An Ethical Analysis of the SUPPORT Trial:Addressing Challenges Posed by a Pragmatic Comparative Effectiveness Randomized Controlled Trial
The SUPPORT trial highlights ethical challenges raised by comparative effectiveness randomized controlled trials (ceRCTs) involving one or more usual care interventions. Debate about the SUPPORT trial has focused on whether study interventions posed "reasonably foreseeable risks" to enrolled infants and, thereby, reflects a preoccupation with U.S. regulations. As ceRCTs are conducted globally, our analysis of the SUPPORT trial is grounded in internationally accepted ethical principles. We argue that the central ethical issue raised by the SUPPORT trial is the following: should the SUPPORT trial interventions be conceptualized as practice, or research? The answer to this question has important implications for "downstream" ethical requirements—including whether the usual care interventions in ceRCTs require research ethics committee review, undergo harm-benefit analysis, and are included in informed consent documents—and it is antecedent to the development of ethical guidance for ceRCTs.
Pragmatic comparative effectiveness randomized controlled trials (ceRCTs) evaluate the effectiveness of one (or more) interventions under real-world clinical conditions (Ford and Norrie 2016; Schwartz and Lellouch 1967). The results of ceRCTs are often directly generalizable to everyday clinical practice, providing information critical to decision-making by patients, clinicians, and healthcare policymakers. The PRECIS-2 framework identifies nine domains that serve to score a trial on a continuum between very explanatory to very pragmatic (Loudon et al. 2015). According to the framework, pragmatic trials may have [End Page 85] one or more of the following features: there are fewer eligibility criteria for participants, in an effort to mirror the target patient population; the implementation of study interventions occurs in real-world clinical settings; the intervention and its delivery do not require specialized training for clinicians or staff, and thus allow for flexibility in how the interventions are delivered; there is no special strategy for monitoring protocol compliance; patient monitoring and follow-up occur as in routine clinical practice; outcomes are clinically meaningful and patient-oriented; and all randomized patients are included in the analysis (Loudon et al. 2015).
The pragmatism movement has gained momentum in recent years owing to a multitude of factors, including: increased attention to waste in medical research; pervasive and persistent unexplained variability in clinical practice; high rates of inappropriate care; and increased healthcare expenditures (Tunis, Stryer, and Clancy 2003; Ware and Hamel 2011; Ioannidis et al. 2014). Historically, the majority of RCTs have been explanatory in design. Explanatory trials are often conducted under tightly controlled conditions—such as with strict recruitment and monitoring practices—in order to determine the efficacy of an intervention under ideal conditions. But treatments found to be efficacious in explanatory settings do not always prove effective in the real world. Accordingly, explanatory trials may fail to provide results that support decision-making by patients, healthcare providers, research funders, and policymakers and they may, thereby, merely contribute to research waste. Recently, there has been increasing recognition that the need for evidence to inform healthcare decision-making requires a shift to pragmatic ceRCTs. Reflecting this insight, ceRCTs have been identified as strategic funding priorities by the Canadian Institute of Health Research's Strategy for Patient-Oriented Research, and the U.S. Patient-Centered Outcomes Research Institute (SPOR 2016; PCORI 2014).
Of particular importance for healthcare decision-making are pragmatic ceRCTs that involve usual care interventions. Pragmatic ceRCTs of usual care interventions are a type of comparative effectiveness trial that evaluate the effectiveness of one or more interventions used routinely in medical practice, head-to-head. Usual care interventions, or medically recognized standards of care, "are treatments or procedures that have been accepted by medical experts as appropriate treatments or procedures for a given type of disease or condition, and are commonly used by healthcare professionals" (HHS 2016).1 Comparative effectiveness RCTs involving usual care interventions offer an opportunity to improve healthcare by [End Page 86] answering clinically relevant questions about the comparative effectiveness of treatments used routinely in clinical practice. Through integration with cost-effectiveness analyses, usual care ceRCTs can also inform decisionmakers about treatment costs in the real world. Comparative effectiveness RCTs involving usual care interventions can therefore help to reduce variations in care, improve uptake of evidence-based practice, reduce treatment costs, and improve patient outcomes.
All research involving human participants must conform to high scientific and ethical standards. Contemporary research ethics is founded upon four internationally accepted ethical principles: respect for persons, beneficence, justice, and respect for communities. These principles serve as the ethical foundation for global guidance documents and national regulatory frameworks.2 The principles aim, first and foremost, to protect the liberty and welfare interests of individual patients (protections for communities are a recent innovation). But, the existing ethical and regulatory frameworks were developed with explanatory trials in mind. The traditional ethical framework assumes a clear distinction between research and clinical practice: the tightly controlled conditions under which explanatory trials are designed make the demarcation of the research components from practice components of the trial relatively straightforward. This is because the traditional research–practice distinction assumes the domain of clinical practice is largely autonomous and self-regulated; hence, not only is third-party review not necessary, but the norms of the physician–patient relationship prohibit such outside interference. Conversely, third-party review is both justified and necessary in the research domain.
But with the move towards pragmatic ceRCTs, traditional ethical guidance is more difficult to interpret. Pragmatic ceRCTs intermingle interventions routinely used in clinical practice with research, and this belies a straightforward distinction between research and practice. The lack of guidance on interpreting internationally accepted research ethics principles in the context of ceRCTs of usual care interventions has left researchers and research ethics committees without a unified and systematic method of ethical analysis for these critically important trials. As a result, research participants are left vulnerable to the unavoidably capricious judgments of individual researchers and research ethics committees. This has, in turn, led to controversies in the wake of a few ceRCTs, resulting in unnecessary and costly delays. The absence of clear ethical guidance for ceRCTs therefore poses a practical threat to the conduct of this socially valuable research. [End Page 87]
2. THE SUPPORT TRIAL
The Surfactant, Positive pressure, and Pulse Oximetry Randomized Trial (SUPPORT) provides a timely and salient example of the kinds of ethical controversies to which ceRCTs give rise. Premature infants are at a substantial risk of mortality and morbidity, including retinopathy of prematurity—an important cause of blindness—and neurological impairment (SUPPORT 2010). Due to the incomplete development of their lungs, premature infants commonly receive supplemental oxygen. The historical use of supplemental oxygen in neonatology dates back to the early 1940s (Wilson et al. 1942). In those early years, unrestricted or "liberal" use of supplemental oxygen (up to 100% inspired oxygen) resulted in drastic reductions in rates of mortality; however, underdevelopment of the lungs and eyes renders oxygen uniquely toxic to preterm infants. Unrestricted oxygen use has since been associated with various morbidities, such as retinopathy, neurological impairment, and lung toxicity (Askie et al. 2009).
Over the past seven decades, numerous studies—both randomized and nonrandomized—have been conducted in an effort to determine the optimal oxygen saturation range that reduces both mortality and morbidity. On the one hand, some studies suggest that the incidence of retinopathy may be lower in preterm infants exposed to reduced levels of oxygenation than in those exposed to higher levels of oxygenation (Patz et al. 1952; Patz 1954, 1957). On the other hand, studies have also indicated that reduced levels of oxygenation may lead to increased rates of mortality (Cross 1973; Bolton and Cross 1974). A 2009 Cochrane Review concluded that "the question of what is the optimal target range for maintaining blood oxygen levels remains unclear, [and thus] further research should be undertaken to resolve this important clinical question" (Askie et al. 2009, 381). The range of oxygen saturation used routinely in neonatology intensive care units at the outset and during the course of the SUPPORT trial fell within a range between 85% to 95% (Askie et al. 2009; AAP ACOG 2007). Nonetheless, as justification for their study design the SUPPORT trial authors say that it was "becoming common practice to use lower target ranges of oxygen saturation with the goal of reducing the risk of retinopathy of prematurity" (SUPPORT 2010, 1967).
