Sirolimus-Eluting Stents vs Vascular Brachytherapy for In-Stent Restenosis Within Bare-Metal Stents:  The SISR Randomized Trial FREE

Coronary stents revolutionized interventional cardiology by greatly improving both initial procedural success and longer-term outcomes, reducing clinical and angiographic restenosis rates by 30% to 50% compared with conventional balloon angioplasty.^1- 3 While the introduction of drug-eluting stents further decreased in-stent restenosis, the absolute magnitude of this reduction depends on specific angiographic and clinical characteristics^4- 16; however, use of the drug-eluting stent remains constrained in some regions because of high cost. Vascular brachytherapy is currently the only approved therapy for restenosis within a bare-metal stent based on its documented superiority to balloon angioplasty and other treatment modalities.^17- 23 The purpose of this trial was to compare the use of vascular brachytherapy with implantation of the sirolimus-eluting stent for the treatment of restenosis occurring within a previously placed bare-metal stent.

This prospective, randomized, multicenter trial complied with the provisions of the Declaration of Helsinki and was approved by the Food and Drug Administration and all institutional review boards. All patients gave written informed consent. This study was designed to compare the safety and the effectiveness of vascular brachytherapy (either β or γ) with the sirolimus-eluting stent for the treatment of in-stent restenosis following bare-metal stent placement.

Eligible patients had a history of stable or unstable angina or documented silent myocardial ischemia. The target lesion was an in-stent restenotic coronary arterial lesion between 15 mm and 40 mm in length and between 2.5 mm and 3.5 mm in diameter by visual estimate. The vessel 1 cm distal to the target lesion was required to be at least 2.5 mm in diameter to allow use of commercially available brachytherapy devices.

Major exclusion criteria included myocardial infarction (MI) within the preceding 24 hours, ejection fraction of less than 40%, prior thoracic radiation or intravascular brachytherapy, total occlusions, unprotected left main coronary artery disease with greater than 50% stenosis, or treatment of a nontarget lesion occurring within 30 days before or planned after the index study procedure. Patients with a serum creatinine level of 2 mg/dL or greater (≥176.8 μmol/L) were excluded as were patients who had undergone initial stent placement in the target lesion less than 4 weeks prior to the index study procedure.

To promote enrollment, patients were randomized in a ratio of 1 to 2 to treatment with either vascular brachytherapy or the sirolimus-eluting stent. Randomization was accomplished at each site using an interactive voice randomization system (Interactive Clinical Technologies Inc, Yardley, Pa). The procedure “PROC PLAN” (SAS software version 6.2, SAS Institute Inc, Cary, NC) was used to generate the randomization list using a block size of 6.

Prior to and following the index procedure, all patients received 325 mg/d of oral aspirin and either clopidogrel (loading dose of 300-375 mg followed by 75 mg/d) or ticlopidine (loading dose of 500 mg followed by 250 mg twice daily). During the procedure, intravenous heparin was administered to maintain an activated clotting time of greater than 300 seconds for patients in the vascular brachytherapy group and greater than 250 seconds for patients in the sirolimus-eluting stent group. Glycoprotein IIb/IIIa inhibitors were administered at the discretion of the physician.

Vascular brachytherapy and sirolimus-eluting stent implantation were performed according to conventional techniques. Predilatation was required in the sirolimus-eluting stent group. “Geographic miss” was defined as a mismatch between the length of the vessel treated with predilatation and the length of the vessel receiving the indicated treatment. In the vascular brachytherapy group, geographic miss was minimized by covering the damaged arterial segment with the radioactive source train extending at least 5 mm beyond each margin of the region treated with balloon dilatation. Placement of a new stent in the vascular brachytherapy group was discouraged.

Geographic miss was minimized in the sirolimus-eluting stent group by selecting a stent that extended more than 3 mm beyond both ends of the region of balloon angioplasty, ensuring stent coverage from the angiographically normal vessel proximal to distal. While it was required that the restenotic segment be fully covered, it was not required that the entire region initially covered by the bare-metal stent be restented.

