Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
3. The Development of Medical Devices Over the past quarter century there has been an acceleration in the development of new medical devices, in part because of rapidly expanding scientific and engineering knowledge (37~. In view of the reciprocal relationship between research and development, we will briefly consider the interactions between the reseach and invention phase of medical devices and the development phase. In addition to basic biomedical and clinical research, bioengineering research, which builds on advances in the physical sciences, mathematics, and engineering in other sectors, provides an important contribution~to the knowledge base . . . .. . . . . ~ . 2, . . : arc ~~oeng~neer~ng research predominantly takes place in universilv and government laboratories. In underlying meolca1 Device development O contrast to some European countries and Japan, funding for fur~damenta1 research in biomedical engineering is relatively small in the United States (e.g., I% of the NIH budget) and is dispersed through a number of agencies (135~. Compared with the U.S., the Federal Republic of Germany has nearly double the amount of space and equipment for bioengineering research (106~. Some recent efforts, however, may ameliorate this situation. The National Science Foundation, for instance, has established a program to fund high-risk fundamental bioengineering research. On the applied research side, the Small Business Innovation Research (SBTR) program was established in the early 19SOs by NIH to provide R&D grants or contracts to small businesses. According to an OTA analysis, 40% of grant applications in 1983 concerned medical devices and 23% of these applications were funded (~13~. Federal support is complemented by private investment, especially in applied research. In 1986, the medical device industry invested on average 7.5% of sales in R&D (~17~. In 1979, medical device firms in the five medical device Standard Industrial Classification (SIC) codes (X-ray and electromedica] equipment, surgical and medical instruments, surgical appliances and supplies, dental equipment and supplies, ophthalmic goods) reported that 3.7% of their 3] Bioengineering research will be defined as the application of engineering knowledge and concepts to the understanding of the human body, its interactions with machines, and to the development of new and improved medical devices. This definition is very similar to a definition provided in a recent National Research Council report (106), except that the scaling-up and production of new products derived from advances in biology (i.e. the engineering aspects of biotechnology) are excluded. Those aspects of engineering are discussed in the previous Chapter. 23
company-sponsored R&D budget was basic research and 23% was applied research (~13,137~. The investment in research depends on the type of device as weld as the kind of firm involved. As Spilker observes, medical devices are a much more heterogenous group of products than drugs in terms of design, use and purpose; many devices never come in contact with patients, some do briefly, and others do permanently (129~. There are roughly 1,700 different types of medical devices and 50,000 separate products (~13~. Clearly, the research required for the invention of disposable needles is very different from that for the invention of a C1~ scanned. Equally, there is much more variety in the kinds of firms that invent and develop medical devices than is the case with drugs. The industry is characterized by a large number of small firms; approximately 50% of U.S. medical device manufacturers have fewer than 20 employees. Large companies, however, dominate the industry in terms of sales (120~. According to Roberts, small firms and even individuals produce most of the innovations in the early stages of developing a new class of medical devices, whereas larger firms play an especially important role later on in the development process (sometimes through the acquisition of small firms). As Roberts put it, the invention of "medical devices is usually based on engineering problem solving by individuals or small firms, is often incremental rather than radical, seldom depends on the results of long-term research in the basic sciences, and generally does not reflect the recent generation of fundamental new knowledge. It is a very different endeavor from drug innovation, indeed" (120~. This observation, however, is not as easily applicable to radical innovations, such as those in modern imaging devices, which require large-scale investments in research and development. Such resource-intensive innovations usually take place in large firms. After a product is invented, a patent application may be filed. While patent protection is extremely important to the pharmaceutical research and development process -- partially because of the long duration of the R&D process and the relative ease with which drugs can be copied -- the value of patent protection in medical device development is much less evident. In the device area, it probably is easier to invent around a patent, and the research and development time is generally much shorter. Furthermore, with devices that require large capital costs, the need for large-scale investments may prevent competitors from entering the market. and small firms mav depend , ~_~ -be ~ -r 32 In view of the heterogeneity of medical devices, the type of device determines if animal research will be undertaken before a device prototype is evaluated in humans. 24