Owing to this uncertainty, and the purported trend in clinical practice at the time, the SUPPORT trial sought to determine the optimal saturation target levels of supplemental oxygen in preterm infants for minimizing retinopathy without increasing adverse outcomes, including death. The [End Page 88] SUPPORT investigators conducted a multicenter randomized controlled trial, with a 2-by-2 factorial design, to compare two target levels of oxygen saturation, and two ventilation approaches (SUPPORT 2010). The oxygen saturation component of the trial compared a lower target range of oxygen saturation, 85% to 89%, with a higher target range, 91% to 95%, on a composite primary outcome measure—a combination of multiple study endpoints—of severe retinopathy of prematurity or death (SUPPORT 2010).3
Eligible infants included those born between 24 weeks and 28 weeks of gestation for whom a decision to provide full resuscitation had been made (SUPPORT 2010, 1961). A total of 1316 infants were randomized. The study was reviewed and approved by 22 research ethics committees (the precise number of study sites is unclear), and written informed consent was obtained from the parent or guardian of each child before delivery (SUPPORT 2010, 1960).
Investigators ensured that masking was maintained with the use of electronically altered-pulse oximeters (SUPPORT 2010). A pulse oximeter is a noninvasive device that enables clinicians to continuously monitor the level of oxygen saturation in the patient's bloodstream. The altered-pulse oximeters showed saturation levels between 88% to 92% for both low and high target oxygen saturations arms, with a maximum variation of 3%. For instance, a reading of 90% corresponded to actual levels of 87% in the low-oxygen (85% to 89%) saturation arm, and 93% in the high-oxygen saturation arm (91% to 95%) of the trial (SUPPORT 2010). Clinicians were therefore told to maintain oxygen saturation levels between 88% and 92% with the use of the altered-pulse oximeters, ensuring that they were unaware as to which study intervention the individual patient had been randomized. In an effort to ensure that infants were not inadvertently exposed to levels of oxygen saturation beyond the range usual care (85% to 95%), the algorithm used for the altered-pulse oximeters gradually reverted back to actual (non-altered) values when readings were less than 84%, or higher than 96%, in both treatment groups. As an additional safeguard, an alarm was triggered when actual oxygen saturation levels reached 85% or lower, and 95% or higher (SUPPORT 2010).
The results of the trial showed no significant difference in the rate of the composite primary outcome (severe retinopathy or death before discharge) between the low- and high-oxygen saturation arms (28.3% and 32.1%, respectively; relative risk with lower oxygen saturation, 0.90; 95% confidence interval, 0.76 to 1.06; P = 0.21). However, a prospective [End Page 89] analysis revealed that infants in the low-oxygen saturation arm were subject to an increased risk of death, while survivors experienced reduced rates of severe retinopathy (SUPPORT 2010). In other words, while instances of severe retinopathy were significantly lower in infants administered low target oxygen saturation (8.6% versus 17.9%; relative risk, 0.52; 95% CI, 0.37 to 0.73; P < 0.001; number needed to treat, 11), mortality was significantly higher in the low target oxygen saturation intervention, when compared with the high target oxygen saturation intervention (19.9% versus 16.2% relative risk, 1.27; 95% CI, 1.01 to 1.60; P = 0.04; number needed to harm, 27).
Given the purported trend in clinical practice toward lower target ranges of oxygen saturation in order reduce the risk of retinopathy of prematurity, the evidence from SUPPORT "adds to the concern that oxygen restriction may increase the rate of death among preterm infants" (SUPPORT 2010, 1967). Indeed, the results of the trial suggest that there is one additional death for each two cases of severe retinopathy prevented. The investigators thus concluded by urging that "caution should be exercised regarding a strategy of targeting levels of oxygen saturation in the low range for preterm infants, since it may lead to increased mortality" (SUPPORT 2010, 1967).
Following the publication of the trial in the New England Journal of Medicine in 2010, an anonymous complaint was filed against the SUPPORT trial investigators, regarding what it claimed were egregious inadequacies in the content of the informed consent documents. In response, the U.S. Office for Human Research Protections launched a formal investigation into the SUPPORT trial. They sought to "evaluate allegations of noncompliance with Department of Health and Human Services (HHS) regulations for the protection of human research subjects (45 CFR part 46)" (OHRP 2013, 1). In conformity with a section of U.S. regulations, known as the Common Rule, researchers are required to provide research-participants with "a description of any reasonably foreseeable risks and discomforts" (45 CFR 46.116(a)(2)). However, only two of the 22 approved informed consent documents included blindness as a risk of participation in the trial, while none of the informed consent documents disclosed death as a reasonably foreseeable risk of study participation.
The investigators appear to have elected not to inform participants of these risks according to the rationale that "all of the treatments proposed in the study are standard of care, [and therefore] there is no predictable increase in risk for your baby" (SUPPORT trial Consent Forms 2011, 211). [End Page 90] The only risk investigators disclosed in the "possible risks" section of the 20 informed consent documents that did not list either blindness or death as research risks related to the use of the pulse oximeter, namely, "the possible risk of skin breakdown at the site" (SUPPORT trial Consent Forms 2011, 4).
Upon investigation, the U.S. Office for Human Research Protections concluded that "the anticipated risks and potential benefits of being in the study were not the same as the risks and potential benefits of receiving standard of care" (OHRP 2013, 9). They argued that, "for the infants assigned to the upper range [. . .] the risk of ROP [retinopathy of prematurity] was greater, while for the infants assigned to the lower range the risk of ROP was lower," and therefore the risks should have been disclosed to the parents or guardians of participating infants. (OHRP 2013, 9). The Office for Human Research Protections therefore determined "that the conduct of this study was in violation of the regulatory requirements for informed consent, stemming from the failure to describe the reasonably foreseeable risks of blindness, neurological damage and death" (OHRP 2013, 2).
3. THE DEBATE ON THE SUPPORT TRIAL
The U.S. Office for Human Research Protections determination incited considerable debate in the literature. While the determination focused on the alleged inadequacies in the content of the informed consent documents, the debate on the SUPPORT trial has focused primarily on whether infants were in fact exposed to increased risk as a result of trial participation. On the one hand, critics of the SUPPORT trial argue that various features of the study posed an a priori (as opposed to post hoc) increase in risk to enrolled infants, when compared with infants not enrolled. On the other hand, defenders of the SUPPORT trial argue that participation in the trial did not pose an a priori increase in risk to enrolled infants compared to those infants treated in clinical practice.
3.1 Critics of the SUPPORT Trial
The SUPPORT trial has garnered much criticism (Macklin, Shepherd, and Dreger 2013). Broadly speaking, critics of the SUPPORT trial raise three substantive objections. First, some critics argue that features of the study compromised the fiduciary duties the physician owes to her patient (Annas and Annas 2013; Carome and Wolfe 2013; Sharav 2013). They argue that aspects of the study design and implementation violate the [End Page 91] physician's duty to exercise her personal judgment in the best interests of the individual patient. George J. Annas and Catherine L. Annas, for instance, argue that "in treatment a patient has a physician who is bound by a fiduciary duty to act in the patient's best interests" (Annas and Annas 2013, 20). These critics argue that the lack of physician judgment exposed enrolled infants to increased risk compared to patients who are not part of a research study: "the primary argument [that] . . . no new risks were introduced to the patient is explicitly rejected by this court" (Annas and Annas 2013, 25).
