The Nature and Extent of Research-Related Injuries 4 In pursuit of an answer to the underlying question at issue--should there be a compensation program for research injuries?--the preceding chapter established that some subjects may "deserve" such compensation if injured. In this chapter the Commission turns to the next part of the overall question and examines how many subjects are injured and how severly. Given the extent of Federal involvement in the research enterprise, surprisingly little is known about either the magnitude of the problem of injury to subjects or the nature of the injuries. The Commission sponsored two studies of the incidence of injuries at research institutions with existing compensation programs. These studies do not provide definitive answers at the national level, but are suggestive of the nature and scope of the problem. Taken together with data from other sources, they provide a basis for several modest, but important, conclusions: (1) the incidence of serious injury and the absolute numbers of people seriously injured are small; (2) most injuries are trivial in nature and require no medical intervention; (3) of those injuries that require intervention, most are only temporarily disabling; (4) most serious injuries and fatalities associated with research are more likely to result from underlying disease than from the research per se; (5) patient-subjects in therapeutic research are more likely than normal subjects in nontherapeutic research to suffer injury; and (6) the existence of compensation programs does not stimulate excessive or unmerited claims of injury. 71. Compensating for Research Injuries: Chapter 4 Before examining the data on research-related in-jury, it is important to note that there is no agreed-upon use of the term "research injury." No attempt is made in this chapter to specify the term definitively. Rather, in reviewing previous studies, the definitions used in each will be noted in order to avoid spurious comparisons. Many quite disparate definitions have been employed for various purposes. In different contexts, "injury" has been used to denote (or, in some instances, has merely not been clearly distinguished from) "adverse effect," "clinical event," significant medical event," complication," and "unanticipated consequence." 1/ Many studies employ some scale of seriousness to distinguish among the consequences of research. Such distinctions tend to be made along a dimension of disa-bility (i.e., trivial, temporarily disabling, permanently disabling or fatal) or to be operationalized in terms of the amount of medical care needed as measured either by the extent of the intervention (no treatment, physician intervention, referral, hospitalizations) or by the "cost" of that intervention (which may or may not include lost earnings as well as medical care costs). 2/ The lack of a uniform definition of injury makes it difficult to compare reports on the consequences of research. Keeping this potentially serious limitation in mind, the Commission analyzed data from many sources in order to try to determine whether injuries of any signif-icance (either in terms of frequency or seriousness) are occurring in research with human beings __________________ 1. In Phase 1 drug trials, for example, it may be appro-priate and necessary to note all "events," including consequences of the drugs and intercurrent illness, but such "events" are not the same as research-related injuries. Furthermore, some of the terms used to connote injury actually precede injury in time. That is, a "complication" may, if it is not handled correctly, become an injury and therefore be only an indirect consequence of the research. Although it is important for informed consent and IRB clearance to indicate what are considered to be possible (hence anticipated) consequences of the research, and it is important for scientific purposes to take account of unanticipated consequences of research, these are not the same as research-related-injuries. For discussions of injury, it is the seriousness of the consequence, not its predicta-bility, which is important. 2. Office of Technology Assessment, U. S. Congress. COMPENSATION FOR VACCINE-RELATED INJURIES (1980). Compensating for Research Injuries: Chapter 4 72 in such a way that a general compensation program for such research would appear to be needed. Correlatively, are the injuries that occur of such an extent that compensation could be paid without overburdening the research enterprise? It would seem that the most relevant data to answer these questions would come from institutions with compen-sation programs already in place. Since no published information was available from these programs, the Com-mission solicited reports from the University of Washing-ton in Seattle and the Quincy Research Center in Kansas City, Missouri. Data from these reports were compared with existing data from other sources, including a previously published longitudinal study of a drug testing facility in Michigan and broad-based data from multiple institutions. Longitudinal Institutional Studies The University of Washington in Seattle and the Quincy Research Center in Kansas City, Missouri, reported their data regarding numbers of subjects covered and the kinds of research in which they were involved, the nature and incidence of injury, the number of claims made (and paid) for medical care or compensation, and the cost of the insurance program. Data from the two programs are not fully comparable because the University of Washington is a general research facility while Quincy is a drug testing facility and because of different definitions of injury. Nonetheless, the programs are large enough to generate data that are suggestive of the extent and nature of research-related injury and these data seem to be corroborated by other reports. General Research Program: University of Washington Seattle. Diana McCann and John Pettit of the University of Washington at Seattle analyzed records for the years 1972-1981 during which an estimated 356,000 subjects on more than 5,300 protocols for biomedical research were covered by the university's compensation program. 