more on trade secrets. Whereas the potential users of new medical devices, i.e. the physician- researchers, may play an important role during the development process, they also may be crucial to the invention of medical device prototypes. Not only do they identify the clinical need for a new device or for improvements in existing devices, but they may also be the innovators and builders of the original prototype. Von Hippe! first described the importance of users in the invention of such scientific instruments as gas chromatography, nuclear magnetic resonance, ultraviolet spectrophotometry, and transmission electron microscopy (68~. He concluded that 80% to 100% of the key innovations in these four fields were originated by users and not the ultimate manufacturers. Von Hippe] and Finkelstein underlined the importance of users with regard to the automated clinical chemical analyzer (69~. For example, the initial prototype of an auto analyzer was developed by Skeggs in the pathology department of Case Western Reserve University; Technicon then made a licensing agreement with Skeggs to patent the auto analyzer and further developed and marketed the machine (147~. Shaw, who analyzed 34 medical equipment innovations in Great Britain presented similar results33 (126~. It follows that close interactions between clinicians and industry are important to the development of medical dev~ces34. Roberts and Peters, however, found that academicians in MIT physics, mechanical engineering and chemical engineering departments and two large research laboratories did not readily transfer their ideas for commercial development (121~. This finding was repeated in an analysis by Roberts (120) of two major medical centers in the Boston area, although this may change somewhat in the present-day climate, where universities and their medical centers are becoming more market-oriented (100~. These considerations affect industrial decision making. In the pharmaceutical industry, the decision to invest in particular research areas involves "potential demand" for a pharmaceutical as an important criterion. If the user-dominance paradigm of Van Hippe] plays an important role in some parts of the medical device industry, manufacturers decisions will be made later in the R&D continuum. The decision whether to pursue development of a prototype involves both technical and market factors35. 33 Shaw found that half of the initial prototypes were produced by users. 34 Allen (2) established the importance of intra-organizational (e.g., between R&D and manufacturing divisions) and inter-organizational communication for R&D performance. 35 With regard to the latter, a recent analysis of the development of devices demonstrated that half of the device firms considered used a formal 25
In most industrialized countries, the development of new medical devices is governed by regulatory schemes, either in the form of standards or extended pharmaceutical laws, which focus mainly on safety. In contrast, the U.S. has passed a specific law governing the development of medical devices. Prior to the passage of these Amendments to the FD&C Act in 1976, the FDA asked Arthur D. Little consultants to provide insight into the safety and efficacy testing practices of medical device firms (5~. This analysis revealed that most devices were tested during their development, but that the extent and nature of clinical evaluation varied considerably among products. In addition, there was considerable variety within product categories. For example, one developer of an artificial knee undertook clinical trials in 200 patients, another performed informal trials with 75, and the third used only 50 patients with no set protocol. Furthermore, in comparison to drug evaluations, the criteria of clinical evaluation may differ with new clinical devices. Criteria more often include user acceptability, either of the design or the reliability and ease of use in the clinical setting, and the competitive advantages of a new device versus alternative devices. Finally, in most cases evaluations did not include the classical randomized clinical trial, review by Institutional Review Boards (IRBs) and conducted with patients who signed consent forms. Because medical devices are a much more heterogenous group of products than drugs, it is understandable that some variation in clinical evaluations exist. The ex~s-tence of considerable variation within device categories, and the fact that half of the clinical investigations had no formal protocol, indicate room for improvement. In addition, the risks associated with some devices, such as certain cardiovascular implants or JUD's, became evident in the 1970s. The Cooper Committee was established to recommend device legislation. It proposed different levels of regulatory control based on the likelihood of risk inherent in specific classes of devices, with more rigorous regulation for devices with higher risk potential. In 1976 the Medical Device Amendments (Public Law 94-295) were passed to ensure that new devices were "safe and effective" before they were marketed (49~. These Amendments divide medical dev~ces36 financial] analysis of the expected returns on investment or at least some form of market survey. Many firms, however, relied on informal decision making processes, usually based on a firm's experience in the market for the product (114). 