Second, some critics argue that the study was not appropriately designed to answer the primary study question (Natanson 2013; Merz and Yerramilli 2013). For instance, Charles Natanson (2013) argues that in cases in which a given standard of care is specified across a range, randomizing patients to the extremes of that range no longer constitutes the medical standard of care. He argues that this creates "practice misalignments which carry risks and do not represent usual care" (Natanson 2013). This concern was also echoed in the U.S. Office for Human Research Protections determination. They argued that enrolled infants had a greater chance of receiving an oxygen saturation at the low, or high, extremes of the range (i.e., 85% or 95%, respectively) that comprised the (then-current) standard of care:
[P]articipating in the study would have significantly increased the chance of an infant being assigned to oxygen levels at both the very low (85% to 88%) and the very high ends (92% to 95%), as opposed to the level they would have received [88% to 92%], had they not been in the study.(OHRP 2013, 4)
As a result, the U.S. Office for Human Research Protections argued that it was "much more likely that they [i.e., enrolled infants] would be within the range in which there were significant concerns about increased mortality," when compared with infants treated in clinical practice (OHRP 2013, 4).
Third, and finally, some authors claim particular features of the study design exposed infants to additional risk over that of routine clinical practice (OHRP 2013; Carome and Wolfe 2013; Natanson 2013). They argue, variously, that randomization, treatment by protocol, and the use of altered-pulse oximeters ultimately exposed infants enrolled in the SUPPORT trial to increased risks of retinopathy, neurological damage, and death, when compared with infants not so enrolled. [End Page 92]
3.2 Defenders of the SUPPORT Trial
John Lantos, who is among the most prolific and forceful defenders of the SUPPORT trial, disagrees with these objections (Lantos 2013, 2014a, 2014b, 2015; Lantos and Feudtner 2015; Wilfond et al. 2013). Lantos argues that infants enrolled in the SUPPORT trial were not exposed to increased risk when compared with those infants treated within clinical practice. Lantos claims that there are ethically relevant differences between ceRCTs of usual care interventions and randomized trials that compare a novel intervention with a control; accordingly, he argues that new regulations are needed for ceRCTs.
First, Lantos claims that the purpose of so-called "innovative therapy research"—i.e., a clinical trial with a novel intervention—is different from that of ceRCTs of usual care interventions. In the former, Lantos says the risks and benefits are "truly unknown," and thus the aim is to "better characterize the safety and efficacy profile of the new treatment" (Lantos and Feudtner 2015, 38). In ceRCTs of usual care interventions, by contrast, he says that "both therapies are in widespread use and can be considered standard" (39); consequently, "much is known about them," including knowledge about their potential harms and benefits (39). Second, Lantos notes that the therapy patients receive within a ceRCT is not appreciably different from the therapy they would receive outside of the study. In other words, if a patient chose not to enroll in a ceRCT comparing usual care interventions, they would nonetheless "get the exact same treatment," namely, the existing standard of care (39). Lantos argues that this is unlike a scenario in which a patient opts out of a randomized controlled trial with a novel intervention, or, as he calls it, "innovative therapy research"; this is because patients that opt out of a trial with a novel intervention would no longer be exposed to the "innovative therapy" but would instead receive the existing standard of care. Finally, Lantos argues that ceRCTs involving usual care interventions do not pose an increased risk to study participants because the treatment arms are not appreciably different from routine clinical practice.
Lantos therefore disagrees with critics who claim that the SUPPORT trial exposed infants to increased risk. To those critics who argue that features of the design undermine the physician's fiduciary duties to her patient, Lantos responds by pointing out that "these criticisms of SUPPORT [. . .] can be generalized as criticisms of any prospective randomized clinical trial" (Lantos and Feudtner 2015, 32). Accordingly, he suggests that we ought to consider how these apparently ethically problematic aspects of [End Page 93] the design of randomized controlled trials have been addressed in the past.
Furthermore, Lantos suggests that concerns about treatment by protocol and the use of altered-pulse oximeters are similarly misplaced. For instance, in their determination letter the U.S. Office for Human Research Protections argued that
[w]hen there is a range of oxygen levels within the standard of care, clinicians (and their institutions) often do, in fact, make their own determinations regarding which oxygen levels within that range to employ in treating their patients. Some physicians, recognizing the particular concerns about risks near the low (85%) and high (95%) ends of that range, might choose to avoid one or both of those regions.(OHRP 2013, 3)
However, citing neonatologist Keith Barrington (2013), Lantos claims that neonatologists do not, in fact, make "decisions about oxygen saturation targets for each patient based on the patient's particular clinical situation and their clinical judgment about whether a lower or higher oxygen saturation should be targeted"; rather, "neonatologists always treat babies by protocol" (Lantos 2015, 33). Lantos thus concludes that the argument that infants were exposed to increased risk as a result of the physician's inability to exercise her personal judgment on behalf of her patient, due to various features of the study design, "are particularly irrelevant to the SUPPORT controversy" (33).
Finally, Lantos rejects Natanson's and the U.S. Office for Human Research Protection's claim that practice misalignments posed increased risk to enrolled infants. He says that the standard of care does not fall neatly within the 85% to 95% range, and so, rather than
imagining that the conventional treatment at the time was to target 85[%] to 95[%], a more accurate statement would be that, in each NICU, there was a different target and that most of those targets were within the range 85[%] to 95[%].
Furthermore, insofar as Natanson's argument suggests that usual care involves the use of a physician's judgment on behalf of the best interests of her individual patient, Lantos simply reiterates that the administration of oxygen for preterm infants in neonatal intensive care units is "provided by a predetermined protocol," precluding the need for clinician discretion (37).
Lantos concludes that concerns about the design and implementation of the SUPPORT trial do not show that enrolled infants were exposed to increased risk. On the contrary, he argues that the risks and benefits [End Page 94] associated with enrollment in the trial were not appreciably different from those associated with routine clinical practice. Indeed, Lantos says, "the primary difference is that, outside of [ceRCTs] [. . .] the choice of a treatment is [made] by idiosyncratic practice variation," whereas in a ceRCT, "the treatment a patient receives is determined by formal randomization" (Lantos and Feudtner 2015, 31). What is needed, according to Lantos, are regulations that "acknowledge the distinctiveness" of ceRCTs involving usual care interventions. He says that the aforementioned aspects of the design and implementation of the SUPPORT trial make it "a paradigmatic case upon which to develop [such] regulatory policy for similar studies in the future" (Lantos and Feudtner 2015, 39).
4. THE CENTRAL ETHICAL ISSUE RAISED BY THE SUPPORT TRIAL
The debate on the SUPPORT trial has been preoccupied with interpreting U.S. regulation. This brief overview of the overarching arguments in the debate demonstrates that proponents and critics alike are predominantly focused on determining whether the SUPPORT trial presented "reasonably foreseeable risks" to enrolled infants, pursuant to a section of U.S. regulation on human participants research, known as the Common Rule (45 CFR 46.116(a)(2)). However, insofar as pragmatic ceRCTs are a global phenomenon, the manifest preoccupation with U.S. regulation has made this debate largely irrelevant to an international audience. This is because proponents and critics alike have failed to ground the debate in internationally accepted ethical principles. As a result, the debate on the SUPPORT trial has been largely ad hoc, in that it identifies a loosely related—and not necessarily comprehensive—set of issues, and fails to trace them to a central ethical principle. As a further consequence, this debate has predictably failed to provide a systematic analysis applicable to a broader class of pragmatic ceRCTs, based on the relevant underlying ethical principles.
