Learning Health Systems - What is Low-Risk, Low-Burden Research

Dr. Mikhail Kosiborod

Great.

Dr. Richard Payne

Thank you. So we thought at this point we'd have Drs. Kosovorov and Dr. Lantos come back up and

Dr. Richard Payne

that we would talk more about these issues related to ethical oversight of

Dr. Richard Payne

They're debating who goes first. Choice is up to you. I'll stay back here for a little. So is that yours? Yeah, let's see. Because I asked them to consider, in the context of comparative effectiveness research and outcomes-based studies, what are we talking about? when we talk about minimal risk, low risk, low burden, anticipating this issue of when is it appropriate to talk about modifying the kinds of consent that would be required with this kind of research.

Dr. Mikhail Kosiborod

Thank you. And I really didn't intend for this to be a formal presentation, so I just put some thoughts together to kind of really maybe stimulate discussion more than anything else. And I guess the argument here is that when you have research activities that provide large benefit to society as a whole and are a minimum risk, they deserve to be handled a bit differently. The opposite, maybe, which would be the activities that may, and let me premise by saying, may provide significant benefit to society, provide the benefit almost immediately, so it's something that's immediately applicable to care that patients receive from doctors and hospitals, and expose patients to minimal risk above and beyond what they normally expose to the part of care. should that be handled differently from something different, which may be

Dr. Mikhail Kosiborod

research that exposes patients to substantial potential risk and may provide huge benefits at some point later, decades from now, but not immediately, and more importantly, exposes patients to more risk. So when you talk about outcomes and comparative effectiveness research,

Dr. Mikhail Kosiborod

just like any umbrella term, it includes many different types of studies, many different types of research, but some of the things that we discussed earlier, for example, the development of large clinical databases, utilizing those databases to develop prediction models, using those prediction models to develop clinical decision support tools that's evidence-based and implementing that as a point of care to improve care for the patients, that kind of a research does have potential huge benefits to us as a society. It generates new hypothesis. It can evaluate effectiveness and safety of treatments, improve process and quality of care, improve outcomes and reduce costs. And again, this is just intended to be provocative rather than making any definitive statements here. But should you consider something... along these lines as a public good that then by default requires public investment, and the public investment would be not just monetary investment in terms of grants and research support, but maybe some easing of regulatory burden, I guess you could say. And probably as importantly, if not more importantly, if this type of research also creates minimal risk of the patient,

Dr. Mikhail Kosiborod

non-negligible, but minimal, above and beyond regular clinical care. So as an example, when a patient comes to the hospital, they typically give a release of information to the hospital. If that data is then de-identified and used for clinical research, that would be considered something along these lines. And then if you are a researcher, then that uses that data, do you need to go through IRB if there is really no risk involved to the patient as in using the de-identified data? So the conditions, so I guess the idea here is should these types of studies be exempt from IRB review? And if they are, what should the conditions be? Should one of the conditions be that personal health information is de-identified and whoever is doing the de-identification to demonstrate that this is not something superficial, but that the data is truly protected. And if you have centers that have demonstrated experience in handling the data safely, storing the data, analyzing the data, and having the expertise both in data safety and data analysis, so both the infrastructure

Dr. Mikhail Kosiborod

expertise as well as methodological expertise, should these centers have blanket approval for studies that involve the identified patient data instead of having to go to the IRB for every time that you have a research question, which may delay things substantially in some institutions more than others, and may, as a result, add to the speed with which these innovations can be implemented in clinical care, but also adds to the cost of doing research.

Dr. Mikhail Kosiborod

Again, just some thoughts and potentially provocative suggestions here from the perspective of somebody who does this type of research frequently, as well as other types of research for which I think there should be an IRB review, such as comparative effectiveness clinical trials that randomize patients to novel therapies for which efficacy and safety has not been fully established, for which there absolutely should be full informed patient consent. and IRB review. So I'll stop here and pass this on to John.