36 According to Kennedy (78), the term medical devices includes al] of the Items readily identified as devices as weld as in vitro diagnostic devices used in clinical laboratories and some products previously regulated by the FDA Bureau of Drugs, such as lUDs, or by the Bureau of Bio~ogics, such as arterial grafts. 26
into three classes (40,79~. Approximately 30% of all types of medical devices are in Class I. Class ~ devices include such instruments as tongue depressors, which do not support or sustain human life and do not present a potentially unreasonable risk of illness or injury. They are subject to the general controls used before passage of the Medical Device Amendments, such as regulations regarding registration, pre- marketing notification, record keeping, labelling, and GMP regulations. About 60 percent of devices are in Class Il. They may involve some degree of risk and are subject to federal~v defined nerfc~rmance standards couch ns X-rav devices). To date, however, no performance standards have been issued by the FDA, and existing national or international product standards apply. Finally, all devices that are life supporting or sustaining, that are of substantial importance in preventing impairment of health, or that have a potential for causing risk of injury or illness are in Class IlI. For these, the sponsor needs to demonstrate safety and efficacy before the FDA grants marketing approval. Approximately 10% of medical devices are in Class Ill, such as the artificial heart, DNA probes or laser angioplasty devices. According to the law, devices introduced since 1976 are automatically placed in Class Ill, unless the sponsor successfully petitions the FDA to reciassi~ it as "substantially equivalent" to a device that was on the market before the amendments took effect. The substantial equivalence provision has provoked uncertainty, as the law did not specify if this equivalence referred to safety and efficacy, or to equivalence of the physical characteristics of a device. FDA regulations issued in 1986 state that devices with new intended uses require pre-marketing approval. Post-amendment devices with intended uses similar to those of pre-amendment devices may be found to be substantially equivalent only if the new technological features of a -device can be shown not to decrease its safety and efficacy (79~. This may be demonstrated through descriptive, performance and even clinical data. This is called a 510(k) submission. If a device is found to be substantially equivalent, the manufacturer may rely on pre-marketing notification. This route to the market is much more expeditious than the pre-market approval route, Hand the impression is that sponsors wild attempt to change the design of devices accordingly. Indeed, a recent GAO report found that roughly 90% of medical devices reviewed by FDA were marketed through SiO(k) review, while 10% underwent the full pre-marketing approval process (55~. 27
To support a marketing approval decision, or in some instances a 510(k) submission, a sponsor is required to conduct clinical studies. Clinical investigations of devices are subject to the two basic elements governing clinical research in general: informed consent and institutional review. In comparison with drugs, however, Institutional Review Boards (IRBs) play a more important role in device evaluations. They review all clinical device studies, decide if the device poses a "significant risk" and approve clinical studies for their institutional. The IRB determines if a device poses a significant risk on the · · ~ e . ~ Cal basis ot an ~nvest~gat~ona~ plan. line plan Includes a description of the device, the objectives and duration of the investigation, the investigational protocol, a risk analysis, monitoring procedures, informed consent materials, and also identifies all involved IRBs. if a device poses a significant risk a request for an Investigational Device Exemption (IDE) is submitted to the FDA (73~. Such an IDE application contains the investigational plan, information on prior investigations, the manufacturing process, and the amount to be charged for the investigational device. In comparison to the 1,250 pages of an average IND, the average size of an IDE is 150 pages. After an IDE has been approved clinical investigations can be initiated38. Data from a random ten percent sample