Hello, and welcome to the SHE research podcast. I'm your host Diego Silva. Before introducing our guests, I want to acknowledge that we're recording on the unceded territory of the Gadigal people of the Eora nation. This is and will continue to be Aboriginal land. I want to pay my respects to those who have and continue to care for country. So today I'm joined by Dr. Gabe Watts to discuss his paper that he's co written with Professor Ainsley Newson. Is there a duty to routinely reinterpret genomic variant classifications? Gabe, welcome.
Thanks, Diego.
Okay, so just to begin, I was wonder if you could provide a quick summary of your paper,
I think in In summary, there are two points. One is that genomic data has what we call diagnostic durability, which is to say that compared to other diagnostic tests, you can get a valid diagnosis out of it for almost indefinitely, there are some, you know, small cases where you might get changes, but more or less, you don't need to retest a patient to still get a valid diagnosis. So compared to say, an x ray, where you take the x ray, and maybe in a few months, you'll need to take another x ray, in order to get a valid diagnosis from that, we basically you can't get a valid diagnosis from an old one, whereas genomic data, you can do it indefinitely, because it doesn't change. So that's the first point. And the second point would be that classifications for genomic variants, which is to say there's a classification system that ranges from pathogenic
likely, likely pathogenic variants of uncertain significance. And then there's likely benign and benign. That's the basic classification system, and classify those classifications for genomic variants often change and in quite short periods of time, so you've got data that you can go back to almost indefinitely. And you've got classifications as to do with pathogenicity of variants that changes. And so that raises ethical questions about should you go back to the data? How often what responsibilities do you have as a service? who administers tests that produce data that we don't know much about? That can change very often, but we can go back again, things like that. That's the that's the core of the issue.
So what are some of those ethical issues that you're referring to? What are the ethical challenges? And I guess also, what are some of the conclusions that you draw?
Well, the key the key challenges that we look at, broadly concern, what would be called reanalysis. So going back to genomic data that you have, and analysing it, again, is reanalysis. And re analysis has a number of different elements. And one of those elements is reinterpretation of the variant classifications. And the ethical issue is, if you will, this is this was brought up in a paper by
Appelbaum et al, which is what we're mainly responding to, is they they claimed that by virtue of providing a test that is quite likely to produce information, which we are not certain about whether it's pathogenic or not now, but likely enough to be so in the future, does providing that test bring a responsibility to actively reanalyze the data. So that's one well actually their point is actively reinterpret the data. And the key point there is that reanalysis is good, it brings new, it has what we call increases diagnostic yield, the more you do it, the more diagnoses you get. But it's not feasible. It's technically very difficult. And we simply don't have the workforce to do it. It would take years and years and years. Without some sort of assistance from some sort of breakthrough from AI it just won't happen.
Reanalyzing genomic data increases diagnostic yield it it brings, the more you do it, the more diagnoses you get. The problem is that it's not feasible. We simply don't have the workforce. But there are many other reasons as well. And then the idea within that is that well reinterpretation you're not looking into everyone's data, you're looking into these classifications, so maybe by just seeing if certain classifications have been changed over time, you'll get some new diagnosis. And it is a fairly, it brings about, I don't I mean, don't quote me, but it's something in the order. It's in the order of 20% between 20 and 30. I forget exactly what and that's significant enough to think hey, maybe even though reanalysis isn't feasible to do regularly, we could reinterpret the classifications, and then find ways to communicate any changes with patients. And this may well be something that we are ethically obligated to do on the basis of providing such a test.
And so in terms of that sort of question, whether there is that obligation, what is your response? Is there an obligation to retest?
Well,
Well, one of the reasons why we like this as an issue is because I think that intuitions can go both ways, relatively easily. We did speak with people, like clinical genomicists to get a picture of things. And one way you could see is that, well, we administer these tests, that one, I remember someone saying that there's no other diagnostic tests that we administer, that we know so little about. And, and yet, we give a test like this. And so insofar as we do that, yes, I feel an obligation to my patients in order to continue to reanalyze or reinterpret the data or do anything, basically, to, eventually to, to give us the best chances of getting a diagnosis, even if that is looking at it every six months. So be it. But then, the other side of the equation is is that well, if we really don't understand this technology, and if the fact that we are providing patients with information that is ultimately uncertain, or not ultimately uncertain, uncertain at the time, but may well be, we may well have further knowledge, diagnostic knowledge about it in the future. Well, if this is such a problem, maybe we shouldn't be giving them the test. Now, I don't think you've got to find basically a middle ground between those two things, because doctors will say no, no, this is the best test we have. We have to give it to them and say yes. But does that imply that you have to therefore keeping keep on reanalyzing, or reinterpreting the variant classifications in order to get a diagnosis or give yourself the best chances? Our answer is ultimately, a mixed one. But no, there's no general obligation and we analyse three, well, we analyse three sources, which may give us an ethical obligation we. And so one of them is an ongoing duty of care, pay doctors or medical professionals have ongoing duty of care to their patients. Another one is what we call systemic error risk. And the last one is diagnostic equity. So we look at this situation through these three frames. And we argue that insofar as ongoing duties of care are concerned, no, there's no general requirement that you continuously or regularly or actively and regularly reinterpret genomic varying classifications. However, in certain cases, there are.
