Jonathan S. Kahan
- THE MAKING OF A REGULATORY PROFESSION
- THE EVOLUTION OF BIOMATERIALS LIABILITY
- THE CHANGING ROLE OF EDUCATION IN FDA'S REGULATION OF MEDICAL DEVICES
- A BRIEF HISTORY OF FDA GOOD MANUFACTURING PRACTICES
- FDA ENFORCEMENT POLICIES: A 20-YEAR PERSPECTIVE
- THE DEMISE OF RULE MAKING AND THE RISE OF GUIDANCES
For a good portion of the past 20 years, FDA worked in partnership with industry to ensure that safe and effective devices proceeded expeditiously to market while unsafe and ineffective devices were the subject of agency enforcement action. Unfortunately, the 1990s--a period some have derisively termed the "Kessler era of regulation"--have been extremely detrimental to the growth and profitability of the medical device industry. Many small companies simply could not survive the slowdown in device clearances and the additional layers of regulation imposed by FDA under Commissioner David Kessler. Only now, after years of regulatory dysfunction, is FDA entering a more rational and moderate era of device regulation, partly because of pressure from Congress and industry.
While changes at the highest levels of the government over the past 20 years have led to periods of "deregulation" and charges of "overregulation," there has been one constant for both FDA and industry: an emphasis on the safety and efficacy of medical devices. Different FDA commissioners and directors of its Center for Devices and Radiological Health (CDRH) have approached their jobs differently and have defined safety and effectiveness differently. Nevertheless, the overall performance of FDA in regulating devices--and the performance of those devices in the market and in the clinic--reflect a record of success. This article will discuss some of the more significant regulatory events and initiatives over the last 20 years and look briefly at what may be on the horizon for the next 20.
Prior to May 28, 1976, there was no federal premarket clearance requirement applicable to medical devices. Under the Federal Food, Drug, and Cosmetic Act (FD&C Act), FDA could take action against adulterated and misbranded medical devices--including the many quack devices that were on the market--but could not require companies to submit their products for review prior to marketing. In 1970, a government panel chaired by Theodore Cooper, then assistant secretary of the Department of Health, Education, and Welfare, was assigned the task of reviewing whether a federal program requiring the premarket clearance of medical devices was needed. The Cooper committee reported that an extensive literature search had disclosed 10,000 injuries related to medical devices over a 10-year period, of which 751 resulted in death. The panel additionally reported that 512 deaths and 300 injuries were attributed to heart valves, 89 deaths and 186 injuries to pacemakers, and 10 deaths and 8000 injuries to intrauterine devices (IUDs).
Notwithstanding the Cooper Committee report, it took six years of effort in Congress for the Medical Device Amendments of 1976 to be enacted. Under the new 1976 law, Congress assigned FDA the task of classifying the thousands of devices then on the market into either Class I, Class II or Class III, depending upon the risk presented by the device. The underlying concept of the law was that only the most risk-laden devices, those in Class III, should require the highest level of government review and approval. The notion that devices should be regulated according to their risk was conceived to avoid the problems that Congress and FDA had encountered in the drug approval process, where an infamous "drug lag" in approvals had developed.
The medical device classification process has essentially served its purpose over the past 20 years. Most devices that have presented little risk have expeditiously proceeded to market through the 510(k) premarket notification process rather than through the more burdensome premarket approval (PMA) process. The Congress of 1976 would be a bit surprised, however, about how Class II devices have been treated by FDA. FDA essentially ignored the congressional mandate that performance standards be drafted for Class II products. Of the approximately 1100 devices that were placed in Class II by the medical device classification panels, only one performance standard, the infant apnea monitor standard, has ever been promulgated by the agency. Consequently, for the past 20 years Class I and Class II devices have essentially been treated the same, subject only to the general control provisions of the FD&C Act. Not until it passed the Safe Medical Devices Act of 1990 (SMDA) did Congress seek to remedy this anomaly. Nevertheless, all in all, the classification scheme devised in 1976 has worked well, and the idea of stratifying regulation according to risk has survived the test of time.