Our diagnosis is that the central ethical issue, or question, raised by the SUPPORT trial is the following: should the SUPPORT trial interventions be conceptualized as practice, or research? To be clear, while it is uncontroversial that the SUPPORT trial as a whole is research, the novel question posed by pragmatic ceRCTs involving one or more usual care interventions is whether the usual care intervention(s) should be conceptualized as practice, or research. The answer to this question will have important implications for the ethical analysis of ceRCTs. Indeed, how these study interventions are conceptualized will have implications for [End Page 95] various downstream ethical requirements, including whether the usual care interventions in pragmatic ceRCTs (i) fall under the purview of research ethics committee review, (ii) undergo harm–benefit analysis, and (iii) are included in informed consent procedures.
Determining whether the usual care interventions in pragmatic ceRCTs that compare one or more usual care interventions constitute clinical practice or research is therefore antecedent to questions pertaining to research ethics review, analysis of study benefits and harms, and the content of informed consent procedures. In other words, questions concerning whether the study interventions in the SUPPORT trial posed additional risk to premature infants are subordinate to determining how the interventions ought to be conceptualized. Determining how the usual care interventions in pragmatic ceRCTs of usual care interventions ought to be conceptualized will therefore enable us to provide a systematic analysis of the SUPPORT trial, grounded in the relevant foundational ethical principles.
5. THE RESEARCH-PRACTICE DISTINCTION
Broadly speaking, research "refers to a class of activities designed to develop or contribute to generalizable knowledge," while medical practice "refers to a class of activities designed solely to enhance the well-being of an individual patient or client" (Levine 1986, 3). Contemporary research ethics, and corresponding research regulation, have evolved in response to the historical abuses perpetrated by researchers against research participants—including the German wartime experiments on prisoners, the infamous Tuskegee Study, and the Jewish Chronic Disease Case—in the name of furthering generalizable knowledge. Research involves the pursuit of interests that compete with those of the individual patient: namely, the research-related interests of investigators who seek to "develop or contribute to generalizable knowledge," and thereby also the interests of future patients, and society in general. These competing interests have the potential to undermine the trust relationship between the physician and patient, and thereby warrant prior third-party review and oversight.
It is thus widely accepted that regulation, review, and oversight are needed in research, because the medical interests of the patient are no longer the sole interests at stake. Third-party review by research ethics committees, and research regulations, are put into place in an effort to ensure that the interests of patient participants are protected in the face of competing interests. [End Page 96]
The domain of clinical practice, by contrast, is self-regulated and autonomous. In clinical practice, "patient consent to care and professional autonomy are intended to ensure that the intervention in question is the product of joint deliberation and agreement between patient and doctor" (Freedman, Fuks, and Weijer 1992, 653). Such dialogue might include patient values, preferences concerning the administration of treatment, medical history, and any other relevant patient factors. Importantly, no third party need be consulted during the course of this deliberation process, or the ensuing treatment plan. While the physician–patient interaction may be reviewed retrospectively by a hospital, a council of peers, or the courts—as in the case of a negligence or battery suit—it does not require prior review or oversight by a third party.
The distinction between research and practice is often understood as applying to whether whole study protocols ought to be considered research or practice, and, consequently, whether they ought to be subject to review and oversight by third parties. But it is claimed that pragmatic ceRCTs that compare one or more usual care interventions introduce a new wrinkle on this historical problem, because they comprise elements of both practice and research. Indeed, as briefly mentioned above, no one denies that the SUPPORT trial as a whole is research, and thus ought to be subject to research review and regulation. Additionally, it is uncontroversial that the various features of the study design, such as randomization, the use of masking with altered-pulse oximeters, and data collection (except when using solely anonymized routinely collected data) are plainly and unabashedly research interventions. Rather, the point of contention—and the uniqueness of the question posed by ceRCTs, if any—is whether the usual care intervention(s) in pragmatic ceRCTs constitute practice or research, and, accordingly, whether the usual care interventions ought to be regulated by the norms of the physician–patient relationship, or those of research.
6. CONCEPTUALIZING THE SUPPORT TRIAL INTERVENTIONS AS PRACTICE
With this framework in mind, we can then ask: what arguments might be proffered in favor of conceptualizing the SUPPORT trial interventions as practice? To begin, both the low- and high-oxygen saturation interventions in the SUPPORT trial fell within the range recognized as "standard of care" for the treatment of extremely preterm infants (Askie et al. 2009). Recall that a standard of care is a routine or prevailing practice pattern within a given medical community (Levine 1986). Insofar as the study [End Page 97] interventions in the SUPPORT trial comported with the then-current range routinely used in neonatal intensive care units, they appear to constitute clinical practice.
As we saw above, clinical practice is self-regulated, and thus enjoys a degree of professional autonomy that precludes prospective third-party review by research ethics committees. Again, no one disputes the contention that the study as a whole is research, or that other features of the study protocol—such as randomization, the use of masking with altered-pulse oximeters, and data collection—are research interventions, and therefore ought to be conceptualized as such in the ethical analysis. Rather, the claim is that the SUPPORT trial interventions are usual care, and therefore ought to be conceived of as practice. It follows straightforwardly, according to this view, that both the low- and high-oxygen saturation arms in the SUPPORT trial ought to be conceptualized as practice and thus governed by the rules and norms of the physician–patient relationship.
This is not an implausible view. Indeed, it is evident that Lantos, the SUPPORT trial investigators, and the 22 research ethics committees that approved the trial, appear to have conceptualized the low- and high-oxygen saturation interventions in precisely this way, that is to say, as clinical practice. At the very least, they may be reasonably understood—for the sake of ethical coherency—as being committed to such a view. This interpretation is supported by the fact that the investigators elected not to disclose risks of mortality, retinopathy of prematurity (at least in 20 of the 22 informed consent documents), and neurological damage as risks of study enrollment; ostensibly, because these are risks of routine clinical practice.
Indeed, the investigators appear to adopt just this rationale when they state in the template used for the informed consent documents that "all of the treatments proposed in the study are standard of care, [and therefore] there is no predictable increase in risk for your baby" (SUPPORT trial Consent Forms, 211). In fact, with the aforementioned exception of two study sites, the only risks investigators elected to disclose in the "possible risks" section of the informed consent documents relate to the use of the pulse oximeter, namely, "the possible risk of skin breakdown at the site" (SUPPORT trial Consent Forms, 4). Furthermore, it appears that all of the participating research ethics committees endorsed this rationale as well, at least tacitly, given that they approved the trial along with the informed consent documents that expound this reasoning. This also includes the two sites that did include blindness as a possible risk of study participation, as [End Page 98] they nonetheless elected not to disclose death or neurological damage as risks of study participation; again, ostensibly reasoning that such risks are inherent to clinical practice. This framework makes it plain that Lantos may also be more clearly, and productively, understood as arguing that the low- and high-oxygen saturation interventions in the SUPPORT trial ought to be considered clinical practice, and therefore ought to be exempt from third-party review and research regulation.
Several implications follow from the contention that the SUPPORT trial interventions ought to be conceptualized as clinical practice, and not research. First, because the administration of low and high supplemental oxygen in preterm infants is governed by the rules and norms of the physician–patient relationship, these components of the trial are appropriately exempt from research ethics committee review. Second, the SUPPORT trial interventions do not require a systematic harm–benefit analysis. Harm–benefit analysis is compulsory for all research protocols; if the study interventions are practice, and not research, they do not enter a research ethics committee's deliberations on study benefits and harms. All that remains for the research ethics committee to determine, then, is that the risks of the research interventions are minimized consistent with sound scientific design, and that they are reasonable in relation to the scientific knowledge to be gained (Weijer and Miller 2004). Third, and finally, no information about the low- or high-oxygen saturation interventions need be disclosed to research participants. These interventions fall within the norms of the physician–patient relationship, and, as a result, neither the nature of the interventions, nor their benefits and harms—including retinopathy, neurological damage, lung toxicity, or death—need be disclosed as part of the research informed consent procedures (of course, any "material risks" would need to be disclosed as part of the clinical informed consent process).
