|Home | About | Journals | Submit | Contact Us | Français|
Drug-eluting stents (DES) reduce the need for repeat revascularization, but their long term safety relative to bare metal stents (BMS) in general use remains uncertain. We sought to compare the clinical outcome of patients treated with DES vs. BMS.
All adults undergoing percutaneous coronary intervention (PCI) with stenting between April 1, 2003 and September 30, 2004 at non-US governmental hospitals in Massachusetts were identified from a mandatory state database. Patients were classified according to stent types used from the index admission. Clinical and procedural risk factors were collected prospectively. Risk-adjusted mortality, myocardial infarction, and revascularization rate differences (DES-BMS) were estimated through propensity score matching without replacement. 11556 patients were treated with DES and 6237 treated with BMS with unadjusted 2 year mortality of 7.0% and 12.6% respectively (p<0.0001). In 5549 DES patients matched to 5549 BMS patients, 2 year risk-adjusted mortality rates were 9.8% and 12.0% (p=0.0002), myocardial infarction, 8.3% vs. 10.3% (p=0.0005), and target vessel revascularization, 11.0% vs. 16.8% (p<0.0001).
DES treatment was associated with lower mortality, myocardial infarction, and target vessel revascularization compared with BMS treatment in similar patients in a matched population based study. Comprehensive follow-up in this inclusive population is warranted to identify if similar safety and efficacy remain beyond 2 years.
The adoption of drug-eluting stents (DES) for percutaneous coronary intervention (PCI) has been based on randomized trials demonstrating significantly lower rates of recurrent stenosis compared with uncoated or “bare metal” stents 1, 2. Pooled data from randomized trials have shown similar rates of myocardial infarction (MI) and mortality between the treatments in extended follow-up 3, but some observational studies have raised the concern that beyond the first year of treatment, drug-eluting stenting may be associated with higher rates of thrombosis 4 and mortality 5. Randomized studies designed to evaluate DES have had limited power to detect small differences in mortality 6, and have also been limited to a less complex patient population than those currently treated in the general population 7, 8. In contrast, observational studies have been larger and more inclusive, but may be limited by completeness of follow-up and control of selection bias.
We have recently described lower rates of recurrent MI and mortality associated with DES in the setting of stenting for myocardial infarction in a regional population-based study9. We sought to evaluate the complete patient population treated in Massachusetts with PCI to determine whether stent choice was associated with MI and mortality overall. Because stent choice was not randomized, we used a propensity score matched analysis in order to minimize bias.
All adults (18 years of age or older) undergoing PCI with stenting for any indication between April 1, 2003 and September 30, 2004 at all acute care non-US governmental hospitals in Massachusetts were included. Because data are used to assess the quality of all cardiac surgeries and coronary procedures as required by the Massachusetts Department of Public Health, no patient consent is required. Study procedures, study investigators, and data security procedures are reviewed annually by the Harvard Medical School Internal Review Board. Subjects were excluded if they were not Massachusetts residents at time of the procedure or if they could not be linked to hospital discharge billing data (see below).
We used clinical data submitted to a data coordinating center (Mass-DAC) located in Harvard Medical School. Data are prospectively collected by trained hospital personnel using the American College of Cardiology's National Cardiovascular Data Registry's (NCDR) instrument (http://www.accncdr.com/WebNCDR/Common) for PCI and using the Society of Thoracic Surgeons National Adult Cardiac Database instrument (www.sts.org/sections/stsnationaldatabase) for cardiac surgery with detailed supplemental patient and operator identifying information collected for quality assessment 10.
We linked data from the Massachusetts Registry of Vital Records and Statistics, the Social Security Death Index Interactive Search website (http://ssdi.rootsweb.com/cgi-bin/ssdi.cgi), and hospital discharge billing data collected by the Massachusetts Division of Health Care Finance and Policy to the Mass-DAC registry. 9
According to the stent types used from the index admission, defined as the first admission during our observational period, patients were assigned to one of three groups: (1) DES if all stents utilized during the index admission were drug-eluting; (2) BMS if all stents were bare metal; and (3) “both” if both types of stents were used.
We defined death from any cause within 2 years of the index procedure as the primary outcome. The Massachusetts Registry of Vital Records and Statistics included two-year mortality data for all study participants at the time of analysis. We used the Social Security Death Index Interactive Search to confirm death dates and mortality status.
Myocardial infarction was determined from the mandatory recording of in-hospital events in the Mass-DAC database (admission or post-procedure), or from the occurrence of any subsequent hospital admission with a principle diagnosis ICD-9 code of 410.×1 in the hospital discharge billing data. Repeat revascularization was defined as any CABG or PCI following the index procedure.