3/ During that period, investigators were required to submit detailed reports to the Human Subjects Review Office on the subjects involved in their research and any adverse effects experienced by those subjects that were possibly ____________________ 3. Diana McCann and John R. Pettit, A Report on Adverse Effects Insurance for Human Subjects (1980); see Appendix H to this Report. The data were updated by Ms. McCann in February 1982. 73. Compensating for Research Injuries: Chapter 4 related to their participation in the research. 4/ McCann and Pettit reported that for the year prior to the establishment of the insurance program, survey data indicate adverse effects in 4.6% of the studies (5 out of 110) involving 0.07% of the subjects (10 out of 14,942). None of the subjects were partially or perman-ently disabled and no deaths were reported. This exceptionally low rate of injury could be accurate or may be partly an artifact of the retrospec-tive survey. Since the insurance program began, the university has had two means of reporting adverse effects internally. An "Adverse Effect Report" is used when there is a liklihood of a claim against the university, while a "Status Report" is required on an annual basis for all adverse effects. The former has been used and are summarized for nine years; the latter have been in effect for only four years. Although the reports provide details about the nature of the adverse effects, they are often vague with respect to the numbers of subjects who experienced particular adverse effects, making it very difficult to calculate incidence rates. Fortunately, for purposes of this Report, most of these imprecisions appear in reports of effects that were merely transient and trivial. A further difficulty in interpreting the data reported by McCann and Pettit arises from the ambiguouos use of the phrase "adverse effect." Whether an adverse effect means both "anticipated" and "unanticipated" con-sequences or just unanticipated consequences has not been resolved by the institution. At present, current policy allows individual investigators to report whichever they prefer. In the adverse effects and status reports, therefore, the term is apparently used in both ways. The Commission staff reviewed the descriptive information accompanying the data for each of the status reports; the incidents listed there as "adverse effects" were more numerous than those reported separately on "adverse efffects" forms. Omitting "trivial" injuries (such as headache,nausea, dizziness, soreness, and other 4. It is noteworthy that even where a compensation program exists at a single institution, the precise number of research subjects is not known. It should also be noted that subjects participating in behavioral re-search have been excluded from the compensation program since the second year of the program. Compensating for Research Injuries: Chapter 4 74 transitory symptoms not requiring medical intervention) the staff calculated that in the first eight years 144 subjects (or 0.04% of the estimated total of 356,000) experienced temporary disability; none were permanently disabled and two patient subjects died. Most of those who experienced injury were patient-subjects with existing disease. Reports on 34 "adverse effects" of research (invol-ving 42 subjects) were filed with the university's Human Subjects Office in the first nine and one-half years of the compensation program, but only 18 claims for compen-sation were submitted. The small size of this figure is probably due to a combination of factors. First, al-though subjects are told that medical care is available in case of injury, they are not specifically informed of the existence of the formal compensation program; second, claims are normally initiated by investigators; and fin-ally, there are informal mechanisms for handling some injuries. About the final point, McCann and Pettit report: It is known that some medical care and other professional services are provided to subjects who experience both anticipated and unanticipated adverse effects for which no claim is made against the adverse effects compensation program.... How such care is financed or accounted for is not reported to a central office, therefore this information is not available. The following means are likely: time donated by the health professional; costs absorbed by a grant or contract, as appropri- ate, e.g., the Clinical Research Center, and reciprocal services provided within the health professions. 5/ It has taken some time for the program to become well established at the university, and the rate of claims has picked up in the last several years. In all cases in which an injury is found to have been related to the research procedure, some payment has been made. Most have been very modest (from $2 to $249 for expenses); until recently, the largest payment made was $1,550. A claim for an injury that resulted from contamination of the equipment used in an experiment has now been settled for $10,000. Six of the 42 injuries were reported as "not related" to the research project; only in one case did the subject file a claim, which is ____________________ 5. McCann and Pettit, supra note 3, at 27. 