Originally published January 1996
An Interview with Larry Kessler
Director, CDRH Office of Surveillance and Biometrics, Rockville, MD
Larry Kessler has never ducked complicated questions. In 10 years at the National Cancer Institute, the new director of the Office of Surveillance and Biometrics (OSB) in FDA's device center documented differences in incidence, survival, and mortality rates using data gathered by the National Center for Health Statistics. He also studied the effect of cancer diagnosis on Medicare payments, finding that the earlier the diagnosis, the higher the total payments. The reason, Kessler concluded, was survival time. These and other studies resulted in a number of peer-reviewed publications, each founded on modeling and statistics applied to the understanding of cancer rates and trends, health economics, and the effects of health-care services.
Since June 1995, Kessler has been applying his talents in quantitative analysis to bolstering the device surveillance programs at FDA. Whether a medical device is involved directly or indirectly as a cause of death or serious injury, manufacturers must report the incident to Kessler and his staff. Under the Safe Medical Devices Act of 1990 (SMDA), user facilities--including hospitals and nursing homes--must also make certain reports to FDA and to the manufacturers. SMDA also requires device distributors to file reports with FDA. As if that were not enough paperwork, OSB gets reports from a voluntary program called Medwatch, which encourages physicians, patients, and consumers to report problems with medical devices as well as drugs.
In this interview with MD&DI, Kessler describes initiatives that are on the horizon in adverse event reporting and postmarket surveillance to enable OSB to get a better handle on the performance of medical devices now on the market.
What kind of workload does the Office of Surveillance and Biometrics handle?
In the past calendar year we received approximately 112,000 adverse event reports. By contrast, the United Kingdom received on the order of 2000 such reports last year, and Australia received approximately 500 to 600. Some of the reports we get are very simple, describing a device-related problem at a specific institution; others summarize the experience of a device across several or even many individuals. Last year, roughly 40,000 of the 112,000 reports were about breast implants. So we're receiving an enormous number of reports about that one topic.
Why is there such a big difference among the United States, the United Kingdom, and Australia?
There's a difference in philosophy. If a device is experiencing a problem, the Europeans and Australians prefer to have the manufacturer conduct a comprehensive and thorough investigation of the problem before making a report. Those reports may cover many instances of a problem, and can determine whether it is truly a device problem that needs to be fixed or a user problem that might be solved by an education program.
In the United States, the statutory requirements have set relatively narrow time frames. For example, there is currently a five-day window within which the manufacturer must report to FDA any death or serious injury that may be life threatening. That is barely enough time to gather the information and send it to FDA, let alone to conduct a thoughtful investigation. As a result, in some ways we get too many reports, and in some ways we get too few reports. Fortunately, the new Medical Device Reporting (MDR) Act extends that time period. We believe the new user forms, which are about to be published for comment, will cut down on needless reporting.
What else have you learned by watching the experiences of regulatory agencies in other countries?
We believe it would be useful to conduct statistical and trend analyses of rare events. That's not something currently done in many other medical device arenas, but there are some models. For example, in Norway there is a device registry for joint implants. Information about every implanted prosthesis--knees, hips, elbows--is entered into the registry. The surgeons follow these implants longitudinally through time and, using epidemiological methods, have discovered some very uncommon device problems.
Right now we don't have a system like that, but we'd like to consider developing one that would permit us to get large numbers of reports so that we can assess whether there's any value to this kind of trend analysis. If you get one, or two, or three reports of a problem, that's not going to help very much. But if you get a series of reports over a substantial period of time, they can help you make an assessment of whether there's a problem. In effect, we're developing a new science for studying the epidemiology of problems related to medical devices.
How would you go about establishing such a tracking system?
We have several different directions we're taking and I'm quite excited about some of them. One of the things we're trying is to use large secondary databases to determine whether they can reveal device problems that have not otherwise been detected. We have just started a collaboration with the Health Care Financing Administration to examine what its large Medicare databases have to offer, and we've already begun using them to study the use of home oxygen equipment. The patients described in these databases are very heavy users of certain medical devices, so the databases may prove to be a rich resource.
We've also begun discussions with research groups associated with managed-care organizations to see if they can provide population denominators. So far, what we have been dealing with is numerator counts, but no denominators. Ten reports of failure associated with a certain device might be important if they are 10 out of 1000, because that's a high failure rate. But 10 out of a million may be meaningless. We want denominators so that we can establish a baseline. Knowing roughly how many devices are out there would give us some public health context for the frequency of failure reporting.
What kind of obstacles do you expect to encounter in trying to establish that data baseline?
There are at least two difficulties. First, there is the difficulty of knowing how many of a particular device are actually in use. Except for certain subclasses of high-risk devices such as heart valves, manufacturers are not required to track their devices. So it's hard to know when a device is actually put into use. But tracking devices is enormously resource intensive, and we certainly would not think that it is an appropriate thing to do for every medical device in the country.
Second, some manufacturers may resist releasing the information that they do know, because the numbers they provide will reflect their market share. That's proprietary information that device manufacturers guard somewhat closely, and I don't blame them. We have no interest in making public such denominator information to the detriment of any device manufacturer. Our need is merely to have an idea of how often a device is experiencing a problem relative to the total number of the devices in use.
