MMAI Model Risk Stratification for NCCN Guidelines for Low, Intermediate, and High-Risk Prostate Cancer - Neal Shore, Daniel Spratt, & Alicia Morgans

June 4, 2024

Zach Klaassen hosts a multidisciplinary roundtable discussing the NCCN Prostate Cancer Guidelines with a focus on the ArteraAI prostate test. Joined by Dr. Dan Spratt, Dr. Alicia Morgans, and Dr. Neal Shore, they explore the importance of risk stratification in prostate cancer, highlighting the superior predictive and prognostic capabilities of the ArteraAI test compared to traditional methods. The test uses AI to provide more accurate risk assessments, significantly improving decision-making for treatment options. They discuss the benefits of using AI in clinical practice, such as rapid result turnaround and tissue preservation. The panel also examines the implications of new data on biochemical recurrence post-radical prostatectomy, emphasizing the potential of AI to refine treatment approaches further. The discussion underscores the transformative impact of AI in oncology, setting a new standard for personalized cancer care.


Neal Shore, MD, FACS, Director, CPI (Certified Principal Investigator by the Association of Clinical Research Professionals), Medical Director for the Carolina Urologic Research Center, AUC Urology Specialists, Myrtle Beach, South Carolina

Alicia Morgans, MD, MPH, Genitourinary Medical Oncologist, Medical Director of Survivorship Program at Dana-Farber Cancer Institute, Boston, MA

Daniel Spratt, MD, Chair and Professor of Radiation Oncology, UH Cleveland Medical Center, Seidman Cancer Center, Case Western Reserve University, Cleveland, OH

Zachary Klaassen, MD, MSc, Urologic Oncologist, Assistant Professor Surgery/Urology at the Medical College of Georgia at Augusta University, Well Star MCG, Georgia Cancer Center, Augusta, GA

Read the Full Video Transcript

Zach Klaassen: Hi, my name is Zach Klaassen, and I'm a urologic oncologist at the Georgia Cancer Center. I'm pleased to be hosting the NCCN Prostate Cancer Guidelines, with a specific focus on the ArteraAI prostate test. This is a multidisciplinary roundtable discussion, and I'm delighted to be joined by three experts in the field who are familiar with the AI platform: Dr. Dan Spratt, radiation oncologist at Case Western University; Dr. Alicia Morgans, medical oncologist at Dana-Farber Cancer Institute; and Dr. Neal Shore, urologist at the Carolina Urologic Research Center. Thank you all for joining us today.

So here's our outline for today's discussion. We're going to talk about why risk stratification is important. We're going to jump into the core of the ArteraAI indications. The first one, looking at NCCN low, intermediate, and high-risk prostate cancer. The second indication in NCCN intermediate-risk prostate cancer. Some new data looking at the post-radical prostatectomy biochemical recurrent setting, and then some closing remarks.

So just to set the stage, why do we risk stratify patients, and why is this important? There are several reasons for this, and we want to determine how likely a cancer is to be confined to the prostate or spread to the regional lymph nodes. How likely it is to have cancer progress or metastasize after treatment? And how likely is adjuvant or salvage radiation to control cancer following radical prostatectomy?

If we look at the brand new 2024 NCCN guidelines, this is the table for risk stratification based on clinical variables. We can see there are a number of options in the localized setting. We have CAPRA-S post-radical prostatectomy. A couple of nomograms in the BCR setting, post-radical prostatectomy. And then looking at high and low volume and number of bone metastases in the metastatic hormone-sensitive setting.

But really, the question is, can we do better with advanced tools for risk stratifying these patients? This table is also new to the 2024 guidelines, and we see here specifically AI pathology. The multimodal artificial intelligence ArteraAI prostate test has both predictive and prognostic biomarkers with assignment level of evidence 1B. And really, this is a new update to the guidelines with its own specific table discussing multimodal artificial intelligence.

So looking at the first indication for the ArteraAI prostate test, this is this new table. This is the first part of the table. And again, to highlight looking at NCCN low, intermediate, and high-risk patients, a continuous score. And we'll just do a quick deep dive into the data that led to this recommendation.

This was a paper published in 2022. This was patient-level baseline clinical data, with digital histopathology of pretreatment biopsies of five RTOG studies. This was more than 5,000 patients, more than 16,000 histopathology slides. And excellent follow-up, more than 10 years of follow-up. The treatments given in these five trials were external beam radiotherapy plus or minus ADT. The slides were then digitized using digital pathology and manually reviewed for quality and clarity. Really, the goal here was to assess the role of multimodal AI to provide accessible and scalable prognostication in localized prostate cancer.

Without getting into too much detail, the multimodal architecture is composed of two parts: a tower stack to parse a variable number of digital histopathology slides, and then another tower stack to merge the resultant features and predict binary outcomes.

