Daniel: FDA Does Not Have A Reliable Surveillance System For Medical Devices
Devices aren’t tracked with the same rigor as drugs, because FDA does not have a data system that can reliably track medical devices and identify potential safety problems, according to Gregory Daniel, fellow and managing director of the Center for Health Policy at Brookings Institution.
“Without having such a data system that can be used for active safety surveillance—i.e., safety monitoring that doesn’t rely on reporting of adverse events by providers or manufacturers—it is challenging to quickly identify potential safety issues with devices early on,” Daniel said.
FDA is working on improving its postmarket surveillance system for devices in collaboration with Brookings—an effort that Daniel said would reduce reliance on spontaneous reporting of adverse outcomes.
The initiative, called the National Medical Device Postmarket Surveillance System, or MDS, will use unique device identifiers, insurance claims data and electronic medical records to create a database that can track devices and link them with patient outcomes.
“I think that the major impetus [for this program] is a realization that there isn’t at all an existing sustainable data system in the United States that one can use to track and understand how medical devices are performing across different patient populations, across clinical settings,” Daniel said. “Generally, I think it is safe to say that with better, more robust postmarket surveillance that the MDS can provide, certainly in a lot of cases, this would enable the accumulation of evidence on the devices much more rapidly and in much larger populations than is currently available right now.”
Daniel spoke with Matthew Ong, a reporter with The Cancer Letter.
Matthew Ong: How did Brookings become involved in FDA’s efforts to develop and expand the Center for Devices and Radiological Health medical device postmarket surveillance system?
Gregory Daniel: Brookings has a longstanding partnership with FDA on facilitating the discussions around a lot of the high-priority topics for the agency.
We started working with the CDRH in our role in developing the Unique Device Identifier implementation roadmap, which is the strategy for how stakeholder groups can implement and begin using the Unique Device Identifiers on devices.
That work helped launch our role into convening the planning board, and we did that by essentially responding to an RFA that was released by the agency.
MO: When did CDRH start working on this, and when did you join the project?
GD: CDRH included this in their strategic priorities in2012, which also included establishing the UDI system. We began working on the National Medical Device Evaluation System by convening the planning board in 2014.
MO: Why did CDRH decide to enhance its postmarket surveillance capabilities for medical devices?
GD: It’s multifactorial. I think that the major impetus is a realization that there isn’t at all an existing sustainable data system in the United States that one can use to track and understand how medical devices are performing across different patient populations, across clinical settings.
First, without having such a data system that can be used for active safety surveillance (i.e., safety monitoring that doesn’t rely on reporting of adverse events by providers or manufacturers), it is challenging to quickly identify potential safety issues with devices early on.
Second, without such a system, it is very costly and resource intensive to develop longer term evidence on the effectiveness and impact on patient outcomes of medical devices. Better systems for developing evidence on safety and effectiveness could also help support innovation through enabling more streamlined and routine data collection that would be required for regulatory decisions.
So developing such as system can substantially improve the ability of FDA, manufacturers, providers, and patients to get better and more timely evidence on safety and effectiveness, and help support innovation.
I think the promise of such a system to deliver on more robust and efficient data collection led the FDA to make this such a high priority.
MO: What initiatives are Brookings and CDRH proposing, and how does it work? In a nutshell, what’s the plan, and how will it be implemented?
GD: Over the last year, Brookings convened the National Medical Device Evaluation System Planning Board. The planning board was put together through a public call for nominations—we had an independent selection committee to determine who would be selected for the board membership.
We spent the last year working with the planning board to articulate what the national vision should be for such a data system and what the system should look like, and what functions it should have. That report came out in February earlier this year.
At the same time the National Medical Device Registries Task Force, led by Duke University and the MDEpiNet partnership to develop priorities for improving the use of registries specifically (which are one of the modes for collection important data on devices and outcomes) for developing evidence on safety and effectiveness. This effort was largely focused on methods and data collection.
Now we’re in phase two, which is continuing to work with the planning board on the actual implementation plan and strategy for the NMDES, incorporating the planning board’s report and the registries task force report—so getting to the details of data coordinating center, the function of this center and its governance, and the sustainability and business plan for the system.
MO: What needs and concerns were important to you in the process of developing the blueprint?
GD: That’s a great question. One of concerns was that—neither FDA nor the planning board wanted this to be a brand new, one-off data system that’s built from scratch. We have too many of those right now.
The vision for the system was that it would leverage and collaborate with existing data models and systems that are already out there, like FDA’s Sentinel System, a national electronic data system to actively monitor the safety of FDA-regulated medical products, but most useful for drugs and vaccines and PCORI’s (Patient-Centered Outcomes Research Institute) PCORnet, a national collaborative research infrastructure focused on comparative effectiveness research that matters to patients. There are many great medical device registries that are also up and running. These are important building blocks and data systems that could enable the MDS to function.
So that’s one, the coordination of existing data networks.
