Artera: Personalizing Treatment Decisions in Prostate Cancer and Beyond With Waqas Haque, MD, MPH, and Trevor Royce, MD, MS, MPH, FASCO

In this latest episode of Exploring Artificial Intelligence in Oncology, Fellows Forum member Dr. Waqas Haque sits down with Dr. Trevor Royce, Senior Medical Director of Artera. Dr. Royce discusses the ArteraAI Prostate Test, how it can be used to help make personalized treatment decisions in prostate cancer, and how Artera is expanding this research further into the cancer space.

Waqas Haque, MD, MPH: Welcome everybody, I’m Waqas. I’m an Oncology Fellow at the University of Chicago. Welcome to our next episode of the Oncology Data Advisor AI podcast. We’re joined today by one of our special guests, Dr. Trevor Royce. Thank you for joining, Dr. Royce.

Trevor Royce, MD, MS, MPH, FASCO: Thanks so much for having me. It’s a pleasure to be here.

Dr. Haque: Full disclosure, I worked with Dr. Royce on an op-ed on antiemetics a few years ago that got published in the Journal of Clinical Oncology, but we’re going to talk about a different topic today—the role of AI in prostate cancer. Before we go on, Dr. Royce, can you give us a brief introduction about yourself?

Dr. Royce: Sure, and again, it’s great to be here. My name’s Trevor Royce. I’m a Radiation Oncologist based in Chapel Hill, North Carolina, and I serve as Senior Medical Director at Artera. I also have an academic appointment at the Wake Forest School of Medicine and see patients one day a week at the Veterans Affairs (VA) Hospital.

Dr. Haque: Thanks for sharing that. Just to get things started for everybody, why is it important to use AI-based methods for guiding treatment decisions in prostate cancer?

Dr. Royce: I think there’s such incredible potential with using AI-based methods for managing, diagnosing, and understanding prostate cancer. A big reason for that is there’s so much information that we use across the whole spectrum of clinical care that we can’t even begin to analyze and interpret, and AI gets us so much further in that in terms of the sheer amount of data that can be processed for useful purposes.

Dr. Haque: That makes sense. And I think in counseling patients, older men who are kind of debating if they should get a prostate-specific antigen (PSA) test or not and interpreting those values as they trend over time, there are definitely a lot of roles for more statistical analysis to help make those decisions. So, tell us a little bit more about how the ArteraAI Prostate Test works in terms of who it’s for, the turnaround time, and how widespread it is in clinical practice.

Dr. Royce: Fundamentally, the commercially available ArteraAI Prostate Test is a prognostic tool that uses digital pathology to provide prognostic information and insights for localized prostate cancer. That could be any risk of disease—low-, intermediate-, or high-risk prostate cancer. In addition, it provides a predictive signal for intermediate-risk men who are considering whether they should get androgen deprivation therapy (ADT) in addition to their radiation treatment. We have a prognostic test that encompasses the entirety of the risk spectrum for localized prostate cancer. Put one way, if you’re intermediate-risk, you have a predictive test that helps inform whether you would benefit from ADT. Put another way, you could be spared ADT if you don’t need it. This is an evidence-based approach to that.

The other thing I’ll say, just kind of stepping back into the big picture, is we use a prostate biopsy sample to run this test that’s created during the routine course of care. There’s no additional procedure, no additional staining. It’s a hematoxylin and eosin (H&E) stain slide that is then digitized with a digital scanner. Once you have that digital image, you can use that and incorporate some key clinical variables to have your model process those and have your output for your prognostic or predictive end points.

Dr. Haque: I think it’s really interesting to see that there’s both the prognostic and the predictive side to the test. That’s really cool to hear. Following along with that, I know your team was really busy at the American Society of Clinical Oncology (ASCO) this year, with several abstracts presented. Would you like to tell us briefly what was presented and what was the outcome of that work?

Dr. Royce: Our first foray into this exciting technology of using digital pathology and AI interpretation to provide insights has been in the prognostic prostate space for localized disease. We mentioned the predictive piece for certain therapies, in this case, ADT. Those two indications are in the National Comprehensive Cancer Network (NCCN) guidelines, so we’ve got a pretty solid evidence foundation and basis for those two tests, the prognostic one and the predictive one.

We’re always thinking, well, what can we focus on next that will be of greatest impact and help patients with cancer the most? At ASCO this year, we presented our first non-prostate model, which was a breast cancer prognostic model taking a similar approach with a similar AI architecture, H&E stain slides from early-stage breast cancer. We asked the question; can you have prognostic insights from that? So, that’s one key piece—showing how this approach can translate across diseases. We had all this evidence in prostate cancer, so can we take the same approach and rationale and do it in an entirely different cancer, breast cancer? And it turns out that it works pretty well.