This is really the key take-home slide from this study. If you look at multimodal AI in blue and NCCN prognostication in gray, across all six of these outcomes varying from different lengths of distal metastases and different lengths of time for biochemical failure, 10-year prostate cancer-specific survival and overall survival, we see that across the board MMAI was significantly better than the NCCN prognostication.

If we look at each of these outcomes, all six of these outcomes in the five trials, there was only one area where there was no difference between MMAI and NCCN. This was biochemical failure at 10 years in the RTOG-9910 study, and this may have been due to short follow-up time, low event rates, and all patients receiving hormone therapy, so less likely to have biochemical failure. Again, across all five of these RTOG trials, we see essentially a clean sweep for MMAI versus NCCN guidelines.

Again to highlight, this is across all trials. This is the relative improvement across these six outcomes. We see a 9 to 15% improved benefit for the MMAI model versus the NCCN prognostication. So the statement from the NCCN is, "Given the superior discrimination of the MMAI model for multiple oncological endpoints over the NCCN risk groups, this test may be used to provide more accurate risk stratification to inform shared decision-making regarding absolute benefit from various treatment approaches. However, the specific score cut points have not been published to date for specific treatment decisions."

So I'm going to bring the panel on now to discuss some of these important implications from this guideline recommendation. I'll start off with Dr. Spratt. Given the superior risk stratification compared to NCCN, what are the clinical implications of this MMAI model?

Daniel Spratt: Absolutely. So in localized prostate cancer, the guidelines NCCN low, intermediate, high, and various subdivisions, that is how functionally we personalize treatment based on prognosis. Gleason score, T-stage, and PSA are not predictive. They help give us an estimate of the aggressiveness of the disease, but as you saw, it's not phenomenal, the NCCN risk groups at doing that. And so what that means is, we're probably over or underestimating an individual's prognosis, risk of recurrence or metastasis, and we're applying therapies to them kind of in this one-size-fits-all.

A great example that we're often trying to decide, let's just take an intermediate-risk guy trying to talk about hormone therapy. Let's say in all comers, let's say it's a number needed to treat of 10, but if they have a more indolent prognosis, it might be a number needed to treat of 50. And that's a very important shared decision-making based on their age, preferences, comorbidities, etc. So it lets you have exactly that shared decision-making regarding whether this is a high probability of a benefit or a low probability of a benefit. It's not telling you with a prognostic biomarker that there's no benefit, but I think each man is willing, based upon the side effect profile of that potential therapy, to help personalize that approach.

Zach Klaassen: Absolutely. How about from your standpoint, Neal? How does this change how we discuss with patients?

Neal Shore: Well, I think it's really rather revolutionary. Maybe take sort of a treetop view of artificial intelligence. I remember when I was at ASCO GU in 2022, and then the following year at 2023, and to Dan Spratt's credit, and the team of all those wonderful authors who are on those papers, it was very multidisciplinary, but it was really led by our radiation oncology colleagues. I think really kudos to them. These five RTOG now NRG trials really had an incredible database to look at, okay, what happens to these patients who have androgen deprivation therapy for the intermediate-risk population? And is there a better way through this MMAI, as you pointed out in your introduction, Zach, really well, and maybe we'll talk a little bit more about it. Can we with a continuous variable say who will benefit or not need androgen deprivation therapy? We all recognize the deleterious effects, but at the same time, we want to synergize with the radiation therapy and T suppression as a sort of a radiosensitizer.

So quite remarkable. This came right after COVID was sort of getting to a nadir of our anxiety, and I just thought it was amazing that the presentations, there were several at each one of the Congresses, back-to-back plenary presentations. I think it's really important from a uro-oncologic standpoint, and it's still early days, 2024. I know we're going to talk about some additional data that was just presented at AUA by Todd Morgan, but I think this is revolutionary, and I don't say that in a blithe way. I mean, as Dan points out, and if you pointed out as well, this is both predictive and prognostic, and it's in a continuous variable for our patients. And so it's less clunky. We just say, oh, someone's low-risk, intermediate, or high-risk.

I think we'll talk more about how you order the test and some of the real advantages to it, which I'll stop. But I think this is really quite remarkable. I think a presentation like this, as you point out, in a multidisciplinary way, will show how we're all going to better use this test now and later on for other potential indications once trials and evidence, more evidence-based, as this is, as you pointed out, this is the highest level of evidence, Simon 1B. So I think this is great for anyone who takes care of patients with any sort of scale with newly diagnosed prostate cancer. Especially if they're considering radiation as a prudent decision.

Zach Klaassen: Yeah, absolutely. Alicia, we've got some really important endpoints from a medical oncology standpoint: overall survival, prostate cancer-specific survival. Just looking at these numbers and how much better it is than the NCCN, what are your thoughts?

