Author + information
- Received August 27, 2007
- Revision received January 10, 2008
- Accepted January 15, 2008
- Published online April 22, 2008.
- Mitchell W. Krucoff, MD, FACC⁎,⁎ (, )
- Dean J. Kereiakes, MD, FACC†,
- John L. Petersen, MD⁎,
- Roxana Mehran, MD‡,
- Vic Hasselblad, PhD⁎,
- Alexandra J. Lansky, MD§,
- Peter J. Fitzgerald, MD, PhD, FACC∥,
- Jyotsna Garg, MS⁎,
- Mark A. Turco, MD¶,
- Charles A. Simonton III, MD, FACC#,
- Stefan Verheye, MD, PhD⁎⁎,
- Christophe L. Dubois, MD††,
- Roger Gammon, MD‡‡,
- Wayne B. Batchelor, MD, MHS§§,
- Charles D. O'Shaughnessy, MD∥∥,
- James B. Hermiller Jr, MD¶¶,
- Joachim Schofer, MD##,
- Maurice Buchbinder, MD, FACC⁎⁎⁎,
- William Wijns, MD, PhD†††,
- COSTAR II Investigators Group
- ↵⁎Reprint requests and correspondence:
Dr. Mitchell W. Krucoff, Professor of Medicine/Cardiology, Duke University Medical Center, 508 Fulton Street, Room A3006, Durham, North Carolina 27705.
Objectives The aim was to compare safety and effectiveness of the CoStar drug-eluting stent (DES) (Conor MedSystems, Menlo Park, California) with those of the Taxus DES (Boston Scientific, Maple Grove, Minnesota) in de novo single- and multivessel percutaneous coronary intervention (PCI).
Background Paclitaxel elution from a stent coated with biostable polymer (Taxus) reduces restenosis after PCI. The CoStar DES is a novel stent with laser-cut reservoirs containing bioresorable polymer loaded to elute 10 μg paclitaxel/30 days.
Methods Patients undergoing PCI for a single target lesion per vessel in up to 3 native epicardial vessels were randomly assigned 3:2 to CoStar or Taxus. Primary end point was 8-month major adverse cardiac events (MACE), defined as adjudicated death, myocardial infarction (MI), or clinically driven target vessel revascularization (TVR). Protocol-specified 9-month angiographic follow-up included 457 vessels in 286 patients.
Results Of the 1,700 patients enrolled, 1,675 (98.5%) were evaluable (CoStar = 989; Taxus = 686), including 1,330 (79%) single-vessel and 345 (21%) multivessel PCI. The MACE rate at 8 months was 11.0% for CoStar versus 6.9% for Taxus (p < 0.005), including adjudicated death (0.5% vs. 0.7%, respectively), MI (3.4% vs. 2.4%, respectively), and TVR (8.1% vs. 4.3%, respectively). Per-vessel 9-month in-segment late loss was 0.49 mm with CoStar and 0.18 mm with Taxus (p < 0.0001). Findings were consistent across pre-specified subgroups.
Conclusions The CoStar DES is not noninferior to the Taxus DES based on per-patient clinical and per-vessel angiographic analyses. The relative benefit of Taxus is primarily attributable to reduction in TVR. Follow-up to 9 months showed no apparent difference in death, MI, or stent thrombosis rates.
Currently available drug-eluting stents (DES) reduce restenosis by inhibiting fibromuscular hyperplasia through targeted delivery of cytostatic drugs, such as sirolimus or paclitaxel, from surface coatings using durable polymers (1,2). Concerns with such first generation DES include infrequent but catastrophic late stent thrombosis (3–5). Although the cause of late stent thrombosis is likely multifactorial, durable polymer surface coatings may play a role (6–8).
The CoStar (Conor MedSystems, Menlo Park, California) stent is a novel cobalt chromium alloy DES platform designed to elute paclitaxel without the use of a surface polymer coating via multiple laser-cut reservoirs within the stent struts, which are filled with a bioresorbable poly–lactic-co-glycolic acid (PLGA) polymer. After drug delivery and subsequent complete polymer bioresorption, only the biologically inert bare-metal platform remains (Fig. 1). Reduction of tissue exposure to polymer on stent implantation and elimination of long-term (>6 months) polymer exposure compared with durable polymer surface-coated stents, such as Taxus (Boston Scientific, Maple Grove, Minnesota), in theory might favorably influence both short- and long-term inflammatory and thrombogenic events. Drug dose, direction (luminal vs. abluminal), and kinetics of delivery are varied by drug-polymer mixing on a reservoir-by-reservoir basis. The precision of these variations in drug delivery has been shown to reach biologically meaningful proportions in several independent human trials, from which the 10 μg/30 day drug–PLGA reservoir load was selected based on the associated angiographic late lumen loss at 4 to 12 months (9–11). Such precision in a lower range of paclitaxel dosing in theory might provide similar efficacy but greater safety margin in settings such as provisional stent overlap compared with the higher dose delivered with the Taxus stent.