This is a research-only model, so it’s not commercially available yet, and we’ve got a lot of work in terms of validation and evidence generation. But just to give you a hint of stuff coming down the pipeline, that’s one piece. The other thing is that in prostate cancer, we started in localized disease, but as you know, there have been so many new ways to look at prostate cancer in the last 10 years in terms of the spectrum of metastatic disease and acknowledging that metastatic disease is not one-size-fits-all. There’s a lot of variability in the natural history of metastatic prostate cancer. Obviously, as a radiation oncologist myself, I’m thinking about oligometastatic populations and whether we can have specific insights into that, or into castrate-resistant disease or castrate-sensitive disease and so forth.

So, there were two other abstracts that we presented at ASCO. One was a validation in an oligometastatic cohort pooling the STOMP and ORIOLE clinical trials, showing that the AI score is validated in that population. Also, in the CHAARTED trial, which was metastatic castrate-sensitive prostate cancer in first line, we had an abstract showing that the AI score prognosticate in that cohort as well. So, you can see how we’re pushing the boundaries of developing this technology. It started in localized prostate cancer, and now we’re seeing how it works in other cancers, in more advanced prostate cancer cases, and so forth.

Dr. Haque: It’s really cool to see the same fundamental approach and methodology with the data being applied to a range of different tumor types, especially with all the different targeted approvals in breast cancer over the past couple of years and the role for human epidermal growth factor receptor 2 (HER2)–low and now HER2–ultra low. I think it’s really cool that you all are working in that space. One thing that you mentioned was incorporation into the NCCN clinical practice guidelines. For people who might not be familiar, could you talk a little bit about the significance of that and Medicare coverage as well?

Dr. Royce: Those that live in academics and have the luxury of being able to focus their entire career and clinical practice on a very specific disease or setting of cancer may not really fully appreciate the rigor and value that goes into being in these guidelines, which then can guide cancer care nationally. Without going too much into the details, there’s a big panel of experts who evaluate the evidence and then decide whether the evidence supports a certain approach to cancer management. If it does, then they can incorporate that into their formal and official guidelines.

We’re proud to say that we’re the only predictive and prognostic AI-based biomarker and localized prostate cancer test that’s in the NCCN guidelines. I think that’s a big success story, not only for the development of an AI tool that’s actually practically useful and can be applied in clinic, but also for the effort that we expended in terms of the evidence generation to meet those bars of evidence required to make it into the guidelines. There were two foundational papers, without getting too much into the weeds. One was our foundational model for prognostic and localized prostate cancer, which was published in Nature Digital Medicine, and then our predictive model, which again predicted whether you would benefit from ADT or not, which was published in the New England Journal of Medicine Evidence. Those papers, in addition to other work we did, contributed to the body of evidence that made it acknowledged in the guidelines.

Dr. Haque: Got it. When we think about some of the arms of healthcare, one of course is research. The other is access. It’s really cool to see some of the focus on inequities that Artera has addressed. Of course, with prostate cancer, a lot of men, especially African American men, tend to present later with metastatic disease, which means they have worse outcomes. We also know that the ZIP code that you live in or are born in can affect your access to prostate cancer care. Uninsured men also tend to get fewer PSA tests, which means that they have less access to therapy and those sorts of things. So, talk a little bit about how Artera’s test can be applied to addressing some of these health inequities.

Dr. Royce: There are a lot of different ways to look at access and inequities in cancer care. You touched on one; before we dive into that one, I’m going to highlight two others. One is access to technology. An interesting thing about this approach to biomarker development is, as I mentioned, we use H&E stain slides created during the routine course of care. You don’t need another biopsy, you don’t need additional staining, and most men that are getting prostate cancer and have that diagnosis are going to have that available. It’s less resource-intense from that perspective.

The other is that the turnaround time for the test is quite quick compared with other risk classifiers that might be commercially available. Again, we’re not extracting genomic material from a sample, and we’re not destroying that sample. We’re literally just scanning it, digitizing it, and then running an algorithm on that already existing sample, so the turnaround time for the test is quite quick. Taking together those elements can lower the cost of the test and can increase the access of the test broadly, both in the US and beyond.

Then the final thing to think about is how it works in certain subgroups or populations that have historically been underrepresented, particularly in clinical research. We haven’t really spent much time talking about this, but the way we developed this test was using clinical trial datasets. We used the NRG Radiation Therapy Oncology Group (RTOG) clinical trial datasets that had tissue banks with the clinical outcomes from the trials and the clinical variables from the trials.