Dr. John Lantos

I have no slides, so you can turn those off if you want. At the risk of waking people up, turn the lights up. So I'm going to talk a little bit about the common rule and the current system of regulation and then

Dr. John Lantos

really focus sort of in the second half on one particular aspect of the critiques of proposals to change it. And that aspect is the idea that one of the problems with research is the conflicted loyalties of the physician investigator as opposed to the pure-hearted clinician. But I'll get to that. So the Common Rule, as probably everybody knows, is an extensive set of federal regulations called the Common Rule because it's binding on 17 different federal agencies that originated in the Belmont Report in 1979, which itself was a response to a series of research scandals like Tuskegee and Willowbrook. all those things that are often mentioned in talks about research ethics the way Harry Potter says Voldemort's name. I mean, they just stand for evil incarnate.

Dr. John Lantos

The system that the Belmont Report describes is idealistic, built on the three principles of respect for persons, beneficence, and justice. But the common rule has grown into something of an administrative monstrosity. Over the years, these three simple principles have given rise to an elaborate set of research regulations that you really need an advanced degree to understand and govern how the principles are applied to situations as variable as emergency research, where it's unfeasible without a waiver of consent, research on pregnant women, children, fetuses, neonates, people in treatment for alcohol and drug abuse, military personnel, prisoners, and other vulnerable groups. The Belmont report itself was less than 5,000 words. The common rule is now up to 11,731, with 54 sections and a 3,000 word technical amendment section. The overall goal, I think, was to empower local institutional IRBs. But in the end, the IRBs have become, I think, passive. responders to this or enforcers of this complex set of regulations. And the enforcement procedures themselves are a little bizarre. Researchers need prior approval before conducting any research, even the most minimal risk research. And even research that's exempt from IRB review must be deemed exempt from IRB review by an IRB.

Dr. John Lantos

But the IRBs themselves have very little actual autonomy. They're subject to random audits by the Office for Human Research Protection in which the criteria by which they are being audited are not usually specified in advance. And so in a Kafkaesque situation, IRBs are found guilty of transgressions which had not been predefined. And they lead to draconian collective punishments of institutions so that when IRBs are found guilty of these transgressions. The entire institution's research enterprise is shut down until corrective action is taken. Now, Ruth Faden and Tom Beecham and the group at Hopkins, as Sarah said, have called for a fundamental change in the way we think about these. And the Hastings Center special report on this is a remarkable document that has some One of the most insightful, I think, was Selby and Krumholz, two researchers at Yale, I think, who differentiate the kinds of studies that Michael was talking about, the retrospective observational studies, which they say should simply be exempt, prospective observational studies using big data, which they also think probably in most cases should be exempt although there's a variation of the prospective observational study which includes interventions either individualized to particular patients or block randomized to different centers and that gets a little more complicated. And then this third category of research which is randomized treatment assignment to particular interventions and there again There's a distinction between so-called comparative effectiveness research, where the two interventions to which people are being randomized are in current common use or randomized to a standard intervention compared to an investigational intervention. But in my opinion, even this third category of randomization, particularly comparative effectiveness research,