Alicia Morgans: So thank you for bringing that up. I mean, obviously overall survival is going to be an important endpoint. And I think that when we think about all of the different endpoints that matter to patients, things like metastasis-free survival, and even biochemical recurrence-free survival, can be things that we, maybe as regulators or policy folks, think, "Oh, well, that doesn't necessarily matter if overall survival is the same." But I think, importantly, our FDA has spoken with their feet and actually metastasis-free survival, so the development of metastatic disease, is very clearly something that we all accept now, thankfully, as important to patients. And biochemical recurrence-free survival may be something that actually is in flux. I mean, especially with PSMA PET scans, and some of our intensified treatment opportunities, and metastasis-directed therapy, perhaps we can sort of intervene on this and make it a longer-term problem, even with biochemical recurrence.

But I think that inherently this is very important to patients. And super important to physicians who take care of patients, because we have to walk with them through that process. So these are all really important endpoints. And I think, to Dan's point earlier, a prognostic endpoint is really critical. But what I think we're all eager to talk about, and is important, is really a predictive biomarker to help us make these treatment decisions. Because prognosis is so critical when we think about the risk of patients and having those shared decisions in areas where we don't necessarily know if the treatment is going to be effective or not. We want to think about, is this high enough risk that if we intensify, there's potentially some benefit? But a predictive biomarker is really going to tell us, will the treatment we have actually help this particular patient? So that's ultimately the goal. And early days with this, and you're going to get to the guidance here, but I really feel like this particular modality of AI and these biomarkers ultimately is going to really help us with that important question as well.

Zach Klaassen: Yeah, absolutely. I think I'd like to touch on just the logistics of ordering the test. And so I know, Dan, you've been involved in the very early work of this, and you've used it in the clinical experience. Just take us quickly through what that process looks like. How long does it take to get the results back, and that sort of thing?

Daniel Spratt: Yeah. So I think those who have ordered other, we'll call it prognostic biomarkers, we'll call the front-end workflow will feel similar in that right now you're essentially mailing, but rather than tissue blocks or unstained slides, you only send necessarily a single H&E slide.

So first of all, even at centers that have a lot of, we'll say, privacy IP things, they don't want blocks sent. This is just a stained H&E slide, very easy to do, usually very quick and not much pushback. And once they receive it at the lab, which is really just digitizing and running it through their algorithm, usually within two days or sooner you have it back, right? Because there is not an actual sequencing to extract RNA, DNA, etc. So it's pretty quick. And like many of these companies, it goes to an online portal, where you can readily access these results. It's pretty quick. If you have, for us, the tissue in-house in under a week from ordering, we'll have the result. If you're getting it from an outside center, it just depends how responsive medical records and pathology are at that center. But that's sort of a universal truth.

Zach Klaassen: Neal, how about your experience too? I know you were pretty quickly on board with this technology as well. And I guess maybe answer the question, do we have to send, and this has been asked to me a lot. Do we have to send all the slides? Is it just one slide? Is it the highest grade? How does that work?

Neal Shore: Yeah, that's a great question. So yes, currently we're sending one to three slides. We want it to be as representative as possible. It gets shipped Federal Express, it goes very quickly. If I send it, for example, on a Monday, and it's there by Wednesday, I've had reports back. As Dan points out, there's no need for tissue extraction. So there's complete tissue preservation, and the slides are returned to you. And then their process is really, once they have the slides and they're able to digitize, I mean, that process is in 24 hours. So I've actually sent on a Monday and had results back by Friday.

Zach Klaassen: Wow.

Neal Shore: I say that in that I'm a big believer in next-generation sequencing. We send off tissue and molecular profiling as well, with a lot of the tests that are out there. And I've kind of gotten into trouble because I'll typically tell my patients, "Oh, you just come back in two weeks, and we'll have it all." And my assistants have gone behind my back and made the appointments three weeks to be on the safe side. Because there's nothing more upsetting to a patient than coming back in for a discussion and you don't have the results.

But the turnaround time, just to throw in another acronym, now, the TAT, is really dramatically quick with this, because they're just digitizing. I think in the future, and when I've talked with some of the folks at the company, Dan may know more about this, but it's not a particularly uber-expensive sophisticated camera. And I think there may be even an opportunity for some sites, where you could take the digitized picture yourself and then put that in the cloud, and that would even cut back a couple of days. So instead of five to seven days, you could be getting this back in three to five days. So that's scary positive in terms of getting our... When we want to make a decision and we want to make it quickly. And patients are very anxious, and we are too. It's like, what are we going to do? And we want that information. And we can do it in time. And then the tissue preservation, so that you're not exhausting any tissue. So those two facets are really remarkable. I think they're great breakthroughs for the field. That's the beauty of AI and digitization.

Zach Klaassen: Absolutely. Alicia, you guys have just started using it in your practice. Anything different than what Dan or Neal has said?

