I recently participated in a debate opposing me to Professor Adam Cifu on the topic of “Evidence-based medicine in the age of COVID.” The debate took place on an episode of Dr. Chadi Nabhan’s Outspoken Oncology podcast. Dr. Saurabh Jha was the moderator and he did a great job keeping us on point and asking for important clarifications when needed. It was a fun and cordial moment and I found it intellectually fruitful. You can listen to it here or on any podcast platform. The discussion strengthened my conviction that the central issue about EBM is the conflation of the role of the physician with that of the clinical scientist.
That conflation was quite apparent in a recent online editorial published by Robert Yeh and colleagues on the topic of equipoise during the COVID-19 pandemic. Yeh at al. are accomplished academic cardiologists and outcomes researchers (Yeh was a guest on The Accad and Koka Report a couple of years ago).
I’ll get to their editorial in a moment, but equipoise is a term that I became aware of only in the last few years, mainly from mentions on MedTwitter. From those mentions I developed an intuitive sense of what equipoise must mean: a mental state of uncertainty about a treatment that prompts the medical community to seek a more definitive answer by way of a randomized controlled trial. For example, one might say “I’m not sure if hydroxychloroquine works to prevent or treat COVID-19. Based on the existing collective experience, there is equipoise about it. We need a clinical trial.”
That seems reasonably straightforward, but the editorial by Yeh et al. piqued my curiosity so I decided to look into the origin of the term and its introduction in the medical literature.
It turns out that the term equipoise (“equal weight”) was introduced into the domain of clinical research in 1974 by Charles Fried in a textbook on research ethics.
Fried asserted that there should be an “indifference requirement” when an investigation compares two treatments. If the investigator has reason to believe that one treatment arm of a clinical experiment is inferior to the other(s), then that arm should be eliminated or the trial cannot ethically proceed. The idea was not original to Fried. Other ethicists, such as Lawrence Shaw and Thomas Chalmers, had previously postulated that requirement but without giving it a specific name.
At any rate, Fried’s term remained in relative obscurity until it was put in the title of a special article in the New England Journal of Medicine in 1987. In “Equipoise and the Ethics of Clinical Research,” Benjamin Freedman described equipoise as a state of “genuine uncertainty” in the conduct of clinical trials, but he offered additional clarity about the meaning of the term in this context.
In his article, Freedman cautioned against what he termed “theoretical equipoise,” i.e., a too strict and too abstract understanding of the indifference requirement. According to this stringent interpretation, a researcher can ethically proceed with an experiment only if a “50-50” indifference regarding the treatment arms exists. Any time he or she has reason to believe that one arm may be better than the other, then the experiment cannot ethically proceed further.
This possibility was vividly on my mind a few weeks ago when comments made privately by investigators at the University of Chicago, one of the sites of a multi-site trial of Gilead’s antiviral Remdesivir, were leaked to STAT News.
The UC physicians commented during their internal meeting that the hospital course and length of stay of Remdesivir-treated COVID patients seemed to them to be much improved as compared to what was being reported about hospitalized patients from China, Lombardy, or other centers with severe mortality rates. This “good news” about Remdesivir spread like wildfire, causing Gilead’s stiock to shoot up. But many were quick to dismiss it as just a rumor or a biased impression, especially given that that trial had no concurrent randomized control group for comparison.
Still, the information seemed important, so I conducted a survey among physicians on Twitter about how this piece of news should affect enrollment in the other, ongoing randomized controlled trials of Remdesivir. What should prospective patients or currently enrolled patients be told?
Interestingly the answer to my survey showed…equipoise
In other words, 50% of those surveyed did not think that the additional information had relevance to the conduct of on-going randomized control trials, even though the information was not available at the time the trials were approved or when they were beginning to enroll patients.
But according to Freedman, research ethicists from the 1970s and 1980s clearly thought that this kind of information not only could have relevance for the conduct of trials but could theoretically jeopardize the entire research enterprise.
Freedman mentioned other ways that equipoise can be “disturbed.” Not only can any new experience with a treatment outside of the experimental study tilt the balance one way or another, but once a trial is begun, any accumulated knowledge in the course of the trial can also sway the judgment of physicians in favor or against the treatment. If, say, among the first 12 enrollees of a 100-patient trial the outcomes seem better with treatment A compared to treatment B, shouldn’t patient #13 and beyond be at least made aware of that fact?
Freedman mentioned some solutions that ethicists had previously proposed to deal with this thorny problem. One of them is actually already part of clinical trial methodology: the blinding of the investigators. In the course of a trial, accrued outcomes are not analyzed by the physicians involved in the care of patients but by an independent data monitoring committee. That way, the clinicians are kept from being biased by the data as it accumulates.