The COSTAR (Cobalt Chromium Stent With Antiproliferative for Restenosis) II study was designed to compare the 8-month clinical outcomes of patients with both single- and multivessel coronary stenoses undergoing elective percutaneous coronary intervention (PCI) with either the CoStar or the Taxus DES. In addition, the biological behavior of vessels treated with each of these stents was examined in smaller cohorts, in whom protocol-specified 9-month angiography was performed.
Study design and population
The COSTAR II study design has been previously described in detail (12). In summary, COSTAR II was a multicenter, prospective, single-blind, 3:2 (CoStar:Taxus) randomized study testing noninferiority of the CoStar versus the Taxus paclitaxel DES. In total, 1,700 patients were enrolled from 71 sites in the U.S., Germany, Belgium, and New Zealand. Anatomic eligibility required a de novo single target lesion of >50% but <99% stenosis in 1, 2, or 3 native epicardial coronary arteries, with a reference vessel diameter from 2.5 to 3.5 mm and a lesion length of <30 mm that could be covered with a single stent. Anatomic exclusion criteria included ostial lesions, left main coronary disease of >50% stenosis, bifurcation lesions with >2 mm side branch involvement, total occlusions, presence of a previously implanted DES proximal to target lesion site, or >50% stenosis elsewhere in the target vessel. Clinical exclusion criteria included myocardial infarction (MI) within 72 h, prior revascularization within 3 months, intolerance of clopidogrel or aspirin, known bleeding diathesis, renal failure, left ventricular ejection fraction <25%, and contraindication to coronary artery bypass surgery. Approval by each participating institution's ethics committee and informed consent from all patients were required and obtained.
Pre-specified subgroup analyses included patients with multivessel disease, patients with diabetes, and patients requiring the provisional use of overlapping stents.
Repeat cardiac catheterization was planned 9 months after the index procedure in the first 250 patients enrolled with multivessel coronary artery disease and the first 100 patients with single-vessel disease. Of these 100 single-vessel patients, the first 70 also had planned protocol-specified intravascular ultrasound (IVUS) evaluation at the 9-month catheterization. In addition, all patients with (provisional) overlapping stents also underwent repeat catheterization and IVUS at 9 months. Serial blood sampling for pK analysis was performed in 45 patients, including 16 patients with multivessel stenting. A hemoglobin A1c level was obtained at baseline in all patients.
The CoStar stent is a 0.0035-inch thick cobalt chromium alloy metal platform with nondeformable strut segments that contain multiple laser-cut holes (12). Each hole is individually filled with a bioabsorbable PLGA polymer matrix combined with paclitaxel, creating discrete reservoirs of drug elution. Abluminal and/or endoluminal direction, total drug dose, and kinetics of drug delivery are controlled by programmed alterations in the ratio of bioabsorbable polymer to drug for each reservoir. The PLGA polymer resorption is complete in the porcine model by 180 days, thus leaving only the bare-metal platform in perpetuity. Based on serial comparative human studies of paclitaxel dose, direction, and kinetics of delivery (9–11), the CoStar stent selected for the COSTAR II study was the 10 μg/30 day elution drug-polymer formulation.
Patients were randomly assigned 3:2 to CoStar or Taxus using an interactive voice randomization system. Randomization was stratified by single- or multivessel status. Patients were blinded to treatment assignment until after completion of 1 year of follow-up.
Antiplatelet therapy with a minimum of 325 mg aspirin and a loading dose of at least 300 mg clopidogrel, and intracoronary nitroglycerine for vessel sizing before stent implantation were required. Pre-dilatation of all lesions was also required. Planned use of a nonballoon device (rotational or directional atherectomy, laser, or any unapproved technology) was prohibited. Selection of U.S. Food and Drug Administration (FDA)-approved anticoagulant therapy and use of adjunctive glycoprotein IIb/IIIa inhibitors were at each operator's discretion. High-pressure post-dilatation was recommended but not required. Intravascular ultrasound other than the protocol-mandated use described in the preceding was at the discretion of the operator.
