- IRB: ETHICS & HUMAN RESEARCH
Competing Commitments in Clinical Trials
Scholars of the ethics of clinical trials have long recognized a tension between the therapeutic obligations1 of clinicians (physician, nurses, and other health professionals) and the scientific demands of clinical trials.2 The basis for this concern is clear: clinical care is focused on improving the condition of the presenting patient, while clinical research seeks valid, generalizable information to help future patients. Although in principle both the investigational and control arms of clinical trials are designed to have an equal likelihood of efficacy for the typical patient, in many situations research and clinical goals may nonetheless be in conflict. For example, during the recruitment process for a clinical trial, clinicians may have to decide whether to offer participation to a patient who might benefit from the study intervention, but who does not meet the trial’s eligibility requirements.3 And once a patient enrolls in a trial, clinicians have to decide whether to deviate from protocol requirements (e.g., dosage of study medication, use of adjunctive medications, or procedures to collect required study data) for the purpose of promoting the patient-subject’s best medical interests. Finally, clinicians involved with clinical trials may have to decide whether to keep a participant in a trial who, despite technically meeting a termination criterion, seems to be benefiting from ongoing participation.
Whatever the scope of a therapeutic obligation concerning clinical trials, the potential for the clinical commitments of physicians, nurses, and other health professionals to undercut the requirements of a research protocol has largely been ignored. With few exceptions,4 ethical and policy discussions of this issue have centered on how research imperatives may interfere with clinical care, not how clinical imperatives may lead to decisions that could compromise the validity of research results. The scant literature on gaps between protocol design and adherence to the protocol assumes that protocol violations are matters of research integrity motivated by selfish interests,5 rather than matters related to therapeutic obligation. This is not just a theoretical issue. In some circumstances, clinicians’ decisions that are motivated by concerns about a patient-subjects’ medical interests may threaten the validity of the research, waste the financial resources invested as well as the efforts of researchers and subjects, and lead to invalid data entering the medical knowledge base.
Although there has been considerable theoretical debate about potential conflicts between research and clinical care, there are no empirical data to indicate how clinicians actually perceive and resolve any tensions that arise between their commitments to the best medical interests of their patients and to protocol requirements of a clinical trial. There is, however, evidence from cancer trials that clinicians prioritize patients’ therapeutic needs in deciding whether to offer them participation in a clinical trial. Some studies report that many physicians refer a patient to a clinical trial on the basis of whether they perceive the patient is likely to benefit from participation.6 Winn et al. summarized these findings: “The physician’s perception of the nature of a research protocol and its potential benefits and detriments to his patient’s condition may therefore be a significant factor in determining whether a physician referral will transpire.”7 Though laudable from the point of view of clinical ethics, in some cases such selection factors might lead to selection bias and therefore limit the generalizability of trial results.
These considerations lead to a concern that recruitment biases or clinically motivated protocol deviations may compromise the validity of clinical research. In light of the dearth of empirical data addressing these issues, we sought to determine how often clinicians confront such conflicts and how they address them when they arise.
Study Sample. We used Centerwatch.com, an online listing of over 30,000 clinical trials that, at the time of the survey, was the largest listing of active clinical trials, to obtain a sample of individuals to invite to participate in our study. The trials listed on Centerwatch.com contained the names of 4,682 unique contact persons. E-mail and postal addresses were available for 3,318 of these persons. Telephone calls to listed contacts who only had telephone numbers resulted in e-mail addresses for an additional 352 potential participants, for a total of 3,670 potential survey participants. From this pool we drew a random sample of 1,000 potential subjects. Because the percentage of physicians in this sample was inadequate (11%) to permit assessment of whether physicians’ opinions differed from those of nonphysicians, we selected an additional 250 physicians who were listed as study contacts, for a total initial sample size of 1,250 study contacts.
Survey Instrument and Procedures . In order to inform development of the final survey instrument, we conducted 27 in-depth interviews with 12 physician-investigators, 12 nurse research coordinators, and three research coordinators who were not nurses. In these interviews, respondents described how and when clinical and research obligations conflict and when they felt it necessary to violate a research protocol in the interests of their “patients.” The interviews helped us to craft the wording of the final survey questions and pointed to the importance of nurses and other research coordinators—not just physicians—in day-to-day decision-making about the conduct of clinical research. After pilot-testing the draft survey with eight additional interviewees, we made several minor modifications.