But it had been argued by others that there was this general duty. So essentially, we're saying no, it's not quite like that there are specific cases. Otherwise, there are, there are times when simply the kind of massive distributed effort that is required for classifying genomic variants creates risks of error, which ought to be mitigated against through regulatory interpretation. That's a time when you get it, when when you should, or where we saw an ethical obligation arise. And then the last one was where, okay, so these tests will, are likely enough to produce uncertain information simply because we don't know they are so comprehensive, and we don't know that much, comparatively or relatively about the human genome at the moment.
So that's an intrinsic reason why it would produce uncertain information, but there are for certain communities, there are extrinsic factors like historical under-representation, and which can occur for very many various reasons. But in that in those cases, we suggest that the diagnostic durability the fact that you can keep on going back to data that it has the virtue of it is that you can keep on reanalyzing small data sets as they get added to. Basically, you can mitigate the negative impact of extrinsic limitations on your data set through regularly reinterpreting the data because of its diagnostic durability, then we also see a moral obligation arise to do so.
So yeah, looks like this is a really interesting topic, I think in part because it's not something that is sort of front and centre I think for many people thinking about sort of bioethics in general, like, like you said, there's there's been some stuff written about this. This is sort of isn't the sort of the regular stuff that we usually think about so I'm wondering how did this paper come about what was some of the motivation behind this paper, and I guess the broader project, perhaps?
I think it is something that clinical geneticists think about quite regularly. They're seeing patients all the time, they're quite aware of the limitations of the tools that they're using. And they want to help their patients as best they can. And so to them, and I think to, to us as well, it's not, it's not like we're separating off deliberately, but you think well, of all the options that you have reinterpreting the varying classifications with a, you know, actively at periods of regularity, that could seem to be a good thing.
Where bio ethicist come in is they interrogate the language of duties in such situations and say, I think, as I said, clinical geneticists do think about this a lot. But you can find a shift to a language of duty and obligation that goes very fast. And I think the way Professor Newson and I saw it was it was the job of bioethicists to come in and provide some conceptual analysis of the duties that arise by breaking up all the possible situations, and then suggesting some of the grounds that you might have for doing this, and then providing reasons for and against. Ultimately, what motivates it is the fact that genomics will become mainstream. And so we need answers to questions about how to handle the voluminous amounts of data that we have, within the practical limits that are imposed. And so that's, that's the motivation for a project like this.
You mentioned the motivation, I guess from, from your perspective, from Ainsley's perspective, in terms of, you know, geneticists are using this information, there's a quick sort of move towards speaking about I have this obligation I have this duty, this is what you guys do is sort of bread and butters to think about exactly these issues. I'm wondering, from your perspective, does it matter whether clinicians or patients believe that this sort of genetic tests in this next generation sequencing will be retested in the future in terms of establishing a general duty to retest in the future? So I guess I'm, I'm wondering to what extent does their intention and motivation factor in your thinking?
I think one issue there that I would think about is whether the fact that it's inspected, expected by patients and that practitioners believe they ought to do it creates a duty of care? I think that one of the contributions of bioethics to areas like this is to be precise about what obligations can be rationally defended, despite what people want. Another way to address a question like this is to say, well, the system as we have it isn't, insofar as problems like this even arise, something's going wrong. And what we need to do is we need to work from a system level down to ensure that we are doing everything in our power to make sure that the pipeline's through which we provide people with access to junk genomic diagnostic tools, and genomic diagnosis.
ethically sound, and one of the components of them being ethically sound will be that we have procedures in place in order to reinterpret, I think that a lot of clinical genomics practitioners would support that sort of view. And I think it is, ultimately, how it's my belief that it's how it ought to go. I think in terms of what we've argued, we've argued for something we've argued from within the position of how things are within the position of how things are some of the duties that are being appealed to, arguably overstated. But in terms of providing the best care, then, I think arguments for broad change, but legitimate, but I think that ultimately it's gonna come down to I'll put it this way, in terms of patient desires for kind of continuing reanalysis to the extent that they're, they're there, that's something you should critically interrogate as well, and practitioner beliefs that they are obligated to do so, feasibility is the issue, you know, until something radically changes about whether it becomes feasible or not to do patient desires and practitioners felt obligations. The rubber has to hit the road somewhere and the road is how much can we possibly do? And so I think that those considerations are important. One of the reasons why reinterpretation of variants rather than reanalysis on the whole becomes the issue is because it seemed to be more feasible. But even within that, we argue that, look, it's not as general as you think. So I guess maybe the actual conclusion is, regardless of what patients want, and regardless of what the felt obligations are, there are grounds for thinking that the kind of ethical duties and the ethical duties in the situation only extend so far.