75. Compensating for Research Injuries: Chapter 4 the only claim of of the 18 in which payment has not been made. Nontherapeutic and Therapeutic Drug Testing: Quincy Research Center. A second report was prepared by John Arnold, M.D., Director of the Quincy Research Center in Kansas City, Missouri, a drug testing facility whose subjects have been covered by a workers' compensation program since 1975. 6/ Because much of the research at Quincy is performed to support applications for new drug licenses, careful reports are kept of the number of subjects involved in each protocol, the duration of their participation in each research project and all "clinical events" (including intercurrent illnesses such as colds, flu, appendicitis, drunkeness, toothache, eye surgery, etc.) Data reported by Dr. Arnold are from 151 Phase 1 (nontherapeutic) projects involving 2,596 normal volunteers, and 78 Phase 2-4 (therapeutic) projects involving 2,478 patient volunteers. 7/ Simple frequency counts of undifferentiated "clinical events" as well as "significant clinical events with sequelae" are reported for both study types and subject categories. Both the total number of clinical events and the number of serious events are greater in Phase 2-4 studies with patient-sub-jects than in Phase 1 studies with normal volunteers. Patient-subjects are more likely than normal subjects to withdraw from protocols because of significant clinical events and, although the incidence rates are very low, patients are more likely to be hospitalized (1.43% vs. 0.2%) and to die (0.24% vs 0). It appears that most of the clinically significant side effects occurred because of preexisting disease in patient-subjects rather than as a direct consequence of the experimental drugs or participation in the research. None of the six deaths and only seven of the 36 hospital- __________________ 6. John D. Arnold, Incidence of Injury During Clinical Pharmacology Research and Indemnification of Injured Research Subjects at the Quincy Research Center (1980); see Appendix 1 to this Report. 7. Phase 1 studies are used to determine toxic dosage and pharmacological actions of drugs such as metabolism, absorption and elimination. Phase 2 trials are conducted on a limited number of patients to evaluate specific disease treatment or prevention. Phase 3 and 4 studies involve extensive clinical trials to assess safety, effectiveness and optimal dosage levels for treating specific diseases. The phases proceed sequentially. Compensating for Research Injuries: Chapter 4 76 izations could conceivably have been research related, whereas three of the five hospitalizations for normal volunteers could have been as a consequence of the research. From these data Dr. Arnold concluded that the conduct of Phase 2 through 4 protocols is "inherently beset by greater incidences of primary disease states in outpatient participants" 8/ for three reasons. First, the testing of drugs and devices for "efficacy" requires preexisting disease states in selected participants; second, testing under these protocols involves research participation of greater than average duration; and third, the subjects tend to be of more advanced age. A second conclusion from the data from the Quincy Center derives from the fact that normal volunteers are more likely than patient volunteers to be referred for medical care as a result of significant clinical events. In addition, protocol termination and alteration of protocol designs are more likely in Phase 2 studies than in Phase 3-4 studies. The available data on significant clinical events with participant and protocol sequelae support the conclusion that the incidence of drug-and participation-related events is higher in Phase 1 protocols employing normal participants by reason of heretofore undiscovered drug effects that are toxic to a greater or lesser extent to human subjects. By employing normal adult male subjects for the conduct of Phase 1 protocols, these events were more easily and immediately resolved via therapeutic intervention and additional physician consultation and resulted in a minimization of the amount of actual risk to participants and additional sequelae such as early termination from protocol and/or hospitalization. Thus, Phase 2 data indicate that normal male participants were able to tolerate the ultimate hazards of early Phase 1 drug testing without serious sequelae, as evidenced by findings of no deaths and few hospitalizations but increases in participant referrals for additional medical care and physician consultations. We believe the same drug-related complications would have been fatal or very serious in ill outpatients. 9/ _________ 8. Arnold, supra note 6 at 12. 9. Id. at 30. 77. Compensating for Research Injuries: Chapter 4 A third suggestion from Dr. Arnold's data is that the number of clinical events is directly proportional to the length of exposure. The average number of days on protocols was almost twice as great for patients than for normal subjects (25 vs. 14 days) and on the average, patients had 35% more clinical events per participant. 10/ Obviously, existing illness is a confounding variable in trying to ascertain the relationship between risk exposure and injury; the illness itself can worsen over time thereby leading to its own effects and rendering the patient-subject more susceptible to injury. When injury occurs, it may be difficult to differentiate whether it is due to the preexisting illness or to the intervention. Nontherapeutic Drug Testing: Michigan State Prison. A previously published study of research injuries was conducted by Dr. Chris Zarafonetia and other members of the Protocol Review Protection Committee for the State of Michigan Department of Coirrections. 