But Freedman rightly pointed out that, from an ethical standpoint, blinding clinicians is like sticking their heads in the sand:
If interim results would disturb equipoise, the investigators are obliged to gather and use that information. Their agreement to remain in ignorance of preliminary results would, by definition, be an unethical agreement, just as a failure to call up the laboratory to find out a patient’s test results is unethical.
Clearly, equipoise highlights the tension between the therapeutic aims of medicine and the scientific aims of research.
Other proposals have been suggested. One of them, advanced by Paul Meier, was to stipulate that participation in clinical trials is commensurate with the everyday risks of not always getting one’s best choices in life. Trial subjects, therefore, should not necessarily expect to get the best care possible.
Another idea, proposed by Arthur Schaefer, was to let patients be the judge of whether equipoise is present or not, a form of “shared decision-making” but only sillier than the one currently in vogue. A third one—at the opposite end of the autonomy spectrum—was to simply conscript patients into participating in clinical trials. Ethics be damned!
Freedman wisely considered these proposals to be “counsels of desperation.” The problem, according to Freedman, was that understanding equipoise in the theoretical sense of holding the 2 treatment arms exactly balanced is “conceptually odd and ethically irrelevant.” Such an understanding of equipoise is so “highly sensitive to the vagaries of the individual investigator’s attention and perception” as to make the concept “overwhelmingly fragile.” Equipoise is liable to be overthrown even on a “hunch” or a personal bias.”
Besides, the “theoretical equipoise” could only apply to very “clean” and one-dimensional hypothesis testing of the type “Is A superior to B in producing outcome X.” In reality, Freedman asserted, clinical research is always conducted with a plurality of outcomes in mind. Not only are benefits tallied, but various side-effects as well. In such a situation, theoretical equipoise is impossible to maintain or even hold to begin with. It’s an irrelevant mental abstraction.
Instead, he offers the concept of “clinical equipoise” which he defined as the uncertainty that arises not when a particular physician cannot decide the superiority of one treatment over another, but more broadly when the “clinical community” is divided and when there is “an honest, professional disagreement among expert clinicians about the preferred treatment.”
Clearly, clinical equipoise as stipulated by Freedman is much less liable to be disturbed than the original concept. So long as there is disagreement among a sufficiently large number of colleagues or clinicians, equipoise can be said to exist, and a clinical trial can be justified on ethical grounds:
…clinical equipoise places the emphasis in informing the patient on the honest disagreement among expert clinicians. The fact that the investigator has a “treatment preference,” if he or she does, could be disclosed; indeed, if the preference is a decided one, and based on something more than a hunch, it could be ethically mandatory to disclose it. At the same time, it would be emphasized that this preference is not shared by others.
Freedman’s proposal has some prima facie reasonableness. Medicine is a collegial enterprise. Rare is the physician who holds opinions in isolation, without the influence or agreement of others. In my debate with Dr. Cifu I remarked that if the community of intensivists remains divided on the question of the benefit of early versus delayed intubation for severe COVID-19, then they may agree to conduct a clinical trial and would be justified in doing so. It would indeed be ethical and realistic to tell a patient “look, I believe that treatment A can help people like you, but other doctors disagree or are concerned that the risks outweigh the potential benefits and that’s why we think people like you should enroll in a trial.”
Freedman’s concept of clinical equipoise would be all well-and-good if modern medicine were built on a sound moral and epistemological foundation. But Freedman made his proposal in 1987, only a year before David Eddy coined the term “evidence-based” and just a couple of years before Sackett et al. articulated the principles of EBM that would become the new dogma of academic medicine.
The EBM takeover of medicine has changed the concept of equipoise in the opposite way to that which Freedman feared. If, in the early days, “theoretical equipoise” was conceivably too fragile and under the constant threat of the whims of the hunches and preferences of individual clinicians, equipoise in the age of EBM has become unassailable.
After all, it is the central dogma of EBM that there is no firm knowledge about a treatment unless the evidence comes from a well-conducted, randomized, blinded clinical trial. If there is no randomized controlled trial, there is no substantive knowledge. As long as EBM reigns, the demand for equipoise is self-fulfilling.
If a new treatment is being to be tried or an old treatment repurposed for a new set of circumstances, the default position that any clinician must hold is the agnostic one, no matter what the prior experience is up to that point: “There is no proof!” Otherwise, one practices out of the bounds of evidence-based medicine. With such a default position, it can be unethical for a physician to treat a patient outside of a clinical trial, a claim that’s been made repeatedly by EBM champions in the course of the COVID pandemic.
Equipoise for all?