So then, am I right in thinking that if the issue of feasibility is resolved, so, you know, technology, computing technology, X, Y, Zed comes up in 10 years time? Then is there an obligation to just say, yep, well, now we have the capability, you know, quantum computing, I don't know, I'm saying things, I don't actually know what the the terms they mean. But you know, they people say quantum computing is powerful. So we'll go with that, you know, at some point in the future, that's ubiquitous, we can then reinterpret everything, therefore we ought to reinterpret everything are there is a duty to reinterpret everything, is that still an argument too far?
I mean, one thing I'm wary of is that it's always throw it as well, without, without some sort of breakthrough in the technologies that we use in diagnostics, then we're going to run up against feasibility constraints. And within those scalability constraints, we argue for X, Y, Zed. But say there were different well, then that's something to think about in the future.
But I mean, if you were to kind of posit a counterfactual situation in which you, and we'd have to specify more precisely that I'm about to now but in which you could rely on machines to do a lot of the work and the work load became very, very low, then it is entirely one of the benefits of genomic data that you can just keep on going back to it like it's amazing, right? It's a, it's, you should still question the framework that looks at it as a resource to be mined, right? But you can keep on going back to it. And if you on the supposition that that is, doesn't cost very much and doesn't use up significant resources, then why not? It's one of the questions, then there are still questions about recontacting patients informed consent, which we set to the side in this paper, but they absolutely come up. And then there are issues around you know, for practice, say, say you do have a diagnosis, that's you've you've you've had a diagnosis, and then the classification changes and it gets, it's like, oh, no, we no longer think that's pathogenic, we're actually not really sure what it means that's quite distressing for a patient that's, you know, has gone through that change. And it's distressing for the clinicians who have to, or the whichever medical professional is in charge of recontacting. That person. And if you were routinely reinterpreting your variance, you'd get cases like that would skyrocket. And so there are other ethical concerns that you would have to think of down the track. But if the if the question is kept to well, you can use machines in order to do this stuff regularly, then yes, it should. And maybe that is part of the the situation where we getting the pipeline's as good as possible is the way that we ought to go. But even if that were the case, the ethical issues will just pop up somewhere else. There's, you know, in terms of the distress, there will be distress somewhere that you've got to think about.
Right. So I think what's interesting about that is I guess the the idea that the paper is currently constituted has a lot to do with feasibility and sort of what do you do in the context of feasibility and scarcity. But what I think is interesting is even if you resolve the question of technological feasibility, obviously, you're going to have all these other sort of other ethical questions come up or sort of come to the fore at the very least then?
I think that's absolutely correct. And as we said at the beginning, one of the reasons for looking at this issue is that the fact that genomic data has diagnostic durability that you can get a valid diagnosis out of it almost in perpetuity, even beyond the life of the person who you got, if I can help their relatives and have all sorts of other people. It's phenomenal. Like it's an amazing property that is likely that brings up novel ethical issues. And it is, you know, like the metaphorical Whack a Mole you can knock one down, but another one will come up because it is genuinely novel. And that is that is a reason for bioethicists to be there. Right. It's, it's interesting in that regard, just of itself, and also for other reasons.
So Gabe I'm just wondering what you're working on right now. What can we expect to read of yours in the near future?
I'm Ainsley and I have been looking at the notion of personal utility within the tech the assessment of healthcare technologies, but mainly for genomic diagnostics again. And it's very similar to the the basic pattern described where certain issues arise in clinical practice for genomicists and then certain ethically loaded language gets used and extended very quickly. And the way that I mean, I presume to speak for Ainsley on this point, but I'll speak just for myself, the way I see the work that we do is just plain old, conceptual analysis Be clear about concepts in play. And then normative analysis. This, I've just feel that this is a situation where, I think it's boring, but I also think it's necessary. I mean, I say boring in the sense that I still like doing it. You know what I mean? I think I think
I'll put it this way, I think it's necessary work in the sense of it's, it's an area where language and norms can become muddled very quickly. And there is value in
pacifying and clarifying
pacicifying and concepts and clarifying demands. And so that's what we try and do with various areas of like, clinical genomics, in particular, genomic diagnostics, because these things not only affect a lot of people's lives, but have the potential to and clarity around concepts and normative demands is highly useful for policymakers, but then we will rephrase from before highly tedious to produce but if you have, you know, a disposition towards tedium and a head for it then that's fine.
I like the way you put that. The other thing it reminds me of, you know, it's uh, you know, it's good for policymakers that we do this work I think part of the trick is trying to convince them that they need this work done in the first place. But like, I want to thank you again for taking the time to be with us today. I want to thank you for listening to this episode of The SHE research podcast. You can find the paper we discuss link to the episodes notes along with the transcript of our conversation. SHE pod is produced by SHE network and edited by Regina Botros. You can find our other podcast episodes on Spotify, radio, public anchor, wherever you get your podcasts of quality. Thanks again for listening. Until next time, goodbye.
