19.6 Ethical issues in palliative care research

Recent growth in palliative care research has created a heterogeneous field that encompasses both qualitative and quantitative techniques, and descriptive as well as interventional study designs. Despite the valuable knowledge that has been produced by this research, and the promise of future important advances, its progress has been impeded by a persistent uncertainty about the ethics of these studies (Casarett et al., 2003). For instance, there have been concerns raised about whether patients near the end of life should ever be asked to participate in research (de Raeve, 1994; Annas, 1998) although others have objected to this extreme position (Mount et al., 1995; Casarett and Karlawish, 2000). Nevertheless, the combination of ethical and practical issues can create substantial barriers to palliative care research (Aktas and Walsh, 2011).

This chapter discusses five ethical aspects of palliative care research that investigators and clinicians should consider in designing and conducting palliative care research. These include (1) the study’s potential benefits to future patients, (2) the study’s potential benefits to subjects, (3) the study’s risks to subjects, (4) subjects’ decision-making capacity, and (5) the voluntariness of subjects’ choices about research participation. Although none of these aspects is unique to palliative care research, palliative care investigators can use these overarching principles to enhance the ethics of palliative care research.

Palliative care research should eventually improve care for future patients. These benefits to future patients can be described in terms of validity and value.

To ensure the ethical foundation of palliative care research, investigators should use valid techniques of design and data analysis, and intend to produce knowledge that is generalizable. Indeed, generalizability is the cornerstone of one definition of research used in the United States: ‘a systematic investigation, including research development, testing and evaluation, designed to develop or contribute to generalizable knowledge’ (Department of Health and Human Services, 1991, §46.102(d)). These requirements collectively describe a study’s validity (Freedman, 1987). Indeed, it is unethical to expose human subjects to risks in studies that peer reviewers agree cannot adequately answer a research question (Rutstein, 1970).

Above this threshold of validity, palliative care studies may offer more or less importance, or ‘value’. Broadly, value can be defined as the likelihood that a study’s results will improve the health and well-being of future patients (Casarett et al., 2003). Like validity, value is an important measure of a study design’s scientific quality. It is also a measure of its ethical quality, however, because a central goal of research is to produce knowledge that will ultimately be ‘important’ (Department of Health and Human Services, 1991), ‘fruitful’ (Brody, 1998, p. 213), or ‘valuable’ (Freedman, 1990). In fact, one reason that subjects participate in clinical research is to produce knowledge that will benefit others (Advisory Committee on Human Radiation Experiments, 1995; Casarett et al., 2001). Because subjects are willing to accept risks and burdens of research at least in part in order to benefit others, investigators have an ethical responsibility to maximize the probability that a study will be able to do so.

There are several ways in which investigators can enhance the validity and value of palliative care research. First, a study’s sample size should be adequate to answer the research question that is posed. Problems of underpowered studies, and particularly clinical trials, are both widespread and well described (Meinert, 1986). But they are particularly relevant to palliative care research, in which random variation can be quite large (Moore et al., 1998) and recruitment can be difficult (Aktas and Walsh, 2011). To minimize these problems, it can be useful to establish consortia or collaborative groups that can participate in multicentre studies (Kutner et al., 2008). Efforts to integrate data collection into usual care have been promising (Abernethy, 2010), and use of electronic medical record data has allowed large-scale studies at minimal cost (Casarett et al., 2012).

Second, palliative care investigators can increase the generalizability of its results. These steps might include sample size calculations that permit analysis of groups of patients that have typically not been the focus of investigation, such as patients with non-cancer diagnoses, ethnic minorities, or elderly patients. The generalizability of a study’s results might also be enhanced by recruiting subjects outside the usual academic medical settings through outreach to community clinics and hospices.

Palliative care investigators can also enhance the ethical rigor of a study by maximizing the benefits that it will offer to subjects, either during the study or after the study ends. During interventional studies, for example, a treatment could offer the possibility of a meaningful improvement. These potential benefits can be enhanced by choosing an active control design, rather than a placebo. Even if a placebo is used, a study’s potential benefits can also be improved by altering the standard 1:1 randomization scheme in a placebo-controlled trial in a way that increases subjects’ chances of receiving an active agent (Farrar et al., 1998) or by using a crossover design, if possible. Of course, the potential benefits of study interventions are never certain and a trial is only ethically acceptable only if there is legitimate uncertainty, or equipoise (Freedman, 1987), regarding the relative benefits of an intervention. Therefore, although investigators have an obligation to consider the benefits that a study offers to prospective subjects, this should not be interpreted as a requirement that all interventional studies should offer potential benefits.

Descriptive studies can also offer benefits to subjects who participate. For instance, data gathered during a descriptive study may identify pain that is inadequately treated, dissatisfaction with pain management, or related clinical problems like depression. In anticipation of instances like these, investigators can design standard operating procedures that help to ensure that these problems are identified and triaged appropriately, including communication with the patient’s health-care provider if appropriate (Casarett and Karlawish, 2000).