11/ They examined clinically significant adverse effects in a Phase 1 drug testing program involving normal prison volunteers during a twelve-year period from 1964-1976. Records were re-viewed for 805 protocols involving 29,162 participants over 614,534 subject days. The authors reported 58 adverse drug reactions and six additional "complications" temporally related to the drug study, forming a total of 64 subjects (.2%) who experienced "significant medical events" (i.e., "the associated appearance of objective clinical signs or laboratory abnormalities; and sufficient discomfort, hazard, or potential haard to require physical inter-vention, e.g., to stop test drug, initiate appropriate therapy, and follow-up to recovery or other outcome"). None of the adverse reactions and only one of the com-plications were permanently disabling; one subject on placebo, died of cerebrovascular hemmorrhage while asleep. Thus, a clinically significant medical event occurred once every 9,602 days of subject exposure or 10. McCann and Pettit also reported longer average risk duration for patients than for normal volunteers although it was found to be highly variable (3 to 20 times greater) depending on the particular mix of research projects in any one year. ____________ 11. Chris J. D. Zarafonetia, Philip A. Riley, Park W. Willis, Lawrence H. Power, Judson Werbelow, Leo Farhat, Wendall Beckwith and Bernard H. Marks, Clinically Significant Adverse Effects in a Phase 1 Testing Program 24 CLIN PHARM & THER. 127 (1978). Compensating for Research Injuries: Chapter 4 78 about once every 26.3 years of individual subject participation. Although the data from the three institutional studies are quite consistent, it is important to compare these with data from a broader base in order to determine how representative these particular institutions are of the research universe. Broadly Based Sources of Data Since American society has been providing extensive governmental support for research for many years, a logi-cal source of data to answer questions about the overall incidence of injuries would be the Federal agencies that conduct, support or regulate research. Unfortunately, very little retrospective data is available from Federal sources. An alternative method would be to follow a prospective approach. Rather than count up the number of injuries that have occurred in a particular period in the past, one could estimate the number that can be expected to occur in the future. To do this, one would have to know the total number of human subjects participating in research, the portion involved in various types of research, and the probability and severity (collectively termed "risk") of injury associated with each type of research. Again, the facts needed to carry out this approach appear to be lacking. Government-Reported Incidence of Harm. The Commission has found that data on research-related injuries and on research subjects generally are extremely limited in terms of both the amount of information avail-able and its generalizability. 12/ Despite the very major role of the government in research, there is no comprehensive Federal mechanism for collecting data on injuries. In response to direct inquiries, officials from FDA and NIH testified that neither agency compiles such information. In addition, of the more than twenty other Federal agencies that conduct or support research with human subjects, only one (the National Bureau of Standards) was able to provide the Commission with information on either the nature or the incidence of ________________ 12. See the Commission's Biennial Report on PROTECTING HUMAN SUBJECTS, Government Printing Office, Washington (19761), for findings and recommendations on data on research participation and injuries. 79. Compensating for Research Injuries: Chapter 4 injuries experienced by subjects in research conducted under its auspices. 13/ Neither the government as a whole nor the individual agencies have data on the number or kind of injuries sustained by subjects of Federally conducted, supported, or regulated research. The only Federal attempt to collect and analyze data on research injuries was a special study conducted in 1976 by Philippe Cardon and his associates for the HEW Secretary's Task Force on the Compensation of Injured Research Subjects. 14/ In a telephone survey, investigators were asked to report the number of subjects involved in therapeutic and in nontherapeutic studies, the nature and incidence of injuries "that could be attributed to the conduct of the experimental regimen," and whether those injuries were experienced by subjects of therapeutic or nontherapeutic research. Injuries were classified as: trivial, temporarily disabling, permanent-ly disabling, and fatal. Investigators reported on a total of 132,615 subjects. Overall, 3.0% of the subjects experienced trivial adverse effects, 0.7% experienced temporarily disabling injuries, less than 0.1% were permanently dis-abled and 0.03% died. (All of the fatalities occurred in patient-subjects in therapeutic research.) Of the more than 39,000 subjects participating in therapeutic re-search studies, 10.8% experienced adverse affects or injuries most of which were trivial in nature. Only 2.4% of subjects were temporarily disabled, less than 0.1% were permanently disabled and approximately 0.1% died. Most of the 43 fatalities were not clearly related to the research. In fact, 37 (86%) of the reported deaths were