of IDEs submitted between 1980-1986 indicate that most clinical evaluative studies are concentrated in a few product categories; ophthalmic, cardiovascular and obstetrics/gynecology products account for nearly 60% of all TDE investigations. The range of products requiring an IDE, however, is increasing (17~. In contrast to pharmaceuticals, the final version of a ~nedical device is often not created "de nova"; instead a device prototype is usually modified technically as a result of initial clinical testing (129~. The period of learning necessary before a device can be used properly arid efficiently may be longer than with drugs. Therefore, according to Spilker, clinical testing usually first involves an initial pilot stage during which the prototype's design and materials are further 37 A significant risk device is legally defined as an implant and presents a potential for serious risk to the health and safety or welfare of a subject; is purported or represented to be for use in supporting or sustaining human life and presents a potential for serious risk to the health and safety or welfare of a subject; is for use of substantial importance in diagnosing, curing, mitigating or treating disease and presents a potential for serious risk to the health and safety or welfare of a subject; or otherwise presents a potential for serious risk. 38 Including also those cases where an IDE application is not necessary, but clinical trials~are conducted. 28
developed and tested. The main questions are whether the device produces the postulated effect in humans and whether it seems to be clinically useful. These evaluations usually are based upon non-formal experiments (see section 4 for discussion of the argument to randomize the first patient). In addition to clinical evaluations, this stage also involves technical testing; for example, the electrical and mechanical components of infusion pumps are subject to technical evaluations, and a number of bench tests are performed to determine a pump's accuracy and reliability. After the technical development has become more or less stabilized, a series of safety and efficacy evaluations of the 'final' 'initial' product can be initiated. This decision is sometimes a difficult one, as clinical evaluations usually reflect risks and benefits at a fixed point in time. Too early assessments may not reflect the true risks and benefits of an evolving device and the results of the study may be obsolete before the evaluation is completed, whereas the results of evaluations done too late in the life-cycle may be irrelevant for health care decision makers (see Chapter 4~. In medical device evaluations, a distinction needs to be made between diagnostic and treatment devices. With the former, it is usually not direct patient benefit, but benefits in terms of clinical utility (i.e. its contribution to further diagnosis or therapy) that are to be evaluated. Fineberg has formulated a hierarchy of criteria for diagnostic technology evaluations: technical capacity, diagnostic accuracy, diagnostic and therapeutic impact, and patient outcomes (48~. Generally, evaluations provide information on the technical and diagnostic performance (not the more comprehensive clinical utility or patient outcomes) of a diagnostic device, and possibly on its risks and complications. The main measures of diagnostic performance are sensitivity (ability of a test to detect disease when it is present) and specificity (ability of a test to correctly exclude disease when it is absent)39. In clinical practice, however, the question of interest is, if the patient has a positive test how likely is he or she to have a specific disease? (71) Therefore two additional measures, the predictive value of a positive test result (i.e. number of true positives/ true positives plus false positives) and the predictive value of a negative test result (i.e. number of true negatives/true negatives plus false negatives) play an important role. These measures indicate the likelihood of the presence or absence of a disease in a tested individual from a given population with a particular prevalence of the disease. In order to compare the sensitivity and specificity of two or more diagnostic devices, the receiver 39 Determining technical performance involves replicabili~cy and reliability as important criteria. 29
