving bare-metal stents and patients receiving drug-eluting stents, matched on the basis of the propensity score, our study was not a randomized clinical trial, and there may still be unmeasured confounding factors that contribute to our findings. We did not have access to certain data (e.g., subtypes of myocardial infarction) that would have been of interest.
In conclusion, the experience in Ontario, Canada, is that drug-eluting stents are more effective than bare-metal stents in reducing the need for target-vessel revascularization in patients at highest risk for restenosis after PCI but offer minimal benefit to patients at lower risk. The small absolute difference in mortality in favor of drug-eluting stents in our study warrants further investigation and should be confirmed or refuted through large, randomized clinical trials with long-term follow-up.
Supported by operating grants from the Ontario Ministry of Health and Long-Term Care to the Program for Assessment of Technology in Health, the CCN of Ontario, and the Institute for Clinical Evaluative Sciences; a Canadian Institutes of Health Research Team Grant in Cardiovascular Outcomes Research; a Canada Research Chair in Health Services Research and a Career Investigator Award from the Heart and Stroke Foundation of Ontario (to Dr. Tu); a Clinician–Scientist Award from the Heart and Stroke Foundation of Ontario (to Dr. Ko); and a New Investigator Award from the Canadian Institutes of Health Research (to Dr. Austin).
Dr. Cohen reports receiving operating grants for clinical trials and speaker's fees from Boston Scientific and Cordis–Johnson & Johnson. No other potential conflict of interest relevant to this article was reported.
We thank the staff members of the CCN of Ontario (Dr. Kevin Glasgow, Amanda Dean, Joyce Seto, Saba Ateyah, and Dr. Kevin Teoh), its member hospitals, referring physicians, interventional cardiologists, and regional cardiac care coordinators for collecting the clinical data that made this study possible; Francine Duquette for administrative assistance with the manuscript; and Linda Donovan and Pam Slaughter for helpful comments on earlier versions of the manuscript.
Source Information
From the Institute for Clinical Evaluative Sciences (J.V.T., M.C., D.T.K., P.C.A., Y.H.) and Sunnybrook Health Sciences Centre, University of Toronto (J.V.T., D.T.K., E.A.C.) — both in Toronto; McMaster University, St. Joseph's Healthcare, Hamilton, Ont. (J.B., R.H., J.-E.T., G.B., R.G.); and Trillium Health Sciences Centre, Mississauga, Ont. (C.L.) — all in Canada.
Address reprint requests to Dr. Tu at the Institute for Clinical Evaluative Sciences, G106-2075 Bayview Ave., Toronto, ON M4N 3M5, Canada, or at tu{at}ices.on.ca.
References
| |||||||||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
