Control group complexities, best-case trial designs, questions concerning new requirements for DES studies produce plenty of debate at Cardiovascular Revascularization Therapeutics meeting.
CDU Washington Editor
WASHINGTON — The FDA and makers of percutaneous devices have struggled with a variety of issues concerning clinical trial designs as the interventional market grows and new information technologies push more data into this mix. A particularly difficult issue is the reluctance of a patient population that has more information — much of it via the Interent — to agree to enrollment in current standard-of-care control arms.
Another problem is that the FDA’s humanitarian device exemption allows use in only 4,000 patients. But even if more patients could be treated, many of these populations consist of older, high-risk patients who cannot sustain the trauma of thoracic surgery and have few options. Because they are at high risk of adverse outcomes, sponsors are reluctant to enroll them in disproportionate numbers. Between the reluctance of healthy subjects to enroll in the control arm of a study and the disproportionate representation of high-risk patients, the modern clinical trial is apparently facing an enrollment squeeze.
Other trial issues surround the ongoing debates concerning stenting. Of key concern here is the continuing evaluation of the trial data that was used to promote the approval of drug-eluting stent (DES) devices, and, now, what impact this will have on the regulatory pathways for the new DES devices poised to come online over the next several months.
Heart valve controls difficult
These were just some of the issues that were raised at the 2007 edition of Cardiovascular Revascularization Therapeutics in early March. Members of the FDA staff joined a wide range of physicians and others to discuss requirements for clinical trials for percutaneous devices designed to deal with failed heart valves, including aortic valves.
Julie Swain, the FDA’s special assistant to the director of the division of cardiovascular devices, noted that, at present, there are no devices on the market approved for aortic valves and that “the key to designing any clinical trial is the control group,” because designation of the control group goes a long way toward deciding on inclusion and exclusion criteria, as well as the safety and efficacy endpoints. Another key issue for firms working on clinical trial plans was whether to go with a study hypothesis that assumes superiority of the studied device vs. the control vs. non-inferiority.
A trial that enrolls very sick patients might use conventional thoracic surgery as the control group, but a trial that enrolls patients who are not very sick might employ a trial that compares the aortic closure device against medical therapy. However, that latter type of trial, Swain said, “may take many hundreds or thousands of patients and many, many years to get an endpoint.”
She also warned that surgeons and cardiologists often have different ideas about whether a patient should undergo a procedure because of risk factors, since there is often no consensus on what constitutes very high risk. “You will not find surgeons who agree on [a definition of] high risk,” Swain said, adding that she prefers the approach of assessing “elevated” risk. “The definition of high-risk is a moving target” and is “a term I choose not to use anymore.”
Historical controls difficulties
Sharon-Lise Normand, a biostatistician at the Harvard School of Public Health (Boston), who in an earlier session had discussed the possible use of historical controls in clinical trial design, said that some of the factors that make historical controls difficult to use in a trial include differences in the kinds of pharmaceutical treatments that the controls may have received, and differences in data collection methods. Other factors she cited as confounding the use of historical controls are inclusion/exclusion criteria and safety and efficacy endpoints.
But however difficult such problems might be, they aren’t insurmountable, according to Normand. “One might argue,” he said, “that the exclusion criteria in the historical control group really would not impact the outcome,” but he added that a sponsor would have to make the case concerning the historical controls and the proposed current study group. He also noted the need to develop “similarities in your definitions of endpoints” and the need for establishing some sort of standardized system to do this.
Swain critiqued Normand’s remarks on the use of historical controls, saying that they were “on target,” and she added that the agency would “consider virtually any kind of trial design.” However, Swain declined to go into specifics on alternate kinds of designs, and none of the FDA staffers present would commit to any approach to non-randomized trials.
Bram Zuckerman, the agency’s director of the division of cardiovascular devices, said FDA is “required to consider the least burdensome way to market, but by the same token, we have to establish a reasonable assurance of safety and effectiveness pre-approval, not post-approval,” and asked the panelists how non-randomized, controlled pivotal trials should be built.
Normand said that that the ways in which historical data can inform the trial design process is different from that of whether historical data is useful as a control group. “I think the historical control data would not be a very good option for approval,” he said, but that “if it was to be a non-randomized trial,” the presence of unknown confounders would put the study in the “observational” category.
RCTs a must?
Peter Block, MD, director of clinical trials at Emory University Hospital (Atlanta), said that most clinical investigators prefer randomized, controlled trials (RCTs), but despite discussions of non-randomized trials, “from what I’ve heard today, everyone is pretty well set on what the trial should look like.” And he said that those outside the agency feel that the agency is “foot-dragging” in an effort to come up with the “perfect” trial for difficult products, such as valves.
Mitchell Krucoff, MD, director of the ischemia monitoring lab at Duke University Medical Center (Durham, North Carolina), said that he didn’t see a lot of foot-dragging.