Dr. John Lantos

is itself generally low risk, maybe even minimal risk, because it has to be approved by an IRB, which uses the standards of either minimal risk or a balance of risks versus benefits before the study can be carried out. So today it's most likely true to say that the safest place a patient can be in an academic medical center is in a research study. The riskiest place would be in the same medical center outside of a research study. Why? Because outside of the protocol, most patients get most of the same treatments they would get if they were in the study, but none of the safeguards of being in a study. No detailed consent form, no carefully designed protocols, no systematic monitoring for adverse events, no safety monitoring boards that determine whether one treatment is better than the other. Yet there's a strange almost perverse romanticism that leads many people to think that clinical care is safer than research and that research is risky. And when push comes to shove and when you actually present data as in the recent support study of oxygen therapy for newborns showing that in fact the babies who were in the study had much better outcomes than babies outside the study, The fallback, the trump card that people often use is we don't care about your stinking outcomes. They're at higher risk because the clinician investigator is not primarily loyal to the patient, but is instead primarily loyal to the research. And therefore, the patient has no advocate. This was the view of the Office for Human Research Protection. in regard to the support study, and they wrote, quote, ultimately the issues in this case come down to a fundamental difference between the obligations of clinicians and those of researchers. Doctors are required, even in the face of uncertainty, to do what they view as being best for their individual patients. Researchers do not have that same obligation, unquote. These concerns were echoed by bioethicist Ruth Macklin, who wrote, quote, It is the doctors, not the researchers, who have a fiduciary obligation and a longstanding ethic to pursue the patient's best interests above all other considerations. I will suggest three problems with this view, and then I'll stop. It's obsolete, it's oddly paternalistic, and it's naive about moral psychology. What makes it obsolete? The view of the dichotomy between the physician clinician and the physician investigator assumes that patients get the best care when their doctor exercises individual clinical judgments to decide what treatment is best for that patient. But that view is completely undermined by remarkably robust studies of idiosyncratic practice variation. This is a line of research that was pioneered by Jack Winberg at Dartmouth but now has been carried out by many other people. that show that clinical practice varies widely between doctors, between hospitals, between regions of the country, and that the variation has no plausible basis in evidence of improved outcomes or even a scientific basis for the choices that the doctors were making. Weinberg concluded that these variations were, quote, natural experiments that ought to be studied in order to determine which treatments led to the best outcomes at the lowest cost. then in 1984, essentially imagining the learning health care system in which traditional categories of research and therapy were obsolete. And many health care systems now take this approach. They mine data for quality assessments, quality improvement initiatives, system design projects, both across systems and within individual centers. But the idea, given that data, that the doctor making an individualized decision is predictably going to do what's better for a patient than a doctor who's following a protocol designed to determine what's best for patients, I think, is simply untenable given what we now know about the way doctors make decisions.

Dr. John Lantos

This view that doctors know what's best is also oddly, given that it's coming from philosophers and bioethicists, oddly paternalistic. The physician who assures the patient that doctor knows best in a situation where there is significant professional uncertainty about what is best is misleading the patient by withholding highly relevant information. Imagine a neonatologist caring for a patient at the time of the support study or a cardiologist making a recommendation for antiarrhythmic medication at the time of the CAST study. In both those situations, a consensus among the experts was that nobody knew what was best, and the only way to find out was to do a prospective randomized trial. In those situations, a doctor who assures her patient that she does in fact know what is best is deceiving either herself, her patient, or both. She might, of course, have good reasons for thinking she knows what is best. Most doctors pride themselves on their ability to figure out what's best for their patient. And as the director of OHRP recently noted in that same special supplement of the Hastings Center report, quote, it is likely a relatively rare study where there are genuinely no good reasons for a patient or a doctor to prefer one treatment over the other, unquote. But respect for patient autonomy and the obligation to inform patients of treatment options would demand that a responsible non-paternalistic practitioner would also inform the patient of the disagreement in the professional community that led to the need for a rigorous trial and of the likelihood that outcomes might be better within the trial. Not to do so will result in two things. clinical trials will continue to be viewed as ethically problematic compared to treatment based on individualized clinical judgment. And second, we'll be left with retrospective data from these natural experiments of idiosyncratic practice variation rather than the better data that we might derive from rigorously designed prospective trials.

Dr. John Lantos

And the final problem with this faith in the individualized clinical judgment of the practitioner as opposed to the researcher is the view that clinician investigators have divided loyalties but clinicians do not. This is quite naive about the demands on clinicians. As David Wendler noted, quote, clinicians have a number of appropriate interests that compete with providing the best care possible, including earning a living, helping other patients, conserving the resources of the institutions where they work, and training new clinicians. All of these might lead to decisions that are not based solely on the fiduciary responsibility to do what is best for patients. On the flip side, clinician investigators may easily hold their conflicting loyalties in check by saying, yes, I want to figure out whether treatment A is better than treatment B, but I will not sacrifice my patient's interests in the course of doing a clinical trial. In fact, it's in my patient's interest to be randomized because that gives my patient the best chance of getting the treatment that will ultimately

Dr. John Lantos

beneficial. It also, of course, benefits future patients. In summary, there may be many reasons to regulate clinical research differently than we regulate clinical care, but I think it's time to discard the old canard that the conflicting loyalties of the clinical investigator is one of them. who seeks to enroll patients in a well-designed clinical trial is often doing what is best for his patients in a manner that is more honest, less paternalistic, and no more conflicted than is the clinician who, in the same circumstances, assures her patient that she knows what is best. To believe otherwise today is simply naive. Thanks.