Postmarket Surveillance of Medical Devices: Panel Discussion
After the presentations, workshop speakers assembled for a panel discussion of postmarket surveillance of medical devices. Gardner, Hall, Maisel, Masoudi, Peterson, Resnic, and Varosy were joined by Susan Alpert, vice president of regulatory affairs and compliance at Medtronic, and Larry Kessler, professor and chair of the Department of Health Services of the University of Washington, former director of the Office of Science and Technology of the Food and Drug Administration (FDA), and former director of the Office of Surveillance and Biometrics of the FDA Center for Devices and Radiological Health.
Moderator and committee member Lazar Greenfield provided Alpert and Kessler with the opportunity to make opening remarks. In the discussion, panelists expanded on the topics of predicates; device development, including the conduct of clinical studies; data collection and data-sharing; unique device identifiers; encouraging broader participation in device surveillance systems; and risk communication.
SUSAN ALPERT: INDUSTRY DEVICE SURVEILLANCE
As device manufacturers develop technologies and evaluate them, Alpert said, they have expectations for the performance of the devices in the marketplace. Manufacturers track and trend performance input from the field against those expectations and against previous experience. Each field report is evaluated to determine whether it meets the requirements for reporting to FDA and to global regulators and whether there is a potential need for a modification in design or manufacturing.
Many companies have voluntarily established their own product registries. Alpert highlighted two of Medtronic’s voluntary databases that are being used for studies. The Systems Longevity Study is looking at survival of implanted leads to determine durability and long-term functionality. For over 27 years, she said, Medtronic has been capturing information about its cardiovascular devices. That is important because generally when a lead fails it is not extracted, so the manufacturer cannot gather forensic information on the device. She noted that information on more than 75,000 leads in 14 countries has been evaluated thus far.
The Implantable Systems Performance Registry was created in 2003 to monitor Medtronic’s infusion and neuromodulation devices (such as implanted drug pumps and spinal-cord stimulators). The registry includes data on more than 5,000 patients in 50 centers. Alpert added that Medtronic evaluates and publishes its data in semiannual product-performance reports.
Another example of device surveillance that Alpert described is Medtronic’s CareLink monitor. Many active implantable devices have electronic monitoring systems, often bedside monitors, to which the devices automatically send information about the patients and the performance of the devices every night. Through CareLink, this information can be transmitted to the provider for remote monitoring.
Alpert stressed that both industry and FDA depend heavily on physicians and end users as reporters. She asked the committee to consider how industry could interface differently with the clinical community to achieve better access, not only to information about devices in practice but to the products that need to be returned, so that forensic work can be done to assess defects.
LARRY KESSLER: ADVANCING SURVEILLANCE
Kessler offered several suggestions, directed to three constituencies, for advancing device safety and surveillance.
Congress, he said, should provide additional resources for FDA. Resources are needed to focus on the risk issues related to 510(k) products, not just the traditionally high-risk products. Development of registries may be helpful, he said, as would additional Section 522 studies. The rate-limiting factor, he said, is identification of the problems.
FDA regulation on unique device identification is critical, Kessler said. Although it was required legislatively in 2007, FDA has not yet issued a rule, and it needs to do so.
FDA should also enforce known engineering standards and foster attention throughout the centers to issues of risk, benefit, and quality systems, Kessler said. The risk-management strategies devised by manufacturers when they are developing 510(k) products should be widely available to
both premarket reviewers and postmarket and compliance staff. That is generally not the case at the 510(k) level, he said.
Finally, clinical professional societies should encourage clinicians to ask questions about devices and foster enrollment of patients in device registries and other studies.
A committee member raised the issue of the high percentage of 510(k) product recalls that are attributed to design flaws and raised the question of how the source of a flaw may be associated with predicate devices.
Alpert agreed that a device on the market that is used as a predicate may be somewhat different from what was cleared as a result of multiple small changes. But a 510(k) submission for a new device is more than just a statement that something is similar to something else, she said. A tremendous amount of testing and evaluation is involved in a 510(k) submission, in some cases including side-by-side testing against another device. There are also 100 or so 510(k) device-specific FDA guidance documents that require a manufacturer to conduct specific kinds of testing according to specific standardized test methods.
Alpert said that industry is required to track all changes of products but noted that not all changes rise to the threshold of submission to FDA. Industry uses the FDA guidance document that specifies the kinds of changes that need to be reported to FDA.1
Kessler added that when evidence of a new problem emerges, the agency can require Section 522 studies for previously cleared 510(k) products that are on the market.
There are alternatives to the 510(k) predicate approach, Kessler said, and he urged the committee to look at other models, such as the Global Harmonization Task Force guidance Essential Principles of Safety and Performance of Medical Devices, which is used in the European Union and elsewhere (GHTF, 2005).