Dual antiplatelet therapy
Continuation of 325 mg aspirin and 75 mg clopidogrel daily for at least 6 months was required by protocol. For patients requiring warfarin sodium therapy, an aspirin dose of 81 mg daily was recommended. Decisions on interruption of dual antiplatelet therapy in case of bleeding or urgent surgery or on extension of clopidogrel therapy beyond 6 months were all managed clinically. Details of all such decisions were systematically captured and collected by the data coordinating center.
Pre-specified clinical follow-up included at index hospital discharge, 30 days, 8 months, and 1 year. Patients with protocol-driven angiographic follow-up for any reason were required to have their 8-month clinical evaluation completed before their 9-month angiogram or IVUS.
All clinical data were double data entered into a Clintrials database at a central facility (Duke Clinical Research Institute, Duke University Medical Center, Durham, North Carolina).
Clinical events committee
Events independently and blindly reviewed by the clinical event committee (CEC) (Duke Clinical Research Institute, Duke University Medical Center) included major adverse cardiac events (MACE), stent thrombosis, and total occlusion. In conjunction with FDA approval of COSTAR II as an investigational device exemption protocol, the definition of stent thrombosis was taken from the Taxus IV study (2), as detailed subsequently. The entire protocol was completed before the later publication of the consensus Academic Research Consortium definition (13) for stent thrombosis.
Angiography Core Laboratory
All angiograms were analyzed by an independent core laboratory (Quantitative Coronary Angiography [QCA] Core Laboratory, Cardiology Research Foundation, New York, New York). Angiograms were acquired using standardized instructions. The QCA was performed using the Cardiovascular Measurement System-Gradient Field Transform algorithm (Medis, Leiden, the Netherlands) (2). The minimum lumen diameter and the mean reference diameter, obtained from averaging 5-mm segments proximal and distal to the target lesion location, were used to calculate the diameter stenosis (diameter stenosis = [1 − minimum lumen diameter/reference diameter] × 100). Acute gain was the change in minimum lumen diameter from baseline to final post-stent implantation angiogram; late loss was the change in minimum lumen diameter from the final post-stent implantation angiogram to follow-up. Binary restenosis was defined as >50% diameter stenosis in the index vessel. All quantitative measurements were performed: 1) within the stented segment (in-stent); 2) in-segment, spanning the stented segment plus the 5 mm proximal and distal peristent areas; and 3) in the 5 mm proximal and distal peristent areas immediately adjacent to the stent.
Ivus Core Laboratory
Blinded IVUS analysis of 70 subjects post-procedure and at 9 months was done by an independent core laboratory (Cardiovascular Core Analysis Laboratory, Stanford University Medical Center, Stanford, California). All images were reviewed by 2 independent observers, and adjudication of opinion was based on consensus of these observers. Volumetric measurements were performed using planimetry software (echoPlaque, Indec Systems, Santa Clara, California) as previously described and included persistent plaque volume, neointimal volume, percent neointimal obstruction, neointimal coverage, and neointimal thickness (14). Incomplete stent apposition was defined as 1 or more struts clearly separated from the vessel wall with evidence of blood speckles behind the strut.
Electrocardiography Core Laboratory
Standard 12-lead electrocardiograms (ECGs) were collected before PCI, before hospital discharge, and at 8-month follow-up and forwarded to an independent core laboratory (ECG Core Laboratory, Duke Clinical Research Institute, Duke University Medical Center, Durham, North Carolina). The ECGs were analyzed for new pathological Q waves by an experienced cardiologist blinded to treatment assignment.
The primary end point of the COSTAR II study was 8-month MACE, defined as a hierarchical composite of cardiac or unknown cause of death, Q-wave and non–Q-wave MI, and clinically driven target vessel revascularization (TVR). This time point was selected specifically to allow completion of all clinical outcomes reporting before any protocol (9 month) catheterizations. Q-wave MI was defined as: 1) clinical presentation with signs or symptoms of MI with new pathological Q waves as determined by the ECG core lab or independent review of the CEC in the absence of timely cardiac enzyme data; or 2) new pathological Q waves as determined by the ECG core lab or independent review of the CEC and elevation of cardiac enzymes. Non–Q-wave MI was defined as elevated creatine kinase (CK) ≥2 times the upper limit of normal with elevated CK-MB in the absence of any new pathological Q waves. “Clinically driven TVR” was defined as a revascularization of the target vessel with: 1) anginal symptoms and/or functional ischemia with a ≥50% stenosis by core lab QCA; or 2) revascularization of ≥70% stenosis by the core lab QCA. All deaths and MI events were counted as MACE events unless the CEC unequivocally attributed them to either a nontarget vessel or noncardiac cause. The primary angiographic end point was per-vessel 9-month angiographic in-segment late lumen loss.