But pandemics are not friendly to equipoise. In the age of COVID-19, patients, clinicians, scientists, and even politicians are restless and ready to pounce at the first bit of hopeful news, even if that news is questionable: “We cannot wait for perfect data!” is the motto of the day.
It’s against that tendency to spring into frenzied action during a pandemic that Yeh et al. made their plea in their editorial blog entitled “Equipoise on Covid-19 therapeutics.”
Here, Yeh and colleagues stress the valid point that a pandemic is precisely not a time to send caution to the wind. They rightly warn that an “exuberant embrace of untested treatments may lead to toxicities” and also delay the discovery of effective therapies. They lament that this is taking place because equipoise has not been “maintained.”
But having just read Freedman’s article, I cannot help but think that Yeh et al. may have also sent his cautionary distinction on equipoise to the wind. For, in framing their discussion of equipoise, these authors are not so careful to distinguish the individual practicing doctor from the trialist, let alone the individual clinical scientist from the “community of clinicians.”
Consider the following passages:
Traditional pathways for evaluating new treatments through randomized trials hinge on establishing and maintaining equipoise at a broad public health level, but also at the bedside where wrenching decisions are made in real-time.
Equipoise—that is, whether physicians making bedside decisions genuinely maintain uncertainty regarding whether a given option is helpful or harmful—can erode rapidly for several reasons.
In both of these sentences one gets the sense that the authors do not view equipoise as a condition necessary to allow a researcher to engage in experimental medicine, i.e., the way Freedman and his predecessors had initially conceived of it. For Freedman and for the ethicists, it’s the trialists’ job to decide if equipoise is either present or not in order to proceed with the experiment.
Instead, for Yeh et al. (and for many contemporary clinician-scientists, no doubt), equipoise is a sort of “epistemic virtue” that both scientists and practicing doctors must cultivate and “maintain” until a clinical trial can ultimately adjudicate the value of a particular treatment. Repeatedly in their article the authors speak of the difficulty of “preserving” equipoise in the age of COVID.
Furthermore, they lump the bedside clinician with the trialist. And that conflation is not benign if we consider the coercive policy proposals they recommend: “stewardship committees” to restrict the use of certain treatments outside of clinical trials:
The coronavirus era demands immediate design and adoption of institutional policies, shaped by clear national guidance, to preserve equipoise and allow focused randomized trials to enroll and yield data as quickly as possible….A model for such policies might be the way in which many hospitals strictly limit access to certain antibiotics, requiring approval from a specialist stewardship committee prior to release from pharmacy.
We also call for widespread commitment among clinicians to restrict use of purported coronavirus treatments outside of clinical trial. This must be paired with a reduction of barriers for pooling patient-level data to maximize knowledge generation and, in doing so, honour the selflessness of patients or families who agree to participate (emphasis mine).
Yeh et al. may rightly consider patients who enroll in trials to be selfless, but they should be mindful that if participation in a trial is the only possible way to access a promising treatment, then those patients may very well be acting out of justified self-interest.
Clear thinking is the way forward
As Yeh and colleagues mention in their article, there is an inherent tension between doctoring and experimenting. But we must deal with this tension with clear thinking, not by muddling ethical and epistemological concerns.
One approach may be to toss out the idea of equipoise altogether.
A few years ago, Franklin Miller and Steven Joffe proposed doing just that in the pages of the New England Journal of Medicine. They raised reasonable objections to Freedman’s concept of clinical equipoise arguing, for example, that there is no objective way to determine if equipoise is present. What does it mean that a sufficient number of clinicians should disagree on the value of a treatment? Instead, Miller and Joffe justified the conduct of clinical trials on the basis of their socioeconomic benefits, arguing that population-level considerations trump any right a patient may have to a new treatment, especially if it is a costly one.
I disagree with Miller’s and Joffe’s severe utilitarianism. For one thing, it has no strong philosophical or moral justification. Utilitarianism is just a preference. For another, the same arguments that they have raised against equipoise can be raised against their own economic arguments. When is a treatment costly enough to warrant that it be made available only after vetting by an experimental trial? Who decides? And what if an old but still costly treatment was not vetted by a clinical trial? Should it be taken away from patients until a trial is conducted? Also, Miller and Joffe seem to have given little thought or consideration to understanding why treatments are costly to begin with.
Like Yeh and colleagues, I think that equipoise is a helpful concept that should not be abandoned. But the way out of the dilemma is precisely to remain mindful that physicians and clinical scientists have distinct primary aims. The aim of the physician is to bring about health in a particular patient. That of the clinical scientist is to bring about some general knowledge that will help physicians. It is to the scientist that equipoise pertains. The fact that the physician and the clinical scientist share the same playpen, or that one and the same person may have a dual role, should not obscure that they each have a radically different task at hand.