DEVICE DEVELOPMENT: DESIGN AND STUDIES
Kessler said that there are not enough opportunities for comprehensive dialogue early in device design and development. Design occurs at one point, interaction with clinicians at another, discussions with FDA at yet another, and then discussion with Centers for Medicare and Medicaid Services (CMS) and other contract carriers. Earlier, more comprehensive
Deciding When to Submit a 510(k) for a Change to an Existing Device. http://www.fda.gov/MedicalDevices/DeviceRegulationandGuidance/GuidanceDocuments/ucm080235.htm#page27.
discussions are important, he said, for finding out what questions clinicians and their patients have.
Alpert noted that industry does engage the clinical community in most of the original design of products, and most of the changes that are made later are driven by users.
Kessler said that there is an eagerness to get new devices to the market and into use, but academicians need to get the word out that real-time pragmatic studies are necessary.
Maisel said that there should be discussion about the type and strength of evidence needed to ensure safety and efficacy instead of the assumption that the randomized clinical trial is the gold standard for every 510(k) device.
The primary incentive for any action by a manufacturer is getting a product to market. Peterson said that the reason that drugs companies conduct large randomized trials and device manufacturers do not is that regulations require trials for approval to market a drug. On the positive side, the current 510(k) structure fosters a high degree of innovation in devices; on the negative side, it is challenging to entice industry to do studies because they can get a product cleared for marketing fairly easily without them.
Clinical demand (or the lack thereof) for information also plays a role in performance-data collection. One reason that Medtronic does a lot of studies that are not required by either payers or regulators, Peterson said, is that physicians and patients have demanded data. In many fields of medicine, however, there has been no demand for information.
A difference between devices and drugs, Maisel said, is that conducting a 4-year randomized trial of a device would essentially be holding back the advancement of technology. Our goal and our measurement, he said, should be benefit to public health. That means coming up with a total-product approach that balances the risks posed by bringing a product to market and the need to obtain information.
For many devices, a clinical study is necessary to understand the final benefit–risk ratios and identify problems that might arise in a particular patient population. But for the technology itself, most information can be obtained better at the bench than in large clinical trials. There is a distinct developmental difference between technology and pharmaceuticals.
One of the major roles for registries, Varosy said, is in confirming the validity of translation of evidence from randomized trials into real-world populations. There is still a major role for real-world evidence even after the primary questions have been addressed by randomized trials.
A committee member noted that some 510(k) devices are tools, such as diagnostic radiology equipment and ultrasonography devices that have a plethora of potential clinical applications. A manufacturer could not be expected to test each application with a clinical end point before entering the
market. Are the necessary data the same for tool claims and clinical claims? he asked. What are the problems when a device comes to market as a tool and seeks a clinical indication later?
Hall responded that a lot of European device indications are more “engineering indications.” A particular device is capable of performing a particular task (for example, cut and ablate). That raises the “specific vs general indication” problem. An ablation device ablates tissue. How would one obtain an indication for ablation of heart tissue? Many of the indications are procedural, not clinical, he said.
Once a device with general indications is on the market, Kessler said, there needs to be an integrated approach that is patient-centered and clinician-centered to figure out what is known and not known about the indications and to begin real-time studies (which need not be trials). In the case of trials, Kessler noted that a lot of device studies are done with 30, 50, or 100 patients to be allowed to use the more specific indications (compared with thousands or tens of thousands for a drug).
That comes back, Alpert said, to intended use vs indication. The issue has been challenging for the agency for many years, she said, because these are tools, and we try to define more specifically where they can be used and what benefits they can provide for specific populations.
Alpert reiterated Kessler’s point that in Europe there is no effectiveness requirement; the focus is on safety and performance. Effectiveness questions are related more to reimbursement than to market entry.
A question was asked about whether, from a consumer-protection perspective, it should be made clearer which indications are not supported by data. Alpert noted that device labeling already clearly states what is known about a device and its indications. Hall concurred and raised the issue of interfering with the practice of medicine. There is a well-established and legal practice regarding off-label use of products, he said, and the American Medical Association has an explicit policy statement on this.
No study and no system can provide the whole the answer, Kessler said, or fully achieve the goals of the 510(k) process, which are the continuous evolution of products and continuous assurance of safety and effectiveness. Kessler drew attention to Gardner’s presentation: the hazards and failures most seen with devices are frank failures, use errors, interactions, mismatches, environmental effects, and so on.
Alpert concurred that one size does not fit all. Industry is not resisting clinical trials that are appropriate, she said. Many 510(k) devices are low-risk tools, others pose moderate risks, and for some robust clinical information is useful for understanding uses and expectations. It is a matter of what the appropriate information for specific kinds of products is and how to obtain it.