We first developed a list of confounders that included patient demographic factors, insurance status, history and risk factors, cardiac status of the patient, catheterization lab medications, diagnostic findings, and lesion characteristics and patient-specific lesion-based variables: maximum percent diameter stenosis, any high risk lesion, any restenotic lesion, any use of embolectomy or atherectomy. Because some hospitals in Massachusetts are pilot programs that perform only primary angioplasty without surgery on site, we also included a variable that indicated if the patient underwent the PCI in a pilot program.
Some confounders (neoplasm, gastrointestinal bleeding) that were not collected in the NCDR instrument were created by linking patients to the hospital discharge billing data and identifying relevant ICD-9-CM codes in the index admission. 9
Propensity score matching is a method of adjusting for observed characteristics of patients non-randomly assigned to differing treatments 11. Because patients were not randomized to receive a DES, we performed a 1-1 matched analysis without replacement on the basis of the estimated propensity score of each patient. The log-odds of the probability that a patient received a DES was modeled as a function of the confounders we identified. Using the estimated logits, we first randomly selected a DES patient and then matched the DES patient to the “closest” BMS patient. BMS patients who had an estimated logit within 0.6 standard deviations of the selected DES patients were eligible for matching. We selected 0.6 because this value has been shown to eliminate approximately 90% of the bias in observed confounders 12. We assessed the success of the matches by examining standardized differences in the observed confounders between the matched DES and BMS groups. Small (< 10%) differences support the assumption of balance between treatment groups 13.
Using the matched pairs, we conducted paired t-tests to determine if mortality, myocardial infarction and revascularization were different between DES and BMS patients. Estimates of the adjusted differences in risks are presented with 95% confidence intervals of the difference. All p values presented are two-sided.
Because matching on the propensity score cannot be expected to balance unobserved confounders that are not related to observed confounders, we undertook several sensitivity analyses. First, we examined differences in mortality between DES and BMS patients two days after stent placement. If this difference was clinically large, this would indicate that there are unmeasured confounders since such an early benefit would be unlikely.
Second, we compared DES vs BMS outcomes after adjusting for changes in selection over time (Appendix). Third, we performed an analysis of patients matched after excluding those presenting with acute myocardial infarction. Finally, we estimated the degree of residual confounding that would alter the conclusions of our study (Appendix).
Dr. Mauri and Dr. Normand had full access to the data and take responsibility for its integrity. All authors have read and agree to the manuscript as written.
Between April 1, 2003 and September 30, 2004: 21,045 patients received percutaneous coronary intervention with stenting in Massachusetts performed by 138 operators, at 21 hospitals (Figure 1). We excluded patients who were not residents of Massachusetts or were not linkable to hospital discharge data and Mass-DAC data, and patients who were treated with a combination of DES and BMS from the primary analysis. This yielded a cohort of 17,793 adults: 11,556 (65%) patients treated with DES and 6,237 (35%) patients treated with BMS. Of patients treated with DES, 8,378 (74%) received sirolimus-eluting stents only, 2,950 (26%) received paclitaxel-eluting stent only. Rates of DES use increased over the course of the study from 17% to 92% (Figure 2).
Clinical and procedural characteristics differed significantly between patients treated with DES and BMS (Tables 1 and and2).2). Patients with diabetes mellitus, hyperlipidemia or hypertension, or a history of prior PCI were more likely to be treated with DES. In contrast, older patients, smokers, and patients with myocardial infarction, a history of chronic lung disease, chronic renal insufficiency, cerebrovascular or peripheral vascular disease, neoplasm or gastrointestinal bleeding were more likely to be treated with BMS. Overall, patients treated with DES received more stents and had more lesions treated than patients with BMS.
Patients treated with both BMS and DES were not included in the BMS or DES groups, and had more complex procedures compared to patients with DES or BMS, with overall a greater number of lesions and vessels treated.
Unadjusted mortality, myocardial infarction, and target vessel revascularization rates were higher for BMS than for DES at 2 years (Appendix Table 1). The unadjusted rates of adverse events for patients treated with a combination of BMS and DES were intermediate between results for either stent type alone for mortality and myocardial infarction at 2 years (10.8%, 8.7%) and higher for target vessel revascularization (34.6%).
The propensity score model included 63 variables and had area under the receiver operating characteristic curve of 0.67. Matching by propensity score yielded 5,549 DES-treated patients matched to 5,549 BMS-treated patients. Standardized differences were less than 10% (Tables 3 and and4,4, Figure 3) with the exception of ACS within 6 hours of hospital arrival (percent standardized difference = 10.9%), emergency or salvage procedure indication (percent standardized difference = 11.7%), and GP IIb/IIIa pretreatment (percent standardized difference = 10.5%) - all slightly more common in the DES group, and elective procedure indication (percent standardized difference = -12.1%) – slightly more common in the BMS group.
