Postmarket Surveillance Context

POSTMARKET SURVEILLANCE CONTEXT

For the majority of marketed products, no, or very limited, clinical data are required. Of 783 Class I device regulations (each of which typically pertains to more than one device), 720 (92%) are exempt from premarket notification. Similarly, of the 898 Class II device regulations, 75 (8%) are exempt. For the Class I and II products requiring premarket notification, many applications do not include clinical data. Even when clinical trial information is provided (for Class III devices), these data have some of the same inherent limitations noted in drug trials [i.e. limited size, dura-tion, and select patient population (e.g. restrictions in age, gender, disease complexity)]. In addition, inves-tigators in premarket device clinical trials tend to be those physicians at the ‘cutting edge’ of product devel-opment and who are most familiar with the device’s characteristics and application. Thus, limited informa-tion may be generated on human factor concerns such as optimal design for ease of use, optimal use envi-ronment (e.g. free of electromagnetic interference), labeling that anticipates less sophisticated use or that minimizes maintenance error, or the consequences of re-use on device performance and safety. Once in the marketplace, devices are likely be used by a wide array of physicians and other clinicians of varying skill levels, training, and experience. In addition, less stringent diagnostic and other criteria may be applied reflecting either non-optimal product choice or off-label use, the latter a hallmark of the evolving practice of medicine.

Since no device is free from adverse events and product problems and since premarket clinical data are limited, postmarket oversight is needed as a ‘safety net’ to ensure the continued safety and effectiveness of marketed products. Postmarket oversight refers to both postmarket surveillance (and risk assessment) as well as postmarket enforcement. The former refers to the systematic process of adverse event/product prob-lem reporting, monitoring, and evaluation as well as the subsequent, more formal, assessments of identified potential patient risks. The latter refers to investiga-tions of a device firm’s compliance with statutory and regulatory requirements. Both processes are integral to product development and evolution. This chapter will focus on the FDA programs constituting postmar-ket surveillance.

GOALS

As with drugs, the goals of device postmarket surveil-lance and risk assessment are: (1) identification of previously unknown or not well-characterized adverse events/product problems (‘signals’); (2) identifica-tion and characterization of sub-groups at risk; (3) collection and evaluation of information on issues not directly addressed in premarket submissions (e.g. long-term effectiveness or changes in use envi-ronment, from professional to home use); and (4) development of a public health context to interpret these data. This process ultimately aims to disseminate information regarding newly emerging device prob-lems to appropriate stakeholders (particularly health professionals and the public), incorporate the infor-mation into the device approval process, and provide findings to the device industry to aid in product corrections and improvements. The principal postmar-ket ‘tools’ utilized by the agency to achieve these goals are: (1) adverse event/product problem report-ing [through the Medical Device Reporting (MDR) system and MEDWatch, PMA conditions of approval, the pilot Medical Device Safety Network (MedSuN), and international vigilance]; (2) mandated postmar-ket studies (including condition of approval and Section 522 studies); and (3) applied epidemiology.

ADVERSE EVENT/PRODUCT PROBLEM REPORTING

MDR and MEDWatch

The FDA monitors postmarket device-related adverse events/product problems (AEs), through both volun-tary and mandatory reporting, to detect ‘signals’ of potential public health safety issues. Voluntary report-ing to the FDA began in 1973 and presently continues under MEDWatch (Kessler, 1993), a program created in 1993 to encourage voluntary reporting by all inter-ested parties (but principally among healthcare profes-sionals) as a critical professional and public health responsibility.

It was not until 1984 that the FDA implemented mandatory reporting as per the MDR regulation. This regulation required device manufacturers and importers to report device-related deaths, serious injuries, and malfunctions to the FDA. Additional legislative initiatives in the 1990s resulted in significant changes to mandatory reporting. Under the Safe Medical Devices Act of 1990, universal reporting of adverse events by user facilities (hospitals, nursing homes, ambulatory surgical facilities, outpatient diag-nostic and treatment facilities, ambulance services, and health care entities) and distributors was enacted. Under the FDA Modernization Act of 1997 (FDAMA, Section 213 of the Act), and in response to expe-rience with distributor and user facility reporting, the US Congress mandated that distributor reporting be repealed and that universal user facility report-ing be limited to a ‘subset of user facilities that constitutes a representative profile of user reports ’. The conceptual framework for these ‘sentinel sites’, collectively referred to as the Medical Device Safety Network (MedSun; formerly Medical Product Surveil-lance Network), is discussed below.