7. CONCEPTUALIZING THE SUPPORT TRIAL INTERVENTIONS AS RESEARCH
Nevertheless, we argue here that this is not the correct way to conceptualize the SUPPORT trial interventions. Rather, there are at least three substantive reasons to believe that the low- and high-oxygen saturation arms in the SUPPORT trial ought to be conceptualized as research, and not practice. The first two reasons may be applicable to a broader class of ceRCTs involving one or more usual care interventions, whereas the third, and final, reason is likely specific to a particular type of ceRCT, of which the SUPPORT trial is an examplar. However, we leave [End Page 99] open for future inquiry any substantive answer to the question regarding the utility of our analysis of the SUPPORT trial for conceptualizing other pragmatic ceRCTs.
7.1 The First Reason: Randomization Interferes with the Individualized Judgment of the Physician
The interventions in the SUPPORT trial ought to be conceptualized as research because the enrolled infants are deprived of the benefit and protection of the individualized judgment of their physician. In the neonatal intensive care unit, very premature infants are prescribed oxygen therapy by their treating neonatologist who relies upon her professional judgment to prescribe the best treatment for the child. Professional judgment is a portmanteau, composed of a multiplicity of considerations, including: an individual infant's diagnosis and prognosis; the medical history of the infant's biological parents; the values or preferences articulated by the infant's parent(s), or guardian(s); the oxygen saturation range routinely used in the neonatal intensive care unit; the state of evidence in the literature on the benefits and harms of oxygen treatment; and past experience—including the experience of colleagues—in the treatment of similar infants.4
In the SUPPORT trial, by contrast, very premature infants do not come to receive treatment by means of the individualized judgment of their treating physician. Rather, infants are allocated randomly to either of the low- or high-oxygen saturation arms of the trial. Indeed, one of the primary purposes of randomization is to prevent a physician from exercising her individualized judgment on behalf of the patient (Friedman et al. 1998, 62–3). Randomization is a feature of the study protocol designed to prevent the bias that physician judgment introduces, and thereby enhance the validity of the study. It therefore promotes the interests of research—viz., the promotion of, or contribution to, generalizable knowledge—but at the expense of the individualized care for the patient.
To illustrate, suppose that the SUPPORT trial were non-randomized, allowing neonatologists to exercise their individual judgment when assigning premature infants to either low- or high-oxygen saturation arms of the trial. Neonatologists are aware of the body of research that suggests that a lower target range of oxygen saturation may reduce incidences of retinopathy, but may also be associated with increased mortality (Tin 2004; Askie et al. 2009). How, then, would a neonatologist make such allocation decisions for her various patients? Faced with an infant with [End Page 100] a substantial chance of mortality, it is likely that the clinician will assign this infant to the high-oxygen saturation arm of the trial to maximize her chances of survival. Conversely, faced with an infant with a relatively good prognosis, it is likely that the clinician will assign her to the low-oxygen saturation arm, in an effort to reduce the risk of retinopathy. As this non-randomized trial progresses, it is evident that the sickest infants will comprise a disproportionate number of the high-oxygen saturation arm, and the healthier infants will comprise a disproportionate number of the low-oxygen saturation arm. But, what can be inferred about the low- and high-oxygen saturation interventions at the conclusion of this non-randomized trial? The answer, of course, is nothing: physician judgment completely confounds the study interventions with respect to the study outcome. Individual physician judgment compromises the internal validity of the trial, and randomization is therefore necessary to allow valid inferences to be drawn.
The SUPPORT trial interventions therefore ought to be conceptualized as research because the "dose" of oxygen therapy is determined by means of randomization, not by physician judgment. Randomization is a feature of the trial employed to promote internal validity, and thus further the interests of research. The introduction of this competing interest undermines the physician's duty to exercise her individual judgment in the interests of her patient. This duty, known as the duty of personal care, arises from the nature of the physician–patient relationship. Charles Fried (1974) was among the first to recognize that the physician–patient relationship is one of trust, and is subject to the law of fiduciary relationships. Fried recognized that a physician has a fiduciary duty to provide personal care to her patient. The duty of personal care requires a physician to exercise her individual judgment on behalf of the medical interests of her patient. According to Fried, the duty of personal care is an obligation that the physician "may not compromise in the general pursuit of the common good" (52).
The introduction of the competing interest in the pursuit of scientific knowledge undermines the physician's duty of personal care. As Fried might say, the otherwise autonomous relationship between physician and patient has been "compromise[d] in the general pursuit of the common good" (52). Thus, the physician can no longer be relied upon to remain faithful solely to the interests of her patient. Indeed, it is precisely for this reason that research, as a moral domain separate from that of clinical practice, is governed by its own rules and norms. When the physician, qua [End Page 101] fiduciary, can no longer be entrusted to exercise her individual judgment solely on behalf of the medical interests of her patients, "strangers at the bedside"—in the form of research ethics committees, data and safety monitoring boards, and research regulations—are needed to protect the liberty and welfare interests of those patients participating in research (Rothman 1991). Accordingly, the SUPPORT trial interventions must be considered research, rather than practice.
7.2 The Second Reason: Masking in the SUPPORT Trial Impairs the Ability of Physicians To Exercise Discretion on Behalf of their Patients
The second reason that the SUPPORT trial interventions ought to be conceptualized as research is because the physician's ability to continuously monitor, reassess, and adjust the prescribed oxygen therapy administered to a premature infant is compromised by altered-pulse oximeters. In other words, the use of altered-pulse oximeters in the SUPPORT trial undermines—or at the very least, threatens—the physician's fiduciary duty of discretion to her patient.
In clinical practice, extremely premature infants are prescribed oxygen therapy by means of the treating neonatologist's exercise of her professional judgment, which, as we saw above, takes into consideration a multitude of factors. However, the neonatologist is not thereby relieved of her fiduciary obligations to the premature infant under her care simply through this one-time exercise of her professional judgment. Rather, the neonatologist must continue to employ her discretion as she cares for a premature infant. In practice, she will continuously and closely monitor the medical condition of a premature infant under her care. For instance, an infant may begin to have trouble breathing, or display other signs or respiratory distress, in which case the neonatologist will likely increase the level oxygen saturation. Conversely, aware of the evidence that a higher range of oxygen therapy may be correlated with increased risk of retinopathy, the neonatologist might decrease the level of the oxygen saturation for an infant that appears to be doing particularly well. Extremely premature infants will often require weeks, or even months, of such discretionary monitoring on the part the treating clinician. This obligation for ongoing clinician judgment is called the fiduciary duty of discretion.
The power to exercise judgment demands of the treating physician that she be vigilant in the pursuit of the patient's medical interests; this, in turn, requires her to use discretion as she continuously assesses and reassesses the patient's condition. By entrusting his medical interests to the physician, [End Page 102] the patient is left vulnerable to the risk that the physician will neglect his interests by simply failing to exercise her professional judgment, either in whole or in part. It is the patient's vulnerability to neglect that gives rise to the physician's fiduciary duty to exercise her "discretionary power" on behalf of the patient's medical interests (Gold and Miller 2014, 72). Indeed, as Gold and Miller (2014) explain:
[P]owers are ordinarily considered fiduciary only if they are discretionary. Discretion entails freedom of choice in the exercise of the power. This in turn means that beneficiaries and benefactors are not merely subject to the conduct of fiduciaries; they are subject as well to their will. Fiduciaries decide, as well as act, for others.(72)
In other words, a physician, qua fiduciary, is obligated not only to "decide"—i.e., judge—what the appropriate therapy for her patient might be, but also to act on that judgment, which requires the ongoing exercise of her discretionary power. The duty of discretion implies that the physician must not "utterly neglect" the patient's interests, nor delegate decisions wholly to another, nor otherwise fetter her discretion in any way (Miller and Weijer 2006). Alternatively, the physician's duty of discretion does not merely prohibit the physician from acting in a way that exposes the interests of the patient to neglect, but requires the physician to actively exercise those discretionary powers.