The 510(k) Notification Process. The drafters of the Medical Device Amendments of 1976 might also be surprised at how the 510(k) premarket notification process has evolved over the last 20 years. The 510(k) process has become the primary premarket clearance route for the medical device industry. Each year, approximately 5000 510(k) notifications are filed with FDA, which finds 98% of the devices substantially equivalent and clears them for market entry. However, the 510(k) process has evolved from what was originally conceived to be a true "notice of intent to market" scheme to now include, in many cases, a de novo review of the safety and efficacy of the product.
In the early days after passage of the Medical Device Amendments, 510(k) notifications would consist literally of a few pages describing the device and the predicate product to which it was claimed to be substantially equivalent. Very few, if any, performance data were submitted to support the substantial equivalence of the device except in the in vitro diagnostics (IVD) area, where information regarding the sensitivity, specificity, and accuracy of the product was typically part of the 510(k) submission.
By contrast, the 510(k) notices of recent times are often laden with performance data (including mechanical, animal, and often clinical test data) as well as biocompatibility or electromagnetic compatibility information, depending upon the nature of the technology. There is no specific regulation that authorizes FDA to require that 510(k) notices be supported by clinical data. But not long after the passage of the 1976 amendments, FDA conceived what has often been termed the hybrid 510(k) or mini-PMA. For these hybrid submissions, the agency requires that the 510(k) filing be supported by clinical data demonstrating that the device is safe and effective for its intended use, and that any technological differences between the device under consideration and the predicate device do not raise new questions of safety and efficacy.
In some cases, these hybrid 510(k) notices walk, talk, and squawk like PMA applications. Over the years, typically 24% of 510(k) notices filed were supported by clinical data. That number in more recent years has increased to 10% or more. Notwithstanding the increase in the number of data-supported 510(k) filings and the increasingly detailed amount of data required in the submissions, the 510(k) clearance scheme has served industry and FDA well over the last 20 years. At times, the process has been broken and filings have gotten stuck in a terrible logjam. But overall, despite these bad times (mostly during the Kessler era), the process has worked and the practice of using hybrid submissions has given FDA the flexibility to clear some fairly novel devices through the 510(k) process rather than requiring PMA applications.
The PMA Process. In contrast to the 510(k) process, the premarket approval path to market has become progressively more difficult for industry to traverse since passage of the 1976 amendments. All PMA applications since 1976 have been supported by some manner of clinical data demonstrating the safety and efficacy of the device for its intended use. Rather than becoming a flexible and understandable route to market, however, the PMA process has become a progressively more complicated, expensive, and extended one.
The PMA process has typically been reserved for novel and breakthrough devices that often have the greatest impact on saving or sustaining lives or that lead to new standards of medical care. Obtaining premarket approval for these important devices has been especially difficult during the Kessler era. In 1987 the average approval time for PMAs was approximately 337 days, having dropped from 395 days in 1986. In contrast, the average PMA review time in 1991 was 633 days, and was on its way up to 840 days in 1994. The difficulties experienced in the PMA arena are partly due to the evolution of the clinical trial data requirements necessary to support both premarket approval and the clearance of hybrid 510(k) filings.
One of the most significant developments in medical device regulation over the past 20 years has been FDA's evolving approach to how clinical trials must be conducted in order to provide valid scientific evidence to support both 510(k) and PMA clearances. From 1976 to approximately 1992, most PMAs were supported by studies that were historically controlled, without randomization, and often without the masking of the patient, the treating physician, or other data evaluator. For many years, this standard of clinical support appeared adequate, and there were no widespread failures of devices in the marketplace. Although there were certainly instances of device failures--such as the problems encountered with temporomandibular joint (TMJ) implants, Bjork-Shiley heart valves, and some ventilators--most devices cleared through the PMA and 510(k) processes performed safely and as intended.
Despite the good record of FDA device review and of product performance in the marketplace, when David Kessler became FDA commissioner in 1990 he became extremely concerned about the quality of reviews within the CDRH Office of Device Evaluation (ODE). It was his view that, although ODE might be able to perform appropriate engineering reviews of devices, CDRH lacked the appropriate medical and scientific expertise to review and clear the more sophisticated implants and other new technologies that would be facing FDA in the future. Kessler's concern over the breast implant controversy led to highly publicized hearings and ultimately to a 1991 Federal Register notice requiring manufacturers to provide patients with information about breast implant risks.