To better understand reporting of AEs under the current MDR regulations governing mandatory report-ing [Title 21 Code of Federal Regulations (CFR) Part 803], requirements should be noted and terms defined. Manufacturers and importers are currently required to submit reports of device-related deaths, serious injuries, and malfunctions. User facilities are required to report deaths to the FDA and deaths and serious injuries to the manufacturer. Seri-ous injuries are defined as life-threatening events – events that result in permanent impairment of a body function or permanent damage to a body structure, and events that require medical or surgical interven-tion to preclude permanent impairment or damage. Malfunctions are defined as the failure of a device to meet its performance specifications or otherwise perform as intended. The term ‘device-related’ means that the event was or may have been attributable to a medical device, or that a device was or may have been a factor in an event, including those occurring as a result of device failure, malfunc-tion, improper or inadequate design, poor manufac-ture, inadequate labeling, or use error. Guidance is issued to reporting entities as needed to more clearly define the reporting of specific events, for example implant failures.

Since its inception in 1973, the FDA’s database of voluntary and mandatory reports of device AEs has received slightly more than 1 million reports and currently averages approximately 200 000 per year, with mandatory reports accounting for about 98% of the total. The reports are submitted on the same standardized voluntary and mandatory forms used to submit drug-related events and capture information on device specifics (e.g. brand name, model number), event description, pertinent dates (e.g. event date), and patient characteristics. The reports are also coded (either by reporters or inter-nally) using a coding thesaurus of patient and device problem codes. Manufacturers also supply meth-ods, results, and conclusion codes relevant to their report investigation. To enhance report handling and signal detection, the FDA has established methods of triage:

•   emergency reports (e.g. a cluster of deaths or seri-ous injuries in a dialysis facility) are handled under agency-wide standard operating procedures;

•   pre-designated high priority reports are reviewed within 24 hours of receipt and include, among others, reports of pediatric death, exsanguination, explosion/fire, or anaphylaxis;

•   other individual reports (account for about 24% of all reports) are reviewed within 5–15 workdays of receipt;

•   autoscreen reports (account for about 12%) are those that are computer-screened (by pre-designated device and event) where events are considered to be familiar, but text may be particu-larly valuable in assessing event or events that are coded inconsistently; 10% of screened reports are later individually reviewed; and

•   summary reports (account for about 64%) capture well-characterized and well-known device events and amount to a quarterly submission by manu-facturers of line-listed data. The data elements per event include the manufacturer, model-specific device, event and receipt dates, and patient and device problem codes. A system is being developed to perform automated numerator-only trend analyses looking for month-to-month vari-ation, monthly moving averages, and 12-month trends.

When potential hazards are detected (either based on internal individual or aggregate review) or upon noti-fication by the manufacturer (under voluntary recalls), denominator data can be obtained from manufacturers upon request. The denominator data most appropriate to the analysis tend not to be generic higher-order data (such as number manufactured of that brand during the past year) but typically are model-specific, many times lot-specific (and thus time-specific), and may be sub-group-specific (e.g. pediatric use). Compli-cating the selection of appropriate denominator data are the myriad types of devices (e.g. single-use disposables to multi-component durable medical equipment) and the inherent difficulties in assess-ing potential population exposure (e.g. factoring in multiple uses, average shelf-life, component replacement).

A staff, predominantly of nurses, review the individual reports from a variety of perspectives including the potential for device failure (e.g. poor design, manufacturing defect), use error (e.g. device misassembly, incorrect clinical use, misreading instructions), packaging error, support system fail-ure, adverse environmental factors, underlying patient disease or co-morbid conditions, idiosyncratic patient reactions (e.g. allergy), maintenance error, and adverse device interaction (e.g. electromagnetic inter-ference) (ECRI, 1998). Since many devices involve complex human interaction, great emphasis is placed on human factor considerations. Simply put, these considerations ask: (1) To what extent did sub-optimal device design, packaging, or labeling induce human error? (2) To what extent was anticipated use (and abuse) of the product factored into device design, packaging, or labeling?

