Discussion on Session 3: Clinical equipose, and harm/benefits
Catarina Kiefe [discussant]: I'm delighted and honored to be here. My first slide is sort of a disclaimer slide, which is the purpose of this brief session is to stimulate discussion, so I'm trying to be provocative and I also need to make the disclaimer that I am not now or have ever been an ethicist. I do cluster randomized trials for a living and have done so for a while, so that’s my interest in the area. So I will talk about the topics that were introduced so well by Charles. Namely equipoise, risk benefit assessment and then hopefully raise a bunch of questions and vibrant discussion.

So, first, I would like to say that, get ahead of myself and say, that I completely agree with Charles' conclusion that equipoise is applicable to cluster randomized trials. I do however, disagree with framing it and justifying it in the way that he does and so, I just reworded the definition of equipoise because it suites my purpose and I'm saying it in the affirmative rather than in the negative. Trial is unethical if one arm is known to produce better outcomes than another but I think we'd all agree on that. And the argument that I do not completely agree with that Charles made is that in order to evoke equipoise, we need to evoke the fiduciary relationship between the physician researcher and the state and the research subject. I apologize for subject but at least it's between quotes. So, and I would like, the argument that he's made very well is that there are two reasons to evoke equipoise in the RCT contest, the fiduciary relationship in the asymmetry of power, which is mitigated by trust. And in the case of the RCT, its patients and physicians that being quoted and in the case of RCTs, it's citizens and state. But I do have a bunch of questions that came to my mind when I tried to understand these concepts. Of course, the first one I've already stated, do we really need to anticipate to involve physician/patient relationships to discuss equipoise and RCTs. I'll come back to that. This may sound strange coming from someone south of the Border at this time, but I am not one of those American citizens who thinks the state is the enemy and therefore, I really think that the state should ensure that all research involving humans, individuals as well as groups of humans, should be ethical. I don't think we need to evoke trust. I think that's just a fundamental principle of democracy.

And then the third question is, given those two questions, which I've already told you how I would answer them, I don't see a reason to say that equipoise would not apply to cluster randomized trials. So, the, what the previous questions were really based by my instinctive review or reaction to asking is the physician/patient relationship the right model to discuss a randomized controlled trial? RCT, let alone CRT, to all investigators’ who conduct RCT's are physicians, or even clinicians. And not all research subjects are patients, yet we do lots of RCT's, we don't need to evoke that. The informed consent aspect is a guiding principle for all human experimentation in my mind and I think in the sense of the regulatory agencies and the asymmetry of power in my mind, is actually higher in the clinical encounter than in an RCT. But again, this is an opinion. However, as many of things in ethics, we evoke principles and we make decisions based on opinions. So, what I would like to propose is that RCT's and CRT's are actually ethically close knit to one another than either is to the clinical encounter and the table below focuses on uncertainly which is, in the case of the clinical encounter, controlled and by the physician’s best judgment but in the case of trials, it is imposed, the uncertainty is imposed. Chance is a force in the trial, whatever they are. Informed consent is only implicit in the physician/patient encounter, it is implicit when we decide we need it in both of the trial situation and equipoise is only involved implicitly in the clinical encounter yet explicitly in the trial situation so for these reasons, I would make the assertion at the top of the slide. So I would propose an alternative framework that states that equipoise is imperative for all experimentation effecting human subjects, individuals or groups. Equipoise should be present at both individual and group’s level and indeed, that is what Charles applied to the bed net study and many other examples.

