iwAL 2025 | Debate: the value and utility of MRD in the management of adults with newly-diagnosed AML

Mark Levis:
I’m Mark Levis and I’m here at the seventh annual iwAL meeting in Washington DC and I’m joined by my colleagues Eunice Wang from Roswell Park and Sanam Loghavi from MD Anderson and we are continuing our very civil debate on the role of MRD in the management of AML. And I can start by saying the debate that we had here at iwAL was passionate. I think we can agree. Everybody is interested in MRD and no one quite knows its role in just yet, but we all sense that this is something that we want to use. So I’ll start off with Eunice here, summarize our debate as best you can.

Eunice Wang:
So one of the debate questions during this iwAL meeting is, is MRD a universally accepted standard of care for treatment of younger fit patients with AML receiving intensive induction chemotherapy? And it’s a fairly specific question, but essentially, it’s saying, do we now have a universal standard approach and that incorporates MRD into our treatment algorithms that affects how we manage these patients? And so in addition to my colleague, Justin Watts, we had the side of no. And the rationale was that, for example, for a 66-year-old gentleman newly diagnosed with a secondary AML defined by U2AF mutation, how would this MRD assay affect the treatment of that individual? So in my case study, I pointed out that this gentleman has a mutation, that there is no current MRD molecular assay. So we fall back upon the other standard approach for MRD, which is multi-parameter flow cytometry. Now, as you know, this is very…

Mark Levis: Which itself has standardization problems. And so, Sanam, I think everybody agrees we all…I don’t think anyone in that room thought that it was standardized. Agree. And yet, we all want it. So as the resident flow cytometrist here, you stood up and valiantly defended the only modality we have in many cases.

Sanam Loghavi:
Right. I mean, I think it’s important to start by acknowledging the limitations of, you know, what we do. I understand that flow cytometry is not perfect, but I do think that we have to first acknowledge that AML is not one disease. So it’s, I think it’s, you know, unreasonable to expect that we’re going to have one assay to monitor something that’s not one disease, right? It’s molecularly heterogeneous, cytogenetically heterogeneous. And so we’re never going to have a single, you know, it’s never going to be B-ALL.

Mark Levis:
And in fact, Andrew Wade brought up, I think, synthesized everything.

Sanam Loghavi:
Exactly. And I think, you know, if you remember, the last slide that I showed was a case where you had so-called molecular evidence of persistent disease, which was actually maturation, right? So I think it’s important to recognize that we have different tools and modalities at our disposal, and it’s good to use everything that you have. But at the end of the day, as great as FLT3 and NPM1 are, molecular markers of MRD, they’re only applicable to a subset of AML. You can’t apply them to all AML. And so flow cytometry at this time, is the more widely applicable test that we have. And I agree with Eunice. Not every center does, you know, AML MRD. It’s certainly difficult to do. It’s subject to interpretation. It’s, you know, it’s difficult to set up a flow cytometry lab, even if you have the best pathologist interpreting the data. So I recognize all of these limitations, but I think we don’t want perfect to be the enemy of the good, right?

Mark Levis:
And I think asking too much of any single modality is fairly ridiculous. So Andrew brought up, I think, a great point. Really, well, if you detected a low level of some mutation, how are you going to interpret this without some flow cytometry? You’re not asking the flow to pick out one cell in a million. You’re asking it to complement another assay. And so perhaps, in fact, we should not be seeking for a single standardized assay, but the best way to synthesize all of the different assays that we have and are going to have coming forward.

Eunice Wang:
And I think you brought up that point, and how exactly do we synthesize the conflicting data, multiple mutations, lack of mutations, significance of maybe sometimes maybe not the perfect flow result. And I think that you discussed that the person, the way to synthesize that is the leukemia clinicians having to kind of use all the tools that they have on their disposal to say, is this significant? Is this possible positive flow? How do we interpret that in the setting of intensive therapy versus less intensive therapy versus regeneration, you know, count recovery versus differentiation?

Mark Levis:
And someone brought up, for example, let’s do single-cell, the fanciest thing we can possibly do, the most expensive thing. And while I might suggest that we have to incorporate that into a randomized trial, maybe that actually is too limiting. Because if you do just one assay, then you’re limited. In fact, what we’re doing potentially wrong would be designing a trial with only a single assay expecting that single assay to function. Maybe we should, in fact, have three or four semi-standardized assays.

