Matthew Kurian, MD, and Arturo Loaiza-Bonilla, MD, preview ESMO 2026 developments in artificial intelligence and genomic testing for breast cancer. They explore how AI could refine recurrence risk predictions, guide chemotherapy decisions, and identify patients who may benefit from extended endocrine therapy.

Transcript:

Dr. Matthew Kurian: Hello, everyone. My name is Matthew Kurian. I am a breast medical oncologist at St. Elizabeth’s Yung Family Cancer Center in Edgewood, Kentucky. I’m very thrilled to have a great colleague and a leader in artificial intelligence here with us, Arturo Loaiza-Bonilla, who is also part of Massive Bio as well too. Arturo, could you introduce yourself to our audience?

Dr. Arturo Loaiza-Bonilla: Sure. Yes. Nice to see everyone. I’m a medical oncologist. I specialize mostly in GI malignancies. I’m the network chief of hematology oncology at St. Luke’s University, which is a network of 16 hospitals in the Pennsylvania Lehigh Valley region. I’m also the co-founder, chief medical officer of Massive Bio. As Dr. Kurian had mentioned, I’m very much enthusiast about AI in oncology for several years. I’m very happy and excited to be here and talk about what’s coming.

Dr. Kurian: Absolutely. Thank you for being here as well. We have ESMO 2026 around the corner here. A lot of buzz in terms of some of the artificial intelligence studies that are coming out of there. I wanted to start a little bit with just thinking about how AI is moving from prognosis to treatment decisions in your mind. Art, what do you see moving forward, not even just within breast cancer, but just within all different types of cancers? How are we moving from just prognosis and prognostic measurements to actually making treatment decisions?

Dr. Loaiza-Bonilla: Yeah. We’re at this very interesting time where we have tons of data coming up. AI is the flavor of the day. Everyone’s talking about it. But now we come into the same question: what models are we going to be using? What is the modality that we should be leveraging to make decisions? It comes down always to validation. What is the data that is really showing that that algorithm, that decision support, that model, in that specific context, is applicable to the patient in front of us?

Dr. Loaiza-Bonilla: That is a key part that I’ve seen still in a number of questions that we have. I see some answers in places like digital pathology, where they have enough data sometimes to prove the case. But when it is coming around other use cases, we’re still learning together on how to use these tools.

Dr. Kurian: Absolutely. I think for me as a breast oncologist, there are a few abstracts that I’m particularly looking for, incorporating AI with some of our genomic decision-making tools. For example, there’s one that’s going to be a late-breaking abstract 20 for AI-based MammaPrint as well too. MammaPrint is a 70-gene expression assay to assess recurrence risk in early breast cancer. It helps us, and it’s been really shown in the MINDACT study to help identify folks that may beable to safely avoid chemotherapy, that may have high clinical risk but are actually low genomic risk as well. They’ll be presenting a study regarding their combination of using artificial intelligence and this genomic platform that many oncologists across the country use.

Dr. Kurian: What do you think is interesting about this type of model, Arturo, regarding combining things like our Oncotypes and MammaPrints and genomic-based scores plus artificial intelligence together? What do you see are the benefits or pitfalls within that type of model?

Dr. Loaiza-Bonilla: Yes. I think this is one of the most intriguing components because it makes total sense to start looking at these multimodal assessments in pathology and try to determine, okay, what patient really benefits from adjuvant treatment? Which one really may have the highest chance of response or not to treatment? I see that that is a really unique case that a lot of folks are trying to really tap into. The pros I see is that it is verifiable and is non-falsifiable. We know that there are certain biomarkers, as we combine them for the TAILORx and all the studies that are really relevant to breast cancer, and that they are not only sometimes prognostic, but sometimes predictive of response.

Dr. Loaiza-Bonilla: What I feel is important is to make sure, to see, is there anything else that we may be missing in those gray zones? When we’re talking about Oncotypes and MammaPrints and all these different tests that we use for several years, there was always this gray zone in between scores that we were not sure what to do with. I feel that’s where AI has the highest opportunity to really show us things that are black box to us, that we just make decisions based on looking at the patient or making informed, but not fully understanding the underpinning of that decision to make sure that that is really accurate for that patient.

Dr. Loaiza-Bonilla: It’s all about making sure that all those tools are really proven in a prospective manner, and that always the bare minimum is the current standard of care uses. I don’t think it’s necessarily to replace the whole process, but maybe to make it faster for us to make determinations, and also hopefully becomes better over time as we combine multimodality. It’s not just the pathology, but also the clinical stage, the patient status, the age, the treatment options that we’re using, because also the paradigm is changing as we go through these tools. Now we’re talking about using ctDNA for certain patients, for example, and that’s going to change the way that we prognosticate for the future for certain folks. I think then the implication is to always be on the lookout for what’s next, and don’t settle for the last model because that model is going to be versioned soon, the same way we change software.

Dr. Kurian: I agree. I think you hit on the head there in the sense of that our current modeling, at least in breast cancer, what I do, it’s very binary. It’s high risk, low risk, and we put people into two of those buckets. But you’re right. There’s a lot of those people in between there, younger women that we potentially may de-escalate therapies for and may not expose them to unnecessary side effects or chemotherapy. I agree with you. I think that’s exactly where I see that really being an advantage too. I think one of the things in breast oncology specifically is that we’re really good at telling people who is high risk of recurrence, but who actually benefits from additional treatments and newer approaches? I think that abstract I’m looking quite forward to overall too.

00:07:15 Dr. Kurian: Other things in our field too, within the breast cancer field, is the question of who benefits from extended aromatase inhibitor therapies, or who benefits from five years of endocrine therapies versus 10 years. Now we have new data that’s going to be presented, and probably the final analysis of the NSABP B-42 study, that’s going to help tell us who really benefits from those extra five years of treatment or not. We have another genomic assay, the Breast Cancer Index score, that helps us figure that out and try to identify. But I really struggle sometimes when we get those assays and we tell someone that they may be at increased risk of recurrence, and their benefit is maybe only a percentage and a half or 2% more of using a medication for five more years. That’s something I think that is going to be presented also as another abstract at ESMO too.

Dr. Kurian: But what do you think regarding that approach, really, for looking at way down the line, years down the line, of utilizing artificial intelligence in that manner? We’ve been able to predict five years down the line, and this assay may be saying 10 years down the line. But do you think we can predict even farther down the line with some of these AI tools potentially? Because that’s been a struggle for our field within breast cancer, these late recurrences 10 years later, unfortunately, that we see within our field. Do you think that’s something that could help, you think, overall?

Dr. Loaiza-Bonilla: Yeah. I think it can help the patient today. But I always would like to tell the patients, take this with a grain of salt because things are changing so rapidly that the paradigms that we currently have are meant to be, again, for transformation in a few more years. The drugs that we’re using right now may be helpful for the patients at this generation, but in five years we may have stronger, better, more powerful medications that are going to be probably used, or algorithms that are already going to tell us, “Hey, by the way, we had predicted that the benefit will be five years, but now we run it further on this bigger, larger model database, and you may not need it anymore.”

Dr. Loaiza-Bonilla: I think if we’re going to using these AI tools, it’s okay to commit to a certain sense of ground truth and reality for today. But we also need to make the patients aware that there’s a lot of dynamic changes happening in real time, that even for us oncologists we’re still adapting to all those things happening. I think, yes, they’re going to be helpful, and they’re going to really give us a little bit more clarity. But at the same time, we need to keep that open mind, because I see all these frontier models and the capabilities and the scalability, particularly where things that we already took for granted, they’re really being transformed for many reasons.