Several immediate actions, aside from routine requests for follow-up information, may be taken by the staff and include:

•   Recommending directed inspections of manufac-turers. These may lead to: (a) label changes, includ-ing those affecting device instructions or training materials, (b) product modification/recall, and (c) rarely, product seizure or injunction.

•   Recommending internal expert safety meetings. These may lead to public notifications, recommen-dations for additional postmarket study, or meet-ings with the company to explore issues further.

•   Alerting regulatory authorities outside the United States through the international vigilance program (see below).

Other internal uses of the AE data are widespread and include: input into premarket review (by provid-ing human factor insights and information on prod-uct experience in the general population); input into recall classifications (involving a hazard evaluation based on AE data); monitoring of recalls (and assess-ing reports in similar products); input into product reclassifications and exemptions from premarket noti-fications (based, in part, on a product’s safety profile); use in, and initiating of, standards efforts that establish device performance; educating the clinical community through newsletters, literature articles (peer-reviewed and professional and trade journals), and telecon-ferences; and as a general information resource for healthcare providers and the general public.

A recent example of reports of AEs typifies the system in action. In June 2002, the agency received reports of bacterial meningitis in patients with cochlear implants for treatment of hearing loss. Early speculation by manufacturers and implanting surgeons implicated the implant positioner (a Silastic wedge that is inserted next to the implanted electrode to facilitate transmission of the electrical signal by pushing the electrode against the medial wall of the cochlea). The one manufacturer that made implants with a positioner voluntarily withdrew their prod-uct both in Europe and the US in July 2002. Other manufacturers, however, notified the agency of addi-tional cases of meningitis, principally in children. A nationwide collaborative investigation was begun by the agency and the Centers for Disease Control and Prevention (CDC) that involved several thousand implanted children. These children were found to have far greater risk of developing pneumococcal menin-gitis compared to children in the general population, and those with positioners had over four times the risk of developing meningitis compared to recipients of other cochlear implant types (Reefhuis et al., 2003). Throughout this process, the agency posted periodic updated public health notifications on its website to keep the public informed (Food and Drug Adminis-tration, 2002, 2003). In addition, the CDC Advisory Committee on Immunization Practices added cochlear implant recipients to the list of high risk patients need-ing routine immunizations (Center for Disease Control and Prevention, 2003).

As is typical of passive surveillance systems (including those for drugs), the FDA’s system has notable weaknesses as well as strengths. Among the former are:

•   data may be incomplete or inaccurate and are typi-cally not independently verified;

•   events are under-reported – causes include lack of detection and/or attribution of device to event, lack of knowledge about reporting system, liability concerns, perceived lack of utility in reporting, and limited feedback;

•   data reflect reporting biases driven by factors such as event severity or uniqueness, familiarity with reporting, or publicity and litigation;

•   determination of incidence and prevalence is not possible due to under-reporting and lack of denom-inator data; and

•   causality cannot be inferred from any individual report. [In addition, devices are often not returned to manufacturers for assessment (for a variety of reasons) and therefore failure analyses of data are often inadequate or lacking.]

The system strengths are:

•   it provides nationwide safety surveillance from a variety of sources, thus providing insight into AEs related to ‘real world’ use;

•   it is relatively inexpensive considering the scope of surveillance;

data collected are uniform in terms of a standard-ized form with pre-specified data elements;

•   it is one of only a few means to detect rare AEs; and

•   it is accessible and the information is open to the public.

Supplementing this reporting system are PMA condi-tions of approval (applies to Class III devices). All products with approved PMAs have conditions of approval, one of those being the submission of infor-mation on AEs outside the MDR regulatory require-ments [Title 21 CFR Part 814.82 (a)(9)]. Examples of this include labeled AEs occurring with unex-pected severity or frequency. This requirement helps the agency cast a wider ‘safety net’ in its surveillance of AEs.