In the SUPPORT trial, the study interventions are administered to extremely premature infants in a way that deprives them of the benefit and protection of the treating physician's ongoing discretion. Masking with altered-pulse oximeters was employed to ensure that physicians were unaware of the actual oxygen saturation levels being administered to premature infants in the SUPPORT trial. In other words, the pulse oximeters were altered to impair the physician's ability to accurately monitor, assess, and reassess an infant's medical condition. Importantly, this was done to promote the internal validity of the trial by seeking to ensure a clear separation between low- and high-oxygen saturation interventions. Indeed, a clear separation between the study interventions is crucial to any "good experiment." Without masking it would have been extremely difficult to achieve a separation between the treatment arms. Had masking not been used, physicians may have tended to increase the dose of oxygen for infants in the low-oxygen saturation arm, with the effect of making infants' oxygen saturation levels similar (Cortés-Puch et al. 2016).
In fact, the SUPPORT trial investigators encountered precisely this problem during the early enrollment phases of the trial: [End Page 103]
When 247 infants had been enrolled, enrollment was temporarily suspended on the basis of the recommendation of the data and safety monitoring committee and the decision of the director of the National Institute of Child Health and Human Development because of concern that readings of levels of oxygen saturation often exceeded the target levels.(SUPPORT Study Group 2010, 1964)
Altered-pulse oximeters were employed to ensure that infants enrolled in the SUPPORT trial receive the treatment mandated by the study protocol by hindering neonatologists from adjusting the oxygen infants receive based on continuing assessments of their condition. Evidently, this feature of the SUPPORT trial serves the interests of research. But it does so by depriving enrolled infants of the benefit and protection of the clinician's ongoing discretion, as she continues to care for, and promote, the medical interests of her patients. The use of masking in the SUPPORT trial therefore introduces a competing interest—namely, the production of reliable generalizable knowledge—that undermines the physician's duty to exercise her discretion. Insofar as the norms of clinical practice have been violated in this way, there is strong justification for conceptualizing the SUPPORT trial interventions as research.
7.3 The Third Reason: Low- and High-Oxygen Saturation Interventions Are Not Usual Care
The third, and final, reason that the low- and high-oxygen saturation arms in the SUPPORT trial ought to be conceptualized as research, and not practice, is that one or both of the study interventions are not in fact usual care. Recall that this reason is specific to a particular type of pragmatic ceRCT involving one or more usual care interventions, of which the SUPPORT trial is a token example: namely, in those cases where the usual care treatment is characterized by a range, or continuum, the bifurcation and ensuing randomization of research participants to protocolized subsets of that range, no longer constitutes usual care. 5 Usual care for the administration of supplemental oxygen in preterm infants is a range between 85% and 95% (Askie et al. 2009; AAP ACOG 2007). However, the interventions in the SUPPORT trial did not comprise this full range; rather, enrolled infants were delegated to subsets—i.e., 85% to 89% and 91% to 95%—of that range. The bifurcation of the range of supplemental oxygen routinely used in neonatal intensive care units to low and high subsets of that range constitutes a meaningful deviation from, or alteration to, clinical practice for the ends of research. Indeed, [End Page 104] this is true even if a neonatal intensive care unit follows a protocol that is equivalent to one of the SUPPORT trial interventions (e.g., 85% to 89%), as the unit of randomization is the patient, not the center. Consequently, even if a center's protocolized range of oxygen therapy for preterm infants corresponds with one of the study arms, the infants in that center would be randomized to both ranges, and thus be exposed to a range that is not usual care for that center. Indeed, in their Draft Guidance on Disclosing Reasonably Foreseeable Risks in Research Evaluating Standards of Care (2014), the U.S. Office for Human Research Protections makes a similar point,
When a research study assigns the specific version of the accepted standards of care to be used, it is almost certain that at least some of the subjects will receive a different standard of care than they would have received if not participating in the research. Indeed, in the common study design where subjects are randomized equally between two treatments, approximately half of the subjects will be assigned to a treatment different from what they would have otherwise received.
As argued above, alterations to clinical practice for the sake of contributing to generalizable knowledge—a competing interest—warrants third-party review and regulatory oversight, separate from that of clinical practice.
The bifurcation of the standard of care range of supplemental oxygen introduces potential risks and benefits distinct from those inherent to routine clinical practice. In this case, the risks and benefits of receiving the low-oxygen saturation arm are manifestly different from the risks and benefits of receiving high-oxygen saturation—as over seven decades of preliminary data portend (albeit, evidence not strong enough to change clinical practice). In their determination letter, the U.S. Office for Human Research Protections pointed out that concerns in the research and clinical community about the differential risks and benefits of the low and high target ranges of supplemental oxygen on preterm infants was a "core reason why the study was conducted" (OHRP 2013, 2). Moreover, they stated that it was clear that the SUPPORT trial investigators were aware of such concerns, having "identified the important need for a large randomized study with sufficient power to detect differences in mortality rates of 5% or greater" (2). As previously mentioned, Charles Natanson (2013) provides a similar argument. It is clear that the risks and benefits of both the low- and the high-oxygen saturation arms in the SUPPORT trial differ from the risks and benefits of the full range, or the de facto usual care range. This is because in routine clinical practice neonatologists [End Page 105] are able to employ the full range (85% to 95%) of supplemental oxygen, subject to their professional judgment and ongoing discretion, in response to the medical needs of each individual infant (as illustrated by Cortés-Puch and colleagues, below).
To illustrate this important point, the U.S. Office for Human Research Protections provides a salient example,
It is known that treatment using surgery and radiation has a high likelihood of curing a particular form of childhood cancer, but that the radiation produces a significant risk of other cancers developing later in the child's life. Consequently, some doctors treating children with this cancer use a smaller amount of radiation. Both amounts of radiation are consistent with clinical care guidelines and considered to be within the standard of care. There is little evidence available comparing the outcomes of the two treatments in terms of their cure rates or the development of later cancers. A randomized clinical trial is proposed with subjects to be assigned to treatment with the higher or lower amount of radiation to compare the effectiveness of the two treatments in curing the current cancer and how often later cancers occur.
It is plain that the different ranges of treatment routinely used for childhood cancer in this hypothetical example carry differential risks. Moreover, it is precisely the purpose of the study to empirically validate those risks. In other words, participants are exposed to risks that they may not otherwise have been outside of the study, for the sake of contributing to generalizable knowledge. The Office for Human Research Protections agrees, stating that the "particular risks that the subjects will be exposed to because of being assigned to a specific standard of care are risks the subjects will be exposed to for the sake of the research" (OHRP 2014).
Recall that Lantos claims "neonatologists always treat babies by protocol," which is assigned by each neonatal intensive care unit, "based on their assessment of the evidence about the harms and benefits of oxygen levels that are too high or too low" (Lantos and Feudtner 2015, 33). Accordingly, Lantos claims a neonatologist does not in fact exercise her professional judgment over the medical interests of the infant over whose care she is presiding. He thus concludes that the risks and benefits of protocolization to either the low- or high-oxygen saturation arms in the SUPPORT trial are not appreciably different from risks and benefits inherent to clinical practice.