In the propensity matched sample, rates of revascularization were lower for DES use at 1 year compared with BMS use (-6.2% [-7.4%, -5.1%], p<0.0001) (Table 5, Figure 4c). A similar absolute difference was preserved at 2 years.
Examination of the magnitude of difference in mortality two days following stent placement between the matched pairs showed in a small absolute difference of 0.23% (0.45% vs 0.68% DES vs BMS; p = 0.10).
Diffusion of DES was rapid in Massachusetts (Figure 2). Analysis adjusting for the interaction between time and selection of stent type gave results comparing DES and BMS that were consistent with the primary analysis (Appendix Table 2).
When patients presenting with myocardial infarction were excluded, and the remaining DES and BMS patients matched (2,947 matched pairs), observed differences in 2 year clinical outcome were consistent with the overall cohort (mortality -1.8% [-3.3%, -0.3%], p=0.0174; MI -1.2% [-2.7%, 0.2%], p=0.0982; TVR -6.8% [-8.7%, -5.0%], p<0.0001), although with the smaller cohort, the difference in MI rate at 2 years was no longer significant.
We examined patients from a regional contemporary U.S. practice with mandatory reporting in order to compare long term outcomes after drug-eluting or bare metal stent placement. Our primary goal was to evaluate whether DES were associated with increased rates of myocardial infarction and mortality in long term follow-up; however, we observed small absolute differences in myocardial infarction and mortality favoring DES. Repeat revascularization procedures were also lower.
Previous randomized controlled trials of approved DES vs. BMS have not shown significant differences in death or myocardial infarction in long term follow-up to date 3. While the chief strength of such randomized comparisons is elimination of selection bias, these studies have been limited by power and generalizability to every day practice 7, 8. Some pooled randomized trials have shown a lower rate of recurrent MI for one DES vs BMS 14. Observational studies have had contradictory findings to date 5, 15. Observational studies require more sophisticated methods than randomized studies to minimize bias, but may offer improved power and greater generalizability 16. We captured all PCI performed in Massachusetts over an 18 month period and included a wide range of hospitals and operators – academic, non-academic, with and without surgical back up, lower and higher volume centers – and a wide variety of procedure types – elective and emergent procedures, and low and high lesion complexity. Because all percutaneous coronary intervention procedures in Massachusetts are subject to prospective mandatory reporting of patient and procedural characteristics and outcomes, this a powerful and reliable cohort.
Our study differed in several important ways from other population-based comparisons of DES and BMS. In contrast to populations where DES use may be restricted geographically, financially, or according to patient risk, 5, 15 our US study reflected a majority of DES use overall (65%). As a result our DES sample represents a broader patient population.
To avoid bias due to differential follow-up, we restricted our analysis to those patients who had complete follow-up at a fixed time point – 2 years. Although we have substantially longer follow-up in a subset of patients, we limited the analysis to a fixed time point to avoid inference on an incomplete sample and avoid changing conclusions as greater follow-up accrues 17.
Regarding ascertainment of study endpoints, we observed somewhat higher rates of MI than other studies5, 15, as we included periprocedural MI, even during the same hospitalization. The absolute difference in target vessel revascularization of approximately 5% is similar to recently reported registries in the United States 18, but lower than reported in randomized controlled trials where surveillance angiography was frequent1, 2. One mechanism to explain the reductions in MI and mortality may be that reduction in the occurrence of repeat revascularization procedures related to restenosis in a patient population of higher complexity had an impact on true clinical endpoints. 19, 20
Late stent thrombosis has been considered a potential mechanism for the increase in mortality in some studies of DES21. We did not have sufficient data to accurately estimate rates of late stent thrombosis. This well-powered observational study would suggest that if a small difference in late stent thrombosis is present between DES and BMS, it is not readily detectable in a clinical endpoint (death or MI) because it represents a small proportion of these events. Given the relatively rare occurrence of stent thrombosis (<1% in the first year and rarer in later follow-up) 6, it would seem unlikely that these events would overtake the constant hazard of adverse events related to disease progression 22 beyond 2 years.
Because the choice of stent was not randomly assigned, we used propensity score matching to select a group of patients who were essentially similar according to all measured baseline variables. As compared with other methods that use a propensity model to adjust for selection bias, a matched analysis is more likely to reduce power (by restricting the sample to closely matched patients) but has the advantage of more reliably reducing selection bias (as compared with regression-based propensity analysis). We were afforded this opportunity because the density of DES and BMS procedures performed in Massachusetts was high (>14000 PCI procedures per year per 6 million adults), and because the patterns of use of DES and BMS in our study spanned both low and high risk patients. We confirmed that the matched patients were similar for each of 63 baseline variables, and with this degree of similarity, matching of unmeasured confounders is expected 23. However, we acknowledge the possibility of residual confounding. This was estimated to be small as measured by the 2-day mortality difference. Some sources of residual bias include varying completion of revascularization, limited ability to adjust for angiographic characteristics (lesion length and vessel diameter were not available), and concomitant medical therapy.