Kessler's concerns also led to his creation of the Committee for Clinical Review, led by Robert Temple, a drug evaluation expert from the Center for Drug Evaluation and Research (CDER). In 1993, after reviewing a sample of PMA applications, IDEs, and 510(k) notices, the Temple committee concluded that there were numerous deficiencies in the design, conduct, and analysis of clinical trials performed by sponsors in support of their product applications. The significant criticism expressed in the Temple Report led to an FDA preference for controlled, randomized, and masked clinical trials (where feasible) in support of all PMA applications and most data-supported 510(k) notices. The Temple Report also led to a huge increase in the number of physicians within ODE, reflecting Kessler's emphasis on clinical issues rather than merely on engineering issues in device review.
While no one argues that appropriate scientific rigor should not be applied to clinical trials, the problem encountered by many device companies over the last five years has been that the agency's clinical trial requirements have been constantly shifting. Study sponsors feel as though they are shooting at a moving target. FDA has responded to this criticism by issuing more and more guidances, which, while helpful, have in many cases led to further difficulties. The guidances have often lacked adequate industry input, and some ODE reviewers have applied them as if they were regulations. This struggle to define appropriate clinical data support for approvals has had a large impact on both the nature and timing of device premarket clearances. Today, industry and FDA are still engaged in discussions to resolve issues that are centered on clinical data requirements and guidances.
One of the most significant changes in medical technology over the last 20 years has been the development of device software and its widespread application across different types of products. FDA policies on device software have evolved in response to this technological revolution.
Immediately after the passage of the 1976 amendments, FDA regulated medical device software on a case-by-case basis with no consistent ODE policy about what information was needed for agency review. In the mid-1980s, the rapid computerization of medical devices and the development of expert or knowledge-based systems, artificial intelligence, and other types of computerized products led FDA to rethink its ad hoc approach to software regulation.
The first policy relating to device software was the "Policy for the Regulation of Computer Products" issued by ODE in 1989. That policy applies to stand-alone software products for medical purposes and is still the controlling guidance for determining whether and how such software is regulated by FDA. In determining whether stand-alone software is a medical device or deserves full FDA device regulation, ODE applies the criterion of "competent human intervention." If there is competent human intervention (such as a diagnostic recommendation) between the software output and the diagnosis or treatment of the patient, the software is typically not subject to premarket clearance or strict device regulation. FDA is considering revisions to this seven-year-old policy, and changes are expected to be announced this year.
In 1988, FDA issued another important policy statement in its "Reviewer Guidance for Computer Controlled Medical Devices Undergoing 510(k) Review," a revised draft of which was published in 1991. This guidance was designed to ensure uniformity in ODE's review of submissions for software-controlled devices where the software or firmware is part of the parent device. Although the document specifically relates to software discussed in 510(k) notices, FDA has also applied the guidance to IDE and PMA submissions. The underlying philosophy of this guidance is that the greater the concern presented by the device and its software, the greater the level of FDA scrutiny and the greater the amount of software detail and documentation required by the agency.
FDA's initiatives on medical device software have had a significant impact on the premarket clearance process for medical devices, enabling the agency to regulate hundreds of devices that are controlled by computers or contain microprocessor components. FDA has appropriately set its level of concern to match the risks presented by device software, and has provided industry with useful guidance on the requirements for verification and validation of device software and for failure mode and effects analysis. While the agency's policies are still evolving, there is no question that they reflect an approach to software regulation that has been a resounding success.
The Safe Medical Devices Act of 1990 was the first major revision of the medical device requirements since 1976 and, while it did not constitute a radical departure from the previous regulatory structure, it has had a significant impact on medical device regulation over the last six years.
SMDA was enacted after four years of effort by Congressmen Henry Waxman (DCA) and John Dingell (DMI), who firmly believed that FDA needed new powers to appropriately regulate the medical device industry. While it is not possible in this article to address all of SMDA's provisions, several warrant special mention.
SMDA defined substantial equivalence for the first time and essentially codified the substantial equivalence decision-making structure that had been adopted by ODE in 1986. In order to determine that a device is substantially equivalent--and thus subject to 510(k) clearance--FDA must find that it has the same intended use as the predicate device and, by order, that the device: (1) has the same technological characteristics as the predicate device; or (2) has different technological characteristics but that the 510(k) submission contains information--including clinical data if deemed necessary by FDA--to demonstrate that the device is as safe and effective as a legally marketed device, and does not raise questions of safety and efficacy different from those of the predicate device. This provision also allows FDA to require that the manufacturer provide an adequate summary of any information relating to safety and efficacy as part of its 510(k) submission.