Firstly, Lantos' argument is simply orthogonal to the concern that allocating infants to low and high target ranges carries differential risks. [End Page 106] While it may indeed be true that neonatologists treat infants according to a predefined protocol, this does not alleviate the concern that the bifurcation of the usual care range of supplemental oxygen into discrete (low and high) subsets of that range carries differential risks to enrolled infants, when compared to those infants not so enrolled. In other words, it may indeed be a fact that neonatologists treat infants according to a defined protocol, but no particular neonatal intensive care unit protocolizes infants to both low and high extremes of the usual care range in order to achieve a "quasi-random distribution" (Freedman, Fuks, and Weijer 1992, 658).
Secondly, and more importantly, it does not follow from the claim that neonatologists treat infants in neonatal intensive care units according to protocol that they consequently do not exercise their professional judgment, or that they therefore follow the protocol exceptionessly or mindlessly. On the contrary, even if neonatologists treat infants according to predetermined parameters, this nonetheless does not exclude latitude for the exercise of professional judgment and ongoing discretion. The neonatologist's duty to exercise her professional judgment on behalf of the medical interests of the premature infants under her care obviously persists within the constraints of a predefined protocol. Moreover, physician judgment includes considerations over whether she ought to observe the prescriptions of the protocol tout court.
To illustrate, suppose a neonatal intensive care unit has mandated that neonatologists prescribe a range of oxygen therapy for extremely premature infants between 85% and 89%. Suppose, further, that a neonatologist finds herself entrusted with the care of a baby with a particularly poor prognosis. Aware of the most recent evidence in the literature that suggests an association between low-oxygen saturation treatment and mortality, the neonatologist may reasonably judge that a higher range of oxygen saturation—outside of the protocolized range mandated by her neonatal intensive care unit—is appropriate in this case. The duty to provide personal care to her patient not only permits the neonatologist to act on such a judgment, but obliges her to. This is especially true when the state of evidence with respect to those ranges is in a state of equipoise, or uncertainty. It is perhaps an unhappy reality of clinical practice that physicians are required to make clinical decisions in the face of such uncertainty. Fried recognized this feature of medical practice, arguing that "the doctor must be given considerable latitude as he works in the presumed interests of his patient" (Fried 1974, 26). Similarly, Freedman, Fuks, and Weijer (1992) echo this important point, [End Page 107] noting that medical practice allows for a "reasonable degree of discretion," which extends even to treatments that carry substantial risk, so long as the treatments in question "fall within the acceptable range of standard treatments" (654). The claim that individualized physician judgment and discretion are somehow obviated, or negated, by the protocolization of a treatment is clearly false.
Finally, a study by Cortés-Puch and colleagues found that the low target oxygen saturation arm in the SUPPORT trial (85% to 89%) was not in fact consistent with usual care (Cortés-Puch et al. 2016). Cortés-Puch and colleagues "sought to determine whether each oxygen target as studied in SUPPORT and four similar randomized controlled trials (RCTs) was consistent with usual care" (1).6 They conducted a literature review to establish the range that constituted usual care concurrent to these five clinical trials. This review uncovered a prospective, observational study, known as the AVIOx study. The AVIOx study collected robust data on oxygen exposure in extremely premature infants, who, importantly, would have satisfied the primary enrollment criteria for the SUPPORT trial (Hagadorn et al. 2006). The study enrolled 84 infants across 14 neonatal intensive care units in the U.S., U.K., and New Zealand; eligible infants included those who were born at less than 28 weeks gestation, and for whom oxygen therapy was required (Cortés-Puch et al. 2016).
The comparison of low and high target ranges of supplemental oxygen in the SUPPORT trial with those used in the AVIOx study revealed that while the high target range (91% to 95%) was consistent with usual care, the low target range (85% to 89%) was "lower and narrower than those applied during usual care" (Cortés-Puch et al. 2016, 4). The upper limit of the low target range used in the SUPPORT trial was 89%, while the AVIOx study showed that the upper limit of usual care was between 92% and 98%. Moreover, the low target range in the SUPPORT trial was narrower than comparably low target ranges used in clinical practice (e.g., 85%); this is because, in clinical practice low target ranges are always coupled with higher target ranges (e.g., 98%) than the upper limit of the low-oxygen saturation intervention in the SUPPORT trial allowed (i.e., 89%):
During usual care, similar lower limits (< 88%) were universally paired with higher upper limits (92%) and providers skewed achieved oxygen saturations toward the upper-end of these intended ranges.(6)
In other words, the wider range characteristic of usual care results in achieved (i.e., median) values of supplemental oxygen higher than the [End Page 108] upper limit of the low intervention arm in the SUPPORT trial. The authors conclude that, "the low range (85 to 89%) was not used outside of the SUPPORT trial," and therefore was not consistent with usual care (1).
It is evident, then, that the SUPPORT trial was not a comparison of two usual care interventions, but rather involved one or more novel interventions. We have argued that the bifurcation of a usual care range no longer constitutes usual care, as the benefits and harms associated with each subset of that range differ from the full range. Indeed, the purpose of the SUPPORT trial was to evaluate the benefits and harms associated with low and high ranges of supplemental oxygen. This challenges the idea that the SUPPORT trial involved the comparison of any usual care interventions: rather, it compared two novel interventions. This analysis may apply to other pragmatic ceRCTs that compare subsets of a usual care treatment, characterized by a range. Moreover, regardless of whether one agrees with our contention that the bifurcation of a usual care range no longer constitutes usual care, Cortés-Puch and colleagues make it clear that the low-oxygen saturation arm was not, in fact, routine medical practice. Thus we conclude that the SUPPORT trial interventions ought to be conceptualized as research, not practice.
7.4 Implications of Conceptualizing the SUPPORT Trial Interventions as Research
Several implications follow straightforwardly from the determination that low- and high-oxygen saturation arms in the SUPPORT trial ought to be conceptualized as research. First, both the low- and high-oxygen saturation interventions require third-party review and regulatory oversight by research ethics committees. There is consensus among national and international research ethics guidelines that human participants in research be submitted to, and approved by, a research ethics committee. Research ethics committee review safeguards the autonomy and welfare interests of research participants, and ensures that national and international ethics standards are upheld. As the SUPPORT trial interventions clearly fall within the domain of research, a research ethics committee ought to review and approve them.
Second, the study interventions must undergo a systematic harm–benefit analysis. The need for harm–benefit analysis is grounded in the ethical principal of beneficence, which requires that the welfare interests of human research participants are protected, and, where possible, promoted. Benefit–harm analysis requires that the low- and high-oxygen saturation [End Page 109] interventions satisfy clinical equipoise. Clinical equipoise is an ethical precondition for enrolling patients in a randomized clinical trial (Horn and Weijer 2015); it states that there must exist "an honest, professional disagreement among expert clinicians about the preferred treatment" (Freedman 1987, 144). As discussed above, the optimal oxygen saturation range for the treatment of extremely premature infants is unknown, and a Cochrane review of the evidence base concluded that "further research should be undertaken to resolve this important question" (Askie et al. 2009, 381). Thus, the SUPPORT trial interventions fulfill the ethical requirement of clinical equipoise.
Third, the details of the study interventions must be disclosed in the informed consent process. The principle of respect for persons underlies the doctrine of informed consent, and ensures that the autonomy rights of human research participants are respected; that is to say, that patients are afforded the opportunity to make free and informed choices as to whether participation in a trial is consistent with their values. Disclosure of the SUPPORT trial interventions must include details of the study procedures, any associated benefits and risks of harm, including retinopathy of prematurity, neurological damage, lung toxicity, and mortality.