An amazing thing about the NRG and RTOG prostate cancer clinical trial databases is that they had a large representation of African American or Black men in those cohorts, almost 20% of the population of those trials. That’s much more than the less than 10% representation of Black men in clinical trials across the board, so credit to RTOG and NRG for enriching that cohort. A lot of this work was done at the University of California, San Francisco (UCSF), enriching that cohort with those populations that have historically been underrepresented. But the point is now that we’ve developed this test, not only using this enriched cohort, but we can also see how it performs specifically in that cohort to make sure that it’s an equitable test and that it provides personalized, precision oncology across the board. We presented this at ASCO as an oral abstract showing the performance of the models in these populations, effectively showing equal performance. That manuscript is forthcoming. So, that is a proud part of the development side that we hadn’t talked about before.

Dr. Haque: It’s so exciting to see. Especially over the last 10 to 15 years, we’ve seen terms like time toxicity and financial toxicity being incorporated into how we make decisions about care for patients with cancer, also for prostate cancer specifically. That’s really exciting to hear. As we wrap up the interview, just a couple other quick questions—what are some challenges that are more specific to modeling in prostate cancer for guiding treatment and understanding prognosis?

Dr. Royce: I think it’s both a luxury and a challenge in that prostate cancer is a very common diagnosis. The trials can be large, and you need a lot of data to build and develop these models. We were very fortunate to have these clinical trials of 5,000 to 6,000 patients pulled together that we could use to train and then develop and validate the model on. That is a nice thing about prostate cancer.

On the other side, prostate cancer is not a homogenous diagnosis. You can have men that have very low-risk disease and might not need any treatment at all, or you can have men that have very aggressive disease and need multimodality treatment like surgery, radiation, or hormone therapy. It makes perfect sense to try to develop a prognostic tool that can differentiate which men are in which group, optimize their therapy for that cohort, avoid overtreating patients who don’t need too much treatment, and avoid undertreating patients who need more treatment so that they can also achieve those good outcomes.

Dr. Haque: That makes a lot of sense. Final question, we’ve had a lot of guests over the past year for this AI and tech podcast. I think you might be our first radiation oncologist. We’ve had other medical oncologists, we’ve had some physicians by training, we’ve had some startup founders. One thing I like to ask everyone is, what’s your information diet? How do you keep up with things, especially in AI with all of the new things that are coming out there?

Dr. Royce: It’s tough. There’s such a torrent of interesting science, and the advances in AI are moving so fast that it really is hard to stay on top of things. I have very curated email digests that I get that kind of keep me up to date, not only on what’s in the medical literature in terms of peer review journals and so forth, but also in the lay press. I’m an avid consumer of media across the board, and I’ve found a nice balance that hopefully keeps me abreast of things without overdoing it. You could spend your whole life reading articles coming out, so you do have to be a little thoughtful about what you read.

Dr. Haque: For sure. And I think it’s important to kind of turn pen to paper by also doing some of the work in the space as well, which is what, of course, you’re doing.

Dr. Royce: I’ll also quickly add that podcasts just like this have really helped, because I can do it on my commute to clinic and back and listen to it on two-speed. That’s a whole new way to access media and insights that really didn’t exist 10 years ago, so I appreciate podcasts like this.

Dr. Haque: Definitely, for sure. Well, thanks so much for your time, Dr. Royce. I really appreciate your insights and sharing everything today.

Dr. Royce: It’s my pleasure. It was great to see you again, and I’m always happy to chat. Thanks for having me.

About Dr. Haque and Dr. Royce

Waqas Haque, MD, MPH, is a Hematology/Oncology Fellow at the University of Chicago. He recently completed his Internal Medicine Residency at New York University (NYU) in a Clinical Investigator Track. Dr. Haque has balanced his patient care work with a variety of research projects. During his fellowship training at the University of Chicago, he plans to further his work in innovative clinical trial design, value-based care delivery to cancer patients, and clinical investigation.

Trevor Royce, MD, MS, MPH, FASCO, is the Senior Medical Director at Artera. He is a practicing Radiation Oncologist with an academic appointment in the Department of Radiation Oncology at Wake Forest University School of Medicine in North Carolina. Previously, he worked at Flatiron Health, and he has more than 15 years of experience in clinical and health outcomes research. Dr. Royce is passionate about leveraging AI and technology to improve outcomes for patients with cancer.

Transcript edited for clarity. Any views expressed above are the speakers’ own and do not necessarily reflect those of the Fellows Forum.