Number two was patient privacy and data security. With the way that electronic health data are generated across the health care system, there are a lot of novel ways to appropriately utilize these data to generate medical evidence. This system should make sure that it develops evidence in a way that is compliant and is appropriately protective of patient privacy and data security.
And number three—not in order of importance; probably in reverse order—is the patient should be at the center of this. This is all about improving the evidence to inform patient and provider decision-making about high quality care, and this is about evidence that can help identify what works in the system and what doesn’t.
MO: How will FDA be using the system of networks? Will it be through Sentinel via tracking UDIs?
GD: Sentinel is certainly a good model in being able to partner with large private health plans and other systems that have majority access to the claims data. That will be an important part of the system. I don’t think Sentinel will be the keystone of the system necessarily, simply because Sentinel does not have a lot of clinical data around the particular devices themselves and right now, claims data do not include UDIs so that makes it nearly impossible to use that data to identify specific brands or models of devices. It would be fundamentally enabling if UDIs were included in claims data, but that doesn’t exist today.
So we have to design better, other ways to be able to identify specific devices and link them to long term outcomes, and that will be including claims data, but also registry data and electronic medical record data.
The planning board didn’t envision that FDA would own or lead this system. Rather a coordinating center managed by a public private partnership with an independent governance structure would need to be created to coordinate this system’s development and use. This would enable many stakeholder groups to be an important part of the system.
MO:How would MDS improve FDA’s ability to track medical devices and keep up with reporting of adverse outcomes in the future?
GD: What we’re hoping the system will be able to do is that when there is a potential concern about safety on a particular device, or there are questions about the benefits that a particular device can bring to a patient population—having a system like this will enable FDA or a sponsor or a provider group to formulate what their question is of the data and our system will be able to coordinate the necessary data in order to efficiently ask those questions and get answers. For safety questions, this would be a big step forward because the FDA would be able to evaluate safety issues without relying on providers or manufacturers to report adverse events. This will be an active system in which the data such as the claims, EHRs, and registries automatically and routinely collect safety and effectiveness information as medical encounters occur.
MO:What is the overall approach on how drugs and devices should be regulated? Are Brookings and CDRH using the National Drug Codes system as a reference point?
GD: On the drug side, the NDCs are great, because they are ubiquitous in electronic health data. NDCs are included in claims data, they’re included in electronic medical records, and because of that, it’s very efficient to go to large claims data sources in electronic medical records and quickly identify unique drug exposures and then link those exposure to outcomes.
That doesn’t exist on the device side. Unique device identifiers did not exist—but now the system exists, but the challenge is that just having the identifier on the device itself doesn’t help us better identify devices in the electronic health care data systems.
Providers, payers, patients need to use the UDIs and document them, mostly on the provider and payer side into the electronic medical record in the claims data in order to be able to much more efficiently identify unique devices in the data themselves.
So the NDC is a good example on the drug side of how that can be done. We do have tremendous amounts of drug safety surveillance, comparative effectiveness research, quality reporting, etc.—and we learn a lot about drugs, not only initially when the drugs are on the market, but drugs that have been on the market for 10, 20 years. We have a wealth of data available to really understand how that drug performs in a variety of different patient populations, thanks to the ubiquitous nature of the NDC.
We’d like to be able to get there on the device side, but it will take a lot of investment up front by payers, providers, and hospitals to develop the data capability and infrastructure to be able to document and report those UDIs in the claims in the electronic medical records.
MO: Should there be a difference between how drugs and devices are regulated? And has this evolved over time?
GD: Drugs and devices are different. Drugs, once they’re on the market, their chemical structure is consistent, and it stays the same over the life of that drug.
When a device gets on the market, version A gets approved and is used, but then there’s version B and version C that are slight modifications, for instance, based on surgeons’ input using the devices as part of surgery.
So the device development continues to evolve in an iterative fashion, even after the devices are on the market. They’re different, so the regulatory pathways need to accommodate those fundamental differences in how drugs are developed versus how devices are developed.
I think we don’t have a one-size-fits-all regulatory approach—I don’t necessarily think that that’s right—but an approach that recognizes the differences between drugs and devices, and diagnostics are clearly part of this, too.
It is important that the reviewers and the folks at FDA—whether they’re on the drug side or the device side—that they have the deep understanding of the therapeutic areas of the disease process within the body, the biological underpinnings of disease and an understanding of patient experiences and preferences in order to best be able to evaluate the benefits and risks in the patient population.
MO: Should adverse outcomes reporting be different for drugs and devices, and is it?
GD: If you’re talking about the spontaneous report data requirement on manufacturers and providers, I’m not sure how different they are between devices and drugs.
MO: There has been a lot of controversy over the past two years on whether FDA medical device regulations adequately ensure patient safety. The Class II 510(k) clearance process has been the focal point in the debate, and critics say that a more reliable risk-based system needs to be instituted.
Since the legislation clears products based on predicate devices without premarket testing, patient advocates say that the potential for patient harm is embedded in the 510(k) process, because there is no rigorous risk assessment mandate for subsequent iterations of equivalent Class II devices.