Data were submitted to the data coordinating center (Harvard Clinical Research Institute, Harvard Medical School, Boston, Mass). Clinical end points were adjudicated by an independent clinical events committee blinded to study group assignment. A separate data and safety monitoring board not affiliated with either the study sponsor or the investigators reviewed data periodically throughout the trial to identify potential safety issues and monitor study conduct. The data presented herein represent all angiographic and clinical follow-up data available as of June 30, 2005.

Coronary angiography was performed at baseline, at the completion of the procedure, and at 6-month follow-up; angiograms were analyzed with a quantitative computer-based system (Medis, Leesburg, Va) at the angiographic core laboratory (Brigham and Women's Angiographic Core Laboratory, Boston, Mass). Minimal lumen diameter was defined as the mean minimal lumen diameter derived from 2 orthogonal views by quantitative coronary artery angiography. Late lumen loss was defined as the difference between the minimal lumen diameter at the completion of the procedure and that measured at 6-month follow-up. Net gain was defined as the difference between preprocedure minimal lumen diameter and that measured at 6-month follow-up. Binary angiographic restenosis was defined as greater than 50% narrowing of the lumen diameter in the target lesion. Quantitative angiographic assessment of the target lesion included measurements corresponding to the analysis segment (defined as all portions of the vessel that received treatment within the radiation or stent zones including the proximal and distal 5-mm margins) and the injury segment (defined as the region of vessel injured during treatment by balloon dilatation or stent placement). A subset of 100 patients was asked to undergo intravascular ultrasound assessment.

The primary end point of this study was target vessel failure defined as cardiac death, MI, or target vessel revascularization at 9 months postprocedure. Secondary angiographic end points included postprocedure in-stent and in-lesion minimal lumen diameter and percentage of diameter stenosis as well as 6-month in-stent and in-lesion binary restenosis and late loss by quantitative angiography. Secondary clinical end points included target lesion and target vessel revascularization at 6 and 9 months and the composite of major adverse cardiac event rates (defined as death, either Q-wave or non–Q-wave MI, emergent coronary artery bypass graft surgery, or repeat target lesion revascularization) at 30 days, at 6, 9, and 12 months, and at 2, 3, 4, and 5 years after the procedure. Non–Q-wave MI was defined as elevation of postprocedure creatine kinase levels to greater than 2 times the upper limit of normal with creatine kinase-MB fraction elevated above normal. Target lesion and target vessel revascularization procedures were required to be “clinically driven”⁷ as adjudicated by the clinical events committee. Measurements of stent lumen and stent volume obstruction were obtained with intravascular ultrasound. Economic data were obtained for cost-effectiveness analysis.

This study was designed to demonstrate the noninferiority or superiority of the sirolimus-eluting stent (Cordis, Warren, NJ) compared with intracoronary vascular brachytherapy. Noninferiority was expected based on documented safety and efficacy with both approaches. Superiority was expected due to significant reduction in target vessel failure and improvement in analysis segment net gain and late lumen loss in previous studies with the sirolimus-eluting stent.

Bayesian statistical methods were used for trial design and to perform the formal analysis of the primary end point. Bayesian methods use a modeling approach that allows a reduction in the sample size of the vascular brachytherapy group (control group) by “borrowing” data from previous studies in which individual patient-specific data are available (the Cordis-sponsored GAMMA I²² and GAMMA II [D.R.H., unpublished data, 2006] studies). Angiographic follow-up at 6 months and clinical follow-up at 9 months were obtained in the current trial to match the time of follow-up in the GAMMA trials.

Because Bayesian methods do not provide initial sample size estimates, the original sample size calculation was based on standard statistical methods. Using methods based on Pocock,²⁴ heterogeneity between the current and previous (historical) trials was accounted for in a revised power calculation that reflected the Bayesian analysis planned for these data. It was assumed that the analysis data set would contain 117 patients in the vascular brachytherapy group and 233 patients in the sirolimus-eluting stent group, and an additional 256 control patients from the historical trials. The assumed target vessel failure rates were 30.5% for the historical control groups and 15.3% for the sirolimus-eluting stent group. The assumed amount of variation in the target vessel failure rate for vascular brachytherapy across the trials ranged from 1.5% to 10.0%. The significance level for rejecting the null hypothesis was .05. In all scenarios, the trial had a power level greater than 80% for both noninferiority and superiority.