Alicia Morgans: I would just say we've really appreciated the educational process that the group, Artera group, has really provided for us and for our clinical team. I think when it initially rolled out, there was sort of a need for the doctor to have access to a portal. And honestly, in my practice, there are too many portals, and there are too many things that I can't always keep up with. I need someone to help curate the information and get that information to me. And I know that sounds silly, but when you've got a lot of patients and you're doing a lot of different things clinically, and then also in other worlds, especially with academics for me and administratively, but for everyone, we all have all of our draws. We need a team to help us with this.

In any event, the Artera team has actually made it really easy at this point. And so, my team has been trained, and they are providing all of this support to help my team and our group, which is over 20 physicians, to really engage with this in a meaningful way. And to have our support staff and our nurses really help to support getting the information out that needs to be sent, and then also to get the data back in. And so, I think it's actually a seamless process. And I've so appreciated that there has been this training and educational experience to make sure that that happens in a meaningful way, and that there's support, should there be any challenges as we have submissions ongoing.

Zach Klaassen: Absolutely.

Neal Shore: Can I make one quick comment?

Zach Klaassen: Sure. Yeah.

Neal Shore: This is a real-world comment that I think is important for some of our listeners, probably most of our listeners. And actually again, gosh, accolades and congratulations to the folks who did all this. I wasn't part, I wasn't an author on these papers, and I think they went from this level one evidence and in rapid fire into the guidelines within, I think, record time, and approval by CMS. This is CMS approved, getting coverage. I mean, of course, commercials, there's always a lag with some commercial payers, but this is approved by CMS and in the guidelines in essentially record time. And I always think that when I see research that leads to that accomplishment, it's really laudable in an incredible way. We do a lot of stuff, and then it's years before it gets into the guidelines, if at all, and/or gets a reimbursable pathway.

Zach Klaassen: Yeah, well said. Neal. Before we get into indication two, I just want to do a quick rapid-fire just around the table. What's the initial experience been from the patient standpoint? When you give them another very good data point that says, we're going to do this and this because of this? Maybe I'll start with Neal on this one.

Neal Shore: Yeah. The reports, we've all seen a lot of different reports, and some of them can be incredibly filled with a lot of jargon, a lot of acronym word salad, so to speak. And then the patients find it a little bit overwhelming. I think they've done a really good job. Not only for the patients, but also for the physicians to say, here's what the percentage is, here's what this means in terms of distant metastasis, here's what it means in terms of it being a response to therapy. In the case of the early days now it's, do you need testosterone suppression or not?

And when you're intermediate-risk, and that's a big decision. And many of our colleagues urologists, radiation oncologists, med-oncs, are rather reflexive and haven't even really thought much about this. And you have some patients who say, "Well, I'm just going to do what the doctors tell me." And then there are others who say like, "No, there's no way I want to have my testosterone suppressed." And others who are like, "Oh, yeah, I want to do it." So it's really good to have a report that's extremely interpretable. So what I do is, I explain the report and then I give everybody the report to have them read it, it's very concise, while I'm in the clinic. So I actually find it very, very, very valuable.

Zach Klaassen: Alicia?

Alicia Morgans: I would say the same. And I think the word I would use is grateful for the information, and eager to hear ways that they can continue to have that collaboration and understanding. And when my patients are affected by prostate cancer, they just want to have all the information that they can to make the smartest decision possible, and they want that information to be as precise as possible. And I feel really confident in these reports and in the data. And I think the patients do too, because the physicians feel confident. So having that ability to communicate, and then also of course, having the patient understand the data, which was presented in a really clear way, I think is really valuable for everyone involved.

Zach Klaassen: Yeah. Dan, how about from a radiation standpoint, just the patient feedback?

Daniel Spratt: Yeah. I think echoing what's been said by Neal and Alicia is look, I mean, patients need to know. I say this a lot, my patients hear me say it. The Gleason score, it's probably about 60% accurate. And it is very uncomfortable when you first tell a patient that because they may have thought, we've got it all figured out. We know exactly the cancer you have, exactly how to treat it. And when I say, "Look, you may not even need treatment, even though you have Gleason seven disease." We just really don't know. Gleason is highly subjective, with high inter-reader variability. One of the points that Neal's talked about before, you send this slide 10 times for analysis, it's the same score every time. Versus sending it to 10 pathologists, you will get a different reading. It gives them information, gives me information to give them, again, that absolute risk estimate and have that decision. And I know I personally would want it, and I think my patients want it. So yeah, it's one more tool to the arsenal to try to focus on quality as well as quantity of life to help these patients.

Zach Klaassen: Yeah. Great points. So a great discussion on this first section, and we're going to move into the second indication looking at NCCN intermediate risk, and some really interesting and important data here.

So this is the second half of that new table in the 2024 NCCN guidelines, looking at biomarker positive for NCCN intermediate risk, and really delineating who should be treated with radiotherapy alone, or radiotherapy plus short-term ADT.

