Progressing from “Whether to” to “How to” Conduct Pragmatic Trials

In this issue of the American Journal of Bioethics, manuscripts focus on the obligations of clinicians and researchers in pragmatic clinical trials.^1,2 As clinicians who care for patients enrolled in pragmatic trials, researchers who lead multicenter pragmatic trials, and administrators in a health system in which pragmatic trials are embedded into clinical care, we have unique perspectives on these topics.

Here is the central dilemma: every day clinicians are confronted with clinical decisions for which the best treatment for patients is unknown. In our practice setting, decades-old clinical questions such as the best treatment for alcohol withdrawal, the best vasopressor for septic shock, and the best approach to emergency tracheal intubation have remained unanswered despite millions of patients receiving these interventions each year. In the absence of evidence regarding the best therapy for a given patient, variability develops within clinical care based on factors unrelated to patient characteristics, such as specialty, geography, hospital formulary, or industry marketing. Because a given patient would receive different treatments based on where, when, and from whom he or she receives care, patients are systematically exposed to interventions that may be suboptimal or even harmful.

Traditional, explanatory trials are poorly suited to address this problem.³ These trials and the regulations that govern them were developed around the paradigm of drug and device development. They focus on protecting patients from the risks of experimental therapies – therapies they would not receive in routine clinical care, as opposed to therapies which are already being utilized in the routine clinical care of patients. Supported by industries that stand to profit from their results, drug discovery trials have successfully brought numerous new therapies to market in recent years. In contrast, comparatively few trials have attempted to aid clinicians and patients in choosing among the treatments that are used commonly used in clinical practice. Several barriers contribute to this failure. First, the difference in outcomes between two treatments used in clinical practice is likely to be smaller than the difference in outcomes between an active treatment and no treatment (i.e., a placebo-controlled drug trial). Thus, trials comparing the effectiveness or safety of two treatments commonly used in clinical care must be significantly larger than traditional drug approval trials. Second, the manner in which screening, consent, intervention delivery, and safety monitoring occur in explanatory trials do not fulfil the goal of establishing the effectiveness among all patients likely to receive a treatment in “real world” clinical care and are mismatched with the relatively low incremental risks of participation in a trial examining the same therapies that would be received in routine clinical care.⁴ Third, the time, cost, and personnel required for explanatory trials preclude use of these methods to answer the innumerable comparative effectiveness questions that clinicians and patients routinely encounter in clinical care.

Pragmatic trials address gaps in evidence for commonly used treatments by embedding study procedures within “real world” clinical care. This approach increases generalizability by evaluating treatments as they would be delivered in clinical care among the patients who would receive them in clinical care. This approach may also increase efficiency and decrease cost by avoiding duplicative personnel and infrastructure for research and clinical care (e.g., using data collected from the electronic health record rather than manually collected by a research nurse). Every study procedure in a clinical trial (e.g., eligibility, enrollment, delivery of the trial intervention, outcome assessment) exists on a spectrum from explanatory to pragmatic.⁵ Increasingly, experts have recommended transitioning from a paradigm that starkly segregates clinical care and research to a “learning healthcare system” in which evidence gaps are identified, pragmatic trials are embedded within clinical care to address these evidence gaps, and trial results are implemented into clinical care to improve patient outcomes.^6,7 Achieving this vision will require addressing novel challenges inherent to pragmatic trials embedded within clinical care, including the important ethical and regulatory considerations addressed by articles in this edition of the American Journal of Bioethics.

Pragmatic trials embedded within clinical care involve the participation of clinicians who might not otherwise be involved in clinical research. Clinicians may be asked, in the context of providing clinical care, to facilitate one or more of the study procedures of a pragmatic trial (Table 1). In their manuscript, Garland, Morain, and Sugarman argue that clinicians have an obligation to participate in pragmatic trials based on duties of care and beneficence to their current and future patients and duty of fairness to fellow clinicians.¹ They also highlight hypothetical exceptions to these duties, which they categorize as moral or non-moral objections. Moral objections include general objections to the trial (e.g., a belief that the trial will worsen patient outcomes) or objections specific to the clinician (e.g., the clinician has a personal, conscientious objection to the study intervention). Non-moral objections similarly include concerns general to the trial (e.g., a belief that the trial is poorly designed and unlikely to answer the study question) and those specific to the clinician (e.g., the clinician lacks the training needed to perform the study intervention or the time required to complete study procedures).

Table 1.

Study procedures that clinicians might perform in a pragmatic trial

Study Procedure	Description	Example from a pragmatic trial
Screening	Clinicians may be asked to screen for the presence or absence of eligibility criteria.	In a recent pragmatic trial comparing two commonly used antibiotics for the empiric treatment of acute infection, clinicians completed an alert in the electronic health record confirming that the patient was not allergic to either antibiotic and was potentially eligible for enrollment (NCT05094154).
Consent	Clinicians may be asked to obtain informed consent for participation in the trial.	In a platform trial of therapies for adults hospitalized with COVID-19 in the United Kingdom, consent for participation in the trial was obtained by the patients’ treating clinicians, who received limited study-specific training, rather than by researchers with extensive training in human subjects research protections (NCT04381936).
Enrollment and randomization	Clinicians may be asked to perform enrollment and receive the randomized trial group assignment.	In a recent pragmatic trial comparing the two available methods for securing endotracheal tubes, clinicians opened opaque, serially numbered envelopes to enroll patients and reveal group assignment (NCT03760510).
Intervention delivery	Clinicians are commonly asked to deliver interventions in pragmatic trials to ensure they are delivered in the same way they would be delivered in routine clinical care.	In a recent electronic health record-embedded pragmatic trial comparing two available approaches to screening for gestational diabetes, providers were notified of group assignment at the time the test was being ordered and then delivered the assigned screening test (NCT02266758).
Outcome assessment	Clinicians may be asked to record study outcomes that are not available in clinical records.	In a recent pragmatic trial of emergency tracheal intubation, clinicians recorded real-time values for oxygen saturation and blood pressure that were unavailable in the electronic health record and could not be recorded by research staff because of the emergency nature of the procedure (NCT03928925).
Notification of enrollment	In trials conducted with a waiver of informed consent, clinicians may be asked to notify patients of their participation and provide IRB-approved information about the trial.	In a recent study of the effectiveness of corticosteroids among patients with head injury, patients were enrolled with waiver of consent and enrollment materials included a patient information sheet that clinicians provided to patients and families (ISRCTN74459797 ).