POSTMARKET DATA COLLECTION AND DATA-SHARING
A committee member noted that databases include the fields that are relevant to the researchers at the time they are developed. It is always problematic to scale up the data-collection efforts or to modify the data that are collected in order to address new areas of interest. Resnic pointed out that in the case of the National Cardiovascular Data Registry (NCDR), the problem of tracking patient-identifier information (Health Insurance Portability and Accountability Act regulations preclude direct linkage) has been successfully worked around with probabilistic matching.
Resources are already being spent by industry on postmarket studies, Resnic said, but the data are often kept behind the company firewall. There are some requirements for sharing of postmarket data with FDA, but those data may not be made available to the rest of the community to learn from (including the next manufacturer).
Alpert pointed out that all adverse events data are shared with FDA and are public. However, when a manufacturer is paying to conduct a study and collect data for the purposes of modifying one of its devices and improving it relative to a competitor product, those data are often proprietary. In initiating additional postmarket studies, she said, Medtronic generally seeks to publish the data, but there is also competitive information that may not be released.
A committee member wondered whether it would be possible to have a core set of data in a precompetitive, publicly accessible space with appendixes of private data for specific studies.
UNIQUE DEVICE IDENTIFIERS
Developing standardized nomenclature for devices and a unique device identification (UDI) system will aid the ability to track devices among institutions, Peterson said. The data are already stored electronically somewhere, but they are not easily shared.
Varosy agreed that having a UDI system would make the process of documentation easier, more feasible, and translatable across systems.
Although UDIs are important, Alpert said, several other things also have to happen, including building electronic health records that have the capacity to collect the information. The big impediment is to put a particular identifier on a product rather than whether it is the hospitals, physicians’ offices, or clinics that capture and use the data; right now, they do not do that, she said. Hospitals use the bar code on the device for supply-chain and billing purposes, but it is not used for anything else.
PARTICIPATION IN SURVEILLANCE SYSTEMS
Perhaps the biggest factor in the Department of Veterans Affairs (VA) system that has facilitated reporting, Varosy said, is having a uniform electronic health record as a foundation. Moreover, although a lot of health information technology is “inflicted” on doctors by developers, a key aspect of the VA CART-CL was that it was developed directly with clinicians to be a clinically useful tool.
The question is one of resource allocation, Resnic added. From an information-systems perspective, the reporting process is similar, whether the issue is the number of failures of tire tread in a manufacturing plant or unexpected falls out of a new hospital bed that is cleared through the 510(k) process. The question is whether it is cost-effective for health-care institutions to develop registries that can capture high-granularity data and determine a denominator.
Alpert noted that for one of the two registries she spoke of, Medtronic pays physicians. The payment is minimal, but it helps to ensure that information is entered into the database, she said.
The acceptance of a data-collection tool depends somewhat on the burden of collecting the data, Resnic said. There has to be some consensus early on that the data are worth collecting for the purposes of those participating institutions. There are multiple reasons why institutions participate in the NCDR, he said, one of which is to receive data from the NCDR for performance and quality benchmarking. Another incentive, Resnic said, is that many payer organizations, including CMS, view participation in the NCDR as evidence of a high-quality organization that is monitoring outcomes.
Peterson added that once a product is on the market, payers can require participation in registries as a condition of payment.
Resnic noted that every new data element that one tries to add to an existing system is going to come at some cost to the population of participants, and expansion has to be a collaborative development effort.
Hall pointed out that many 510(k) products are home-use products. The surveillance systems that have been discussed work much better in a hospital setting. One of the challenges is to facilitate data collection outside the hospital procedure-based structure.
A committee member asked, With continuous accrual of information in the postmarket period, at what point do actions need to be taken to communicate the information?
Resnic stressed again that in monitoring of datasets, the signals identi-
fied are really only hypothesis-generating. In the conduct of the DELTA studies in collaboration with FDA, Resnic said, results were shared with the FDA postmarket staff, who shared them with the premarket staff who looked at the internal FDA datasets, including engineering data and preapproval and postapproval studies, to see whether there were concordant signals. Such hypothesis-generating signals must be vetted through a relatively rigorous mechanism to avoid undue alarm in patient communities. Risk communication is a science unto itself, he said, and an initiative on risk communication is under way in FDA.
Maisel said that communication should be guided by ethical principles of what a patient would want to know and would need to know to make an educated decision about care.
Hall added that as a result of modern information technologies, there are no longer separate communication pathways for physicians and patients. The communication challenge is incredibly difficult.
GHTF (Global Harmonization Task Force). 2005. Essential principles of safety and performance of medical devices (SG1-N41R9:2005). Study Group 1 of the Global Harmonization Task Force. http://www.ghtf.org/documents/sg1/sg1n41r92005.pdf (accessed July 2, 2010)