CAPTURE in the Context of Space and EVA-3S
William A. Gray, MD, discusses what lessons can be learned from large-population, high-enrollment, postapproval registries.
Endovascular Today: What is the CAPTURE registry? Why was it started?
Dr. Gray: The CAPTURE registry was a condition of approval that the FDA mandated when Guidant (now Abbott) received its approval for the Accunet and Acculink stent systems. It is a typical postmarket surveillance trial, and there were several goals. One goal was to survey for rare and unanticipated device-related events. The pivotal trials that led to device approval (in this case, ARCHeR) would not have identified these rare events, which occur with a less than 1% incidence. The second goal was to examine the transfer of the technology from the trial setting into the community; this registry was a measure of how well physicians can be selected and trained to do the procedure. Lastly, the original CAPTURE registry was meant to be a 1,500-patient registry but was extended at the request of Guidant/Abbott and approved by the FDA for an extension to allow for data gathering and further analysis of carotid stenting in the community at large.
Before the CAPTURE dataset became available, we really had very small data cohorts; 300 to 500 patient pivotal trials really do not allow for any significant subset analysis, but this 3,500 patient registry has allowed for that. The enrollment is now well over 3,500 patients, but we were able to look at the data and analyze these 3,500.
Endovascular Today: Of those 3,500 patients, is there an argument that these are cherry-picked patients, or are physicians required to submit all of their patient data to the registry?
Dr. Gray: Basically the sites and investigators who were performing carotid stenting and involved in the CAPTURE registry were asked to enroll all their patients undergoing carotid stenting with this embolic protection stent system. That does not mean that all patients in the US were in CAPTURE; actually, based on estimates of the US activity of CAS, it represents only a fraction of the total. To the extent that the physicians were contributing and participating from a research perspective, they were encouraged to enroll all of their patients. I should also say that it was encouraged that patients be enrolled according to the IFU so that enrolled patients should have followed FDA approval language, which means they should have been high-surgical-risk for a CEA due to either prespecified, asymptomatic with stenosis >80%, symptomatic with stenosis >50%, and so on.
Endovascular Today: So the CAPTURE registry also includes the subset of patients that was not covered by the CMS reimbursement?
Dr. Gray: That is correct, and this is one of the reasons the CAPTURE demographics become somewhat self-fulfilling from the symptomatic cohort standpoint. We only had 14% symptomatic patients in this study, largely because symptomatic patients were treated on-label and were not required to be in the study, but the CAPTURE registry and other postmarket surveillance registries have provided some of those asymptomatic patients continued access to this therapy in the face of a CMS noncoverage decision in this group.
Endovascular Today: There are different definitions for the major adverse events when dealing with carotid studies; does the inclusion of any stroke or MI also impact those results?
Dr. Gray: That is a great question. Clearly the inclusion of MI and all stroke (not just ipsilateral but contralateral) affects the outcomes. You will see in subsequent publications that there is actually a 1% rate of contralateral stroke or nonipsilateral stroke, which occurs in this and most other surveys. In other trials that do not include all-stroke and only ipsilateral stroke, they are missing some of the strokes that occur in carotid stenting and will, by comparison, have better outcomes. The issue comparing outcomes across trials is a complex one given the differences in populations studied, that is, inclusion and exclusion criteria, definitions used for adverse outcomes, primary endpoint definitions (the aforementioned stroke example is a good one), and event adjudication through the clinical events committee process. I am coming to the conclusion that the best way to compare outcomes across trials is by using major stroke and death, which are not subjective measures.
Endovascular Today: Is there an argument to be made that those strokes were wholly unrelated to the procedure taking place?
Dr. Gray: That argument is probably not valid; the nonipsilateral stroke should be included in order to assess the risk of the entire procedure. Interestingly, if one looks at the overall procedure, it appears that the periprocedural risk of contralateral stroke is the same, roughly 1%, regardless of whether the patient is over 80 years of age, symptomatic or asymptomatic, or the operator's expertise–all of those features that you would think predict the rate of nonipsilateral stroke actually do not. There is something happening from a procedural standpoint that we need to learn more about, but it clearly does make a difference. Obviously, MI will contribute to the total number of events at 30 days, but more importantly, the process of adjudication–which is having a clinical events committee looking at events–seems to increase the total number of events at 30 days because we are forced to define events more clearly.
Endovascular Today: In order to compare apples to apples, can we compare SAPPHIRE with the EVA-3S and SPACE studies?
Dr. Gray: No, because they compare differing patient cohorts. EVA-3S, SPACE, and SAPPHIRE were all randomized. SAPPHIRE is a high-surgical-risk trial in which only about 25% of the patients were symptomatic; the patients in EVA-3S and SPACE were all symptomatic and were not defined as high-surgical-risk patients. EVA-3S and SPACE are reasonably comparable to each other because they seem to study the same patient population: symptomatic, reasonable surgical risk; but they are not comparable with each other in terms of operator expertise, which was poor for the stenting arm of the EVA-3S.
Endovascular Today: Why was there such a large divergence of adverse events between the CAS and CEA arms in the EVA-3S study?
