Comparison of Paclitaxel- and Sirolimus-Eluting Stents in Everyday Clinical Practice:  The SORT OUT II Randomized Trial FREE

Percutaneous coronary interventions (PCIs) initially were performed using balloon angioplasty and had a restenosis rate of approximately 40%; introducing stent implantation reduced this rate to approximately 20%.¹ The next stage in the evolution of PCI incorporated stents that eluted antiproliferative drugs aimed at reducing neointima formation, which substantially reduced the restenosis rate further, to approximately 10% of treated lesions.^2- 3

Even if the rate of undesirable events is low, different stents might still produce important differences in clinical event rates that require direct and randomized comparison to detect. When devices are compared, the low event rate necessitates randomizing many patients, which is expensive and takes precious time. Small studies with surrogate end points derived from scheduled repeat angiography among highly selected patients are used instead, but the results from such studies may not reflect the clinical situation for unselected patients in the real world.

This study aimed to compare the efficacy and safety of sirolimus-eluting and paclitaxel-eluting stents in a study design reflecting everyday clinical practice by attempting to include unselected patients and by specifically using only symptom-driven clinical end points.

The Danish Organization on Randomized Trials With Clinical Outcome (SORT OUT) performed this study. In brief, SORT OUT is a nonprofit, industry-independent clinical research collaboration between the 5 university hospitals in Denmark that perform PCI 24 hours a day, covering the entire population of 5.5 million. SORT OUT was founded with the aim of conducting randomized trials of intracoronary stents to ensure the efficacy and safety of these devices in everyday clinical practice.

The SORT OUT studies generally do not include mandatory repeat angiography: that is, all repeat angiographies are clinically driven. In this study, however, a subgroup of 102 patients with diabetes was scheduled for coronary repeat angiography after 8 months according to an angiographic substudy (M. Meng, PhD, and L. Okkels Jensen, PhD, written communication, May 2004).

The inclusion period was August 2004 to January 2006, and a follow-up period of 5 years was planned, with the first primary end point results scheduled to be reported after 9 months. We conducted the study in accordance with the Second Helsinki Declaration, and the local biomedical research ethics committee approved the study. All study participants provided written informed consent.

An independent end point committee blinded to study randomization detected and classified all events and assessed all deaths, major adverse cardiac events, and secondary end points. Two authors (A.M.G., N.B.) reviewed the cine loops from all index PCI angiographies to classify target lesions in accordance with the most recent guideline definitions from the American College of Cardiology/American Heart Association/Society for Cardiovascular Angiography and Interventions.²

We sought to enroll consecutive patients undergoing PCI with the planned use of a drug-eluting stent. We excluded patients who were enrolled in other studies or who lived outside Denmark. Inclusion in the study and patient randomization were operator dependent; ie, some operators included the majority of their patients, while other operators did not participate actively in the study. To define the study population, we collected demographic data and 30-day all-cause mortality data for all patients in Denmark who underwent PCI during the SORT OUT II inclusion period.

Patients were randomized 1:1 by the method of permuted blocks to receive 1 of the first 2 commercially available drug-eluting stents—the sirolimus-eluting Cypher stent (Cordis/Johnson & Johnson, Miami Lakes, Florida) and the paclitaxel-eluting Taxus stent (Boston Scientific Corp, Natick, Massachusetts). The study was open-label, and a target lesion was defined as a lesion assigned to be treated with a drug-eluting stent. If the patient had more than 1 lesion requiring treatment, we recommended using the study stent in all lesions, but balloon angioplasty or implantation of bare-metal stents was allowed.

Glycoprotein IIb/IIIa inhibitors were used at the operator's discretion. Dual antiplatelet therapy with aspirin and clopidogrel was recommended for 1 year for all patients. After that period, clopidogrel was discontinued and aspirin continued lifelong, if tolerated. If a new PCI was performed during follow-up, treatment with clopidogrel was prolonged or reinstated for 1 year in accordance with Danish guidelines.

The primary end point was a major adverse cardiac event composite of cardiac death, acute myocardial infarction, target lesion revascularization, and target vessel revascularization.² Secondary end points included individual components of the composite end point, all-cause mortality, and stent thrombosis as defined by the Academic Research Consortium.⁴

The Danish Civil Registration System and the National Patient Registry were used to detect all cases of death, hospital admission, coronary angiography, repeat PCI, and coronary artery bypass graft surgery. Every hospital admission was investigated by reviewing discharge summaries, and if these suggested a major adverse cardiac event, we reviewed serial electrocardiograms, blood sample results, and copies of complete patient records. We examined death certificates and, if deemed necessary, interviewed the deceased patient's general practitioner for further details. In cases of reintervention by PCI or coronary artery bypass graft surgery, we retrieved copies of patient records and identified treated coronary lesions. In cases of uncertainty, the independent end points committee collected and reviewed copies of the coronary angiography findings.