Secondary end points included the individual MACE components, 30-day and 12-month MACE, clinically driven target lesion revascularization, and target vessel failure, as were reported in the pivotal Taxus IV study (2). Secondary technical end points included device success (final stenosis of <50% using the assigned device only), lesion success (final stenosis of <50% using any device), and procedure success (final stenosis of <50% with no procedure-related MACE). Stent thrombosis was categorized as acute (before leaving the catheterization laboratory), subacute (after the index procedure and within 30 days), and late (after 30 days). Acute stent thrombosis was defined as abrupt vessel closure of the treatment site resulting in clinical manifestations of ischemia and angiographic evidence of occlusion or flow-limiting thrombosis in a treated vessel, in which the investigational device was successfully implanted, that occurred after the procedure but before the patient left the catheterization laboratory. Subacute stent thrombosis was defined as abrupt vessel closure of the treatment site producing clinical manifestations of ischemia and occlusion occurring after the patient left the catheterization laboratory but within 30 days of the interventional procedure. Late stent thrombosis was defined as MI attributable to the target vessel, with angiographic documentation of thrombus or total occlusion at the target lesion >30 days following successful implantation of the device.
The primary end point (8-month MACE) analysis (12) was a noninferiority analysis using either the relative risk or the absolute difference in rates between the CoStar stent rates and the Taxus stent rates, depending on the actual MACE rate observed in the Taxus (control) arm. If the actual MACE rate in the Taxus arm was ≥10%, then the observed relative risk was to be compared with a relative delta of 1.5, calculated using statistical software (Proc Genmod, SAS, Cary, North Carolina) using a log link function. If the actual 8-month MACE rate in the Taxus arm was <10%, then an absolute delta of 5% was to be used, calculated using software (StatXact, Cytel, Cambridge, Massachusetts) to provide asymptotic and exact confidence intervals for a difference in proportions.
In conjunction with discussions with the FDA for this pivotal study, a series of consistency analyses were pre-specified to provide “reasonable assurance” of safety and effectiveness of the CoStar stent. As has previously been detailed (12), these consistency analyses were additive and based on confidence intervals specifically calculated with regard to the denominator of each subgroup (e.g., angiographic cohort, single-vessel cohort), not as serial comparisons per se. Thus, if 8-month MACE rates from the primary end point analysis met the boundaries specified, then 2 consistency analyses were to be computed to confirm noninferiority using the pre-specified confidence intervals. The 9-month angiographic in-segment late loss was to be analyzed on a per-vessel basis using the protocol-specified angiographic cohort. An absolute difference in 9-month late loss between the Taxus group and the CoStar group was to be estimated. A 95% confidence interval was to be constructed around this estimate using a statistical analysis procedure (Proc Mixed, SAS) with the repeated statement. If this confidence interval fell completely below 0.32 mm, then noninferiority of the CoStar stent relative to the Taxus stent would be confirmed by the late loss observations. If the confidence interval overlapped or fell completely above 0.32 mm then noninferiority would not be confirmed.
Finally, if the preceding 2 conditions were met, then a third confirmatory test for angiographic late loss was to be performed using only the single-vessel stratum of angiographic cohort patients, using the same analytic methods as the preceding but with the relative delta set at 1.65 and the absolute delta set at 7% to accommodate the smaller denominator of patients.
If all 3 conditions described were met (per-patient 8-month MACE, per-vessel late lumen loss in entire angiographic cohort, and per-vessel late lumen loss in single-vessel angiographic cohort), then the CoStar device was to be declared to be noninferior to the Taxus device.