Eunice Wang:
I mean, I think that would be great. I mean, I think we should prospectively. It’s like, as I think you mentioned and we’ve mentioned during our debate, there aren’t a lot of MRD assays, which hardly any at all, that have been validated prospectively that somebody develops an assay to look at MRD, and then they go and find some retrospective stuff or retrospective samples, which are, of course, biased, and apply the MRD assay and look at outcomes and say there’s an association, right?

Mark Levis:
But I’m watching trials now get developed, and they’re going to do the one assay that’s going to work, and maybe that’s going to be a terrible idea.

Eunice Wang:
Well, I think that might be a possibility because of cost, you know? Yes. Yes.

Sanam Loghavi:
And I think we’re going to find out. I hope that we don’t learn the hard way that even for NPM1 and KMT2A, you know, sometimes early molecular persistence of MRD may not be as meaningful. Again, right? For that purpose. Again, you’ve seen a differentiation effect. But if you have, you know, depending on how, you know, how smart and savvy we are about this and how we write the clinical trials and how we assess response, what is the timing? You know, when are we going to declare the patient MRD negative? And when do we say, oh, this patient is not responding to therapy, right? We don’t want to take the patients off trial early, right? Because we think they’re not responding to therapy.

Eunice Wang:
But that’s the question we want to know. That’s the question that you want to design a trial. And then the question will be not only validating these MRD assays, but then also tying them to some treatment decision. So if you are MRD negative in this trial, you know, is there going to be an outcome? Is it going to be just…

Sanam Loghavi:
Yeah, but define MRD negative, right? How much time do we allow for that patient to become MRD negative?

Eunice Wang:
Well, I think it depends because I think intensive therapy, we’ve always done two cycles. And I think in less intensive therapy, the data increasingly is four or more, right? But I think that one of the things that was brought up, which I think was a great suggestion, is AI. I think when you interpret these MRDs, because there’s so few events, and you yourself said.

Sanam Loghavi:
For flow, yeah, absolutely. AI is going to be helpful. No, no, no. For molecular, it’s not that we don’t see it. So for flow, it’s different. For flow, the problem is that the humans are missing the small population. There is a subjective component. Yeah, there is a subjective component. So I think AI could… Yeah, AI could certainly assist in that. For molecular, you see it, you just don’t know what it means, right?

Eunice Wang:
But I think AI can help us decide when you interpret the data, like what level.

Sanam Loghavi:
The aggregate of data is going to, but I also want to, you know, I alluded to this earlier too. I think, you know, obviously for us pathologists, the cases that are difficult to interpret are the ones that we’re going to remember and stick with us. And I think for you, the reports that are not helpful are the ones that you are going to remember and the hedgy reports. But I think if you look at all your patients in aggregate, right, or even the molecular borderline cases, the vast majority of the cases, you are going to have a black and white positive or a negative. There’s a small subset of flow cases that will have this hedgy, it may be an underlying MDS clone. It may be a pre-leukemic clone. We don’t know what it means. It may be AML with phenotypic shift, right? But the vast majority are going to be positive for MRD or negative for MRD. It’s similar for molecular.

Eunice Wang:
But I think that even aside from the interpretation subjectively, which we’ve talked about, can be overcome maybe with AI. I think for some of our patients for who we have secondary mutations, they have a secondary AML, they’re fit and they get a morphological remission with upfront induction and consolidation therapy. I would argue that if they were FLT3 positive or they’re FLT3, they’re flow positive or flow negative for MRD, I’m still taking them to transplant. So the question is, even now, even in the setting of a more reliable methods or methods or to detect MRD. One of the issues that came up, and I think Dr. Daver mentioned it, he said, look, this is a high-risk patient. They’re going to go to transplant either way as long as I can get some modicum of disease control unless my transplanter says I’m not going to take it.

Mark Levis:
Well, they’re a high-risk patient because ELN has defined them from the beginning, but in fact, MRD has the potential for calling out some non-high-risk patients.