in cancer chemotherapy trials. In the other categories as well, many of the "injured" were cancer patients who experienced familiar side effects of standard treatment. The incidence of injury for subjects participating in nontherapeutic research was even lower. Of 93,399 subjects, only 8.8% experienced injuries, most of which were trivial. Thirty-seven people (0.1%) were ____________________ 13. The agency reported two injuries since 1975; one subject who fell while testing emergency egress from mobile homes and one who allegedly injured his back during research to establish portability guidelines for the FTC. 14. Philippe V. Cardon, F. William Dommel, Jr. and Robert R. Trumble, Injuries to Research Subjects: A Survey of Investigators, 295 NEW ENG. J. MED. 650 (1976) (report prepared for the HEW Secretary's Task Force on the Compensation of Injured Research Subjects). Compensating for Research Injuries: Chapter 4 80 temporarily disabled, one person was permanently disabled and there were no fatalities. The authors correctly point out: "the data are a gross summation of the many interactions and perceptions of patients and subjects, principal investigators and probably others involved in the conduct of their research, the authors of the questionnaire and the telephone interviewers. Other approaches and assumptions might give different results." 15/ Furthermore, it is likely that in a retrospective telephone interview there will be some underreporting of injuries because of incomplete records, problems of recall, and unwillingness to disclose such information. How large a bias this introduces into the data is not known. These data are not, hwoever, inconsistent with those reported by McCann and Pettit for the University of Washington. Prospective Approach. An alternative method for determining the incidence and seriousness of research injuries would be to look forward rather than backward. Each research project entails certain steps or proce-dures. Some of those carry known risks. If one added to those risks an appropriate factor for the risks of any new and untested procedures--and for the risk, if any, of the particular and perhaps novel use of the known proce-dures in combination--it should be theoretically possible to project a risk for each type of research. When multi-plied by the number of subjects in each research activity, the result would be an estimate of the injuries expected. Again, unfortunately, neither basic data on numbers of subjects nor the more sophisticated numbers needed for risk estimates are available. Number of subjects. Very little information is available about the numbers of people serving as research subjects at any given time who are at risk of injury. Neither the funding agencies nor the recipient institu-tions are required to collect such information. Thus, although it is known, for example, that the Public Health Service (PHS) supports approximately 80% of the Federally funded biomedical research that is conducted throughout the country, the number of subjects involved in such studies is not known. This is also true of other Federal agencies supporting research. In the 1977 HEW Task Force Report it was estimated that approximately 800,000 subjets are involved annually in PHS supported clinical trials(i.e., controlled studies _____________ 15. Id at 653. 81. Compensating for Research Injuries: Chapter 4 of new therapies). 16/ Such trials are only a part of the research supported by the Public Health Service; a large amount of research involves studies of basic physiology, normal growth and development, and a variety of other inquiries utilizing normal volunteers. Other partial estimates include the Food and Drug Administration's figure of 375,000 subjects per year participating in research designed to test new drugs and medical devices; this figure is expected to increase as a result of the recent promulgation of regulations governing the testing of medical devices. 17/ It seems unlikely that a firm estimate of the number of subjects at risk for injury at a given time could be prepared from figures currently available; any such conclusions would be based on too many levels of approxi-mations, extrapolations and assumptions to be reliable. Thus, without either the numerator or the denominator with which to determine the incidence of research-related injuries, neither the absolute magnitude of the problem nor the size of the universe from which it emanates can be known with certainty. Subject characteristics. Although the PHS and FDA have not provided direct assessments of the characteris-tics of subjects and of the projects in which they are involved, some data were collected by the University of Michigan's Survey Research Center in the study of IRBs it performed for the National Commission for the Protection of Human Subjects in 1975. 18/ These data provide the best available description of the characteristics of subjects participating in biomedical and behavioral research. Principal investigators were asked to estimate the age, sex, racial and income distributions of their experimental subjects. Projects were then weighted, based on the number of subjects, to produce overall estimates of demographic characteristics of research subjects. _________________ 16. HEW Secretary's Task Force on the Compensation of Injured Research Subjects, REPORT, U. S. Department of Health, Education, and Welfare, Washington (1977) at VIII-2. 17. Information provided by personal communication with John C. Petricciani, M. D. Director, Bioresearch Monitoring Program, FDA (1980). 