The final survey comprised 37 questions on various topics, including respondents’ demographic, educational and training characteristics, the duration of their clinical trial experience, the roles that they had played, and the types of trials in which they had been involved. However, the primary survey items fell into two groups:attitudinal statementsandbehavioral reports. Eight attitudinal items queried respondents’ beliefs about the appropriateness of actions that, although they may conflict with good research standards in clinical trials, might benefit trial participants. The questions were framed as statements, and respondents were given the option to agree, mostly agree, be uncertain/neutral, mostly disagree, disagree, or to judge the statement not applicable.
Because attitudes may not accurately predict behavior in a tightly regulated situation like a clinical trial, we also asked respondents about their behavior during the prior two years. The behavioral items were organized in groups of three questions. For each topic, the first question asked how often they had faced a particular conflict between protocol requirements and the research subject’s best medical interests during the previous two years. After each opening question, two follow-up questions asked how often they had resolved the conflict by following the protocol, versus how often they had decided to do what they believed was in the subject’s best medical interest.
Survey Methods.The survey was conducted utilizing the online survey site SurveyMonkey. An Internet-based survey was appropriate because most clinicians use the Internet and have e-mail access. The survey was conducted in six steps using a modified version of the Tailored Design Method:8 1) a hand-signed, mailed letter; 2) an e-mail with a link to the survey; 3) a reminder postcard; 4) a repeat e-mail with a link to the survey; 5) a personal telephone call; and 6) an e-mail sent from our study e-mail address (rather than the automated online system) to minimize the possibility that recipients’ spam blockers prevented our e-mails from reaching them.
Of the 1,250 study contacts in the initial sample, 216 had an invalid e-mail address. Of the 1,034 who had a valid e-mail address, 744 (72%) responded and completed at least some of the attitudinal or behavioral questions. Sample sizes for individual questions varied due to the nonresponse category. Three respondents who reported having no involvement with clinical trials in the last two years were excluded from the analysis, leaving a final respondent sample of 741. Frequencies, chi-squares, and analyses of variance were computed using SAS and JMP statistical packages.
Participants in the in-depth interviews gave written informed consent after study approval by the institutional review boards (IRBs) at the University of Massachusetts (UMass) Medical School and the Dana-Farber Cancer Institute. Because of its anonymous nature, the Internet survey was judged by the UMass IRB to be exempt from the requirement for human subjects review.
Most respondents were between 30 and 60 years old; 80% were female, and 88% were Caucasian. Nearly half were nurses (46%), whereas 20% were physicians. The remaining subjects mostly had bachelor’s or master’s degrees in biological or social sciences. Three-quarters had been involved with at least six clinical trials in the previous two years. (Tables showing details of the findings are available at http://www.umassmed.edu/uploadedfiles/cmhsr/publications/tablesoffindings.pdf.)
Most of the respondents said they had direct involvement with study participants. For example, 89% of respondents said they always or frequently participated in the consent process; 42% said they always or frequently referred patients to clinical trials; 45% always or frequently decided in which trial a patient should enroll; 78% always or frequently managed the care of trial subjects; and 90% always or frequently participated in data collection. Involvement in the design and reporting of trials was less common. Few respondents (11%) always or frequently participated in designing trials, 21% always or frequently participated in data analysis, and 12% always or frequently participated in writing articles. Not surprisingly, physician-respondents differed from other respondents in the activities they focused on in their clinical trial work. They participated in fewer trials (p < 0.0001) and were less likely to report involvement in collecting data (p < 0.0001). In contrast, they were more likely to participate in designing trials (p < 0.0001), analyzing data (p < 0.0001), and writing manuscripts (p < 0.0001).
Attitudes about Clinical and Research Imperatives . Respondents were presented with eight statements that addressed attitudes toward conflict between scientific and clinical imperatives. Two items addressed researchers’ and medical and research centers’ decisions about participating in clinical trials. Fourteen percent of respondents agreed or mostly agreed that “centers should choose which trials to participate in based on how much the trials contribute to science,” whereas 33% agreed or mostly agreed that “researchers should only participate in trials that are likely to help the subjects who take part.”
Three questions concerned which patients should be recruited to participate in clinical trials. Almost 20% of respondents agreed or mostly agreed that “even if patients are technically eligible for a trial, they should only be recruited if being in the trial will be in their best medical interests.” Likewise, 33% agreed or mostly agreed with the statement that “patients who are not doing well with standard care should be recruited most actively so that being in the trial can help them.” Finally, 33% percent agreed or mostly agreed that “when several subjects at a site do considerably worse than would be expected in ordinary care, that site should stop recruiting for that study.” These attitudes are not statistically significantly related to how often the respondent reports referring patients to trials or deciding which trial to offer to patients.