MEDICAL DEVICE SAFETY NETWORK (MEDSUN)

Although user facility reporting was mandated in 1990, it accounted for only 3% of all reports in 1999. Furthermore, only about 2000 reports came from hospitals in 1999, representing about 800 hospi-tals out of a universe of about 7000. Likewise, only 90 reports came from nursing homes, representing 50 nursing homes out of a universe of about 12 000. This lack of mandatory institutional reporting has many root causes (some alluded to above under weaknesses of AE reporting), but basically reflects a lack of educational outreach coupled with a lack of enforce-ment (with both tied to inadequate resources). Recog-nizing the need for user facility reporting but also the difficulties behind universal reporting, the US Congress mandated under FDAMA 1997 that report-ing be limited to a ‘subset of user facilities that constitutes a representative profile of user reports ’. Since 2002, FDA has been collecting data about problems with the use of medical devices from a sample of hospitals and nursing homes via MedSun. By mid-2005, this interactive Internet-based report-ing program expanded to approximately 350 health-care institutions (mostly hospitals) nationwide. The program’s principal objective is to increase the util-ity of user facility reporting by recruiting a cadre of well-trained and motivated facilities and establish a collaborative effort to better understand device use in its natural clinical environment. It is envisioned that, in addition to enhancing the detection of emerg-ing device problems, the network acts as a two-way communication channel between the FDA and the clinical community and serves as a setting for applied clinical research on device issues. To succeed, the effort must: train staff in the recognition and reporting of AEs, assure confidentiality to reporters, minimize burden of participation, and provide timely feedback. To achieve its mission, MedSun staff have initiated a variety of efforts within the network: monthly newsletters (highlighting device reports, FDA actions, and other notable safety initiatives by other agen-cies); clinical engineering audioconferences; device safety exchanges (highlighting best safety practices and safety solutions); and surveys on high-profile safety concerns.

INTERNATIONAL VIGILANCE REPORTING

The reach of AE surveillance was augmented and truly became global under the auspices of the Global Harmonization Task Force (GHTF) established in 1992. The GHTF was established to respond to the increasing need for international harmonization in the regulation of medical devices (www.ghtf.org). The GHTF is a voluntary international consortium of public health officials, responsible for administering national medical device regulatory systems, and repre-sentatives from regulated industry. The task force acts as a vehicle for convergence in regulatory practices related to ensuring the safety, effectiveness and qual-ity of medical devices and promoting technological innovation as well as facilitating international trade. This is principally accomplished through publication and dissemination of harmonized guidance documents on basic regulatory practices.

One of the four GHTF study groups is charged with reviewing current adverse event reporting, post-market surveillance and other forms of vigilance for medical devices, and performing an analysis of differ-ent requirements with a view to harmonizing data collection and reporting systems. A process for the global exchange of vigilance reports between National Competent Authorities (NCAs) has been established. Standardized reports on potentially high risk issues for which action is to be taken (even if investiga-tions are incomplete) are submitted electronically to a shared listserver. General and specific criteria for categorizing issues as high risk have been established and include: the equivalent of US Class I and high level Class II recalls, all public health notifications, and special public health concerns (e.g. high index of preventability or particularly vulnerable populations). Currently, the program exchanges approximately 150 reports per year.

GLOBAL MEDICAL DEVICE NOMENCLATURE

Part of the information requirements for the vigilance exchange program includes the official name of the device that is the subject of the vigilance report. Only since 2001 has the medical device community had an official international source for such names, the Global Medical Device Nomenclature (GMDN). The GMDN, developed through a major international stan-dards effort, was created largely via the merging and evaluation of six extant naming systems (including the one used by the FDA). Currently, version 3 of the GMDN has 8000 primary terms that abide by speci-fied naming rules and conventions as well as defini-tion structure and content (e.g. incorporating intended use). The GMDN is based on the level of specificity of the ‘device group’, which is best described by way of example, that is pacemaker, cardiac, implantable or gastroduodenoscope, flexible, fibreoptic. It is meant for use by regulatory agencies, but has the potential for wider applications (e.g. inventory control or market-ing) and may eventually be incorporated into admin-istrative and healthcare databases that could be used for public health purposes. When compared with the National Drug Classification coding system, however, the GMDN is more limited in that it does not at present include model-specific information, or other potentially useful data such as material composition, component parts, or size.