The core event laboratory at Gentofte University Hospital classified each death and hospital admission as a major adverse cardiac event, possible major adverse cardiac event, or non–major adverse cardiac event, respectively. The independent end points committee further evaluated the first 2 end point categories. The chair of the end point committee also independently performed additional control of the event classification by randomly selecting 2 different numbers from 1 to 31. The chair then reclassified all hospital admissions, deaths, or both among patients born on 1 of these 2 days in any month of the year. This additional quality control did not yield any major adverse cardiac event that had not already been detected.

All analyses were performed according to the intention-to-treat principle. We compared baseline characteristics with SAS version 7.0 (SAS Institute Inc, Cary, North Carolina) using the χ² test for discrete variables, t test for continuous variables, and Kruskal-Wallis test for nonparametric variables. Cumulative proportions of patients experiencing major adverse cardiac events were analyzed according to the date of the first event for each patient, depicted in a Kaplan-Meier plot. We performed statistical tests using the log-rank test. We calculated proportional hazard risk ratios using PRISM version 3.0 (GraphPad, San Diego, California). If patients emigrated, the date of emigration was used as the censoring date. For the secondary end points, the date of emigration or the date of death was used as the censoring date.

A subgroup of 102 patients with diabetes underwent scheduled repeat angiography after 8 months. If this repeat angiography led to intervention, we censored the patient on the day of the repeat angiography and the event was not classified as a major adverse cardiac event. To estimate the importance of this subpopulation, study results with and without censoring for scheduled repeat angiography were calculated. We continued the observation period beyond all major adverse cardiac events to allow for hierarchical analysis of all secondary end points. P  < .50 was considered statistically significant.

At the time the study was planned, data from comparative stent trials were not available, and published studies of the 2 drug-eluting stents vs bare-metal stents had limited follow-up. We hypothesized that 5-year rates of major adverse cardiac events would be markedly lower with the drug-eluting stents than with bare-metal stents. The SORT OUT II trial was therefore planned to enroll at least 1800 patients to detect a minimum relevant relative difference in rates of 25% (ie, rates of 20% vs 15%) with 5 years of follow-up. An interim safety analysis was prespecified after 9 months. Since the rates at this point were lower, which diminished the study power, we continued randomization beyond 1800 patients and stopped when the next SORT OUT study (SORT OUT III) was ready to begin.

During the randomization period, 11 766 patients were treated with PCI in Denmark at least once; of these, 2098 were randomized (Figure 1). Table 1 shows the baseline clinical and angiographic characteristics of the randomized and the nonrandomized patients. They differed slightly in age, sex, number of lesions, and vessels treated per patient. The most notable clinical difference was in the proportion of patients with ST-segment elevation myocardial infarction, which was much higher among the nonrandomized patients (3177 [32.9%] vs 364 [17.5%], P < .001). Importantly, the nonrandomized patients had higher 30-day all-cause mortality (367 [3.8%] vs 18 [0.9%], P < .001).

Table 2 summarizes the baseline clinical and angiographic characteristics of the randomized patients. The groups did not differ significantly. All patients were observed from randomization to death, emigration, or up to 18 months after randomization. In the diabetic subgroup study (n = 102), 19 patients were censored due to revascularization after protocol-driven repeat angiography after approximately 8 months (10 patients in the sirolimus-eluting stent group and 9 in the paclitaxel-eluting stent group). Figure 2 shows the cumulative proportions of patients experiencing major adverse cardiac events by time in the sirolimus- and the paclitaxel-eluting stent groups, excluding all scheduled repeat angiography. The rates were 98 (9.3%) for the sirolimus-eluting stent group and 114 (11.2%) for the paclitaxel-eluting stent group (hazard ratio, 0.83; 95% confidence interval, 0.63-1.08; P = .16).

Table 3 summarizes the primary and secondary end points. The end points did not differ significantly in the groups, regardless of whether we included or excluded the patients with diabetes who underwent scheduled repeat angiography.

Figure 2 also shows the cumulative proportions of patients experiencing definite and probable stent thrombosis by time, according to the definitions by the Academic Research Consortium for the sirolimus- and paclitaxel-eluting stent groups (excluding all scheduled repeat angiography).⁷ The stent thrombosis rates were 27 (2.5%) in the sirolimus-eluting stent group and 30 (2.9%) in the paclitaxel-eluting stent group (hazard ratio, 0.87; 95% confidence interval, 0.52-1.46; P = .60).

In this large randomized trial designed to reflect everyday clinical practice, we found that the 2 first commonly used drug-eluting stents did not differ significantly in the primary end point or any of the secondary end points. This equivalence may have resulted from a truly negligible difference in the effectiveness and safety of the stents in everyday clinical practice, but it also may be due to insufficient study power. A priori, the planned study power was 80%, but during the course of the study the actual power was reduced, since the mean rate of major adverse cardiac events was considerably lower than expected (10.2% vs the expected 20.0%), as was the absolute difference between the rates in the 2 treatment groups (1.9 percentage points vs the expected 5.0 percentage points). If a new study were planned with the same mean rate and difference between the groups that we found in SORT OUT II, the power of such a study would be only 28.7%.