The last group of items asked about attitudes toward explicit protocol violations in the service of therapeutic objectives. Almost 64% of respondents agreed or mostly agreed with the statement that “researchers should deviate from the protocol if doing so would improve the subject’s medical care”; 83% agreed or mostly agreed that “the protocol should be used as a guideline rather than something to be strictly followed under all circumstances”; and 90% agreed or mostly agreed that “it is acceptable to disregard minor entry criteria if a patient will benefit from being in a trial.”
Respondents’ answers reflected some differences in attitudes between physicians and nonphysicians involved in research. Physicians were more likely to suggest that one should choose trials in which to participate based on the study’s contribution to science (p < 0.0001) and less likely to endorse choosing trials based on whether the study might help patient-subjects (p < 0.003). They more often endorsed ignoring minor entry criteria for clinical benefits (p < 0.005) and were less likely to agree that one should remove subjects from trials on the basis of clinical judgment (p < 0.0001).
Behavioral Responses to Conflict Regarding Clinical and Research Imperatives . Ninety-six percent of respondents reported experiencing at least one situation during the previous two years in which clinical and research commitments conflicted. Two questions addressed conflicts that arise during a clinical trial’s recruitment process. Almost 70% of respondents reported at least one instance in which it did not appear to be in an eligible patient’s best medical interests to enroll in a trial; 6% reported that this had happened 16 times or more. Of the respondents who reported this situation, 55% said there was at least one instance in which trial participation had not been offered to an eligible patient. Similarly, 69% of respondents reported at least one patient who was ineligible to participate in a trial, but for whom participation would have been beneficial. Of these respondents, 22% reported that the ineligible patient was recruited to participate, and 3% reported that this had occurred six times or more. Respondents who more frequently decided which patients should be recruited for a clinical trial were more likely to report both having patients who would benefit from being in the trial but did not meet eligibility criteria (p < 0.0001) and having patients who were eligible but whose medical interest was not served by being in the trial (p < 0.0001).
Two questions addressed medication restrictions. Fifty-two percent of respondents had faced at least one situation in which the trial protocol prohibited the use of a medication that they believed to be in a subject’s best medical interests. Of these respondents, 28% reported that the subject had been given the restricted medication at least once, and 5% reported that this had occurred six or more times. Similarly, 48% of respondents had faced at least one situation in which adjusting dosages of the study medication outside the range prescribed by the protocol seemed to be in a subject’s best medical interests. Of these respondents, 16% reported that the dose had been adjusted outside the prescribed range at least once.
Respondents were also asked how often they had confronted a situation in which breaking a study blind without reporting having done so seemed to be in a subject’s best medical interest. Although 11% reported having been in such a situation, only 1% reported actually having broken the blind.
The final question asked about study participants who met a trial’s termination criteria but whose best medical interests would be served by staying in the trial. Thirty-six percent of respondents reported confronting this situation at least once during the previous two years; 9% reported that a subject had been kept in a trial despite meeting termination criteria. There were no differences between physicians and nonphysicians on any behavioral questions.
In recent years there has been considerable concern about the possibility that when physicians, nurses, and others in the clinical setting are also part of the research enterprise, their role in the research setting may compromise the clinical care they provide to patient-subjects. Many commentators have noted the potential for conflict between the research goals of clinical trials and the clinical goals of maximizing the quality of care for patients who are trial participants. Great effort has gone into both the ethical task of justifying the potential disadvantages of participating in a clinical trial and the practical task of designing trials that minimize those disadvantages.
The findings reported here suggest another potentially problematic aspect of the conflict between clinical and scientific goals. In this survey, many clinicians expressed strong commitments to their “patients.” For example, a third of respondents agreed or mostly agreed that clinicians should only participate in trials that are likely to help the patients who enroll in them, and a third agreed or mostly agreed that patients who are not doing well with standard care should be those who are recruited most actively so they can benefit from trial participation. Larger numbers of respondents endorsed ignoring minor clinical trial entry criteria or deviating from the protocol if doing so would be in the medical interests of patients who want to participate and of the patients who become research subjects.