MANDATED POSTMARKET STUDIES

Another ‘tool’ that the FDA uses to achieve its surveillance and risk assessment goals are mandated postmarket studies, conducted under either PMA conditions of approval (for Class III products) or FDAMA (Section 522) authorities. A sponsor may be required to perform a post-approval study as a condition of approval for a PMA [Title 21 CFR Part 814(a)(2)]. The study questions may relate to longer-term performance of an implant, or focus on specific safety issues that may have been identified during review of the product for which additional informa-tion is felt to be needed, postmarket. Results from these studies may be included as revisions to the prod-uct’s labeling (including patient- and clinician-related material).

In addition to the PMA authority for Class III prod-ucts, the agency may, under Section 522, impose postmarket study requirements on certain devices. The latter provision, originally mandated in 1990 under SMDA, allows the agency, under its discre-tion and for good reason, to order a manufacturer of a class II or class III device to conduct a post-market study if the device: (1) is intended to be implanted in the human body for more than one year; (2) is life-sustaining or life-supporting (and used outside a device user facility); or (3) failure would reasonably be likely to have serious adverse health consequences. Although this discretionary authority overlaps the PMA post-approval authority for some products (e.g. PMA Class III implants), it effectively extends FDA authority to cover non-PMA products as well, that is those subject to premarket notifi-cation. Unless there are unusual circumstances, the Section 522 authority is typically reserved for the latter.

Prior to issuing an order, the FDA will discuss the public health concern with the firm. The concern may arise from questions about a product’s long-term safety, about performance of a device in general use or involving a change in user setting (e.g. professional to home use), or notable AEs. Upon receiving an order, the firm has up to 30 days in which to submit their study plan and, by statute, studies are limited to 3-year patient follow-up (or longer if agreed to by the firm). The FDA recently issued a regulation clearly specifying, among other items, the requirements for a study plan, conduct, and follow-up (Title 21 CFR Part 822).

The FDA has issued guidance on criteria used in considering order issuance as well as possible study approaches (October, 1998; www.fda.gov/cdrh/ postsurv/index.html). Briefly, the criteria include: the public health issue must be important; other postmar-ket mechanisms cannot effectively address the issue; the study must be practicable (i.e. feasible, timely, not cost-prohibitive); and the issue is of high priority. The possible study approaches vary widely (designed to capture the most practical, least burdensome approach to produce a scientifically sound answer) and include: a detailed review of complaint history and the literature; non-clinical testing of the device; telephone or mail follow-up of a patient sample; use of registries; observational studies; and, rarely, random-ized controlled trials.

Generally speaking, these mandated postmarket studies (both via PMA conditions of approval and Section 522) require the participation of both firms and the clinical community. Problems, however, may arise in the conduct of these studies if, for instance, it is difficult to recruit physician investigators or accrue patients or if industry lacks incentive. These issues particularly resonate with rapidly evolving technolo-gies, where rapid device evolution may make stud-ies of prior models obsolete by the time they are completed.

Although there may be difficulties in study conduct, an example of a Section 522 study reveals the author-ity’s public health importance and its risk assessment role. In 1991, FDA scientists demonstrated that it was possible for polyurethane to break down under labo-ratory conditions to form 2,4-toluenediamine (TDA). TDA had been shown to be an animal carcinogen. Prior to this it was thought that breakdown could only occur at very high temperatures and pH extremes. The firm that manufactured polyurethane foam-coated breast implants ceased sales in 1991 and agreed to a clinical study under Section 522. The study involved comparing TDA levels in urine and serum samples from women with and without the implants. Although minute amounts of TDA were found in the majority of women with the implants, the increase in cancer risk was determined to be vanishingly small (1 in 1 million) (Hester et al., 1997; DoLuu, Hutter and Bushar, 1998). The FDA issued a public health corre-spondence (FDA Talk Paper) on the results and their reassuring implications (Food and Drug Administra-tion, 1995).

