Vidit Sharma: Yeah, thank you so much. It's a pleasure.
Neeraj Agarwal: So tell me first, and again, please feel free to answer. I'll start with Chloe first. So Chloe, congratulations again. Tell us about the title of the abstract first you are presenting, and then we can talk more about the steps you have taken to characterize the muscle loss in these patients.
Chloe Shi: Yes. So the title of our abstract is, "An Artificial Intelligence Algorithm Characterizes the Deterioration of Body Composition In Patients With Advanced Prostate Cancer." So we wanted to look at patients that are on different systemic therapies in addition to ADT and the way that their body composition changes longitudinally. And we did that with an artificial intelligence algorithm that automatically segments PET-CT scans to extract body composition metrics.
Neeraj Agarwal: Why artificial intelligence?
Vidit Sharma: It's a good question. If we were to do this manually, which is the standard and established way to do this, it takes over 15 minutes per individual slice on a CT or PET scan to identify and trace, which we call segmentation, every component. And there is significant inter-observer accuracy in doing that, and reliability. And so using an objective algorithm to identify the components of your body composition just makes intrinsic sense, and also speeds up the timing so that we can actually get the information in our clinical workflow.
Neeraj Agarwal: And you mentioned 2D versus 3D. Could you please elaborate more on that? What is the advantage of 2D versus 3D?
Chloe Shi: 3D assessment will give you a more comprehensive look at a patient's body when you want to look at body composition metrics. So instead of just using 2D scans or segmentations, like looking at a single slice of a PET-CT or a CT scan, looking at the 3D metric of evaluating a patient's body composition will give us a more accurate and also a better way to clinically apply this body composition algorithm.
Neeraj Agarwal: Thank you. So we have traditionally focused on PSOAS muscle. Always in literature going back two decades, people look at the CT scan and look at the PSOAS muscle. What is the difference in this project compared to all those initiatives which have been taken in the past?
Vidit Sharma: Well, this builds on that. So the PSOAS muscle, I think, is a easy to identify muscle that often correlates with generalized sarcopenia, but there's still differences in terms of, it's not 100% correlated. So you can lose abdominal musculature, rectus sheath musculature, rectus muscles, but maintain your PSOAS muscle. So there's differential muscle loss in different domains of muscles. And so a comprehensive look at the 3D muscle mass of the torso actually captures more variation in muscle mass than just looking at the PSOAS muscle. So I think it's more applicable, and in clinic we often have an eyeball test to see if patients are losing muscle. And just anecdotally, the 3D changes correspond more with the eyeball test than just PSOAS muscle.
Neeraj Agarwal: Very interesting. So please tell us about the methodology, how exactly things are being done.
Chloe Shi: We used this algorithm that was created by our friends in Mayo Clinic Florida, and we used our advanced prostate cancer registry to identify patients that had PET-CT both before the initiation of systemic therapy and after. And we used this algorithm on the serial PET-CT scans that these patients received and then categorized their PET-CTs into the four different systemic therapy cohorts that we looked at, which were ADT, ARPIs, chemotherapy, and lutetium-177. And then we were able to use linear mixed effects models to measure, how did these patients' body composition changes such as muscle density, muscle area, subcutaneous fat area, and visceral fat area change over time while they were on these different systemic therapy cohorts?
Neeraj Agarwal: And what were the results?
Chloe Shi: So we found interestingly, first looking at just the longitudinal trajectory, that patients lost muscle density and muscle area typically over the first 500 days that they were on systemic therapy. And then looking more closely at the numbers, we found that yes, patients lost muscle density and muscle area while they were on ADT. But interestingly, they continued to lose this muscle mass going beyond ADT into the other systemic therapy cohorts that we looked at with patients in the lutetium cohort. So the furthest escalation of therapy, we observed around a 20% decline in their muscle density compared to their baseline scans before they were on systemic therapy.
Neeraj Agarwal: That's worrisome, isn't it? That patients continue to lose muscle mass even after maybe stopping ADT, which may have been given to them for, say, two or three years in conjunction with radiation therapy.
Vidit Sharma: Yeah, it's definitely worrisome to see that patients are losing muscle mass. When we looked at the time course of muscle mass loss, we actually observed a plateau in the raw data. So about 500 days is where the plateau happened, where if you're on systemic therapy for 500 days, you've probably reached your plateau of where you're going to be for muscle mass. And that I think intrinsically makes sense. There's probably a certain amount of muscle that is more dependent or more susceptible to loss with these medications. And then once you're closer to whatever your genetic setpoint is, you're going to maintain that muscle. So the first 500 days is that critical period, but it definitely is concerning that patients are losing a significant amount.
Neeraj Agarwal: And in the clinic, I treat these patients with ADT all the time, and the perception is just opposite. Perception is, "Oh, we should do something after two or three years off ADT therapy," and by the time it looks like they've already lost a significant amount of muscle mass. So with it going forward, what are the implications of these data for our patients, for our colleagues out there in the community?
Vidit Sharma: Well, I think there's several implications. I think prevention is better than regaining muscle. So preventing muscle loss is, in my opinion, the way to address this. And so when you start someone on ADT, when you start someone on ARPIs, really emphasizing that, "Hey, we now have data to show that you can be at risk of significant muscle loss. Let's start a strength training program, let's focus on large muscle groups, compound exercises that are safe, let's optimize nutrition to prevent muscle loss." Just saying that from a physician is important, and that often, I think, really influences patients to make some lifestyle changes.
Neeraj Agarwal: Very well said. When the resources are limited out there, we want to recognize people at the highest risk of losing muscle so that we can tailor, not only triage them first, but also tailor interventions around them or for them. Any last comments, Chloe?
Chloe Shi: I think that this algorithm is an important first step in seeing, how can we better identify these patients that are at risk to have earlier implementation of these important prevention strategies? And then also looking further down at what type of muscle loss, or how do we quantify muscle loss in association with poor outcomes, and seeing, is there an association with these patients losing muscle in their oncologic outcomes with prostate cancer?
Neeraj Agarwal: Absolutely. Metabolic syndrome, diabetes, increased risk of heart disease, stroke risk, and all sorts of things which you are trying to prevent.
Vidit Sharma: Yes.
Neeraj Agarwal: Well, congratulations for your fantastic presentation, and we look forward to hearing more from you, Chloe, in the next coming years, great things to happen.
Chloe Shi: Thank you very much.
Vidit Sharma: Yeah, thank you very much. And I'll just add one thing. A lot of the credit really goes to our collaborators. Alex Weston, he's an R01 funded investigator to create this algorithm. And so in terms of how the algorithm works, how accurate it is, he's done so many studies on it, really an expert in this domain, so just want to give full credit to him for creating this awesome tool for clinicians.
Neeraj Agarwal: Thank you for doing that.
Vidit Sharma: Yeah, it's great. Thank you.