SMDA also sought to resolve the problem of Class II devices for which no performance standards had been drafted by the agency. Under the new law, FDA was given authority to order special controls for Class II devices beyond the performance standards allowed under the 1976 amendments. The available special controls include performance standards, postmarket surveillance, patient registry, development and dissemination of guidelines (including guidelines for the submission of clinical data and premarket notification submissions in accordance with section 510(k)), and other appropriate actions as FDA deems necessary.
SMDA also included provisions enabling FDA to implement postmarket surveillance, civil penalties, temporary suspension of PMA applications, required reports of removals and corrections, user and distributor reporting, and medical device tracking. Some of these provisions will be discussed in sections below.
The thrust of SMDA was to shift some of FDA's emphasis from premarket reviews to postmarket regulation and surveillance. The underlying theory was that this streamlining would enable the agency to expedite premarket clearances and approvals. Unfortunately, SMDA was enacted at the same time that Commissioner Kessler was increasing scrutiny of clearances at ODE. The combined result was a slowdown in premarket clearances at the same time that postmarket requirements were being increased.
One of the major regulatory requirements imposed upon the medical device industry is the obligation of medical device reporting (MDR), which became effective on December 13, 1984. Under the MDR regulation, manufacturers and importers are required to submit a report to FDA whenever they become aware of information that reasonably suggests that a death or serious injury has occurred and that one of their marketed devices may have caused or contributed to that event. Additionally, a report must be filed if one marketed device has malfunctioned and a recurrence would likely cause or contribute to a death or serious injury. Pursuant to this regulation, hundreds of thousands of MDR reports have been processed and reviewed by the agency. In fiscal year 1995 alone, for example, FDA processed 87,884 reports from manufacturers.
SMDA extended similar device reporting obligations to distributors and user facilities. SMDA requires user facilities to file a report with FDA within 10 days of learning of a reportable death and to file a report with the manufacturer upon learning of a reportable death or serious injury. The user facility reporting requirement went into effect by operation of law on November 28, 1991. The "new" final regulations relating to both user and manufacturer reporting were published on December 11, 1995, and will become fully effective on July 31, 1996.
When they were first proposed, in 1990, the user facility reporting requirements were very controversial. Many in industry and at FDA argued that these requirements would simply result in more paperwork and constituted double reporting of information that was already in FDA hands through manufacturer reports. Congress persisted, however, and enacted the requirements. The general consensus appears to be that user facility reporting has been a notable failure, providing little helpful information to FDA. Moreover, the requirements are frequently ignored by most regulated facilities; in fiscal year 1995, for example, only 4779 user facility reports were received by the agency, with another 3793 reports coming from health-care professionals and private citizens.
Similarly, the SMDA requirements for distributor reports, which became effective by operation of law on May 28, 1992, have also been essentially ineffective and duplicative. In fiscal year 1995, FDA received only 2459 distributor reports. On December 11, 1995, the agency announced that it intends to revoke the distributor adverse event reporting requirement and replace it with provisions based on notice-and-comment rule making.
While controversial and burdensome, the MDR requirements have served their purpose of providing FDA with information concerning the postmarket performance of devices. Nevertheless, many have questioned whether FDA has optimally processed, analyzed, and used this information. In a 1989 hearing, for example, Congressman Henry Waxman stated that FDA's handling of MDR reports was "floundering," with only a fraction of medical device firms that would be expected to report problems actually doing so. Since that time, however, it appears that device industry compliance with the MDR regulation has improved considerably. The MDR manufacturer reporting system appears to be a device program that, while flawed in many respects, has essentially served its purpose over the last 12 years. The same cannot be said for the user reporting requirements, which should be repealed.
Over the past 20 years, FDA's enforcement initiatives in the medical device area have been fairly uniform and predictable--though there are a few notable exceptions. FDA has consistently sought to seize or enjoin the manufacture of products that it believed presented a risk to the public health or constituted an economic fraud. Some have argued that FDA often sues device companies even when no serious public health risk is apparent or when there have been no true infractions of the law. Upon review of 20 years of enforcement cases, however, it can be fairly said that FDA has actually been quite selective in the cases it has brought. Typically, the agency has not brought suit unless good manufacturing practice, fraud, or safety issues have raised serious concerns about the ability of a company to produce and market a safe and effective product. This is not to say that FDA has not litigated some cases that were ill-considered and without adequate support in law and fact. But the agency's overall enforcement record remains pretty good.