APPLIED EPIDEMIOLOGY

Postmarket surveillance and risk assessment would not be complete without epidemiology, a discipline that provides the means and methods to further elucidate a device’s postmarket safety and effective-ness in a population context. Through employing methods of observational (as opposed to experimen-tal) study, epidemiologists help refine AE signals, characterize sub-groups at risk, test hypotheses, and evaluate device performance and use. The epidemiol-ogy program serves a vital postmarket function at the agency and works to inform Center and agency device policy, address relevant scientific questions, assess the effectiveness of regulatory approaches, provide risk assessments, develop new postmarket surveil-lance and other data resources, and provide important public health information (e.g. through peer-reviewed publications). Importantly, as of 2005, the program has been given oversight of post-approval studies (i.e. those as a condition of approval of PMA prod-ucts). It is now the program’s responsibility to help design, implement, track, and oversee completion of these studies of high-risk devices. To accomplish this, the program works collaboratively with product manufacturers and the premarket staff.

To accomplish its overall mission, the epidemiol-ogy program makes use of a variety of databases (e.g. the National Inpatient Sample to evaluate in-hospital mortality associated with heart valve replace-ment; Astor et al., 2000) and develops device-specific supplements to nation-wide surveys (e.g. US National Mortality Followback Survey to assess characteris-tics of persons receiving pacemakers in their final year of life; Hefflin, 1998). In addition, the program explores new means of surveillance [e.g. through a nation-wide surveillance network of emergency departments operated by the US Consumer Prod-ucts Safety Commission (CPSC); Hefflin, Gross and Schroeder, 2004], explores methods of active surveil-lance (in a large tertiary hospital; Samore et al., 2004), develops and expands existing device registries (e.g. exploring device safety using the American College of Cardiology National Cardiovascular Data Registry; Tavris et al., 2004), reviews and assesses observational literature (e.g. studies of cellular phones and their relation to brain cancer), and conducts applied research (e.g. breast implants and rupture rates) (Brown et al., 2000).

The ability of drug or device epidemiologists within the agency to address issues, however, is at times limited for both practical and regulatory reasons. There may be practical resource limita-tions (e.g. limited staff or limited funding) or time constraints (i.e. issues requiring immediate resolution may not lend themselves to observational study). Limits imposed by the regulatory environment are most apparent when mandating postmarket studies. The agency levies these studies on specific manu-facturers of specific products. In doing so, there is no intent for comparative analyses, or pooled analy-ses, amongst manufacturers of similar products. Nor is there any intent on assessing cost effectiveness, or conducting other economic analyses, since this is not within the agency’s mandate.

Other practical limitations, with regard to medi-cal devices, have to do with the type of infor-mation available from extant data sources. Many of the data sources used by pharmacoepidemiolo-gists (e.g. hospital-based, public health-based as in Saskatchewan, or health maintenance organization-based) may not have device-specific information, whether at the ‘device group’ level such as an ultrasonic rigid laparoscope or carbon dioxide surgi-cal laser or certainly not at the model- or brand-specific level. Other data sources, such as medical care claims records, often collect procedure-specific, but not device-specific, information, leaving one to infer device use. Compounding this situation is the relative lack of data sources for assessing device exposure and difficulties in deriving the most appropriate denomi-nator data (as noted previously with regard to AEs) (Bright, 2000).

These limitations not withstanding, epidemiology continues to play a vital role in addressing agency device concerns. The role of epidemiology is exem-plified by the following two cases. On the basis of concerns about use and performance of transmyocar-dial revascularization, a new and not fully understood technology, the program undertook a collaborative effort with investigators who oversee the Society of Thoracic Surgeons National Adult Cardiac Surgery database (Peterson et al., 2004). The study findings noted large scale off-label use and higher operative risks in patients with a recent myocardial infarction and unstable angina. Potential reduction in mortal-ity was suggested through optimization of timing of the procedure. The epidemiology program was also involved in assessing the public health impact of the only marketed continuous glucose monitoring system in the US (Tavris and Shoaibi, 2004). A thor-ough review of the literature suggested that use of the system could result in a substantial reduction in morbidity and mortality associated with diabetes.