Dr. Gray: EVA-3S had the lowest rate of complications with CEA ever reported in a symptomatic patient population; it also had one of the highest rates of complications ever reported in the stenting arm of a carotid trial. There are several reasons why we believe that occurred. First, operator experience was quite limited in the EVA-3S stent arm, whereas the experience level in the surgical arm was quite good–enrolling surgeons had to have performed 25 or more CEAs per year. In the CAS arm, operators could have done as few as five carotid stents or as many as 12, with no reports as to whether they used embolic protection, the particular technique, or the outcomes. There was no central organization within that trial that vetted operators for the CAS arm. It was possible to be an EVA-3S CAS investigator, having never done a carotid stent, and performing your first randomized case being proctored by a physician who had done five carotid stents, who may or may not have used embolic protection before. The results of that first case are included in the pivotal data. This obviously represents a learning curve function reminiscent of the original Wallstent trial, which had very similar results at 1 year with a 12% complication rate in the stent arm and a 4% to 5% complication rate in the surgical arm, largely the result of a lack of trained operators, by today's standards. The inclusion of nontrained or poorly trained operators is a major criticism of EVA-3S.
Endovascular Today: CAPTURE showed that there was little difference between the results obtained by operators at all levels of experience. What accounts for the similarity of outcomes in CAPTURE versus what we are seeing with EVA-3S?
Dr. Gray: There are selection criteria for operators that were followed in CAPTURE: participants had to be peripheral interventionists of good standing who had previously performed a certain number of cerebral angiograms and rapid-exchange procedures, and had experience with .014-inch-based systems. There was no indication that a selection of operators like that occurred with EVA-3S. There was no mention of mandated training in EVA-3S, just the proctorship. In CAPTURE, operators with no prior experience underwent a 2-day intensive training course that included wet lab models and simulation training, and they were offered proctors for their first several cases. There was a careful rollout of selected operators for this technology, with training geared toward previous operator experience.
Endovascular Today: Outside of any training or experience required in some of the studies versus EVA-3S, what other factors would account for the differences between the CAS and CEA arms?
Dr. Gray: There was initially a nonmandated use of embolic protection, so of the first 80 patients, only a portion received embolic protection. The Data and Safety Monitoring Board got together after analyzing the data and, realizing that the outcomes for nonembolic protected patients were poor, mandated embolic protection, but not before there were four or five events related to nonprotected procedures (highly unusual in itself), which raised the overall rate of complications in the stent arm. Second, there was no standard equipment use mandated in the EVA-3S study; there were seven embolic protection devices and five different stents listed in the manuscript. That amount of variation in equipment in a population of operators with limited experience probably led to some complications. Lastly, the overall number of randomized cases performed per site per year was 1.7 in EVA-3S. The poor rate of enrollment per site speaks to the issue of low-volume operators, who probably have poorer outcomes in general, especially when initially poorly trained.
Endovascular Today: The conclusion of SPACE was that CAS failed to demonstrate noninferiority to CEA. Many people might read that to mean that it demonstrated inferiority. What in fact happened in the study?
Dr. Gray: The trial did not enroll enough patients to demonstrate noninferiority. Just because something is not noninferior does not mean it is superior or inferior. It means that it has not been proven to be noninferior, but that doesn't mean that it cannot be.
Endovascular Today: Why was it unable to enroll enough patients?
Dr. Gray: There was an interval analysis performed that was prespecified in the SPACE trial that occurred at 1,200 patients, which was meant to look at safety and outcomes. The Data and Safety Monitoring Board decided that, at the current rate of complications and the current difference between the two arms, more than double the number already enrolled would be needed. The null hypothesis that needed to be rejected was that CAS is not noninferior to CEA, that that there is actually some difference between the two therapies. At 1,200 patients, they found that they would need another 2,500 patients for a total of approximately 3,500. At that point, the trial sponsors withdrew support. So, after they decided to pull the plug on the trial, which may have made sense from an economic standpoint, we are unfortunately left with these results. If you look at the results, with almost 600 patients in each arm, there was a nonstatistically significant difference of four events in 30 days: 37 in the surgery arm and 41 in the stent arm. It is also important to note that in SPACE, there was no mandate for embolic protection, and only 27% of patients actually received embolic protection. Despite of some of these issues, I do not view SPACE as a negative trial for CAS. Stenting actually had very comparable outcomes to CEA even without the standard use of embolic protection.
Endovascular Today: Do you believe that the treatment of patients in the US will change as a result of these studies?
Dr. Gray: Honestly, if we thought that SPACE or EVA-3S were representative of outcomes in CAS, then we would have to stop trials like CREST, or at least the symptomatic arm of CREST; it would be unethical to continue to randomize patients if we believed that there was a difference between the therapies. However, I think the conduct, execution, and design of EVA-3S was so poor that it makes it difficult to analyze the results in a meaningful way, so we continue to randomize patients with normal surgical risks in the US in hopes of answering these important questions. SPACE and EVA-3S clearly did not give us the definitive answers.
William Gray, MD, is from the Center for Interventional Vascular Therapy, Columbia University Medical Center, Cardiovascular Research Foundation, in New York. Dr. Gray may be reached at firstname.lastname@example.org.