However, the small difference between 2 treatments is unlikely to be of clinical importance. The area between the curves showing cumulative proportions of patients experiencing major adverse cardiac events from the 2 treatment groups reflects an estimate of the net health gain with the superior device. Analysis of the size of this area showed that patients treated with a sirolimus-eluting stent would experience a “possible” major adverse cardiac event an average of 6.3 days later than patients treated with a paclitaxel-eluting stent. Because 89.8% of the patients did not experience any major adverse cardiac events during the first 18 months of follow-up, the health gain with the sirolimus-eluting stent should be divided solely among the patients who actually developed such an event; thus, a “de facto” major adverse cardiac event could be delayed by a mean of 62.0 days. This effect is arguably of clinical importance; there may be a clinically relevant difference favoring the sirolimus-eluting stent.

Published randomized trials comparing the sirolimus- and the paclitaxel-eluting stents have included up to 1355 patients, making our study 55% larger than the largest of these previous studies.^7- 17 Two recent meta-analyses summarized the results of these trials; as Table 3 shows, these results are comparable with our data.^5- 6 The major difference between our study and the previous studies was that all of our repeat angiographies were driven by clinical events, such as angina pectoris or acute myocardial infarction. This feature of SORT OUT II is important, because evidence indicates that preplanned repeat angiography may lead to unnecessary revascularization.^11,18- 20 The difference in target lesion revascularization between the 2 groups did not reach statistical significance (hazard ratio, 0.72; 95% confidence interval, 0.48-1.07; P = .10). The meta-analyses detected comparable but significant differences in target lesion revascularization favoring the sirolimus-eluting stent, and the absence of statistical significance in our study may be a type II error (Table 3).^5- 6

Moreover, the groups did not differ in rates of cardiac death, acute myocardial infarction, or stent thrombosis. The latter finding contrasts somewhat with the result of one of the meta-analyses,⁵ in which the paclitaxel-eluting stent group was associated with a statistically significant increased risk of stent thrombosis (Table 3). Indeed, concern has been raised recently about the safety of drug-eluting stents. In the SORT OUT II trial, the rate of serious adverse events, cardiac death, and acute myocardial infarction was 6.2%, which is less than the 8.3% reported in one of the meta-analyses.⁶ Our results therefore indicate that drug-eluting stents may be safe in a broader spectrum of patients than those examined in the previous studies, because the SORT OUT II patient population had considerably more complex coronary lesions. Seventeen percent of the SORT OUT II population presented with ST-segment elevation myocardial infarction, more than 50% presented with acute coronary syndromes, and 33% had more than 1 lesion treated. This case mix is different from those of the previous studies, which usually involved stable patients with single-vessel disease and 1 simple lesion. Our results therefore suggest that using drug-eluting stents for expanded off-label indications is safe.

Late stent thrombosis has been a particular concern in the era of drug-eluting stents. In our study, the stents did not differ significantly in rates of either early or late stent thrombosis (Figure 2). The mean stent thrombosis rate after 18 months was 2.7%, which is more than the 1.6% found in a recent meta-analysis.⁵ This finding may be explained by a different case mix in our study, and our data therefore suggest that using drug-eluting stents for off-label indications will increase the stent thrombosis rate. After the first few months after stent implantation, the mean stent thrombosis rate was 1 stent thrombosis per month for every 1000 patients treated. Because dual antiplatelet therapy was planned for only 1 year, increased stent thrombosis rates were expected thereafter. Although our study was not planned to specifically address this issue, it is notable that the slope of the curves showing cumulative proportions of patients experiencing stent thrombosis remained constant after 1 year of follow-up, indicating that withdrawing clopidogrel after 1 year is safe (Figure 2).

The SORT OUT II trial demonstrated that complete follow-up of all patients undergoing PCI in Denmark was feasible, allowing for valid comparison of randomized and nonrandomized patients in national clinical practice and thereby enabling qualified estimation of the extent to which the study population was representative of everyday clinical practice. The sole criterion for inclusion in the study was the use of a drug-eluting stent, and a limitation is that the criteria for using drug-eluting stents were not specified and that they changed considerably during the randomization period. It is therefore impossible to determine more precisely the subsets of patients or lesions to which the results may be extrapolated. We cannot compare the randomized patients to the proportion of the nonrandomized patients who were also treated using a drug-eluting stent, but we believe that our study is the first to compare all the randomized and nonrandomized patients nationwide in the entire potentially eligible Danish population.

In conclusion, the SORT OUT II trial found no statistical significant differences in the primary or secondary end points between the sirolimus-eluting stent and paclitaxel-eluting stent in everyday clinical practice among patients undergoing PCI for ST-segment elevation myocardial infarction, non–ST-segment elevation myocardial infarction or unstable angina pectoris, and stable angina. The rates of serious adverse events, cardiac death, acute myocardial infarction, and stent thrombosis were low, suggesting that, at least when considering 18 months of follow-up, the use of drug-eluting stents in the general population may be safe.