In the absence of any previous trial comparing the CoStar stent with an FDA-approved bare-metal stent, a final step was included in the primary statistical analysis plan: an imputed placebo calculation (12). For this analysis, the primary end point was estimated for a bare-metal stent against the experimental CoStar DES in the single-vessel disease population, because bare-metal stent data versus Taxus DES was only available for single-vessel disease patients at the time COSTAR II was designed. This analysis used the formulas and logic previously detailed by Hasselblad and Kong (15). The hypothesis was tested in the relative risk space (CoStar vs. placebo) regardless of the COSTAR II Taxus arm outcome rate. The estimate and variance used for the effect of standard treatment relative to placebo was taken from previously published pooled data for the Taxus IV (2), V (16), and VI (17) studies. If the confidence interval for the estimate of experimental treatment relative to placebo was completely to the left of 1.0, it would be taken to indicate that the CoStar stent was better than a bare-metal stent would have been.
Continuous data are presented as means, and categorical variables are presented as percentages, unless otherwise noted. Selected baseline characteristics and clinical and angiographic outcomes are compared between treatment groups by the chi-square test in case of discrete variables and t test in case of continuous variables. A p value of <0.05 was considered to be statistically significant. No statistical adjustment was made for multiple comparisons. All analyses were done using SAS version 8.0 or higher statistical software.
Patient and lesion characteristics
Of the 1,700 patients randomly assigned, 1,675 (98.5%) were evaluable and constituted the primary study population (Fig. 2). The 25 derandomized patients were excluded because: 1) angiographic inclusion criteria could not be confirmed at the time of the index PCI procedure; or 2) the study-assigned stent was never removed from the packaging or advanced beyond the guide catheter. Of the 1,675 patients, 989 were randomly assigned to receive treatment with the CoStar DES and 686 with the Taxus DES in the asymmetric 3:2 design. Clinical descriptors and target lesion characteristics were evenly distributed across treatment groups (Table 1). Of the 1,675 patients, 1,330 (79%) had single-vessel intervention and 345 (21%) had multivessel intervention. In total, 2,058 target lesions were treated, 1,212 (59%) with CoStar and 846 (41%) with Taxus. Procedural glycoprotein IIb/IIIa inhibitors were used in 20.5% of all patients (20.2% for CoStar, 20.9% for Taxus).
On a per-vessel analysis, device success was analyzable in 2,049 (99.6%) of 2,058 lesions. Of these 2,049, device success was achieved in 1,173 (97.3%) of 1,205 lesions with CoStar (single-vessel 794 of 813 [97.6%], multivessel 379 of 392 [96.7%]) and in 825 (97.7%) of 844 lesions with Taxus (single-vessel 539 of 552 [97.6%], multivessel 286 of 292 [97.9%]), and lesion success was achieved in 100% and 99.9%, respectively. On a per-patient analysis, procedural success was achieved in 957 (97.4%) of 983 patients with CoStar (single-vessel 770 of 786 [97.9%], multivessel 187 of 197 [94.9%]), and in 672 (98.2%) of 684 patients with Taxus (single-vessel 532 of 539 [98.7%], multivessel 140 of 145 [96.6%]). None of these differences were significant.
Periprocedural/index hospitalization hierarchical MACE was 2.6% with CoStar (0% death, 2.4% MI, 0.2% clinically driven TVR) and 1.6% with Taxus (0% death, 1.6% MI, 0% clinically driven TVR); p = 0.160. The rate of MACE at 30 days was 3.4% with CoStar (0% death, 2.8% MI, 1.0% clinically driven TVR) and 1.9% with Taxus (0% death, 1.6% MI, 0.2% clinically driven TVR); p = 0.063. In patients with single-vessel disease, periprocedural/index hospitalization hierarchical MACE was 2.0% with CoStar (0% death, 1.9% MI, 0.1% clinically driven TVR) and 1.3% with Taxus (0% death, 1.3% MI, 0% clinically driven TVR); p = 0.3131. The rate of MACE at 30 days was 3.1% with CoStar (0% death, 2.4% MI, 1.0% clinically driven TVR) and 1.5% with Taxus (0% death, 1.3% MI, 0.2% clinically driven TVR); p = 0.0711. In patients with multivessel disease, periprocedural/index hospitalization hierarchical MACE was 5.0% with CoStar (0% death, 4.5% MI, 1.0% clinically driven TVR) and 2.8% with Taxus (0% death, 2.8% MI, 0% clinically driven TVR); p = 0.2977. The rate of MACE at 30 days was 5.0% with CoStar (0% death, 4.5% MI, 1.0% clinically driven TVR) and 3.5% with Taxus (0% death, 2.8% MI, 0.7% clinically driven TVR); p = 0.4794. Table 2 shows MACE at 8 months, the primary study end point, and each of its components. The rate of MACE was significantly lower with Taxus than with CoStar, a difference due predominantly to the difference in clinically driven TVR rates. At 8 months, both MI and mortality rates were not significantly different. The MACE to 8 months is shown discretely for both the single- and the multivessel subgroups (Table 2). The absolute difference between observed event rates between Taxus and Costar groups with multivessel PCI is numerically larger than in the single-vessel cohort, but with the smaller number of patients does not reach statistical significance.