How does the MDS address those concerns?
GD: In the specific example that you’re talking about, I think that with the 510(k) and the PMA processes, there is some form of a risk-based system.
The question is: How well are we doing in identifying what products go through which process? And how well are we accumulating and generating the best highly reliable and quality data to support both pathways?
The MDS doesn’t change the regulatory pathway, but we certainly need to look at the pathways and make sure that they are the most appropriate for improving quality and reducing the risk of harm.
For the MDS, that does and will improve the availability of high quality data for regulatory decision-making, whether that’s premarket decisions or postmarket decisions.
Leveraging the constant availability of clinical and administrative data from real patient experiences with these devices can help much more than before identify safety concerns much earlier in the process. It helps companies and providers better understand the right patient population where the benefits are outweighing the risks the most, and guide care in that context. Most importantly these data do not rely on reporting. The data are automatically collected as part of routine care.
So there are a number of improvements in the decision-making ability of regulators, providers and patients that these kinds of data—that are generated from actual patient experiences—can bring into the system.
MO: I’ve got a proposal from patient advocates that I’d like to bounce off you. Some advocates say that perhaps the most effective way to protect patients is to require manufacturers to track the first wave of high-risk devices via a registry—sample size to be determined—and report outcomes to FDA.
Can you juxtapose that proposal against the MDS and weigh the pros and cons?
GD: The proposal you’re talking about is requiring high-risk devices to automatically have a registry that continues to collect postmarket data on these devices. Is that right?
MO: Right, as an active way of reporting back to FDA instead of waiting for adverse outcomes to be reported.
GD: This is a prime example of how valuable something like the MDS could be. Because, requiring a company to automatically start collecting this data—that’s pretty expensive to do for the company and the health system.
Generally, what that means is a brand new registry needs to be created. And to get the data into the registry, the providers and clinical staff need to manually add data to the registry at the point of care and typically during any follow up visits. That’s actually very burdensome on providers and their clinical staff.
For some medical device implant procedures, it can take longer to fill out the data forms for the registry than performing the procedure itself.
The vision for the MDS is that it will enable coordination and use of ongoing national data systems and registries to eliminate or minimize the need to start a brand new registry from scratch, thereby enabling the evidence to be developed more efficiently.
For some questions on safety and effectiveness, a registry may not be needed because we’ll be able to use claims and EHR data, potentially as they are collected and made available for Sentinel and PCORnet, for example. However, for other questions that require specific clinical data elements, registries will be an important component—but by also leveraging claims and EHRs, the burden on the providers can be reduced.
There’s a lot of work right now exploring how to more efficiently automatically include claims data and electronic medical records into a registry so that you alleviate the burden on the provider allowing them to spend more time with their patients.
If there is a requirement on a company to capture specific outcomes data, the MDS could be the data infrastructure backbone, eliminating the need to build a new registry from scratch.
So it’s a longwinded way to say that this system could be a much more cost effective and efficient and flexible way to meet regulatory requirements on collecting postmarket safety and effectiveness evidence of medical devices. That’s a major goal of this system.
MO: What do device companies think of the MDS? What have you heard from industry?
GD: We have some industry members on the planning board. Generally, they are very supportive of having this national system that brings stakeholders around the table to develop and coordinate the right kind of data systems that we need to generate better evidence.
There are, as you’ve mentioned, a lot of regulatory requirements for collecting data—they see the value in a system like this being able to do that more efficiently.
It’s sort of a blueprint right now, it’s not a reality yet, so support from FDA and federal partners to launch this system to begin generating evidence can really then provide back to industry proof that this system can be valuable and is worth investment.
MO: What are the milestones for the blueprint, going forward?
GD: Generally, what we’re trying to do over the next year is identify the characteristics and functions of the coordinating center, including the organizational and data a governance models and to stand this coordinating center up. Then further refinement and development of the data strategy, business model, and operating policies will need to be supported along with early evaluations to demonstrate the value of the system. A full-blown system operating as envisioned by the planning board will take at least five years to really develop. A lot will need to go into bringing disparate data together and using it for medical devices.
MO: Circling back to the genesis of this story, if the MDS had been in place, say, in 2010, how would it have prevented patients from being harmed by power morcellators in recent years?
Also, how would it have impacted adverse outcomes reporting, considering that these devices were on the market for two decades before patients—not manufacturers and user facilities—reported harm to the FDA?
GD: Generally, I think it is safe to say that with better, more robust postmarket surveillance that the MDS can provide, certainly in a lot of cases, this would enable the accumulation of evidence on the devices much more rapidly and in much larger populations than is currently available right now.
You get a better understanding of what’s happening sooner. If there is a safety issue that’s happening out there, this would improve the accumulation of evidence in terms of speed and quantity. Also the system will not rely on reporting of serious adverse events as the data that will be leveraged will automatically include them as they are identified from claims, EHRs, and registries, not on spontaneous report data.