Leslie Ballas: Hi. I'm Leslie Ballas, a radiation oncologist at Cedars-Sinai in Los Angeles, and I am honored to be joined by Dr. Timothy Showalter, who is the chief medical officer of Artera AI and a professor of Radiation Oncology at the University of Virginia. Tim, thank you so much for joining today.

Timothy Showalter: Leslie, it's great to be here. So great to see you. And thanks for inviting me to talk to you today.

Leslie Ballas: Let's start by having you explain to our viewers, what is Artera AI or Artera MMAI?

Timothy Showalter: It's a great question. So Artera is the company that was founded about four years ago, Felix Feng, who's maybe well-known to a lot of members of the audience, was one of the two co-founders. And the mission was really to use computer vision, artificial intelligence, and digital pathology to create really strong-performing tools that can provide prognostic information for patients and predictive insights that can help personalize cancer therapy.

Our first test is for prostate cancer. It's called the Artera AI prostate test. And it's a risk stratification and predictive tool that is based upon a patient's biopsy slide. So we actually run the test on an H&E-stained biopsy slide, digitize that slide, and run an algorithm that provides personalized information for that patient's likelihood of having a metastasis at 10 years, as well as predictive information for whether that patient is likely to benefit from hormone therapy.

And clinicians are using this now for prostate cancer patients who've been diagnosed with prostate cancer to help make risk-based treatment recommendations for how to approach their prostate cancer.

Leslie Ballas: And so I know that it's referred to as a multimodal artificial intelligence-based test. What is the multimodality of it?

Timothy Showalter: So the multimodality refers to our test incorporating both information from the digitized pathology image. So we have computer vision, machine-learning model that is analyzing features that are seen on a high-resolution image of the biopsy slide, as well as clinical factors.

So our test includes patient age, tumor stage, and PSA level. And all of that information goes in together to generate the Artera MMAI, or multimodal AI score, which is designed to risk stratify patients and predict distant metastasis, as well as prostate cancer-specific mortality.

Leslie Ballas: And how robust is the data that comes out or is on those printouts for each specific patient?

Timothy Showalter: Yeah. It's pretty remarkably robust. And as a clinician who came to this work several years ago, I am still amazed at how well this performs. So I'll tell you that we have validated this repeatedly. It was initially developed and validated on a set of five randomized controlled trials from the RTOG group, now part of NRG Oncology. And then we've gone on to validate performance of this test in several more RTOG radiation therapy trials, as well as data sets for patients who were managed with radical prostatectomy.

So the ones that are in print right now are the PLCO and the UPCA cohorts with Anders Bjartell. Those are available in PubMed. And then we've expanded even more. So the test performs consistently even into metastatic prostate cancer. So we've shared results publicly for validation in STAMPEDE and CHAARTED.

And what's really amazing is that this is based on the original prostate biopsy sample, even for these metastatic patients, years later, that signal carries through. In terms of the overall precision or accuracy of the test, we generally think about for tests like this as being measured based on an area-under-the-curve approach, where one is perfect prediction, down to the day of when that patient is likely to have a recurrence.

And 0.5 would essentially be a coin toss. And the Artera test generally, depending on the patient population, you'll see AUCs for that cohort for predictions that are in the high 0.7s, 0.8s type range. And that's pretty similar to what we see with the genomic risk classifier-type approaches. So it's in the general ballpark. And it's pretty amazing that it's working ultimately directly with a digitized whole-slide image.

Leslie Ballas: That's interesting. One thing that you said is that even in the metastatic patient population, you're going off of the biopsy specimen. And so I'm always curious. If I were to give a biopsy specimen, send it to Artera AI, and then the patient develops mets and we happen to biopsy one of the mets to document it, we don't need to resend to Artera, or you do want that resend.

Timothy Showalter: Well, I should first clarify that our test that's often recommended by national guidelines and commercially available is for localized prostate cancer. So I primarily mentioned the information about the metastatic disease because I think it's so interesting from a biological perspective that the test performs that well.

But to your point, Artera is in the process of making, because we validated so well in metastatic prostate cancer, we are in the process of, later this year, making our test available for patients who have metastatic prostate cancer. And for that test, there is no need to biopsy metastasis. We're actually still working with specimens from their original prostate biopsy years before.

The first time we extended our validation of our test into later-stage disease and published it was with the SPARTAN trial, which was a collaboration with Johnson & Johnson. And in that trial, those are actually patients with castrate-refractory prostate cancer. And we showed that working with their original prostate specimen generally years later after their original diagnosis, the test not only still predicts for prostate cancer-specific mortality and time to next recurrence, but also the prognostic score was useful as a predictive tool to identify patients who are likely to benefit most from apalutamide.