A Bayesian regression model was used to analyze target vessel failure at 270 days using the concurrent and historical data. The model accounted for unobserved heterogeneity between the trials by accounting for the clustering of observations within trials and controlled for the following observed variables: reference vessel diameter, lesion length, sex, history of diabetes, left anterior descending artery disease, and the number of diseased vessels. To test for noninferiority, the posterior probability that the target vessel failure rate for the sirolimus-eluting stent group was less than the sum of the target vessel failure rate for the vascular brachytherapy group was calculated; the “Δ” corresponded to 7.5% on the probability scale. To test for superiority, the posterior probability that the target vessel failure rate for the sirolimus-eluting stent group was less than the target vessel failure rate for the vascular brachytherapy group was computed. If the posterior probability exceeded 0.95 (required for significance at the .05 level), the null hypothesis was rejected in favor of the alternative (noninferiority or superiority).

The effectiveness analysis and safety evaluation were performed on the intent-to-treat study population. Unless otherwise specified, all frequentist (ie, non-Bayesian) statistical tests and/or 95% confidence intervals (CIs) were performed with a 2-sided α level of .05. Treatment group comparisons on continuous measures were performed using the 2-sample t test. The 95% CI of the mean difference between groups was calculated. Treatment group comparisons on categorical measures were performed using the Fisher exact test and excluded the “unknown” category. The 95% CIs of the difference in percentages are presented for each treatment group using the normal approximation to the binomial distribution. The relative risks (RRs) between the treatment groups and the 95% CIs of the risks were also calculated.

Computations for all frequentist results were performed using SAS version 8.2 (SAS Institute Inc). For Bayesian analysis, the BUGS (Bayesian Inference Using Gibbs Sampling) software was used: WinBUGS version 1.4.1 (MRC Biostatistical Unit, Cambridge, England).

From February 12, 2003, through July 27, 2004, 384 patients were enrolled and randomized to treatment with vascular brachytherapy (n = 125) or the sirolimus-eluting stent (n = 259) (Figure 1). Follow-up continued until June 30, 2005.

Clinical characteristics did not differ between the 2 groups except for chronic renal insufficiency, which was more prevalent in the sirolimus-eluting stent group (Table 1). Overall, mean (SD) age was 62.9 (11.1) years, 67.4% (258/383) were male, 47.1% (173/367) had a history of prior MI, and 48.3% (154/319) had unstable angina. Treated lesions were predominantly type B2 (33.5% [128/382]) or type C (41.4% [158/382]) using the modified lesion classification system of the American Heart Association and the American College of Cardiology²⁵ and most often were located in the left anterior descending artery (46.9% [179/382]). Compared with the vascular brachytherapy group, total occlusion of the lesion was more common in the sirolimus-eluting stent group (1.6% [2/125] vs 6.7% [17/255]; P = .04).

Procedural outcomes were excellent (Table 2) with no significant differences in in-hospital complications between the groups. The mean (SD) radiation length in the vascular brachytherapy group was 39.7 (11.0) mm. Final mean (SD) stent length in the sirolimus-eluting stent group was 32.49 (12.3) mm while the total stent length to lesion length ratio was 1.9 (0.8).

Baseline dimensions including lesion length, minimal lumen diameter, diameter stenosis, and reference vessel diameter were similar between the 2 groups (Table 2). However, compared with the vascular brachytherapy group, the postprocedure analysis segment minimal lumen diameter (mean [SD], 1.87 [0.39] vs 2.06 [0.48] mm) and diameter stenosis (28.91% [10.44%] vs 23.53% [12.09%]) favored the sirolimus-eluting stent group (both P<.001) and resulted in a significant improvement in acute gain in vessel diameter (Table 3).