18. Robert A. Cooke and Arnold S. Tannenbaum, A Survey of Institutional Review Boards and Reserch Involving Human Subjects. Compensating for Research Injuries: Chapter 4 82 Table 1. Demographic Characteristics of Biomedical and Behavioral Research Subjects SEX Female 51.0% Male 48.0% RACE White 71.2% Black 20.7% Other 7.1% INCOME Higher 11.7% Middle 51.3% Lower 31.5% AGE Newborn 5.8% 3 months-6 years 9.2% 7-12 years 5.3% 13-18 years 7.6% 19-40 years 42.0% 41-64 years 20.8% 65+ years 7.4% Additional figures on subjects' characteristics are divided according to percent of projects rather than per-cent of subjects. Figure 1 shows the subjects participa-ting in therapeutic and nontherapeutic research and their source (by percent of project). Note that both patients and nonpatients are involved in therapeutic as well as nontherapeutic research. This apparent inconsistency is easily explained. First, research in preventive medicine is usually conducted on normal volunteers yet is consid-ered therapeutic (e.g., vaccines, controlled diet to pre-vent heart disease,flouride in toothpaste to prevent cav-ities). On the other hand, some studies are undertaken to understand a disease process although the subjects must necessarily be patients suffering from the disease, since the research provides them with no tretment, these studies are considered nontherapeutic. These data do not indicate the numbers of subjects involved in each kind of project. Although one might assume probability of injury occurring when a specific procedure is performed repeatedly "under similar circumstances...[I]njury occurs as a result of one or _______________________________________________________ Appendix to REPORT AND RECOMMENDATIONS: INSTITUTIONAL REVIEW BOARDS, National Commission for the Protection of Human Subjects, U. S. Government Printing Office, Washington (1978). 83 FOR REFERENCE SEE (16bb07.gif) 84 FOR REFERENCE SEE (16bb08.gif) 85 FOR REFERENCE SEE (16bb09.gif) 86. Compensating for Research Injuries: Chapter 4 complications associated with the performance of an invasive medical or surgical procedure." Most published reports of research projects in which these procedures were used either did not discuss compli-cations and injuries at all, or did so only in descrip-tive terms. Reports that did exist were incomplete and inconsistent thereby making it impossible to establish meaningful estimates of research risks. An attempt to extrapolate from the clinical literature to the research literature revealed further difficulties. Not surpris-ingly, retrospective reports of injury tended to be lower than prospective reports for the same procedures because of underreporting, low response rate, and insufficient data in existing medical records. Prospective studies tended to give a wide range of estimates. Finally, the circumstances under which procedures are performed affect the amount of risk involved. The skill of the investigators, familiarity with the proce-dures, equipment availability, institutional policies and the existing disease state of the patients are just some of the variables which influence the occurrence not only of the initial complications, but also their subsequent resolution. For example, it has been suggested that at least some procedures requiring sophisticated technology are best performed in hospitals that do many such procedures each year. 25/ Conversely, at least one study has raised the question "whether access to the latest obstetrical savvy and gear may lead to overmuch intervention, pos-sibly boosting the risk as well as the cost of having a baby." 26/ Thus although some procedures are inherently more dangerous than others, how much more depends on ___________________________________________ Twenty Invasive Proceedures used in Human Experimentation and Assessment of Reliability of Risk Estimates (1980); see appendix E to this Report. 25. Harold S. Luft, John P. Bunker and Alain C. Enthoven. Should Operations be Regionalized?, 301 NEW ENG. J. MED. 1364 (1979). 26. Small Hospitals Found Better for Normal Births, 22 MEDICAL WORLD NEWS 25 (December 7, 1981). Compensating for Research Injuries: Chapter 4 87 circumstances. Hence, injury can only be predicted in terms of estimated ranges of probable events. Conclusions It is evident from the preceding discussion that full data do not exist with which to answer the questions posed initially. One has neither broad-based retrospec-tive data on overall incidence of injury nor adequate data with which to calculate the extent of expected injury in the research universe because none of the components is known with precision. Federal agencies do not know how many subjects there are nor how they are distributed across the different kinds of research endeavors. What studies there are do show that most research involves minimal or no risk of physical harm. Yet it is also apparent that risk is a composite of many factors, such that for those procedures which do entail risk its magnitude cannot be specified. Although the evidence consistently suggests that the incidence of serious injury is small, nonetheless, it is clear that at least some subjects sustain injuries as a result of their participation in Federally funded or regulated research. For them, as for those who invite them to undertake risks on behalf of society, the question of compensation is real and of immediate importance. Furthermore, the studies of the Quincy and University of Washington programs demonstrated that compensation was paid at those institutions without overburdening the research enterprise. Finally, it is likely that only when there are functioning compensation programs--even on a limited, pilot basis--will the lack of data be remedied.