Will OSB be sharing the data it obtains from the Medicare and managed-care studies with individual companies?
Absolutely. We hope that this will provide valuable feedback to manufacturers who can't track the many devices they produce. I think they would probably benefit from knowing how often their devices are used and what their failure rate is. I think we can learn collaboratively and together.
Have you made any progress with this initiative?
I've approached the head of a network of managed-care organizations who does research on these topics and we're going to try to collaborate with the medical device industry and other managed-care organizations to get population-based data. Research is a part of what managed-care companies do. And I think the managed-care organizations have an opportunity here to get in on the ground floor of some new medical device technologies. We'll be approaching major trade associations such as the Health Industry Manufacturers Association and the National Electrical Manufacturers Association to ask whether they can work with us on presentations to managed-care groups that have been conducting research in a variety of areas but haven't really touched the device industry.
Will some technologies receive a higher priority for study than others?
Manufacturers are already tracking certain types of devices, such as heart valves, because it's a statutory requirement that they be able to contact patients within 10 days. So manufacturers know exactly where every single one of those goes, and I think those counts are already done. But there's less information about other, less-critical types of products. The need and fit between the various groups collaborating on these studies will determine what gets looked at first.
What are some examples of some less-critical devices that might be studied early on?
Patient direct-use devices that people don't have a good handle on, such as blood glucose test meters. Those are being used very frequently, and their failure can be life threatening. Over the past couple of years, we have received MDR reports of a number of deaths associated with diagnostic test strips and blood glucose meters that failed or were misused.
Are there any other initiatives, for example in postmarket surveillance?
We are just now beginning to work with medical device manufacturers to develop a series of postmarket studies for certain devices, as required by Congress. We want to make these studies informative both for the public health need and for the manufacturers, but we also want to conduct them as efficiently as possible because we recognize that they can be somewhat burdensome to device sponsors who may never have thought that they were going to be required to do them.
We've also begun a limited number of discretionary postmarket surveillance studies of products that we think need this kind of surveillance. The general direction of this work is to establish collaboration between OSB and the Office of Device Evaluation (ODE). ODE is moving cautiously toward identifying products that have been judged safe and efficacious enough to be allowed on the market, but for which there may be some lingering questions in areas such as long-term safety or how they are actually used in the community. For these products, ODE is turning to the postapproval and postmarket surveillance arena and asking if we can conduct systematic studies that would enable the manufacturers to get their products onto the market quickly while we develop information to help refine or improve the technology.
Isn't that pretty much what the adverse event reporting system was designed to do--to give feedback on device function?
That system is really not designed to answer research questions that remain at the conclusion of a postmarket approval (PMA) application. When a PMA is approaching a final FDA decision, for instance, the agency or an advisory committee may feel that answers to some specific clinical questions are still needed, but that these can be comfortably investigated in the postapproval arena.
About 10 years ago there was a proposal to establish groups around the United States that would provide in-depth information about medical devices. It was called the sentinel system. Is this a variation on that idea?
Well, these are fundamentally two separate issues. The postapproval studies that we are talking about, and some of the required and discretionary postmarket surveillance studies, are almost always device-specific or product-line-specific. So, for example, the investigators would be looking at specific hip implants or at a specific group of in vivo diagnostics.
Proposals to develop a sentinel system stem from the fact that we know the adverse event reporting system has a number of built-in biases. People make reports because they think they're supposed to, and although the reports come from the entire nation, they come unevenly from different areas. The sentinel system would provide either a hospital sample or area sample with a known population denominator, from which we would try to get comprehensive and thorough reporting on every adverse event. From a scientific standpoint, this is a very attractive option.
Is the idea still alive at FDA?
It's coming back into vogue as we speak, and OSB is at the forefront. Over the next year, we will be making some additional analyses and proposals to try and move toward a reporting system that is more targeted, because we think we can probably do a better job.
What other activities do you foresee for OSB in the coming year?
This winter we'll be conducting conferences on electronic data interchange. As the MDR system currently operates, when an adverse event occurs at a hospital someone writes a report and sends it to the manufacturer. The manufacturer then transcribes it and sends it on to FDA, where we record it. Basically, we're entering 100,000 reports onto our computers each year, and it doesn't make any sense because most of those reports have already been entered electronically somewhere else.
It would make much more sense for an adverse event to be recorded using a uniform data description. The device user could then electronically transmit that information to the manufacturer, who would then conduct an investigation and add the results to the report. The whole file could then be transmitted to FDA. This could make everybody's life easier and more efficient. So we are moving toward trying to work with the device industry to develop electronic data interchange specifications.
Another area in which we expect to be very active over the next couple of years is in the development of uniform international requirements for the postmarket surveillance arena. I'm now chairing a study group on that topic for the global harmonization task force, and I'm very excited about it because I think this effort could also make life easier for manufacturers. If a heart valve is failing in Denmark, there's no reason that we shouldn't be able to find out about it before it starts failing in Poughkeepsie or Milwaukee. We're trying wherever possible to be extremely collaborative with the medical device industry.