And so, the evidence for this came from Dan in 2023, again looking at four RTOG trials with eight or more years of follow-up. There were some nuances in terms of what trials they used, but essentially had a developmental model and a validation model with NRG/RTOG 9408. This was men with low, intermediate, or high-risk prostate cancer, with radiotherapy treatment plus or minus four months of ADT. Primary objective here was to develop a validated and AI-based prognostic or predictive model, that could identify a differential benefit from the addition of short-term ADT to radiotherapy in localized prostate cancer.

The primary endpoint was time to distant metastases. Key secondary endpoints included prostate mortality, metastasis-free survival, and overall survival.

The next two slides really do highlight what this model can do in this setting. So these are all for distant metastases. In the overall cohort, we see a differential benefit for radiotherapy alone, versus radiotherapy with short-term ADT, favoring short-term ADT along with radiotherapy. Where it really gets interesting is, in the predictive model, if you were positive biomarker, you had a significantly improved distant metastasis rate for radiotherapy plus short-term ADT. And importantly as well, if you were biomarker-negative, there was no additional benefit to adding ADT to your radiotherapy. And I think what's important here with the sample size, you can see here that roughly almost two-thirds of patients based on this analysis would not benefit from short-term ADT.

And just to emphasize again, this is prostate cancer-specific mortality, with very similar-looking shapes of these curves. Again, a benefit in the overall cohort. A significant benefit for adding short-term ADT if you're biomarker-positive. But no benefit to prostate cancer-specific mortality, if you are biomarker-negative.

This is a nice plot to sort of just summarize exactly what we just looked at. We can see, again, to highlight biomarker-positive, you have a benefit for distant metastases and prostate cancer-specific survival if you have short-term ADT added to your radiotherapy, but no benefit to the addition of short-term ADT to your radiotherapy if you're biomarker-negative for both of these endpoints. The synthesis statement from the NCCN for this section, "Patients with intermediate-risk prostate cancer planning to receive radiotherapy, those with biomarker-positive disease, and especially those with unfavorable intermediate-risk disease, should be recommended for the addition of short-term ADT regardless of radiotherapy dose or type, notwithstanding contraindications to ADT." And importantly, "Those with biomarker-negative tumors, especially tumors with more favorable prognostic risk, may consider the use of radiotherapy alone."

And so, I'll bring back the panel again to discuss these next couple of points. And so the first point here, and I'll start with Dan, because you were the lead author on this paper. How does this information translate to counseling these patients in the clinic, specifically with this indication?

Daniel Spratt: Yeah. So this is a huge, huge team effort. So I'm extremely honored to be listed first, but it was so many people's work that made this possible.

I think this is kind of an uncomfortable space right now for many of us, and it's because in prostate cancer we don't basically have real predictive biomarkers. So in other diseases, whether it's EGFR, or ER positivity, or certain gene expression signatures in breast cancer, they have a test, and they rely on that test to tell them what to do. We like to say, "This is how I've been practicing. I'm used to giving hormones. It's intermediate risk." To now have a test to say, "Well, it's an intermediate risk, but you don't need it." It's kind of like, "Oh my god, can I trust this test?"

Well, I mean, I didn't make the Simon criteria, some brilliant people did many years ago, and this is, other than doing the equivalent of what's called the TAILORx trial, which will take decades, and obviously, more data is always better. This is about as high-level data we can, of course, continue to validate, which is always fantastic. But this is validated in an independent trial. We know if back in your plots of distant mets, the number needed to treat was about 25 in all comers. That's telling you 24 of your 25 guys you're treating do not benefit from hormone therapy. We know the majority are not benefiting. But yet, we have this uncomfortability with now saying, "Oh, you're biomarker-negative." Two-thirds of them don't give ADT. But it should be pretty obvious, the majority of our patients are not benefiting in this specific intermediate risk context.

So I think it has huge implications, and I think it's, again, one more tool to the arsenal that I think patients greatly appreciate, knowing no data is perfect. And I'm sure there will be continual refinements as more data comes in. Like in breast cancer with the oncotype test, how they've now further refined this sort of intermediate risk category to find better cut points to who benefits from chemotherapy. I assume that'll continue to be done in prostate cancer.

Zach Klaassen: Yeah, absolutely. And I think, when I first saw this paper, I couldn't decide what I was more excited about. The fact that you've got a biomarker-positive patient that clearly benefits from ADT, and then a biomarker-negative patient, which is two-thirds, like you said, that has really no benefit. So Neal, as a urologist, when you're getting ready to send this gentleman to Dan, how are you able to use this information to already sort of set the stage for what the conversation may be like with a radiation oncologist?

Neal Shore: Yeah. I think that number one, it's important for our colleagues who are listening who may be very new to this entire concept. I love the evolution of the word biomarker. We've talked about biomarkers that are blood-based, plasma, serum, tissue-based, urine-based. Urologists, we like urine-based biomarkers as any opportunity to say the word urine and we're happy.