At follow-up angiography, despite no significant difference in late loss between the 2 groups, minimal lumen diameter, percentage of diameter stenosis, and net gain were significantly better in the sirolimus-eluting stent group (Table 3 and Figure 2). At the 6-month follow-up, the mean (SD) minimal lumen diameter in the analysis segment was 1.52 (0.63) mm in the vascular brachytherapy group and 1.80 (0.63) mm in the sirolimus-eluting stent group (P<.001) while the net gain was 0.68 (0.60) mm and 1.00 (0.61) mm, respectively (P<.001). For the injured segment, the net gain was 0.96 (0.68) mm in the vascular brachytherapy group and 1.29 (0.70) mm in the sirolimus-eluting stent group (P<.001). The binary angiographic restenosis rate was 29.5% (31/105) for the vascular brachytherapy group and 19.8% (45/227) for the sirolimus-eluting stent group (RR, 1.5; 95% CI, 1.0-2.2; P = .07).

Edge restenosis, defined as greater than 50% narrowing in the 5 mm immediately proximal or distal to the treated region, was apparent in the vascular brachytherapy group, with numerically greater late loss in the proximal edge and significantly greater late loss in the distal edge. This observation is supported by the distinct patterns of restenosis observed in both groups at follow-up (Table 4).²⁶ Lesions treated with vascular brachy therapy demonstrate several morphological characteristics of restenosis predominantly involving the margins of the stent including type 1b (margin restenosis), type 1c (focal restenosis), type 2 (restenosis extending to the stent margin), and type 3 (diffuse restenosis extending outside of the stent margins), while the pattern of restenosis for lesions treated with the sirolimus-eluting stent were predominantly type 1c (21/45 [46.7%]).

There were 4 major in-hospital adverse clinical events in the sirolimus-eluting stent group (Table 5). Major adverse cardiac events in or out of the hospital were markedly different at 270 days (19.2% [24/125] for the vascular brachytherapy group vs 10.0% [26/259] for the sirolimus-eluting stent group; RR, 1.9; 95% CI, 1.1-3.2; P = .02). There were no deaths in the entire trial and only 6 patients had an MI; all 6 patients were in the sirolimus-eluting stent group and 4 occurred while the patients were in the hospital. Of the 4 in-hospital MIs, 1 was symptomatic and the other 3 were serum marker elevations only. Thus, the significant difference was predominantly due to a difference in the rate of target lesion revascularization (19.2% [24/125] in the vascular brachytherapy group vs 8.5% [22/259] in the sirolimus-eluting stent group; RR, 2.3; 95% CI, 1.3-3.9; P = .004). There also was a significant difference in the primary trial end point of target vessel failure (21.6% [27/125] in the vascular brachytherapy group vs 12.4% [32/259] in the sirolimus-eluting stent group; RR, 1.7; 95% CI, 1.1-2.8; P = .02) (Table 5 and Figure 3). Only 2 patients experienced stent thrombosis; both patients were in the sirolimus-eluting stent group and the thrombosis occurred after 30 days (P > .99).

Bayesian regression models were used to analyze the primary outcome (target vessel failure at 270 days). The primary analysis used data from the 3 trials (SISR, GAMMA I,²² and GAMMA II) but excluded minimal luminal diameter. The 95% CIs for the parameter estimates for the sirolimus-eluting stent group are below 0, indicating that the sirolimus-eluting stent has a target vessel failure rate that is significantly less than that of vascular brachytherapy (Table 6). The inclusion of minimal lumen diameter in the model has minimal effect on the results, suggesting that little of the effect was due to the mechanical properties of the stent. The posterior probabilities of both hypotheses (noninferiority and superiority) were both highly significant indicating that the sirolimus-eluting stent is superior to vascular brachytherapy.

To support the conclusion of superiority, a sensitivity analysis was performed. When only the data generated from patients treated in the Sirolimus-Eluting Stent for In-Stent Restenosis (SISR) trial were analyzed, the posterior probabilities were further from 1 but were still significant (at the .01 level for noninferiority and at the .05 level for superiority). Therefore, the results remain the same regardless of whether the historical control trials were used in the analysis.

Predictors of target lesion revascularization to 270 days were assessed in the entire cohort using univariable and multivariable techniques. Nine factors were found to be significant on univariable analysis. However, only 3 of these remained significant with multiple logistic regression analysis: modified American Heart Association and the American College of Cardiology lesion classification score, diameter stenosis postprocedure within the injured segment, and specific treatment group.