The temporal distribution of hierarchical MACE and its component events are shown in Kaplan-Meier curves to completed 9-month follow-up in Figure 3. The impact of 513 protocol catheterizations (100 single-vessel, 250 multivessel, 168 patients with overlapped stents [60 with Taxus, 108 with CoStar]) on TVR events from 8- to 9-month follow-up can be discretely appreciated. In this 1-month interval, additional MACE events produced an absolute increase in event rates of 2.4% in Taxus patients and 3.7% in CoStar patients, numerically increasing the difference between the 2 cohorts. The Kaplan-Meier curves of death and Q-wave MI show no significant differences at any time, whereas the clinically driven TVR curves begin to separate between 30 and 90 days after index PCI.
Odds ratios of 8-month hierarchical MACE are shown for all pre-specified subgroups in Figure 4. Differences do not reach statistical significance in many of the subgroups owing to small denominators; however, numerical differences and trends consistently favor the Taxus treatment group.
Per-vessel angiographic outcomes
Angiographic results at 9 months from 456 lesions in 286 patients (262 lesions CoStar, 194 lesions Taxus) are shown in Table 2. Per-vessel in-stent and in-segment late loss, diameter stenosis, and binary restenosis were all significantly better in the Taxus group compared with the CoStar group.
Both overall and time-related incidence of protocol-defined stent thrombosis are shown in Table 3. The 1 patient with acute thrombosis had coronary perforation during the index procedure, which was treated with a stent graft (Jomed International, Helsingborg, Sweden). Recurrent instability while still in hospital led to recatheterization, which showed thrombus on the stent graft that was treated medically without further complications. Of the 4 CoStar patients with subacute stent thromboses, 1 (thrombosis on day 3) was found to be dual antiplatelet therapy resistant, 2 (thromboses on days 6 and 9) were clopidogrel noncompliant, and antiplatelet therapy compliance was indeterminate in 1 patient. Both patients who suffered late thrombosis did so within days of stopping their clopidogrel (day 177 for CoStar, day 232 for Taxus). All stent thromboses were associated with MI, none with death.
The relative risk of 8-month MACE outcomes with the CoStar stent versus the imputed placebo of the Express bare-metal stent is shown in Figure 5. Confidence intervals of CoStar 8-month MACE cross the unity line, suggesting that use of the CoStar stent produced no significant difference from the imputed outcomes with a bare-metal stent placebo.
In this prospective, randomized, multicenter study of 2 paclitaxel-eluting stent platforms, per-patient primary clinical outcomes at 8 months and per-vessel angiographic end points at 9 months demonstrated significant differences between the Taxus DES versus the CoStar DES, with the result that it cannot be concluded that CoStar DES is noninferior to Taxus DES. Divergence between the clinical event curves begins around 3 months, is numerically exaggerated by oculostenotic events at 9 months, and is predominantly driven by TVR (MI and death rates remain essentially the same between devices over that time). This strongly suggests the mechanistic hypothesis that fibrointimal hyperplasia resulting in in-stent restenosis is greater with the CoStar stent than with the Taxus stent. The per-vessel angiographic findings at 9 months also support this mechanistic explanation. Finally, the imputed placebo calculation implies that the CoStar stent is not superior to bare-metal stents, which have been shown to have significantly higher in-stent restenosis rates than the Taxus stent in randomized controlled clinical trials (2).
Secondary analyses of multiple subgroups, including the 3 pre-specified groups (single- vs. multivessel disease, diabetes, and patients with overlapping stents), were highly consistent with the primary study results. No significant increase in the risk of death, MI, or stent thrombosis was appreciated in any subgroup.