So, now I'm going to transition to the second part of the conversation which is the harm/benefit analysis. Equipoise, as we know, focuses on lack of non-superiority for the main question posed by a study, but the superiority is not along a single scale, it's not one-dimensional. We need to understand and hopefully, although frequently, we cannot, quantify the possible harms and possible benefits for everything we do and traditionally in the RCT, we do that for each individually study, i.e., we use the individual perspective. I ask, should we also use the perspective of the group, of each of the groups that is randomized or should we also use the perspective of society at large. So these are questions that I just throw out for discussion. I do think that the component analysis that was proposed is a very useful framework in attempting to distinguish between therapeutic and non therapeutic and as Charles said, sometimes it's not so clear which one, what is what. And also, the other thing that we need to keep in mind is that most studies, although they test an underlying question or several, but they involve many different procedures, so, I was trying to understand the reach of this component analysis approach and I tried to understand things by making tables, as you may have noticed, so I tried to understand what exactly was being proposed and it was very clear to me that, in terms of the intervention group of say, two armed trial, the therapeutic procedure should be evaluated by whether their potential benefit of it outweighs their harm. It was also clear to me that in the control group, the non therapeutic procedure should be evaluated as to whether the harm has been minimized by individual benefit was not on the table. What was not so clear to me was how we were going to handle non therapeutic procedures for the intervention group or more importantly, the therapeutic procedures of the controlled group, for which there are some. So those are additional questions to think about.

And, so, I would like to add a few more questions about this concept of component analysis and how it’s applied. Minimizing harm is a very tricky term, in my opinion. It's in the eye of the beholder. And whether we use component analysis or not, I bet different ethics boards will come up with different answers to what minimizes harm. Implicit in what Charles did, actually, quite explicitly, when we were talking about non therapeutic procedures minimizing harm, there was also the issue of Andrew's resulted in a better social good so it changed, it shifted the perspective from the individual to the social group, good. Which in my mind, correct, that should be done but it's not clearly articulated yet, I would propose to make that clearer. And then another question I would like to ask is can the rest of the non therapeutic procedure be balanced at the individual level, but the future benefit to that individual was undergoing the procedure. I really don't know.

And finally, is the cost of the procedure part of the harm benefit discussion. This issue was raised this morning, I think Kathy raised it and several other people came back to it and I think it's something we have to deal with. So, this is my final content slide and so, I would just like to add a few more questions to the pot. I do a lot of implementation trials so that is near and dear to my heart. So one of the questions I would like to raise and I would like people to pay attention to is in the individual randomized controlled trials, there's a lot of, it's all vogue right now and justifiable so, to talk about non-fiduciary trials. I think we need to do that as well in cluster randomized trials but I haven't seen it happen. Maybe somebody here has done one but I haven't seen it. The other question is, the statistic inefficiency that was eluded to this morning, was clearly there and is well justified, is it an ethical issue? Do we need to talk about the cost of the statistical inefficiency? And finally, there was a question raised this morning, I made these slides yesterday, so, there, I do think that monitoring boards have an important role to play in cluster randomized trials. I've been a member of a couple of these and I do think we need to pay a lot of attention to stopping roles, which as was said this morning, have not been developed yet for the cluster randomized trials. And that’s the end of my slides.

Charles Weijer: That was incredibly helpful, thank you. Tremendous set of questions to motivate discussion. I only want to really make two comments and then see which of the other questions you've raised, people in the audience want to pick up on. So the first point related to grounding clinical equipoise and so I think we're in broad agreement that clinical equipoise is a useful and important concept that has broad application, including application to cluster randomized trials. We're definitely in agreement there. I was just trying to tell a story about where clinical equipoise comes from and emphasize their work in different ways. There is broadly two ways that you can talk about moral obligations. One is in terms of free floating moral obligations. There's a general moral obligation to ask and everybody has it sort of right, and there are thousands of these things floating out there and it can be really unclear who exactly has a duty to do what, when. I'm interested, as a result of that, of talking about moral obligations as emerging from relationships and that has an advantage in terms of the fact that when you talk about obligations actually emerging from defined relationships, it's clear who has the duty and to whom they have that duty and it becomes suddenly something that's enforceable. You know who has the duty and you know who has to do what when. So that's my interest in talking about relationships, whether they are between the physician researcher and patient subject or between the state and the research subject. It seemed to me the state and research subject did the work that you wanted it to do of demonstrating the broad applicability of clinical equipoise. You could, for sure, tell a different type of story that strong political institutions within a liberal democracy need to have a mechanism like this in order to maintain the satisfaction of citizens - to quote Angie White. And we'll hear more about that, so there's a different kind of story you can tell. Anyway, so I think we agree, I think there's interesting stuff there about the foundations though.