Eunice Wang:
But right now, for a secondary AML patient, you would take that patient to transplant regardless of whether their flow was positive or not because we don’t have an MRD eraser, right?

Sanam Loghavi:
But again, I think this goes back to what is the purpose of doing the MRD? Is it risk stratification at baseline to decide who to transplant or not? or is it, are you using it? Like for instance, we have the one arm of Intercept open at Emmy Anderson right now and it’s based on flow conversion, right? So are you using this as a surveillance tool to say that someone is relapsing?

Eunice Wang:
I think I’m going to measure it. But right now, if I took a patient that had a secondary mutation, U2AF mutation, fit, and they got induction and they got consolidation, that achieved a morphological remission, regardless of whether their flow MRD was positive or negative, I would still send them to transplant, and that would be whether they were higher risk or lower risk. That’s still the path that I’m taking, right?

Mark Levis:
I think at the beginning of the patient’s diagnosis, we make the decision for what’s going to happen to them going forward. And therefore, I think the answer to your question from my perspective is I’m using it to monitor the therapy re-op right now. It’s a practical tool right now. It’s not yet one that I’m quite comfortable making a decision on for therapy. Are you going to transplant or not?

Eunice Wang:
And the issue is that there are right now a few validated MRD markers for NPM1, FLT3, that we feel confident enough and that we might have interventions enough that we might treat those patients a little bit differently, right? But for the greater majority of AML, which is, as you mentioned, an incredibly heterogeneous disease, this is the best that we have. So, of course, the question is people are like, well, are you going to use flow MRD? Of course I’m going to use flow MRD. But you have all these problems. I do have all these problems with it, but it’s the best that we have. Absolutely.

Mark Levis:
I do wonder and I’m tempted because the single-cell modalities that use both DNA sequencing and flow to combine those lovely plots, you do wonder if that wouldn’t help us getting a baseline and using that in MRD give us greater clarity, you know, or do we just fall into the trap of greater and more complex technologies that are more and more expensive?

Sanam Loghavi:
You’ll get greater and more complex. So I mean, but, you know, I’ll pose a scenario. Let’s say in an ideal world we have the single-cell sequencing, the, you know, combined protein and DNA, and you get a sample, and you have a residual IDH mutation with some abnormal immunophenotype that’s associated with it, how do you interpret that? It’s on a single cell.

(Overlapping conversation)

So I think the issue is not that we’re able to detect a lot of these abnormalities and mutations. We just don’t know what they mean and what to do with them.

Eunice Wang:
I think any other issue is cost. Can you imagine every single AML patient, you have to draw a sample, do single-cell sequencing, you have to do the flow. So even the turnaround for that.

Mark Levis:
And yet each, I would envision that AML being the heterogeneous disease that it is, we’re going to come up with a whole series of different MRD assays. Each one needs to be validated in a well-designed, randomized trial. And there isn’t any other way to get around that.

Eunice Wang:
And if you do all of this sequencing, all of this MRD testing using all of these methods, and you come up with like 10 different patients that had 10 different ways…

Mark Levis:
Then it’s up to the leukemia doctor to synthesize that.

Eunice Wang:
Right. And then what do you do? Then you just have data on like 10 different patients, you know, like…

Mark Levis:
Well, but we’re used to putting together that there is the complexity of our field. It’s hard, as Andrew pointed out.

Eunice Wang:
I mean, I think it’s going to end up because our disease is so diverse that we’re going to end up with a multimodality approach. We’re going to have to figure out, just like we, right now, when we diagnose AML, we do cytogenetics, we do flow, we do cytology, we do morphology, and we integrate all of that information.

Sanam Loghavi:
That’s perfectly fine.

Eunice Wang:
So I think that’s what we’re going to have to do as we get more and more of these targeted therapies. We’re going to have to come up with a multimodality MRD assay, where we might need to run five different modalities and see. But again, it’s complex. It’s not as simple.

(Overlapping conversation) I was told today that if I wanted something very simple and straightforward and have a straight MRD that’s already validated, I should go into myeloma.

Sanam Loghavi:
It’s never too late to change.

Mark Levis:
Okay, I think on that note, we’ll end the mini debate that we’re having here. Thanks. Thank you. Thank you.

This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.