An interventional or descriptive study also might benefit a subject after it is over. Subjects may benefit from learning about a study’s aggregate results. For example, if a study comparing two pain medications found that one resulted in fewer side effects, this knowledge may help a subject make a more informed choice among available medications. Others may benefit if the study permits continued access to an intervention after the study ends, often through reduced rate programmes or open-label extension phases. This benefit may be particularly important in palliative care research, if subjects are unlikely to live long enough to see a study medication’s approval for clinical use, or to see a study’s results published and translated into improved care.

Investigators can also enhance a study’s ethical rigor by taking steps to minimize a study’s risks (e.g. medication side effects) and burdens (e.g. additional clinic visits or time required for surveys). The criteria by which study risks and burdens are identified and evaluated apply the concept of incremental or ‘demarcated’ risks imposed by participation in a study (Freedman et al., 1992). For instance, a study that compares two commonly used sustained-release opioids for pain may not present a significantly increased (‘demarcated’) risk to the patient because treatment with one of the drugs would have occurred off protocol. In contrast, a study that evaluates a sustained-release opioid in a condition for which it is not typically used (e.g. pruritus) poses risks that must be justified in the study’s design. In either case, however, the risks of any medication in a clinical trial should be disclosed in the informed consent process.

Perhaps one of the most contentious and emotional questions in palliative care research (Hardy, 1997; Kirkham and Abel, 1997), and indeed in research generally (Rothman and Michels, 1994; Macklin, 1999; Temple and Ellenberg, 2000), is whether a placebo or sham control arm is ethically appropriate. Broadly, placebos can be defined as interventions that are ‘ineffective or not specifically effective’ for the symptom or disorder in question (Shapiro and Shapiro, 1998, p. 12). The consensus that all subjects in a clinical trial should have access to the best available standard of care (World Medical Association, 2000) highlights the potential for unethical use of placebo. For example, all subjects with meningitis must have access to an antimicrobial agent that has proven effective; a placebo control in a trial of a new antibiotic would be unethical. In some types of research, however, such as symptom research, the placebo response can be substantial and mimics an active intervention. Moreover, the symptom under study may be transient, such as incident pain. In these situations, a placebo control may be ethically acceptable, particularly if one or more of several other conditions are met. First, placebos are acceptable if subjects receive a placebo in addition to the standard of care, such as a design that randomly assigns patient to receive an opioid plus an adjuvant agent for pain or an opioid plus a placebo. Second, a placebo arm is justified if the symptom under study has no effective treatment. Third, a placebo control is justified if subjects have adequate access to breakthrough, or ‘rescue’ treatment. This may in turn alter a trial’s end points, and may require the inclusion of these doses as a study endpoint (Dhaliwal et al., 1995; Silverman et al., 1993; Broomhead et al., 1997).

Although sham procedures are difficult to define, the term generally is applied to a procedure that is administered in a way that makes it ineffective (Polati et al., 1998). These create ethical concerns because some subjects are exposed to the risks of the procedure without any hope of its benefits (Macklin, 1999). Like placebo controls, though, the use of a sham procedure may be ethical in some studies if the risks to subjects are minimized, some subjects are likely to benefit from non-specific effects, and the research question has high value. For instance, Leonard Cobb’s research in the 1950s effectively debunked a widely used cardiac procedure that, if it had been widely disseminated, would eventually have put thousands of patients at risk. Designing a sham-controlled study so that the procedure itself (whether sham or real) poses few if any additional or ‘incremental’ risks above and beyond usual care also can reduce concerns about ethical acceptability. For example, investigators might insert a sham epidural catheter that would then be used for postoperative analgesia (Haak van der Lely et al., 1994) or a crossover design can be used to ensure that all subjects who bear the risks of the sham procedure also have access to the real procedure’s potential benefits. This crossover sham design has been used in other settings (Hahn et al., 1996), and might be appropriate for pain research when the risks or discomforts of the sham procedure are substantial.

For the most part, opportunities to minimize burdens are readily apparent. For instance, it seems reasonable wherever possible to minimize surveys, interviews, and additional study visits (Bruera, 1994). The importance of these actions may be accentuated when study subjects are very ill and an additional appointment or the requirement to complete another measure may increase distressing symptoms, such as fatigue, Investigators who conduct palliative care research should be particularly attuned to strategies that minimize these burdens, such as telephonic data collection.

Palliative care investigators may also need to consider the burdens that a study creates for friends and family members who often take on substantial burdens as caregivers (Steele and Fitch, 1996; Emanuel et al., 1999). By building flexibility into a study design (e.g. use of brief telephone interviews, multiple options for timing of clinic visits), investigators may be able to reduce the burdens of research participation on others.

Patients who consent to participate in research should have adequate decision-making capacity, which refers to subjects’ ability to understand relevant information, appreciate the significance of that information, and reason through to a conclusion that makes sense for them (Grisso and Appelbaum, 1998). Concerns have been raised about capacity in research involving patients with dementia (Marson et al., 1994) and psychiatric illness (Elliott, 1997), and patients in the intensive care setting (Lemaire et al., 1997), among others.