To obtain compliance, FDA typically relies upon warnings to manufacturers rather than attempting to litigate every controversy. When the device company has understood the warning, it has usually been able to work out its differences with FDA outside the courtroom. Unfortunately, FDA and device manufacturers do not always communicate well or understand one another.
In the last 20 years, one of the major enforcement policy changes that have reduced understanding between FDA and industry has been the demise of the notice of adverse findings (NAF) and the regulatory letter. In 1991, FDA's regulatory procedures manual (RPM) was revised to replace the NAF and the regulatory letter with a single warning letter. This step was taken because Commissioner Kessler believed that industry was ignoring the agency's NAF letters and that greater compliance could be obtained by using a document that did not telegraph whether FDA would take action or not. Kessler wanted enforcement to be unpredictable.
Over the years, the impact of this change has been that companies have had to guess about whether a particular warning letter is serious. The importance of the warning letter has been devalued, and a greater number of misunderstandings and failures in communication have resulted. Moreover, FDA district offices now have greater discretion in the issuance of warning letters, which has led to less consistency within the agency about what conduct should be the subject of such a letter. Indeed, many warning letters appear to focus on trivial infractions. Far from meeting Kessler's expectation when the RPM was revised, the new procedure has led to less industry respect for the warning letter and has probably won less industry compliance than the old procedure enjoyed.
Although FDA enforcement policies have been fairly constant over the 20-year period since the passage of the 1976 amendments, the arrival of Commissioner Kessler led to a "ratcheting up" of the agency's enforcement activity. In 1991, the first full year after Kessler's arrival, the number of seizures increased by 109% from 32 to 67, and the number of injunctions sought by the agency increased 133% from 3 to 7. Over the last couple of years enforcement ac-tions have tended to return to their historical levels, and no new radical device enforcement policies appear to be on the horizon.
One of the agency's most controversial enforcement initiatives, severely criticized by both industry and Congress, has been the "reference list." First implemented in 1992, this policy enabled FDA to place a hold on the 510(k) clearance of any device manufactured in a facility where it had found GMP or other violations. Those violations are typically reflected in warning letters or FDA-483 observations.
In response to devastating congressional criticism of the reference list policy, FDA recently announced that it had essentially eliminated the list. In its place, however, the agency has implemented a restructured program that can still stall 510(k) clearances--so long as there is an "appropriate linkage" between the "serious" GMP or other deficiency and the pending 510(k) application. Although many in industry have argued that the reference list program--under any name or structure--is illegal and not supported by statute or regulation, FDA still appears to be wedded to the concept of linking 510(k) clearances to GMP deficiencies. Hopefully, through either court action or a future change in policy, this illegal linkage will be eliminated.
The power to levy civil penalties was another enforcement tool granted to FDA under SMDA. Since 1990, several civil penalty proceedings have been initiated under the new provision, but only one has actually been brought to fruition. In November 1993, Infusaid, a subsidiary of Pfizer, agreed to pay a civil penalty of $290,000 for shipping a modified infusion pump without first obtaining PMA supplement approval. To date, none of the other civil penalty initiatives brought by the agency has been resolved, including actions brought against SensorMedics and Lexcor Medical Technology in 1993. As many predicted when the civil penalties provision was first enacted, it remains likely that the penalty procedures will not be widely invoked by the agency because of the burdensome hearing process attendant to bringing such actions.
It is difficult to foresee what the next 20 years of FDA medical device regulation will bring. There are numerous proposals in Congress for FDA reform, ranging from minor modifications of the agency's present structure to the adoption of far-reaching changes. Some proposals would impose new deadlines on FDA for device reviews, while others would take device regulation out of the agency's hands entirely. In the end, Congress will likely be reluctant to drastically restructure a program that has essentially worked well for 20 years. Moreover, there is a strong argument that the problems that have arisen have been due not to the enabling law but to the way that CDRH and FDA have carried out their statutory responsibilities under that law.