As my second comment, which will be briefer, just to talk about this table, I think actually, the work of the moral rules actually runs across on both the intervention and control groups. So the clinical equipoise question is not merely a question about what we do to the intervention group. It's a question really that says, is the comparison of bed nets to a no treatment control, a comparison that's consistent with clinical equipoise. So it, the reach of that moral concept, is across both groups. Similarly, when we're asking about non-therapeutic procedures, the risks of those procedures need to be minimized both for participants in the intervention group and those for within the control group. The real risk with not separating out therapeutic and non-therapeutic procedures is that you can mix in these therapeutic risks with the therapeutic benefits that may come from treatments and as a result, you may justify a lot of risk which really doesn't hold out the prospect for direct benefit and I think that's the justification that we don't want. Anyway, enough from me. Thank you, that was great.

Question/Comment from the audience:
I'm Sarah Edwards London, UK. A quick point of clarification. I'm assuming that you were using the idea of equipoise to justify randomized cluster trial, yeah? So in the bed netting case, would you be trying to argue that a prospective randomized cluster trial would be justified despite previous evidence from elsewhere or would you be simply saying there needs to be some test of generalized ability and if so, could that be done through these sorts of design? Charles Weijer: Yeah, I was trying to demonstrate how one might apply clinical equipoise looking at that trial prospectively. Could you use equipoise to examine whether that comparison of bed nets to no intervention was an ethically permissible thing to do in that context and I think making that case, whether one could convince a research ethics committee, would appeal to things like the broad evidence base but also, biological and sociocultural differences in the Cambodian setting that perhaps haven't been adequately addressed by the evidence base. I think clinical equipoise actually situates a variety of designs; it's not merely a concept for randomized controlled trials of various sorts. So a variety of designs might be consistent with the equipoise requirement.

Question/Comment from the audience:
Merrick Zwarenstein: from the Institute for Clinical Evaluative Sciences in Toronto and about clinical equipoise, I like the concept of therapeutic and non-therapeutic, but I wonder if other terms might not be more widely useable. I think non-therapeutic might be easily replaced by scientific purposes or evaluation purposes and I think that would be an assistance because firstly, it's not negative and secondly, it enables the concept which I think the discussion raised about the benefits of evaluation measurements on patients, for example, in the bed netting study, the system tested controlled patients for whether or not they have malaria, and treated them for malaria that was so detected, which would be a health benefit of being in the controlled group due entirely for the research procedures and for the therapeutic procedures, I think that therapy has such medicalized connotations that we should probably replace it with the word intervention or some other such generalized word that would also assist with distinguishing between the clinical intervention which is often the subject of knowledge translation trials and the knowledge translation intervention whose aim is to try and implement the clinical intervention, the two level activity which I struggled with when I first met the collaboration in Cochrane. Charles Weijer: We've been struggling with those terms and that's very helpful contribution, thanks. Catarina Kiefe: I like it and perhaps experimental and evaluative would be two words to use.

Question/Comment from the webcast: So we have a question from Laura-Lee Balkwillat the Secretary for Research Ethics here in Ottawa and the question is, "By taking clinical equipoise out of the doctor/patient relationship to make it more applicable to other types of trials, is it conceivable that other clinical equipoise would become a criteria for non health related studies? And several speakers referred to cluster trials and education settings, how far would proposed cluster trials guideline reach beyond typical health care settings and would any study involving randomization of participations at the individual or community group level be considered a trial to which clinical equipoise might apply?" Jeremy Grimshaw: I think that's a very valued point. I think when we started the project, and we did our key informative interviews, we did include various people, particularly doing cluster trials in the social spectra for this and it's clear that they face similar kind of issues in those sorts of sectors. I think when we went on to do the work we've done, we largely restricted ourselves to health, largely just for making it tractable at this point in time, but I've already had discussions with several people in the room that as we work through the health issues, we really didn't seem to think about the application of this to other settings, social sector, school based research, or international development work. So I think there's been some value in keeping this thing fenced in for now but I think a lot of the issues we raise are probably likely applicable with some modification to a whole range of sectors. Allan Donner: For there, for example, there are trials that have randomized litters of mice to different dietary interventions and technically, that's a cluster randomization trial as well so one could expand the use of this design and I think almost infinitely.