operating characteristic (ROC) curve analysis can sometimes be used40. However, ROC analyses require simple models and large numbers of cases, and Friedman observes that they are often very difficult to undertake (53~. Thus, most diagnostic tests are evaluated only in terms of their technical and diagnostic performance before marketing. Furthermore, according to Schwartz, the range of Datients tested mav he inadequate as they usua1Iv invasive those ·, ~ ~ ~ ~ e ~ ~. ~ ~ ~ ~ ~ _ ~ _~ ~ ~ e - wltn advanced disease, and a tew young healthy controls (123). A diagnostic test, However, may not perform as well with patients with earlier disease, which indicates the need for more comprehensive evaluations (454. As with drugs, the question here concerns what endpoints should be evaluated in pre-approva] and/or post-approva] trials. Traditionally, most device evaluations lack randomized control groups (129~. While this may in part be due to less sophistication in clinical research on the part of many device manufacturers, it also may be a result of inherent characteristics of device development that make the classical RCT more difficult to perform. The statutory standard recognizes this and is less rigorous than with regard to new drugs; i.e. safety and effectiveness information for devices may be provided through "well-controlled scientific studies" or through "valid scientific evidence". The randomized placebo-controlled, double-blind clinical trial, optimally suited to provide pre-marketing efficacy information on drugs and biologicals, indeed has more limitations with new devices. This holds especially for diagnostic but to a certain extent also for treatment devices; for example, a placebo may be unethical (as with heart valve replacements) or certain situations may not be amenable to observing a placebo effect (as when the patient is unconscious or the interaction between patient and device is minimal). Another essential characteristic of RCTs as used in drug evaluations, patient and physician blinding, may also cause more difficulties with devices. However, creative techniques to eliminate bias are emerging. For example, one physician may insert an implant device while another physician evaluates its benefits and risks. Thus if RCTs are possible at this stage their use or that of otherwise well-controlled study designs, such as parallel study designs or crossover designs, should be stimulated. Generally the sample sizes used in these clinical studies are considerably smaller than is the case with drugs. Ophthalmic IDEs, for example, called for an average of 280 patients, while all other IDEs involved about 150 patients (17~. 40 The ROC analysis allows one to compare the technical performance of diagnostic tests over a range of different cutoff points or reference values that denote a positive test result. This test displays the true positive ratios and the false positive ratios for these different cutoff points. See (92~. 30
ticket" devices. But after a decade some of the drawbacks of such planning laws surfaced, in part because the numbers of new devices expanded beyond the scope of regulation. At the same time, policy attention increasingly turned towards the payment method as an important too] for influencing the adoption and use of devices (67~. During clinical studies, much industrial effort may be directed towards 'scaling up' for production. The necessary production capacity may vary widely, ranging from 10 to 100,000 devices a year. Depending on the kind of devices, specific manufacturing requirements may exist, such as the need for sterility or for a- certain shelf life. Good Manufacturing Practice (GMP) Regulations4~ govern the manufacturing process in general. International and national standards may also exert an important influence on the manufacturing process, for instance those set by the Association for the Advancement of Medical Instrumentation or the International Electrotechnica] Commission (~13~. On the basis of the results of clinical investigations, a device may be approved for marketing42. In contrast to drug regulation, the device amendments require that advisory committees participate in the pre-marketing approval (PMA) decision for class Ill devices (9~. In general, the PMA is an individual license to the developer for a particular device. Other developers of similar types of devices need to submit a separate PMA, with adequate clinical data. Data of previously approved PMAs cannot be used, unless they are published and generally accepted by the medical community. This policy protects each manufacturer's investment in the development process, but it also may stimulate the duplication of investigational efforts, including the performance of unnecessary trials. However, the next model of a medical device often differs in materials and/or design, and these differences may affect clinical risks and benefits. Recently proposed Amendments to the 1976 medical device legislation would allow the FDA to waive data requirements for PMAs following that of the innovator. Adoption of these amendments could lessen the incentive for innovative R&D. An important decision point in the course of development concerns the adoption of a new device by physicians and hospitals, which is influenced by a complex set of medical, economic, regulatory and social factors (63,124~. In the 1970s a number of health planning laws, such as Certificate of Need (CON) laws and rate regulation, were enacted to control the adoption of "bip- C7 41 One needs to distinguish between critical and non-critical devices. Most rigorous GMP regulations apply only to critical devices. 42 On average the FDA takes a year to approve a PMA (79~. 31