But we also have imaging as a biomarker, and now we have AI as a biomarker. And that's really, really important to not lose sight of that. That this AI is now a legitimate very level one evidence biomarker. And for my patients who have high-risk, intermediate-risk localized prostate cancer, they all see a surgical oncologist, they all see a radiation oncologist. And we give them literature to read through. And part of doing that, they're going to say, "Okay, I see that there is data that says I should have androgen deprivation therapy." I immediately say, "Yes, but now we have this other important AI biomarker that's going to help better inform us. Mainly you and the radiation oncologist, as to whether or not you do or do not need androgen deprivation therapy."

And as we all know, and everyone listening, there's no free lunch with androgen deprivation therapy. We want to make sure that we invoke it and encourage patients that will see the benefits in distant metastasis and biochemical relapse declines, and of course, mortality and overall survival. So that's very incredibly refreshing, to be able to tell a patient, "I'm very comfortable that I don't want you to have androgen deprivation therapy." And of course, there's all these various economic issues regarding doing that in our healthcare system. But I think it's really, when you can look at somebody eye to eye and you're knee to knee in the clinic and say, "Look, I'm very comfortable. We're going to just go with the radiation therapy alone." That is a very positive experience for the patient.

Zach Klaassen: Yep. It's a great segue to my next question. I'm going to bring Alicia in for this one, because nothing's free when it's happening to you, right? Six months ADT might seem like nothing to us because we prescribe it all the time. But we know there's a cumulative dose effect that leads to not just metabolic issues, but dementia, depression, all these things that this patient may need lifelong ADT going forward in three or four years. So Alicia, what does this mean from a survivorship standpoint when we're talking about not having to start ADT at this point? Maybe hopefully never needing it, but if we do need it down the road, you're not setting that cumulative clock to start quite yet.

Alicia Morgans: The way I think about this may be a little differently than you're phrasing it, and perhaps it's just my interpretation. I think that this allows us to give ADT in situations where we might be able to prevent lifelong ADT, because that it's a major change in life, major. It's just a lot for someone to take on. It is the right thing to do, and we definitely prolong life and maintain quality of life and improve quality of life when we need to. But if we can prevent lifelong ADT by giving short-term ADT, we want to do it.

It is also important for us to avoid ADT at all costs if we can do that. And I think it's interesting, because as a medical oncologist, obviously, I have given many things that are more toxic, or feel more toxic, than hormonal deprivation therapy, or androgen deprivation therapy. However, we can set into motion with ADT effects that are cardiometabolic and physiologic, and changes in body morphology even, that can affect a person physically and mentally for the rest of their lives. And if we do not need to do that, or if we can keep that to a short six months, then it's really, really of our best interest.

ADT causes changes in physical composition, certainly loss of muscle mass, gain of mass around the middle. Patients lose their ability or drive sometimes to kind of get up, go and have that physical function. They feel fatigued, they're mentally tired, they're physically tired, and then there are all of the cardiometabolic changes that happen as well.

Importantly too, patients don't always recover testosterone production, especially if they're older. And so sometimes, and I have had this even in modern days, where I've tried a certain duration of therapy, even in therapies that I think are going to be more rapidly reversible, they don't always reverse. And then I have patients who are trying to pull themselves back from something that I thought was short-term, but ends up being long-term. So any way that we can select patients, reduce the number of patients exposed, prevent long-term ADT when we can use short-term, and really ensure that we're kind of matching that right treatment for a disease control standpoint with the quality of life and physiologic standard of life that somebody is used to is incredibly, incredibly important.

Zach Klaassen: That's very well said.

Daniel Spratt: Zach, one of the things that I tell my patients usually about radiation, about surgery, about hormone therapy is, if there were no side effects, I have my two kids, I would just radiate my prostate tomorrow. I'd take hormone therapy. I'd just get done with it. But I don't care what anyone will tell you, there are potential side effects, and there will be some degree of side effects. I think overtreatment in general, it isn't just about hormone therapy, I think we need predictive biomarkers for everything. I think that's the concept of active surveillance, even of treatment in general has its own morbidity, and if we can spare them in that. This will be, as I said, I think a very uncomfortable feeling for many providers, because it is a test guiding them what to do, which in many other cancers, they've been using this for a long time. PD-L1 expression, are you going to give it? Not give IO? But we are going to have to move past the sort of gut instinct medicine and start using technologies that are validated. And I think this is just the beginning.

Alicia Morgans: Can I just add one thing here too?

Zach Klaassen: Yep.

Alicia Morgans: The favorable/unfavorable intermediate risk distinction has been an analysis that's been more post-hoc on some studies. I actually feel more comfortable with the Artera platform, and helping me understand the distinction between whether or not I should use hormonal therapy or not, than I do with that analysis. And I know I have colleagues who are on today who have worked in that analysis, and I appreciate that analysis, and use it every day in my clinic. But I appreciate too, the Artera distinction, and think of it just as strongly, if not more so, than those analyses.

