A focused discussion on best practices for MDS pathology and risk status reporting in both community and academic center settings.
Rami Komrokji, MD: I think there’s a difference in this based on the pattern of practice. At least in the United States, where you have academic centers where we do all those exercises, calculating 4 risk models, vs a community oncologist who is seeing so many other types of diseases, and they get the patients. I think the message is that we all should have a unified system. The prognostic models can impact treatment availability for patients, like in Europe, for example, and for us, obviously, enrollment in clinical trials and eligibility. Now, going back to hematopathology, do you help with this? I know it’s different in academic centers, that you may have to calculate the risk for the patients. But do you think that’s a good idea to have those prognostic models in the hematopathology reports, particularly for a community oncologist?
Sanam Loghavi, MD: I think the information should be provided to the clinician so they can use the prognostic models. For instance, in my report, or the way we do hematopathology at The University of Texas MD Anderson Cancer Center, we try to integrate the reports. Obviously, when you’re making the diagnosis at first, you only have the morphology part. You don’t really have the molecular part of it. But I think the gold standard of practice should be that you do include that information in your report, because your patient may be taking that report elsewhere, too. So, Dr Garcia-Manero has access to the molecular testing that’s done at MD Anderson through our Epic medical records system, but if that patient goes elsewhere, and they just take their bone marrow biopsy report with them, if I don’t include that information in the report, then the other clinician may not have access to that information.
Rami Komrokji, MD: I think, for the community, particularly in the United States, it’s very helpful when the hematopathology report includes the risk stratification. Because, as I mentioned, most community clinicians have a busy practice, they don’t have time to calculate the risk, look at 20 genes, and put that together. So it would be helpful to have that. And I think, again, in the end, we are just trying to put patients into 2 major categories—higher risk, where we’re thinking of transplant immediately, and lower risk, where we could go stepwise, managing cytopenias.
The other thing I always find a bit challenging from the hematopathology is, what’s your opinion about the blast percentage, like 3%, and above 2%? In most of the reports we see, the blasts are just reported as less than 5%. What’s your thought about this blast cutoff, in the revised IPSS [International Prognostic Scoring System], at least?
Sanam Loghavi, MD: There were a couple of abstracts at this ASH [American Society of Hematology] meeting showing that maybe 5% to 10% doesn’t matter as much as more than 10% does. I think you have to realize that counting blasts is not a very reproducible practice, right? I think we have to be honest. I do think there’s a biologic difference between cases that have more than 10% blasts. It’s a more advanced form of the disease. Those are usually associated with worse cytogenetics, with worse molecular features. But I still think there is value to counting blasts. There may not be much of a difference between 10% and 12%, or 2% and 3%, but I do think it’s an additional piece of information that you can use. And it has been shown to reproduce in different models.
Uwe Platzbecker, MD: Can I ask a question?
Rami Komrokji, MD: Sure.
Uwe Platzbecker, MD: Do you think that maybe AI [artificial intelligence]-based technologies can help us to count 5000 cells, instead of 200 cells?
Sanam Loghavi, MD: I think, yes, definitely. It’s one of the more mundane tasks we do that is still required, and I’m sure a computer algorithm, if it’s trained appropriately, is going to be much better than a human, right? Because the input is going to be much larger, and the reproducibility is going to be more. But I can tell you from experience, we have a high-resolution scanner right now that has the ability to count cells, kind of like what CellaVision does. And we’re training it. Right now, it’s not very good at recognizing cells, it makes mistakes. But you have the ability to train it, saying, “Oh no, you’re wrong, this is not an eosinophil, it’s actually a dysplastic neutrophil,” or something like that. But I do think that in the future that’s probably what’s going to happen.
Rami Komrokji, MD: That’s a very interesting concept, and I totally agree.
Transcript edited for clarity.