We collected information regarding use of and contraindications to dual antiplatelet therapy with aspirin and thienopyridine therapy because of its known relationships to mortality, myocardial infarction, and stent thrombosis 24-26. Although we did not have pharmacy data to compare the duration of therapy and compliance in patients treated with DES and BMS, during the period under study, the existing cardiology society guidelines recommended a longer duration of thienopyridine therapy in patients with DES (3-6 months minimum) than with BMS (1 month minimum) 27. We cannot exclude the possibility that a longer use of thienopyridine treatment that is recommended with DES selection could have contributed to a mortality and myocardial infarction benefit favoring DES over BMS. As the most recent national society recommendation is for 12 months dual-antiplatelet therapy for DES 28, the associated effect of dual-antiplatelet therapy with DES selection compared with BMS may increase in future studies.
We combined patients treated with paclitaxel-eluting and sirolimus-eluting stents within the DES group. Differences in safety between drug-eluting stents remain controversial, and other population-based studies have not shown significant differences in mortality between these two stents 29.
In conclusion, in a large population-based study reflecting a majority of DES use across a cohort with comprehensive follow-up after PCI, we found that DES were associated with lower mortality, myocardial infarction, and target vessel revascularization relative to BMS use at 2 years of follow-up. Longer term follow-up will be required to provide further reassurance of preserved safety in the broad patient population currently treated with these coronary devices.
The authors wish to acknowledge Paul Dreyer, Ph.D. of the Massachusetts Department of Public Health and the members of the Mass-DAC PCI Data Adjudication Committee: Kurt Barringhaus, M.D., Clifford J. Berger, M.D., David Cohen, M.D., Angela Corey, R.N., Jean Crossman, R.N., Daniel Fisher, M.D., Joseph Garasic, M.D., Jean-Pierre Geagea, M.D., Gregory Giugliano, M.D., Kalon Ho, M.D., Alice Jacobs, M.D., James Kirshenbaum, M.D., Josh Krasnow, M.D., Anthony Marks, M.D., Theo E. Meyer, M.D., Ph.D., Kathy Minahan, R.N., Zoran Nedelijkovic, M.D., Barbara Oxley, R.N., Thomas C. Piemonte, M.D., Kenneth Rosenfield, M.D., Pinak B. Shah, M.D., Samuel J. Shubrooks Jr., M.D., James Waters, M.D., and Bonnie Weiner, M.D. for their assistance. The authors also wish to thank Manu Varma for assistance with preparation of tables and figures.
Funding Sources: Study funded by Massachusetts Department of Public Health Contract 620022A4PRE (Dr. Normand, Ms. Lovett, Silverstein, Zelevinsky, and Mr. Wolf). The study sponsor study sponsor did not influence the study design, collection, analysis, and interpretation of data, the writing of the report or the decision to submit the paper for publication.
Disclosures: Dr. Mauri has received honoraria from Abbott Vascular, Boston Scientific, Cordis, and Medtronic Vascular. Dr. Zhou had been employed by Boston Scientific one year prior to commencing this work. Dr. Resnic has consulted for or received honoraria from Abbott Vascular and Cordis Corporation.
Clinical Perspective: The late safety of drug-eluting stents (DES) has been a matter of recent controversy. Randomized trials have not been powered to detect differences in mortality compared with bare metal stents (BMS), and have been limited to lower risk patients. This population-based study collected prospective clinical and procedural data on all patients who underwent percutaneous coronary intervention (PCI) with stents from April 1, 2003 through September 30, 2004 at non government hospitals in Massachusetts. Rates of mortality, myocardial infarction and repeat revascularization procedures on the treated vessel were ascertained over the 2 years following stent placement. The goal of the study was to determine long-term patient outcomes by stent type in a population representative of current United States medical practice.
Because the choice of stent was not randomly assigned, a propensity score-matched analysis was performed compare DES and BMS-treated patients with similar characteristics. One to one matching was performed using a logistic regression model created from 63 variables, to identify unique pairs of DES and BMS patients with similar baseline characteristics. A total of 17,793 patients that underwent placement of DES or BMS were identified; patients that received both stent types were excluded. Sixty-five percent (11,556) of patients received DES compared with 35% (6,237) who received BMS. 5549 DES patients matched to 5549 BMS patients were analyzed for each of the 2-year outcomes. 2 year risk-adjusted mortality, myocardial infarction and target vessel revascularization rates were each lower for patients treated with drug-eluting stents compared with bare metal stents.