Question/Comment from the audience: David Forster, Olympia, Washington, USA. That last question in your comment almost stole my thunder but I think I'm going to phrase it differently enough that it’s still relevant. I think the tools of clinical equipoise and component analysis are great and I'll consider you the tool maker, especially you Charles. So I'm someone who’s been out using these tools for 10 years and my findings over the years is that clinical equipoise is a wonderful tool where there's a therapeutic intervention. Whether it's medical, psychological, education, but the minute you get away from research involving a therapeutic intervention, it’s not very applicable. It's not easy to use and I quite frankly don't use it. And so, I think the applicability of component analysis to these cluster trials has to do with whether there's a therapeutic component or not and if there's not, it doesn’t apply. Now contrary to that, component analysis is a much more useful tool for a broader setting and so the component analysis works. And so I think that's the way we should be thinking about application of the cluster trials.

Question/Comment from the audience: Nicholas Birkett: Another comment about equipoise. I'm assuming you don't mind if we're a bit controversial from the floor as well. I guess the question; I understand conceptually the role of clinical equipoise and clinical trial ethics. Where I get into problems is that I'm not convinced that it ever actually applies and so the question, I guess I'm going to ask, the question would be who is the judge of whether we have clinical equipoise and should the opinions of the research subjects be included in the situation, the question of whether there is clinical equipoise. I think it would be very hard, I'd be very hard pressed to come up with any situation where somebody has succeeded in getting millions of dollars to do even a non clustered randomized trial where you could honestly, with an objective non biased committee, come in and say my prior probability of a success is 50/50. I just don't think so. I think the evidence is always going to be there. We can always invent a story in the way you did; I don't mean that in a negative way cause they were always inventing stories that will pull out some part that will justify why we need to do yet another study. But at some point, that becomes unnecessary and wasteful so I guess my real question is who is going to be making that decision, which should make that decision?
Charles Weijer: Great, so there's a really interesting literature on equipoise and what on earth it means and who has to be in equipoise and it’s become a bit of a cottage industry I think, so there's patient equipoise, there's doctor equipoise, there's researcher equipoise, there's community equipoise, there's industry equipoise, not sure what that's supposed to mean, so who is supposed to be in this sort of ill defined state of equipoise. I think there are broadly 2 approaches to thinking about equipoise and one is sort of thinking about it as a state of uncertainty that a patient experiences and if you understand it that way, so a lot of people interpret equipoise in basic terms and there's a nice sort of connection there. But I think the difficulty with that approach is that it just doesn't seem to be psychologically true that investigators doing seemingly ethical studies are in fact in equipoise in that way. People doing studies reasonably enough often really believe in the experimental intervention. I mean, its lots of work doing a study. I mean, of course they do, right? So, that sort of seams like, for me, that makes that interpretation of equipoise, that way of thinking about it, not as useful. I like to think about it really as a property of the evidence that exists at the beginning of a trial and it might be something that might emerge out of a systematic review that is motivating or part of the preparatory work for a cluster randomized trial that is part of building the justification so it's a property of the state of evidence that really just says something like we don't already know the answer to this question, right? Because if we already know the answer to this question, it's unethical to randomize people in the study, so I think that's what it's getting at and in this context, the people who make that judgment are the research ethic committees.
Nicholas Birkett: In the bed net example that you've used, if you get informed consent from every participant in that study, you would have to provide them with the informed consent with the existing evidence of the effectiveness of bed nets. Under that situation, I would be surprised if there would be many or any people who would agree to take part in a study where they wouldn't get a bed net, I certainly wouldn't. And the fact that mosquitoes eat at a different time, to me would be relatively minor and would look at it being done a different way, so again, the question is, who’s judging that equipoise because the participants should have a right to make that decision through informed consent and yet, the scientific review committee or the ethics boards are making it divorced from their context of those individuals.
Charles Weijer: So, which is just another way of making the point and I think correctly, that equipoise is a necessary condition but not a sufficient one, right? That there are for important reasons, requirements of informed consent and pragmatic, if people are just going to vote with their feet, for pragmatic reasons, you may need to rethink the design. Catarina Kiefe: And I would agree to what Charles said and I would add that to me, equipoise is a framework to think about things and in most real circumstances; we can't make a categorical decision about whether or not there is equipoise. There are instances when there has been a Cochrane review that says we cannot decide between that and the other, then the evidence is pretty strong that there is equipoise but most of the times, that is not and that's why we have ethics boards that need to help us with it.
Allan Donner: I just wanted to respond to something Charles said about ... Charles mentioned that the investigator is rarely in a state of true equipoise in a sense that the investigator is naturally enthusiastic about the intervention that they work perhaps for years on. I once posed that question to a prominent investigator and what he told me was interesting. He said he distinguishes between emotional equipoise and intellectual equipoise. Emotionally, he definitely wants the intervention to work but intellectually, he feels he has to stand away and let the scientific process take course. Anyway, I thought that was interesting.