It is also clear from this analysis that questions concerning whether enrolled infants were exposed to additional risk when compared with infants not so enrolled are not central to, or determinative of, the substantive ethical issues to which the SUPPORT trial gives rise. The answers to questions about risk, disclosure, and consent are secondary to—and follow from—an antecedent determination as to whether the usual care interventions ought to be conceptualized as practice or research.
The debate surrounding the SUPPORT trial has focused on determining whether the low- and high-oxygen saturation interventions posed "reasonably foreseeable risks" to study participants. This preoccupation with U.S. regulation has not only made this debate inapplicable to an international audience, it has also failed to identify and engage the SUPPORT trial's central ethical issue.
The central ethical issue raised by the SUPPORT trial is the following: should the SUPPORT trial interventions be conceptualized as practice, or research? To address this question, we have attempted to provide a systematic analysis of the SUPPORT trial. In so doing, we have attempted to reframe the debate about the SUPPORT trial in a way that is both [End Page 110] historically informed and ethically relevant. We have provided three reasons for conceptualizing the low- and high-oxygen saturation arms in the SUPPORT trial as research, and not practice. First, the enrolled infants are deprived of the benefit and protection of the individualized judgment of their physician. Second, the physician's ability to continuously monitor, reassess, and adjust the prescribed oxygen therapy administered to a premature infant is compromised by altered-pulse oximeters. Third, in those cases where a usual care treatment is characterized by a range, or continuum, the bifurcation and ensuing randomization of research participants to protocolized subsets of that range may no longer constitute usual care. An important implication of the third argument is that while the SUPPORT trial is a pragmatic ceRCT—as it compares two interventions, head-to-head, under real-world conditions—it is not a ceRCT that compares two usual care interventions. We propose that the first two reasons are likely applicable to a broader class of pragmatic ceRCTs involving one or more usual care interventions, whereas the third reason is likely applicable to those ceRCTs where the treatment used routinely in clinical practice is characterized by a range; however, we leave questions regarding the applicability of our analysis for the broader class of pragmatic ceRCTs open for future inquiry. Finally, we have demonstrated that how we conceptualize these interventions has implications for which components of a study protocol fall within the purview of research ethics committees, undergo benefit–harm analysis, and are appropriately disclosed in informed consent procedures.
Nevertheless, with Lantos, we "acknowledge the distinctiveness" of pragmatic ceRCTs involving one or more usual care interventions, and thus recognize that important questions remain regarding how these socially important and, broadly speaking, low-risk studies ought to be regulated. For instance: should those pragmatic ceRCTs that are deemed low-risk undergo an expedited review process, proportionate to the risks and benefits involved? Ought there be modifications to the informed consent process? Addressing these and other difficult questions, and the ethical challenges they give rise to, requires further investigation.
Our analysis makes it clear, however, that any proposed solution to these important questions must proceed in a clear and systematic fashion, grounded in internationally accepted ethical principles. The ethical principles upon which contemporary research ethics was founded aim first and foremost to promote the liberty and welfare interests of individual participants. Based in these ethical principles, the analysis provided in [End Page 111] this paper has cleared a critical conceptual hurdle for pragmatic ceRCTs involving usual care interventions: namely, how we ought to conceptualize study interventions involving treatments used routinely in medical practice.
Austin R. Horn is a PhD student in philosophy at Western University. His doctoral research focuses on the intersection of biomedical ethics and the philosophy of fiduciary law.
Charles Weijer, FRSC, is Professor in the Departments of Philosophy, Medicine, and Epidemiology and Biostatistics, and he holds the Canada Research Chair in Bioethics at Western University in London, Canada. He is a leading expert in the ethics of randomized controlled trials.
Jeremy Grimshaw, MBChB, PhD, is a Senior Scientist at the Ottawa Hospital Research Institute and Full Professor in the Department of Medicine at the University of Ottawa. He holds a Canada Research Chair in Health Knowledge Transfer and Uptake.
Jamie Brehaut, PhD, is a Senior Scientist with the Ottawa Hospital Research Institute, and Associate Professor in the School of Epidemiology and Public Health at the University of Ottawa. Much of his work focuses on the application of psychological theory to facilitate health care practice change and the ethics of health care decision making.
Dean Fergusson, MHA, PhD, is a Senior Scientist and Director, Clinical Epidemiology Program, at the Ottawa Hospital Research Institute. He is also a Full Professor, Departments of Medicine, Surgery, and of the School of Epidemiology and Public Health, University of Ottawa. Dr. Fergusson holds the position of Endowed Chair, OHRI/uOttawa, Clinical Epidemiology Program.
Cory E. Goldstein, MA, is a philosophy doctoral student at Western University in London, Ontario, Canada. He received his BA from McGill University (2014) and his MA from Western University (2016). His academic interests are primarily in research ethics, particularly the ethical design and conduct of medical research.
Monica Taljaard, PhD, is a Senior Scientist with the Ottawa Hospital Research Institute, and Associate Professor in the School of Epidemiology and Public Health at the University of Ottawa. She is a biostatistician with a research interest in the design and ethics of cluster randomized and pragmatic trials.
1. The concept "standard of care" is derived from the legal lexicon, and the use of the concept commonly fails to disambiguate between its normative and descriptive meanings. On the one hand, "standard of care" might refer to what healthcare providers ought to do, given the best available evidence (the normative meaning); on the other hand, it might refer to what healthcare providers in fact do in clinical practice (the descriptive meaning). For the purposes of this paper, we have in mind a descriptive meaning of the term "standard of care." We use language such as "usual care" and "routinely used in practice" to highlight this descriptive meaning.
2. Internationally accepted ethical principles are articulated in several documents, including the Council for International Organizations of Medical Sciences (CIOMS) International Ethical Guidelines for Health-Related Research Involving Humans (CIOMS 2016), CIOMS International Ethical Guidelines for Epidemiological Research (CIOMS 2009), and the World Medical Association Declaration of Helsinki (World Medical Association 2013).
3. This paper focuses entirely on the component of the SUPPORT trial that compared the two target levels of oxygen saturation. However, others take issue with the aspect of the trial that compared two different ventilation approaches, continuous positive airway pressure and surfactant (cf. Carome, Wolfe, and Macklin 2013; Macklin and Shepherd 2013).
4. It is worth emphasizing that, pace Lantos, neonatologists do—and, indeed, must—exercise their professional judgment whether or not that judgment is constrained by a defined protocol. As just illustrated, the range of oxygen saturation routinely used in a neonatal intensive care unit is but one of many factors that neonatologists must consider when treating extremely premature infants. Furthermore, neonatologists must use their professional judgment to determine what subset of that defined range is appropriate for an infant, in light of all other considerations. This argument is expounded in more detail below.
5. As a point of clarification, it is worth emphasizing that we nonetheless maintain that the SUPPORT trial is a pragmatic ceRCT, as it compares two interventions head-to-head under real-world clinical conditions. Nonetheless, it is an implication of this third argument that the SUPPORT trial may not be a specific type of pragmatic ceRCT, namely, one that compares one or [End Page 112] more usual care interventions. The contention that the low- and high-oxygen saturation interventions are not usual care therefore provides direct, and independent, support for our thesis that the SUPPORT trial interventions ought to be conceptualized as research, not practice. Of course, as previously mentioned, we acknowledge that there is a separate question regarding the utility of our analysis of the SUPPORT trial for pragmatic ceRCTs more broadly; a thorough investigation into this further question is, however, beyond the scope of this paper.
6. The four other trials included the Benefits of Oxygen Saturation Targeting trials (BOOST II)—in Australia, New Zealand, and the U.K.—and the Canadian Oxygen Trial (COT).