The CoStar DES incorporates a number of unique features with theoretic appeal compared with the Taxus stent (12). Whereas the Taxus DES was developed by applying a surface coating of durable polymer to the bare-metal Express platform, the CoStar stent was designed specifically for drug delivery via individual wells filled with drug and polymer (12). Initial studies that evaluated the optimal drug load and kinetics of paclitaxel delivery using the CoStar platform were encouraging. The PISCES (Paclitaxel In-Stent Controlled Elution Study) study examined 6 different drug, direction, and duration of elution configurations, and the COSTAR I study independently evaluated dose optimization using 3 configurations (9,10). Both early (4 months) and late (12 months) angiographic surrogate measures suggested that the abluminal unidirectional 10 μg/30 day and 30 μg/30 day stent loads effectively preserved in-segment and in-stent late lumen loss (9,10). These 2 formulations were then tested head to head in the EUROSTAR (European Cobalt Stent With Antiproliferative for Restenosis) trial, where the 10 μg/30 day platform was found to result in lower levels of late lumen loss in human subjects (11). Based on this series of human dose-finding investigations, the 10 μg/30 day load was selected for the pivotal COSTAR II study.
Despite theoretic design novelty and the consistent legacy of previous small trial performance with 4- to 12-month in-segment late loss values of <0.3 mm with the 10 μg/30 day formulation, the results of the pivotal COSTAR II trial serve as a reminder of the importance of adequately sized and well designed trials conducted in contemporary patient cohorts (9–11). Notably, in COSTAR II, the CoStar stent performed worse than in preceding trials and the Taxus stent performed better than in its pivotal study (Taxus IV), despite the more complex patient cohort in COSTAR II. Potential explanations for under- or over-performance are speculative, but include small previous study size with wide confidence intervals, differences in device manufacturing process (CoStar), and operator familiarity and learning curve (Taxus).
The COSTAR II study represents the actual implementation of a different set of key principles outlined in the early anticipation of the challenges of active-control studies of DES. Instead of emphasizing the use of surrogate angiographic measures in simple patients, COSTAR II enrolled a more complex (“enriched”) patient population in whom a higher density of clinical events would be expected (18). This strategy supports a logistically feasible clinical trial that also provides pre-market data with greater relevance to real-world post-market use (18). Furthermore, by enriching the angiographic cohort with multivessel patients, the ethical and logistic issues associated with mandated repeat catheterization were minimized, which is a key feature when patients have the alternative of receiving an approved DES without participation in the research study. Indeed, the COSTAR II trial has the smallest percentage of patients subjected to protocol recatheterization of any published pivotal DES study. In addition, by completing the clinical outcome evaluation 30 days before protocol-specified catheterization, the impact of oculostenotic events could be discretely appreciated. A number of these study design features are among recommendations for future DES pivotal trials (19).
The statistical analysis of an enriched population in an active control noninferiority study design is unique as well. At the time COSTAR II was being planned, multivessel stenting with the Taxus DES was widely practiced in post-market use, but it was not an approved indication for the device and had not been reported in a randomized clinical trial. Thus, the COSTAR II control group also represented an “exploratory” population (e.g., one without a clearly definable predicate) (12). The statistical analysis plan developed a noninferiority delta as a relative risk to the actual observed event rate in the control group over a range of possibilities and supported the relative risk delta with an imputed placebo calculation to ensure that the confidence intervals at the high end of the delta would not reach the lowest event rates expected from bare-metal stents.
There are a number of limitations worth noting from this report. Initially, the hope was to enroll 50% multivessel patients; however, this proved to be unfeasible. Nonetheless, the 20% enrolled represent the first advance to more “enriched” populations in pivotal DES studies. In addition to the multivessel group, there are many other key subgroups that remain relatively underpowered for noninferiority comparisons, despite the fact that COSTAR II is larger than any previously reported pivotal DES trial. The IVUS findings in the very small number of single-vessel and overlap cases acquired are not included in the present report, and they might offer further mechanistic insights into the behavior of both CoStar and Taxus stents in these populations. However, the clinical and angiographic findings are all so consistent that it is unlikely that IVUS would affect any of the conclusions from this primary study report.
Although clearly not noninferior, safety issues between Taxus and the CoStar DES platforms remain of interest. By 9 months of follow-up no differences in safety parameters were observed between devices. The COSTAR II study was designed before widespread awareness of infrequent but catastrophic events, such as late stent thrombosis, and is not powered for such comparisons, particularly within such a short follow-up period. Further insights may be gained when longer-term follow-up is completed and integrated with the detail of actual duration of dual antiplatelet therapy, therapy interruption, and so on. Finally, although the definition of stent thrombosis used in COSTAR II was developed primarily for comparability with Taxus IV, the subsequent emergence of the Academic Research Consortium stent thrombosis definitions may make readjudication of these events useful to support comparability with other DES platform experiences (13).