But he cited the “dangers” of “combining too many issues into one clinical trial design, as opposed to thinking more systematically about the technical features” of the behavior of valve closures. He said that one approach might be to ask what sort of randomization patients will agree to and “work backward” from there.
“The notion that we can create one master trial design that can do all these things,” such as technical considerations and design features that are not fixed, “and answer the correct clinical questions” is the reason that “these discussions are so slow to move forward.”
Krucoff said that if a sponsor could construct a trial that would allow evaluation of some of the questions about the device that “don’t need randomized answers . . . you could build in a longitudinal trial design that ends with a randomized piece but is not completely constructed as a randomized piece.” Under those circumstances, a sponsor and FDA could move a product along, but in the end, some of the technical questions simply have to be hammered out before going to an RCT, Krucoff said.
Zuckerman acknowledged that the agency has not determined the “least burdensome” approach to trial design in this sector, and “we need to appreciate certain realities when we’re talking about a really new design.” He said, for instance, that valves installed via thoracic surgery are not “totally predictive of what’s going to happen with a percutaneous-type valve that is subject to different loading forces.”
He added: “the basic engineering is not as well understood as we would all want.”
Zuckerman said that some of the rate-limiting steps for the clinical trial process are well managed if a sponsor will “come in as early as possible to the agency to understand our clinical requirements. By far, the majority of IDE [investigational device exemption] disapprovals are for inadequate preclinical data” concerning safety and efficacy.
“There is a lack of recognition that this is an extremely high risk procedure,” he said, and a significant learning curve for both surgeons and makers of the devices. “We’re looking for ways whereby we can minimize the problems of confounders and chance.”
Block said that in some cases, “we’re dealing with people who essentially have a death sentence within a year,” and he pleaded for some other chance for very elderly patients. “Please let me have an option for this 91-year old lady,” Block said.
Krucoff said that one possibility is randomization of patients by trial centers, a scenario in which a center would enroll only either controls or study subjects, based on the training and expertise of the surgeon at a particular center and by a surgeon trained in the installation of the percutaneous device.
Mitral valve repair
The panel also discussed trial issues related to the study of devices for percutaneous mitral valve repair (MVR). Laura Mauri, MD, associate physician at the cardiac catheterization lab at Brigham & Women’s Hospital (Boston), said that percutaneous interventions of this sort will usually be compared to surgical valve repair in trials. She added, however, that “even successful trials have been controversial because of challenges to validity and generalizability,” driven by the fact that “they are two very different types of interventions.”
Mauri said enrollment presents a “major challenge,” partly because of referral patterns, and even in situations where therapeutic advantage is not established, “strong physician and patient preferences” can make randomization difficult. And she said that another source of statistical “noise” is that most trials involve some of the very best surgeons in the business, whereas the conventional surgery control procedures are performed by surgeons with a broader, real-world range of skill.
In case of ‘bailout’
Mauri added: the role of the randomized trial is not just to determine safety and efficacy of the procedure or product, but also to determine which patients would most benefit from the device.
Wolf Sapirstein, associate director for clinical trials at FDA’s Center for Devices and Radiological Health, said that efficacy endpoints for MVR must demonstrate comparability to conventional surgical approaches and also establish the degree of correction of the hemodynamic failure as well as durability. “The safety factors . . . must include the consequence of bailout surgery,” plus the impact of the procedure on any subsequent corrections to the valve, he said.
Any patients who do not qualify for surgery under the guidelines written jointly by the American College of Cardiology (Washington) and the American Heart Association (Chicago) “should be controlled to medical treatment,” and this group will be subject to stringent evaluation for clinical improvement, according to Sapirstein.
In follow-up discussion, Zuckerman said that the agency is “really looking for . . . data to define this high operative risk.”
Donald Glower, MD, a surgeon at Duke University School of Medicine (Durham, North Carolina), said that mitral valves “are much more difficult to score out for operative risk because of the fact that there are so many different diseases that are captured under mitral valve repair or replacement.”
He said the database provided by the Society of Thoracic Surgeons (STS; Chicago) includes an even split between valve repairs and replacements, but 70% of the repairs “are annuloplasty only.”
“The problem is that all of that is for functional regurgitation,” which does not extrapolate readily to [other] mitral valve disease.” He mentioned a pair of single-center studies that might be more informative as to risk, but these are the experiences of surgeons more skilled than the typical surgeon.
Weakness in endpoints
Block said that in his experience, the ratio of patients who apply to a trial for MVR is “at least 15 to 1” to those who are actually enrolled in either arm of a study. He said that one difficulty in establishing endpoints for procedures to treat end-stage ventricle with regurgitation, “none of the [endpoints] is going to be good,” and that FDA should consider the possibility that relative mortality rates for the two procedures applied to this population might not demonstrate a clear difference.