Most industrialized countries are moving away from a cost-based, essentially open-ended reimbursement system to a prospective payment system (PPS). This transition has probably been most prominent in the United States with the establishment of Medicare's prospective payment system for hospitals, based on diagnosis-related groups (DRGs). Under PPS hospitals have a strong financial incentive to provide the least resource-intensive treatment. The system promotes a significantly lower level of growth in service intensity than traditionally has been the case, and the recalibration of DRGs is lagging behind changes in medical practice (4~. Although the price system is intended to be neutral under PPS, this is not always the case. For example, the lithotriptor was covered as a medical treatment for kidney stones under DRG 323. But this DRG pays only half as much as DRG 30S, for the surgical treatment of kidney stones. Thus, although lithotriptors may improve the quality of care and may be cost-effective for some indications, hospitals have less financial incentive to invest in these machines (104,3~. Also, because PPS deals only with payments for inpatient hospital care, there is an incentive for hospitals to utilize technologies that are cost-effective over the short term of hospitalization. There is little incentive for hospitals to use technologies which have long term benefits, even though they may ultimately have a greater impact on the efficiency of the system as a whole. As the existing reimbursement system affects the market for new medical products, changes in this system may exert strong feedback signals to the development process, e.g. it has been observed that medical device manufacturers react to the demand for products that are cost-effective over the short term and neglect R&D projects dealing with products that are cost-effective over the longer run43. With these changes in reimbursement, the coverage decision, e.g. by the Health Care Financing Administration (HCFA1. has become a more important factor in the rl~velnnment nrnr`~44 ~ ,, .~_,~^ a. i.,_ ---it Air . Traditionally, the coverage decision making process was based on generally subjective evidence provided by medical expert panels; increasingly, however, forma] evidence becomes the basis for these decisions (105~. This evidence includes safety and efficacy considerations 43 As mentioned before, the economic environment in general and cost analyses of devices in particular are outside the scope of this paper. 44 The statutory provision indicates that this decision should be based on whether a device is considered "reasonable and necessary", which has been translated to mean "accepted by the medical community as a safe and efficacious treatment for a particular condition". Based on 13 technologies that completed the full Medicare coverage process (including technology assessments by OHTA) from the 1983-~8 period, it took 2.4 years from the time that HCFA received the initial inquiry to the final disposition date. 32
narrowly defined, but there is a tendency to consider the effect of devices on the quality of life of patients (including their preferences for certain outcomes) and their cost-effectiveness. As such, the change in decision-making provides an important incentive to undertake evaluative studies after pre-marketing notification or FDA approval. Phase [V studies include a number of nost-marketin~ surveillance mechanisms to rlPt~rt arts - rem - ri`~`nr" r - artery - Q45 .~ ~_.__. ~~_~`h~ ~~- A~lVllO . Me FDA maintains a Device Experience Network (DEN) that receives reports on device hazards from health professionals and manufacturers. Device manufacturers are required to keep records of complaints as part of GMP regulations. On the basis of adverse reaction reports, the FDA may require removal of designated devices from the market or restrict their sale or use. In comparison to drugs, acute injuries are probably more easily associated with a particular device. A major issue, which needs to be examined, is whether the adverse reaction or event is a consequence of the skill of the professional or inadequate maintenance of the device, or can be attributed to a defect in the device itself (24~. In addition to these surveillance mechanisms, a number of epidemiological methods may be used to detect possible risks of device use. As discussed in Chapter 2, the potential of using observational methods for risk detection is increasing. In addition, information on effectiveness is needed; such information can be provided by experimental or observational studies. Because the life cycle of a device is short and next-generation versions of a particular device may emerge relatively quickly (as with diagnostic pregnancy kits, for instance) the applicability of RCTs may be more limited. An advantage of using modern observational data bases is that they represent continuous monitoring of the use of devices in practice and their outcomes. Uncertainty, however, remains as to the strengths and weaknesses of these methods in providing reliable evidence (62~. 45 A condition of the approval for new Class IT! devices is that information received by manufacturers on device defects or adverse reactions should be reported to the FDA within 10 days. 33