We also found that commitments to research subjects’ medical interests can lead some clinicians to deviate from the letter and spirit of a trial’s protocol. For instance, 26% of respondents reported that research subjects had been given medications that were prohibited by the protocol, 18% said that medication dosages had been adjusted outside the permitted range, and 9% reported that a subject had been kept in a trial on clinical grounds despite the fact that she or he met termination criteria. The effects of such deviations are unpredictable, but given the right circumstances, they could compromise the validity of trial data. Although a large majority of respondents did not report being aware of such deviations from protocols, the number of affirmative responses suggests that such behavior is not rare and raises questions—which this study was not designed to answer—about the impact of these actions on the findings of clinical trials.
Potentially just as serious for the generalizability of the results of clinical trials was the frequency with which survey respondents reported that patients had been included or excluded from trials based on the likelihood of direct benefit. Fifty-four percent of respondents reported that they were aware of situations in which trial participation had not been offered to patients because participation was thought not to be in the patient’s best interest. Such decisions may, of course, be ethically justified when viewed from the perspective of a particular patient. However, recruiting patients who are doing poorly on standard care more vigorously than those who are doing well might bias the results of a trial in two ways: 1) the resulting sample may be more resistant to treatment interventions in general, or 2) if the control group is receiving standard care, the difference between the control and experimental interventions may be exaggerated. Although decisions not to offer trials to eligible patients do not usually constitute protocol violations, it is important to be aware of the potential for such decisions collectively to impact the generalizability of study results. Conversely, 21% of respondents reported that trials had been offered to patients who did not meet recruitment criteria because participationwasperceived to be in their medical interest. Unless approved by sponsors and IRBs and described in study reports, such deviations from the prescribed eligibility criteria clearly represent protocol violations and may directly affect the generalizability of study findings.
We note several limitations to these results. First, as with all surveys, we cannot be sure that the respondents correctly interpreted the survey items. Although we based the questions’ phrasing on detailed preliminary interviews and pilot-tested the survey instrument, the questions may have confused some respondents. Second, the population sampled was drawn from contact persons listed on a consumer-oriented Web site and does not represent perfectly the population of clinicians involved in clinical trials. Third, the incidence of behaviors reported may reflect underestimation of their actual occurrence, as respondents may not have been aware of or recalled all of the instances that arose in the studies with which they were involved. Fourth, we do not know how often these deviations from the protocols were reported to sponsors and IRBs, and the subjects involved subsequently excluded from the data analysis. Fifth, the survey method prevented us from collecting data on the various trials that the respondents had worked on; thus, we were unable to link the specific behaviors with the details of the clinical trials. Finally, because we asked about the number of the behaviors of interest that occurred but not about the number of subjects with whom the respondents had had contact, we cannot offer a base rate for the occurrence of these behaviors. Thus, it is not possible to gauge directly the ultimate impact of these protocol deviations on the results of clinical trials. If the deviations reported represent a sufficiently small proportion of clinical decisions in each trial, their impact might not have substantially altered trial results.
In spite of the above limitations, the survey findings raise questions about the impact of decisions grounded in clinical commitments on the validity of data derived from clinical trials. Most of the respondents admitted they are aware of selection biases and explicit deviations during the recruitment process and of protocol violations due to what they perceive to be the best medical interests of their patients. This finding suggests that conflicts between therapeutic obligations and the scientific demands of clinical trials may adversely influence the validity of findings from clinical trials. Thus, there is a need for additional data on the prevalence of protocol deviations, the causal role that clinical imperatives play in prompting these behaviors, and especially the impact of such behaviors on clinical trial results.
It is important to note that many respondents did not experience conflicts between the medical best interests of their patient-subjects and the requirements of the research protocol. This suggests that it is possible to design and implement studies that do not raise such conflicts. However, given the frequency with which conflicts are reported, there is a need to consider further actions to reduce clinically motivated protocol deviations. Investigators developing clinical trials should anticipate situations in which clinical and scientific goals may conflict, and where methodologically feasible should make design choices that minimize the frequency and intensity of those conflicts. Whatever the scientific virtues of tight restrictions on entry criteria, adjunctive treatments, and dose modifications, they are of little value if they are not followed.