He said that it might make more sense for the agency to ask sponsors to simply clearly define a reasonable set of endpoints “and then let’s go for it.”
One panelist asked why FDA allows a high-risk registry of patients to be used to round out data for controls for other device trials, but not for mitral valve replacement studies employing a percutaneously installed device.
Zuckerman said aortic disease “is not equivalent to mitral disease” and “there is a critical lack of data for precise risk stratification” for MVR. He added: “one of the central points is that we’re just now starting to get the type of data that show the limitations of STS” data, which hopefully will eventually tell doctors which patients would be most appropriate for mitral valve replacement.
Going the resorbable route
The third day of the 2007 edition of Cardiovascular Revascularization Therapeutics included a session on the latest technology in bioresorbable stents, being pitched by their developers as important alternatives to both bare-metal and drug-eluting stents.
Besides having to work through the normal R&D and clinical trial difficulties, they are likely to have to cope with a somewhat altered regulatory landscape. This is likely not only because of the ground-breaking nature of these devices but because — given the current debates concerning DES in particular and stenting in general — the agency wants to make sure the new devices will present no additional surprises for patients and physicians.
Early in the conference, Andrew Farb, a medical officer at the interventional cardiology branch of the CDRH, said that that the agency wants enrollment populations “closer to actual populations in practice,” such as those with diabetes. And he said that is a recommendation applying to all DES, current and in development. And he promised that FDA intends to release a guidance for clinical trials for DES “later this season.”
Representatives of several firms gave briefings on their stents, and afterward, Ashley Boam, the chief of the interventional cardiology devices branch at CDRH, said that the discussions of the bioresorbable stents had made for “a very interesting session.” But she cautioned that there will be “challenges in characterization” of the interaction of these stents with human physiology.
The new materials going into stents, Boam said, “adds a level of complexity to the review” compared to bare-metal stents, and is further complicated with the addition of an eluted drug. She said that the agency wants to “emphasize the need to fully understand the stent degradation profile at the outset,” adding that this is “critical.”
Some of the information FDA will expect sponsors to be able to answer early on include the characteristics of the expected degradation profile, the nature of the degradation products, and how the design balances the need for mechanical integrity following deployment with the stent’s degradation.
Boam remarked that some of the traditional bench tests, such as fatigue testing, “will no longer be as relevant” due to the degradant nature of this class of stents. “We still need to have an understanding of the mechanical properties” of the bioresorbable stent, he said.
FDA reviewers are likely to be interested in seeing multiple sets of mechanical tests to establish whether sponsors can characterize how the stent behaves over time, not just after deployment “when you expect integrity to be maintained, but [also] into the ‘spongy period,’” according to Boam.
Other information the agency will want includes the nature and timing of any particulate matter that remains in the body, especially any other than what was expected.
New standard tests coming?
Standard biocompatibility tests may need to be altered for differences in extraction conditions and time of exposure. If the agency has questions about the degradation products and the mechanism of degradation, sponsors may need to “separately test the degradation products” for any toxicities. Boam warned that stent makers may be subject to a “case-by-case, stent-specific test paradigm.”
On the subject of animal studies, Boam said that some of the traditional test markers “may be modified or added [in order] to have a full understanding of the safety parameters,” and that data concerning degradation will also be needed, including data on migration of the degradation products into the vessel wall and beyond both during and after complete degradation.
“One of the things that may be on the table” is a possibility of tracking the effects on animal models even after degradation is complete to check for late remodeling of blood vessels, but she did not speculate on the duration of such follow-up.
“Additional parameters that we’re thinking about . . . are the biologic effects of the degradation products on inflammation” and other tissue reactions, including endothelization time and vasomotor function, Boam said.
The duration of Phase II and III trials and measurement time-points will hinge largely on the degradation profile of the stent’s materials, and the agency may recommend some modification of typical endpoints, depending on the characteristics of the stent.
Other issues include whether the stent will be visible to fluoroscopic imaging equipment, which could affect the need for additional imaging. For follow-up imaging, sponsors will have to be able to predict how well angiography and intravascular ultrasound (IVUS) will pick up the stent as it degrades over time. Boam also recommended that sponsors attempt to incorporate endpoints that are unique to the stent in question.
“In conclusion, we agree that bioresorbable stents might represent the next great leap forward, but the development process really should be careful and deliberate. Depending on the intended use, whether for coronary use or for peripheral use, there may be other factors to be weighed that make getting in touch with us early all the more critical.”
During a Q&A session, Boam responded to a question on how much animal model data a sponsor would have to submit before moving into human trials for a stent that is projected to degrade over two years.
Boam said that the answer “will depend on what the animal studies tell us about healing and how they look at six and 12 months, for example.”
She said that the agency would not automatically take the position before the fact that clinical studies cannot start before the animal models go to 24 months in this scenario, but a sponsor should not assume that FDA will allow people to enroll before the animal models have cleared the intended duration.