Daniel Spratt: That paper was another Zach, Zach Zumsteg, and he was first author, and I was a resident at Sloan Kettering with Michael Zelefsky as the senior author. We made that as residents, never thinking anyone would adopt this in the national guidelines from a single institution retrospective study. So just to give you an idea of the origins of that.

Zach Klaassen: So the residents listening out there, you can make guidelines, specific studies currently, which is great. So great discussion. Thank you, everybody.

We're going to move to our last section, which is, as Dr. Shore mentioned earlier, hot off the press. This is post-radical prostatectomy biochemical recurrence. This was presented by Todd Morgan at the AUA in San Antonio, Texas. And again, we're using RTOG studies for this. This was two RCTs in the post-RP biochemical recurrence setting. With all of these trials, excellent follow-up for this one, 9.3 years, split between a training dataset of 70% and a validation dataset of 30% of the patients. And what we can see here, this is the distant metastases data. When we cut this between 40% high-risk and 60% low-risk, we see an estimated 10-year cumulative incidence of distant metastases of 27% for high-risk patients versus 8.7% for low-risk patients.
And we can see here that again, with some of these other outcomes, including distant metastases, all the way down through overall survival, we see the ability of this ArteraAI prostate test to differentiate for high versus low-risk patients.

The really two key take homes from this study are the following ones. This is the high-risk group. And when we look at the addition of hormone therapy to no hormone therapy for radiotherapy in the salvage setting, we see a 10-year cumulative risk of metastases of radiotherapy alone of 41%, versus 18% for radiotherapy plus hormone therapy. So in these biomarker-positive patients, a clear benefit to adding hormone therapy to the radiation.

And similarly, as we've seen previously in this discussion, if you are biomarker-negative, the 10-year risk of metastases for radiotherapy alone was 10% versus 8%. So no benefit to receiving hormone therapy to your salvage radiotherapy if you're biomarker-negative.

So some discussion points from this new data. This is not yet commercially available in this setting. But initial thoughts, I'll start with Dan for this one, based on the RP BCR data.

Daniel Spratt: So in the BCR setting, which is now a rapidly evolving setting, there are many ways we can risk stratify BCR, but in sort of the typical prototypic discussion of early salvage radiation, there have been four big trials, creating a lot of confusion of do we add systemic therapy? How long? Clearly in the localized setting, we showed all these ways we risk stratify even with clinical systems. In the post-op setting, we kind of historically say it's all BCR, when we know these are a very heterogeneous group. So to me, this is a very valuable tool to be able to personalize that risk estimate. And as you can see, even though it's trained to be prognostic, it clearly can differentiate patients that are more or less likely to have absolute benefit differences. So again, similar conversation to how that can help personalize therapy for patients.

Zach Klaassen: Neal, how about you? We both take out prostates, and you're sitting there, you've got a biochemical recurrence, not comfortable for you, not comfortable for the patient. How does this data sort of differentiate that conversation in lieu of sending to Dan versus what we have now?

Neal Shore: Well, I think it's very important. As you said, this is hot off the press, and really, I got an inkling of it for the first time. And when we were at APCCC and Dan presented this data with confidentiality, it was embargoed. And now we got to see it with Todd Morgan's presentation, literally less than a week ago at AUA 2024. I'm looking forward to reading the paper in greater detail, but I mean, I think the data that has been presented is extremely compelling for our patients who are biomarker-positive, who will benefit with early salvage radiation therapy and ADT, versus those who clearly will not. And for all the other reasons that we described, avoiding the obvious toxicities and the adverse reaction profile of unnecessary T suppression is key.

And then at the same time, providing folks the opportunity to get curative results, or ideally curative response, with early salvage radiation in combination with a short-term ADT.

So those are the goals. That's how we want to really optimize therapy. As you said, it's not yet commercially available. I look forward for it to become commercially available. And as Dan said, and I completely agree, we have so many, this field of BCR and how we think about it, not just with RT and RT plus minus ADT, but based upon EMBARK and that trial, and how that's now creating some really important educational opportunities to think about how to, with systemic therapy, ADT in combination with enzalutamide, or even enzalutamide monotherapy. This is clearly, it's not evolving, it's evolved.

And so, for all of our colleagues who are listening, read this paper that's going to get out. Certainly, look at that AUA presentation by Todd Morgan. And start to understand where AI is now, and where it's ultimately going to go to, in many other applications, clearly.

Zach Klaassen: So Alicia, these patients probably aren't coming to see you, but if I think about this disease space and it's really evolved, we now have guidelines based on the AUA that just came out of AUA as well. We've got AI based on this data, we've got genomic classifiers, PSMA PET scan. If we take this all together, we're really personalizing medicine for specifically these patients. Maybe you can comment on what all that means if we take it together.