The SISR trial enrolled patients with one of the most challenging conditions in interventional cardiology, namely, restenosis following previous implantation of a bare-metal stent. Placement of bare-metal stents is associated with improved acute gain in luminal diameter but neointimal hyperplasia produces in-stent restenosis in 30% to 50% of patients who may be recalcitrant to treatment.^27- 30

Vascular brachytherapy is currently the only approved treatment for in-stent restenosis. β and γ radiation appear to achieve similar reductions in restenosis in comparable lesions.^17- 23 The widespread application of vascular brachytherapy has been limited due to logistic challenges, radiation safety concerns, and evidence of both edge restenosis and late loss of efficacy (late restenosis) after 3 years.^31- 32 Because of the significant reduction in restenosis rates in de novo lesions with drug-eluting stents, there has been an interest in their application for the treatment of in-stent restenosis.

Initial nonrandomized studies of the use of the sirolimus-eluting stent and the paclitaxel-eluting stent for in-stent restenosis appeared favorable.^33- 38 Both angiographic and clinical indices of restenosis compare well with historical controls, reporting 6-month to 12-month in-lesion late loss values of 0.08 to 0.36 mm and binary angiographic restenosis rates of 0% to 9.7% for sirolimus-eluting stents and 0.42 to 0.54 mm and 16% to 20% for paclitaxel-eluting stents.

Superiority of drug-eluting stents over balloon angioplasty was demonstrated in the ISAR-DESIRE (Intracoronary Stenting and Angiographic Results: Drug-eluting Stents for In-Stent Restenosis) study³⁹ in which 300 patients were randomized to balloon angioplasty, the paclitaxel-eluting stent, or the sirolimus-eluting stent. Net gain at 6 months was 0.41 mm for patients randomized to balloon angioplasty, 1.02 mm for patients randomized to the paclitaxel-eluting stent, and 1.12 mm for patients randomized to the sirolimus-eluting stent; 6-month binary angiographic restenosis occurred in 44.6%, 21.7% (P = .001 vs balloon angioplasty), and 14.3% of patients (P<.001 vs balloon angioplasty), respectively. Target vessel revascularization was 33.0%, 19.0%, and 8.0% for the 3 groups, respectively, favoring the sirolimus-eluting stent vs both balloon angioplasty (P<.001) and the paclitaxel-eluting stent (P = .02).

The current trial demonstrates a marked reduction in target vessel failure with the sirolimus-eluting stent driven predominantly by a reduction in the rate of target vessel revascularization. We found similar results using both Bayesian and classic frequentist analytical approaches. Angiographic findings in the current investigation provide a clear mechanistic explanation for the superiority of the sirolimus-eluting stent: (1) acute gain was markedly improved by the use of a stent compared with balloon angioplasty; (2) late loss was similar in both the vascular brachytherapy group and the sirolimus-eluting stent group; and (3) margin restenosis or edge effect was observed on angiography in the vascular brachytherapy group.

No evidence of late catch-up in restenosis has been observed when the sirolimus-eluting stent is used to treat de novo lesions. Longer-term follow-up of patients in this trial is critical to document rates of late thrombosis and whether late catch-up of restenosis occurs as has been observed with long-term follow-up of vascular brachytherapy trials. While the trial was not formally powered for safety outcomes, the safety outcomes and the event rate trends are consistent with the findings in the other trials that studied the sirolimus-eluting stent.^{4- 5,10- 13}

Asymmetric randomization was used to promote enrollment in this study. During the course of the study, enrollment slowed due to a reduction in both the occurrence of in-stent restenosis and a decrease in the number of sites performing vascular brachytherapy. Trial participation bias cannot be controlled for but is thought unlikely due to the similarity of patient clinical and angiographic characteristics with other reported restenosis trials.^33- 38

In conclusion, in-stent restenosis following bare-metal stent placement remains a significant clinical problem. While vascular brachytherapy remains the only approved therapy for this condition, the results of this study indicate that the sirolimus-eluting stent is superior to vascular brachytherapy at 9 months. Angiographic measurements indicate that while both methods are effective at suppressing neointimal hyperplasia, the sirolimus-eluting stent yields greater benefits from acute gain due to the stent component of the device and from the absence of edge restenosis. This study suggests that the sirolimus-eluting stent is a safe and effective treatment for in-stent restenosis occurring within bare-metal stents.