Jeremy Grimshaw: I think the bed net example also heightens Catarina's point about these issues not necessary always about the individual benefits or harms at the individual levels. The key issue in that example would be whether health care systems or donors should actually invest in insect nets and so as an individual, if someone was going to give me a free bed net, I would say great. But actually, the issue is the opportunity costs that if we give out bed nets free, which are going to be ineffective, the opportunity costs in those countries and to those donors are great and also brings in this idea about from whose perspective, so the health policy cues 6 or 7 is really quite important.

Question/Comment from the audience: I'm Abha Saxena from the World Health Organization from Geneva. One is very minor given all the discussion that has happened. The paper and the presentation was focusing on clinical equipoise and it's not very clinical and especially in the context of cluster randomized trials, most of the times its not clinical, it could be social behavior, it could be preventative, diagnostic, so to me, just to say equipoise, rather than clinical equipoise makes more scientific sense. The second question is more related to the cluster randomized trials and the control group and the exemption being that having a control which gets nothing is ok because we are providing, we don't whether, in that context, in the context in whether cluster randomized trial is being done, the standard of care is our control. But if you go to some of the guidelines, which are saying the standard of care should be the best proven therapist or preventative standard and how do you ... should you do, could you do no action at the control level, would that be ethical justifiable. It's difficult cause I think most CRT's are actually not using control as the standard of care but I sometimes feel troubled because I think randomized controlled trials, we have a different approach versus cluster randomized trials.
Charles Weijer: I may say something, that's an enormously difficult issue. There are special and pressing concerns that emerge out of globally inequities and the distribution of health and access to health care, obviously and so, the question is what, we say that the various treatment arms in a trial need to be broadly consistent with component care but component care where exactly, right? Is it the care that's available in a developing country or is it a developed country standard of care, that's a really difficult question. I think part of the concern there is one that individuals in developed countries might be particularly susceptible to exploitation in health research and that there ought to be procedures in place to guard against that possibility. I think there are really important protections that need to be a part of guidelines that are addressing country context, including a requirement that health research ought to address local health needs and priorities and furthermore, that there should be some reasonable prospect of access to interventions once the study is complete. There at least needs to be some kind of plan about how in fact people in that country might benefit from a treatment, an intervention, sorry. So I think that's just mandatory and ethical clear requirements. What you need beyond that in a developing country setting is contentious I think and there's a debate as you well know about the requirements of ancillary care in developing country context and I'll just flag that as a really hard question, so thanks.

Question/Comment from the audience: Hi, I'm Alison Paprica from the Ontario Ministry of Health and Long-term Care and I have a comment that I think builds on the question 11 and 12 that have to do with societal benefit. Just sort of put in there how I see these questions come up in the world I work in. So where we are, we talk a lot of policy interventions, complex strategies at the ministry level and basically, the benefits and harms are discussed and agreed on by elected officials or through governments, they trickle down to regional authorities hospitals, others who have authority to make these decisions. So, really, when it comes to the question of are we going to do something like a cluster randomized trial or coupled research with the intervention, a lot of the benefits and harms are already set in terms of what the intervention is. And the discussion is then about what are the benefits and harms of the evaluation aspect. So when I think about your two arms there, I would say that really I see us focusing on the side that doesn't have to do so much with ethnicity, people have already weighed the benefits and harms but there is this question about what are we going to learn generally seeking a societal benefit in terms of new knowledge and what will is take us to get there in terms of burden and cost.