Mikkael Sekeres, MD, MS: Do you restage patients? If so, when do you do that?
Rami Komrokji, MD: As I alluded, I think those models have not been looked at as dynamic. It’s very difficult to apply them, but I think there is still some common sense. If somebody had low blasts originally and the blasts are increasing or acquired new mutations—further cytopenias—I think that the patient is not going to do well. One issue, unfortunately, is that many of the studies sometimes will still require the IPSS [international prognostic staging system] as an eligibility criterion. For example, after hypomethylating agent [HMA] failure, that doesn’t stand.
We showed that neither the IPSS nor the revised IPSS predicts outcome after HMA failure, and yet we still do it for every study. There are different models of looking at it. But I think what makes common sense is still—if you see a change in counts, increased blast count, new cytogenetic abnormalities—that would make a difference. Sometimes when patients change, especially if I saw a clinical path in the change, I think it’s reasonable to recheck for somatic mutations. This is because, every now and then, we’ll see patients acquire new mutations that could be targeted. If somebody became proliferative—suddenly the count is high—it’s rare but you may acquire a FLT3 mutation. You could acquire IDH1 [isocitrate dehydrogenase 1], IDH2 [isocitrate dehydrogenase 2] mutations, things that nowadays could be potentially actionable and targetable.
There is no good restaging system. Usually, if patients are doing well, I don’t repeat regular frequent bone marrows, even if we make the argument that if someone is not going to transplant and you started them, for example, on hypomethylating agents and they have hematological improvement, you don’t need to repeat the bone marrow unless you see clear failure of the treatment by dropping…counts later on.
Jamile M. Shammo, MD: I totally agree. I don’t have a set point where I would stage patients. However, I will do it if I feel like the results might influence my decision as to how to treat these individual patients. For example, if someone becomes cytopenic and is not recovering in time for their next cycle of therapy, I might stage and see if this is disease progression or if this is drug-related cytopenias—perhaps people may need to have their cycles extended a little bit—or if I, as you said, think that there may be progression. So, there’s no set time for me.
Mikkael Sekeres, MD, MS: It seems like we do tend to do it based on a patient’s blood counts and the appearance of blasts. There are usually clear clues that somebody is starting to progress. As we move forward with prognostic scoring systems, we even have examples of using machine learning, artificial intelligence, to incorporate all the many factors that go into prognosis for a person—from all the molecular abnormalities and their interaction and hierarchy to clinical characteristics—and I think that in the future we’re going to have to go to an online—some kind of app—enter a lot of information in real time about a patient and then have an output about prognosis.
Ellen K. Ritchie, MD: I’m glad that you mentioned artificial intelligence. I think the most interesting session that I went to at the 2019 ASH [American Society of Hematology] Annual Meeting was the artificial intelligence session where Kun-Hsing Yu, MD, PhD, from Harvard presented one of the most beautiful overviews of what the use of AI [artificial intelligence] is going to be. He was even able to show that, morphologically, machine learning can actually pick up p53-mutated cells and was able to fit the curve just to the p53-mutated patients and predict what their outcome would be with the morphological data with machine learning.
I think that as we start to incorporate the complexity of this disease…, as he said at the end of his presentation, physicians will not be replaced by AI, but physicians who are able to intelligently use AI will replace those physicians who are not able to use it. So, I think that we are entering one of the most exciting phases of medicine. I probably won’t live long enough to see the full outcome of this, but we’re going to enter a phase where we’re able to use machine learning to better stratify these patients right off the bat to how they should be treated. It’s very exciting, I think.
Jamile M. Shammo, MD: We may not have a job.
Mikkael Sekeres, MD, MS: We will be one of those people with the 25-year-old muscles forever.
Ellen K. Ritchie, MD: I’m trying, but I don’t know about that.
Transcript Edited for Clarity