Concerns about the potential for impaired decision-making capacity are reasonable when research involves a population with advanced illness. The prevalence of cognitive impairment is high in this setting (Breitbart et al., 1995; Pereira et al., 1997), occurring in 10–40% of patients in the final months of life and in up to 85% of patients in the last days of life. Cognitive impairment may be difficult to identify because decision-making capacity varies over time (Bruera et al., 1995) and because impairment may result from the experimental or therapeutic medications themselves, such as opioids, benzodiazepines, or corticosteroids (Bruera et al., 1989; Stiefel et al., 1989). Investigators who conduct trials of medications will encounter these challenges even more frequently if trials are designed to evaluate treatments for delirium, for which impairment is an inclusion criterion (Breitbart, 1996).

Concern about impaired capacity also is increased by the potential comorbidity of clinical depression, which occurs in between 5% and 25% of patients near the end of life (Derogatis et al., 1983; Brown, 1986; Kathol et al., 1990; Massie and Holland, 1990). Clinically significant adjustment disorders may be even more common (Derogatis et al., 1983), and it is possible that these disorders may impair either comprehension or decision-making, or both (Elliott, 1997).

Decision-making capacity also can be undermined by severe symptoms, the experience of which may impair comprehension if patients are unable to concentrate on the information offered in the informed consent process (Kristjanson et al., 1994). This concern is particularly relevant in clinical trials that require the presence of one or more poorly-controlled symptom as an inclusion criterion.

The challenges encountered in ensuring decision-making capacity among research subjects may be magnified in longitudinal studies. Even if patients have the capacity to consent at the time of enrolment, they may not retain that capacity throughout the study. Days or weeks after patients give consent to participate, they may be unable to understand changes in their condition clearly enough to withdraw. The result can be a ‘Ulysses contract’, in which research subjects find it easier to enrol than they do to withdraw (Dresser, 1984).

None of these challenges is easily remedied. Indeed, it is obstacles like these that lead some authors to argue that patients near the end of life should not be allowed to enrol in research (de Raeve, 1994; Annas, 1998). To address the concern and enhance the ethical rigor of palliative care research when decision-making capacity is uncertain, it is important to briefly assess understanding in either open-ended or multiple choice format whenever studies involve subjects who may be near the end of life (Penman et al., 1984; Miller et al., 1996). In some situations, investigators may wish to assess decision-making capacity more formally using validated instruments (Grisso and Appelbaum, 1995). This need not be employed in all studies, but rather considered whenever the prevalence of cognitive impairment is likely to be high and the study design includes risks that must be balanced against benefits (Casarett et al., 2001). For instance, when palliative care research involves only interviews or behavioural interventions that pose minimal risks, informal capacity assessments are generally sufficient. When research poses greater than minimal risks, but offers potential benefits, a structured assessment of understanding may be appropriate. Finally, when a study poses greater than minimal risks but does not offer potential benefits, a formal evaluation of capacity should be considered.

If a patient does not have the capacity to give consent, a legally authorized representative may be able to give consent for research. This approach is justified by the argument that surrogate decision-makers should be allowed to consent to research just as they are allowed to consent to medical therapy, using either a substituted judgement of the patient’s preferences or an assessment of what would be in the patient’s best interests. If a patient does not have the capacity to consent, but is still able to participate in decisions, investigators should obtain assent from the patient and informed consent from the patient’s surrogate (High, 1993; High et al., 1994). This ‘dual consent’ ensures that patients are as involved in the decision as possible, yet provides the additional protection of a surrogate’s consent.

If a patient has decision-making capacity intermittently, or is expected to lose capacity, investigators may obtain advance consent. This innovative approach has been used in a study of treatment for delirium, in which informed consent was obtained from patients while they had decision-making capacity (Breitbart, 1996). Advance consent should be obtained only for specific studies, and should be obtained close to the planned start of research (e.g. at the time of hospitalization or enrolment in a hospice or palliative care programme).

Another way that investigators can enhance the ethical soundness of a study’s design is to examine ways in which subjects’ voluntary participation can be protected. In general terms, a choice is voluntary if it is made without significant controlling influences (Faden and Beauchamp, 1986, pp. 241–268; Beauchamp and Childress, 2001, p. 123). Voluntariness can be protected by ensuring that a subject’s choice to enrol is made with full knowledge of available alternatives and with the understanding that he or she can withdraw at any time.

Recruitment of subjects from an environment with excellent standards of palliative care enhances the likelihood that decisions about research participation will be voluntary. If patients receive excellent care, they will be best able to make a free and uncoerced choice about research participation. Patients without access to excellent care may view research participation more favourably out of desperation, or may be less willing to withdraw from research because of the perception that effective palliative interventions (e.g. opioids) are not readily available off protocol.

The field of palliative care, and the standard of care that it represents, depends upon rigorous research to provide data that will guide clinical care. Although this research raises ethical questions, these questions need not curtail what promises to be a valuable and highly productive area of research. Ethical questions can be addressed through careful attention to the study’s consent process and scientific design. With proper attention to strategies that ensure the ethical conduct of studies, missteps that have produced scandals in other fields will be avoided.