We share Garland et al.’s views regarding both the duties of clinicians and potential exceptions to these duties. As noted by the authors, engaging the clinicians who will participate in a pragmatic trial in its design can mitigate many of the hypothetical objections to participation. Engaging with clinicians and health system leaders through surveys, listening sessions, and collaborative leadership of learning activities within the health system can identify both questions best suited for a pragmatic trial and the sites best suited to conduct such trials. A site may choose not to participate in a trial because of a lack of equipoise regarding the research question, a lack of experience with one of the treatments examined, or other local factors. Importantly, a site may reasonably choose not to participate due to lack of clinical equipoise locally even if the trial experiences broad acceptance and clinical equipoise at numerous other sites. Similarly, a clinician may have equipoise for a given trial generally but still encounter specific patients for whom, or settings in which, equipoise is not present because of specific patient or situational factors. Thus, a foundational aspect of the design of all pragmatic and explanatory trials must be allowing clinicians the flexibility to either exclude patients (individual-level trials) or provide non-assigned therapies (cluster-level trials) when felt to be required for the optimal care of the patient. Such exclusions or crossovers should be tracked closely during a trial and reported with trial results; if sufficiently numerous they can threaten the trial’s validity. However, we have found that with an appropriate research question, careful site selection, and deliberate clinician engagement, such events are relatively uncommon and occur at a rate similar to the rate in traditional, explanatory trials. We have found that many clinicians find satisfaction in participating in pragmatic trials to answer questions of importance to them and their patients.

In their manuscript, Morain and Largent address the ways in which the traditional model of a researcher-participant dyad is inapplicable to pragmatic clinical trials.² Through specific case studies, they highlight 6 key differences in the relationship between the researcher and the patient in traditional, explanatory trials vs pragmatic trials: 1) pragmatic clinical trials may be conducted with a waiver or alteration of consent, 2) pragmatic researchers may not have direct interaction with study participants, 3) addressing individual-level findings in a pragmatic clinical trial may be prohibitively costly to institutions and research enterprises, 4) pragmatic clinical trials are embedded within real-world settings where gaps in care are common, 5) individual-level findings in pragmatic clinical trials may have implications for others receiving care in that setting, and 6) the tools used to address individual-level findings in explanatory trials (e.g., prescribing a new medication) may not be available in pragmatic clinical trials. The authors conclude that the traditional paradigm used for researcher-patient dyads cannot be applied to pragmatic trials. We agree with the authors’ arguments and their conclusion that we must look beyond the researcher-participant dyad and explore the extent to which the institutions in which pragmatic clinical trials occur should bear additional responsibilities. As institutional leaders, we would welcome further scholarship and guidance in these areas.

Finally, we believe that the US healthcare system has a basic choice to make: allow arbitrary variation in clinical care and continue to systematically expose patients to suboptimal or harmful therapies indefinitely or structure that variation through pragmatic trials to generate knowledge, reduce variation, and improve outcomes over time. During the COVID-19 pandemic, the US made the wrong choice. Rather than revising regulations to facilitate the conduct of pragmatic trials to quickly prove the effectiveness of available treatments, the US FDA maintained regulations for research and relaxed the regulations for the use of unproven treatments in clinical care. For example, the FDA set up the Mayo Expanded Access Program to facilitate the use of convalescent plasma, an unapproved therapy, in clinical care. The program included nearly all the study procedures of a pragmatic trial embedded in clinical care (Table 1). Treating clinicians (not researchers) identified eligible patients and obtained consent for the unapproved therapy. When consent was infeasible, patients were enrolled with the agreement of a second physician (21 CFR 50.23).⁸ Treating clinicians delivered the intervention and collected outcome data in a research database. This was essentially a 100,000-patient pragmatic trial without a control arm. Because of the lack of a control arm, the program did not generate the information clinicians needed to inform the effectiveness of convalescent plasma. Worse, relaxing the regulations for the use of convalescent plasma in clinical care coupled with strict regulations for the evaluation of convalescent plasma in clinical trials made it challenging for trials of convalescent plasma in the US to enroll patients.⁹ Ultimately, the RECOVERY trial, a pragmatic trial in United Kingdom, used an embedded consent model similar to that of the Mayo Expanded Access Program to randomize 11,558 patients hospitalized with COVID-19 to convalescent plasma or usual care and found that convalescent plasma – the treatment administered to hundreds of thousands of patients in the US in clinical care – was ineffective.¹⁰

We are hopeful that one lesson of the COVID-19 pandemic in the US will be the imperative to develop the ethical and regulatory framework required to successfully embed pragmatic trials within clinical care. The FDA recently endorsed the conduct of highly efficient “pragmatic point-of-care” trials like the RECOVERY trial.¹¹ Recent bioethics articles, like those in this issue, have progressed from addressing whether pragmatic trials should be conducted to addressing how they might best be conducted. As clinicians, pragmatic trialists, and healthcare leaders, we welcome this progress.