Alicia Morgans: So thank you so much, Zach. So I actually do see a lot of these patients, and I'm fortunate too, because I think there are a lot of opportunities for us to help these patients as they're considering hormonal therapy in the context of salvage radiation. And certainly, there are a lot of clinical trials that are trying to answer questions in this space. And of course, we have data on high-risk biochemical recurrence, particularly after we've kind of completed a salvage radiation strategy as we think about intensified therapy. But in this particular context, I think this is going to be so important as we think about, of course, ongoing trials, but how we think about using hormonal therapy to intensify the effects of salvage radiation. If I can avoid ADT with salvage radiation, I will do it, and will appreciate doing it, and so will my patients in every single setting.

What I look forward to in future data, is understanding if I need to do six months or 24 months? Because there is data regarding that, or from the RADICALS-HD setting, where I am still thinking that two years, which is not something I want to prescribe, of ADT, is actually preferred versus six months, especially in patients with the highest risk disease. And so I really want to understand, can we use AI in the future to help make that decision? Because there is no setting in which I want to give someone two years of ADT when I can get away with none, or with a shorter course. So this is definitely a step in the right direction, and I look forward to future work to come.

Zach Klaassen: No, it's well said. And I think if you look at the goal of this discussion that we're having, as Dan mentioned, it's going to be some uncomfortable conversation, and it's educating based on what this data means, and hopefully this has been helpful.

In closing, I want to do the true round table, give everybody a minute or two to sort of summarize their thoughts. Maybe look for a couple of other areas in the prostate cancer landscape where ArteraAI may be beneficial. So I'll start with Neal.

Neal Shore: Well, thanks very much, Zach. Clearly, this is the now biomarker that is going to progress into many other opportunities, and not just in prostate cancer and GU-oncology, but clearly in bladder cancer and in renal cancer, and work is going forward in that. So I want all of our colleagues to be aware of that. I think the fact that we've seen this evolution, it's the rapidity of getting the result. There's obvious clear tissue preservation. And it's now being reimbursed.

And it continues to get better, I think, as Dan said. The algorithm just only gets better over time, and the sensitivity in the area under the curve continues to improve over time, as opposed to the forever art of interpretation of the radiologist and our pathologist. No matter how good they are, there's always going to be some discordance. But that is sort of the wonderment of artificial intelligence. We're now getting involved in some interesting bladder cancer trials to see who's a BCG responder or not, things of that nature. And I know that in all solid tumors, in breast and lung, there's just a panoply of trials going on. So I think it's really wonderful for all of us who take care of patients with cancer and in the field of oncology.

Zach Klaassen: Excellent. Alicia, you mentioned the difference in between hormone therapy, duration, any other areas, particularly in more advanced, where you could see us being very helpful for our patients?

Alicia Morgans: There are so many areas, Zach. I mean, we are trying to figure out whether we should use chemotherapy. Whether we should intensify in that way, whether we can get away with ADT and an AR signaling inhibitor. We would love to understand in some patients, if we can get away with ADT alone. Because right now it seems like we can't, but maybe we can. Are there patients where we can intensify in the short term, and then stop all treatment if we use multi-modality, metastasis-directed therapy, and have a better chance of curing our patients, even though they look like they have oligometastatic disease? I think this window is just broadening for where we can use something like this.

And I have to say, it is okay with me that I don't understand, as a human being as compared to a computer, all of the aspects of artificial intelligence. Because one of the benefits of using artificial intelligence is that it can interpret some of the shades of gray that I can't even see with my eyes. So I don't always care why it can tell me why one patient will benefit from this or from that. I just want to know for that patient sitting in front of me, that that's going to be helpful or not helpful, so I can give them the best answer. And there are so many areas where this can be useful.

Zach Klaassen: That's well said. Dan, you've been key in driving a lot of this with your team, and you've got the last word.

Daniel Spratt: Yeah. I think this is truly the beginning. It is revolutionary, the times we are in and where we are going. I think we're going to see that just as there are companion type efforts, whether it's in the diagnosis of cancer, whether it's in radiology imaging, I think we're going to find that these tools, as we can amass just never before seen amounts of data, to use kind of as Alicia would say, the being uncomfortable, not knowing what's necessarily is seen. It's human uninterpretability, right? And I don't think we understand humans. Why do we like someone? Why do I like certain food? Why do I like this? I don't think someone can give me a concrete biological mechanism to a lot of our nuanced traits about ourselves. And I think there's a lot on a pathology slide, a radiology image of these things that we cannot appreciate, but with enough data, it can be trained to help guide us.

And so I view these as, again, these are tools to help us and across the landscape. They've already started in the advanced setting, so they've looked in, I think Dr. Parker may have presented some work from the STAMPEDE using this, and I think it'll only expand. So very exciting times.

Zach Klaassen: That's great. Well, thank you all three of you for your time, your expertise. It truly is a great time to be in the GU oncology space. And thank you all again.