Clinicians involved in the research enterprise retain values and commitments to the clinical context; thus, trials must be designed so that clinicians are comfortable carrying them out. This may involve a move to something closer to the way effectiveness trials are currently conducted, i.e., by allowing clinicians more freedom to use the study intervention in the way in which it would be used in ordinary practice.9 It could also involve new approaches to the analysis of the results of trials, since as a consequence of the increased clinical flexibility, more variables would need to be statistically controlled.Equally important, there is a need for renewed attention to the education and training of people who conduct clinical trials. We cannot take for granted that clinicians—whether they are investigators, nurse coordinators, or other health professionals—truly appreciate the importance of the research methods they are charged with implementing and the impact that failure to adhere to those methods may have on the validity of the research findings. Such training must deal directly with concrete situations in which the dual commitments to research and clinical care might conflict and how clinicians in the research context should respond to such tension.
No authors had conflicts of interest with this study. This research was primarily supported by Grant #NS049595 from the National Institute of Neurological Disorders and Stroke. No persons other than the authors participated in the research.
Charles W. Lidz, PhD,is Research Professor of Psychiatry, Center for Mental Health Services Research, University of Massachusetts Medical School, Worcester, MA;Paul S. Appelbaum, MD,is Professor of Psychiatry, Columbia University College of Physicians and Surgeons and New York State Psychiatric Institute, New York, NY;Steven Joffe, MD, MPH,is Assistant Professor of Pediatrics, Dana Farber Cancer Institute, Harvard University Medical School, Boston, MA;Karen Albert, MA,is Project Coordinator, Center for Mental Health Services Research, University of Massachusetts Medical School, Worcester, MA;Jill Rosenbaum, BA,is a graduate student, Department of Psychology, University of Alabama, Tuscaloosa, AL; andLorna Simon, MA,is Biostatistician, Center for Mental Health Services Research, University of Massachusetts Medical School, Worcester, MA.
1. Miller FG, Rosenstein DL. The therapeutic orientation to clinical trials.NEJM2003;384(14):1383-1386.
2. Marquis D. Leaving therapy to chance.Hastings Center Report1983;13(4):40-47; Helman S, Helman D. Of mice and men: Problems of the randomized clinical trial.NEJM1991;324:1585-1589. Gifford F. Community equipoise and the ethics of randomized clinical trials.Bioethics1995;9(2):
3. Fisher B, Constantino J, Redmond CK, et al. Fraud in breast-cancer trials.NEJM1994;330(20):1458-1462.
4. Oberle K, Allen M. Ethical considerations for nurses in clinical trials.Nursing Ethics2006;13(2):180-186.
5. LaFollette M. Research misconduct.Society1994;31:6-10; Rankin M, Esteves M. Perceptions of scientific misconduct in nursing.Nursing Research1997;46:270-276; Polk HC, Bowden TA, Rikkers LF, et al. Scientific data from clinical trials: Investigators’ responsibilities and rights.Archive of Surgery2002;137:639-640; Healy D, Cattell D. Interface between authorship, industry and science in the domain of therapeutics.British Journal of Psychiatry2003;183:22-77; Lagnado M. Increasing the trust in scientific authorship.British Journal of Psychiatry2003;183:3-4.
6. Antman K, Amato D, Wood W, et al. Selection bias in clinical trials.Journal of Clinical Oncology1985;3:1142-1147; Lee JY, Marks JE, Simpson JR. Recruitment of patients to cooperative group clinical trials.Cancer Clinical Trials1980;3:381; Taylor K, Feldstein M, Skeel R, et al. Fundamental dilemmas of the randomized clinical trial process: Results of a survey of the 1,737 Eastern Cooperative Oncology Group investigators.Journal of Clinical Oncology1994;12:1796-1805.
7. Winn RJ, Miransky J, Kerner JF, et al. An evaluation of physician determinants in the referral of patients for cancer clinical trials in the community setting.Progress in Clinical and Biological Research1984;156:63-73.
8. Dillman DA.Mail and Internet Surveys: The Tailored Design Method2nd ed. New York: Wiley, 2000.
9. Peto R, Baigent C. Trials: The next 50 years. Large scale randomized evidence of moderate benefits.British Medical Journal1998;317:1170-1171; Peto R, Collins R, Gray R. Large-scale randomized evidence: Large, simple trials and overviews of trials.Journal of Clinical Epidemiology1995;48(1):23-40.
Charles W. Lidz, Paul S. Appelbaum, Steven Joffe, Karen Albert, Jill Rosenbaum, and Lorna Simon, “Competing Commitments in Clinical Trials,” IRB:Ethics & Human Research 31, no. 5 (2009):1-6