The COSTAR II study demonstrates that it cannot be concluded that the CoStar DES is noninferior in clinical and angiographic performance compared with the Taxus DES. The relative benefit attributable to the Taxus stent is predominantly due to lower rates of clinically driven TVR, with no differences observed in the incidences of death, MI, or stent thrombosis by the end of 9 months' follow-up.
Duke University received significant grant revenue from Conor MedSystems as the enrolling site; none of this revenue went to Dr. Krucoff or his salary. Duke Clinical Research Institute received grant money for clinical trial operations. Duke Clinical Research Institute also receives significant grant support from competitive companies including Terumo, Cordis Johnson & Johnson, Abbott, Medtronic, and Boston Scientific. Dr. Krucoff has received consulting fees and honoraria (modest) from Conor MedSystems, Cordis Johnson & Johnson, Abbott, Medtronic, OrbusNeich, Affinergy, Biosensors, and Terumo. Dr. Kereiakes has received research grants (modest) from Conor MedSystems, Pfizer, Cordis Johnson & Johnson, Boston Scientific, Medtronic, and Daiichi-Sankyo; received consulting fees (modest) from Conor MedSystems, Cordis Johnson & Johnson, Core Valve, Eli Lilly & Co., Boston Scientific, and Abbott Bioadsorbable Vascular Solutions; and served on the Speakers' Bureau of Eli Lilly & Co. Dr. Peterson has received a significant research grant from Conor MedSystems. Dr. Mehran has received modest research grants from Tyco Mallinkradt, Geurbet, and Cordis Johnson & Johnson and served on the Speakers' Bureau (modest) of The Medicines Company, Cordis Johnson & Johnson, and Boston Scientific. Dr. Fitzgerald received significant consulting fees from Conor MedSystems and Cordis Johnson & Johnson and honoraria from Cordis Johnson & Johnson. Dr. Turco has served as a consultant and on the Speakers' Bureau of Boston Scientific, Cordis, Medtronic, and Abbott Vascular; he has received research grants from Boston Scientific, Cordis, Medtronic, Abbott Vascular, Sanofi-Aventis, Lumen Biomedical, and EV3. Dr. Simonton is an employee of Abbott Vascular. Dr. Dubois served on the advisory board of Boston Scientific and received a research grant from Conor MedSystems. Dr. Batchelor received consulting fees from Conor MedSystems (significant) and Boston Scientific (modest). Dr. Hermiller has served as a consultant (minor) for Abbott, BSC, and St. Jude. Dr. Buchbinder has equity interest and is a stockholder in Conor MedSystems and served on the Scientific Advisory Board of BSC, Cordis, and Abbott. Dr. Wijns received research grants from Bristol-Myers Squibb, GlaxoSmithKline, and Therabel.
For trial structure and participating sites, please see the online version of this article.
Supported by a research grant from Conor MedSystems. For full author disclosures, please see the end of this paper.
- Abbreviations and Acronyms
- clinical events committee
- drug-eluting stent(s)
- Food and Drug Administration
- intravascular ultrasound
- major adverse cardiac events
- myocardial infarction
- percutaneous coronary intervention
- poly–lactic-co-glycolic acid
- quantitative coronary angiography
- target vessel revascularization
- Received August 27, 2007.
- Revision received January 10, 2008.
- Accepted January 15, 2008.
- American College of Cardiology Foundation
- Camenzind E.,
- Steg P.G.,
- Wijns W.
- Laskey W.K.,
- Yancy C.W.,
- Maisel W.H.
- Joner M.,
- Finn A.V.,
- Farb A.,
- et al.
- Finn A.V.,
- Nakazawa G.,
- Joner M.,
- et al.
- Serruys P.W.,
- Sianos G.,
- Abizaid A.,
- et al.
- Conor MedSystems, LLC. COSTAR 1: The COSTAR INDIA Trial—Cobalt Chromium Stent With Antiproliferative for Restenosis in India trial. Corporate communication. Paper presented at: Transcatheter Therapeutics; October 21, 2005; Washington, DC.
- Cutlip D.E.,
- Windecker S.,
- Mehran R.,
- et al.
- Kataoka T.,
- Grube E.,
- Honda Y.,
- et al.
- Dawkins K.D.,
- Grube E.,
- Guagliumi G.,
- TAXUS VI Investigators
- Kereiakes D.J.,
- Kuntz R.E.,
- Mauri L.,
- Krucoff M.W.
- Krucoff M.W.,
- Boam A.,
- Schultz D.G.