Leslie Ballas: That's really amazing. You hit on a topic that I really wanted to discuss with you, which is the difference between prognostic and predictive. I know that Artera is predictive in the intermediate-risk prostate cancer space. Can you tell me a little bit about—well, one, inform our listeners what is the difference really between prognostic and predictive, and where does this test fit along that spectrum within prostate cancer generally?

Timothy Showalter: So yeah. We spend a lot of time thinking about prognostic versus predictive. I think in terms of how clinicians have approached risk stratification, I think we're all used to using prognostic tools for the most part, and thinking about a risk-based treatment decision. And we all do that in the clinic when we're seeing a patient where we're basically using clinical factors to think about what's the right treatment for them.

The statistical definition of predictive essentially is, is the biomarker or test result independently associated with the response to that treatment? So we are always looking to have a predictive claim or to say that this test is predictive. We restrict that using that sort of terminology to instances where there's a significant interaction p-value.

So where we're basically running a statistical test to show, yes, it is the status of that test result that can be used to identify patients who are likely to benefit from that therapy. So we have an algorithm that's part of our commercially available test right now that's been validated for NCCN intermediate risk, so including favorable and unfavorable intermediate-risk patients, that does have the performance characteristic of that score result can predict for benefit from short-term ADT.

That algorithm was developed specifically for that indication, so it was developed with a predictive application in mind, and it meets the statistical definition of a predictive tool. There are also other situations where the Artera prognostic score, so the main risk stratification test which was developed with a prognostic purpose in mind, so to predict for likelihood of distant metastasis, has been shown to have a predictive performance, meeting the same statistical test.

So it's significant interaction p-value. And a couple examples of those cases are I mentioned the instance with apalutamide in the castrate-refractory prostate cancer. We've also worked with investigators including Phil Sutera and Phuoc Tran, who you know, where we've looked at the performance of our test in an oligometastatic population and shown that the Artera prognostic score performs as a predictive tool for benefit from metastasis-directed therapy.

And then we're really excited that at this year's ASCO annual meeting, there's another application of our core prognostic test for a predictive purpose. And that's for NCCN high and very high-risk non-metastatic patients predicting which patients are likely to benefit most from adding abiraterone to ADT. And that's something that we're planning to make available right after the ASCO meeting. Really excited about that. And that's going to extend the clinical utility of our tests to men with higher-risk prostate cancer.

Leslie Ballas: Was that validated on the STAMPEDE data set?

Timothy Showalter: That was. Yes. So the STAMPEDE investigators will be presenting that at ASCO. And our validation data set includes patients with node-negative high-risk prostate cancer, as well as node-positive but all in the M0 population.

Leslie Ballas: That's really exciting. Yeah. So you mentioned that that test would be commercially available for that particular population after ASCO. Tell me a little bit. How is Artera used commercially right now versus in clinical trials?

Timothy Showalter: So right now, commercially, our test is orderable and indicated for patients who have been diagnosed with prostate cancer who are being evaluated for primary management, be that active surveillance all the way through surgery, and radiation with hormone therapy. And it's indicated across all risk groups of localized prostate cancer, so your very-low through very-high risk.

Our prognostic performance is validated across all those risk groups. And so the same risk stratification that folks may be used to with the tissue-based genomics test, you can get with the Artera test across all risk groups. We also, on top of that, have the predictive test result for benefit from short-term hormone therapy added to radiation, which is restricted to patients with NCCN intermediate-risk prostate cancer.

And so we find that most of our orders are coming from the low-risk but maybe like higher PSA or more biopsy cores positive all the way through NCCN high-risk with a real concentration around NCCN intermediate risk. And depending on different practice patterns, you're seeing clinicians use it in different ways.

We do have an active surveillance-directed box of information on our test report that's based on our prognostic algorithm. And so a lot of urology practices are using that to help inform active surveillance decisions. And then we're seeing a lot of folks use it for hormone therapy decisions in the intermediate-risk space, as well as treatment intensification or deintensification decisions in the high-risk space as well.

Leslie Ballas: What about in the post-surgical space? There's, obviously, use for genomic biomarkers there. Is Artera useful in that setting as well?

Timothy Showalter: Yes. So we're really excited that we have validated a post-prostatectomy test. So it basically was developed and validated initially in two trials through a partnership with RTOG and presented by Todd Morgan at AUA 2024. And so that's indicated for patients who have had a PSA failure after prostatectomy, and it predicts for distant metastasis in that group.

There was a signal that there's some trend towards identifying which patients are likely to benefit from hormone therapy, but we've still got some more work to do in that space. That is now currently being offered through an early access program. So we're moving towards making that broadly commercially available. It's certainly available for ordering if folks are interested in reaching out to Artera.

But I think what we're really happy to be working with right now is the PROSTATE-IQ trial. So this test is being used for all patients enrolled on an investigator-initiated trial that's being led by Karen Hoffman and Paul Nguyen that is being managed by the MD Anderson Cancer Center. This is a study that is ultimately sponsored by Johnson & Johnson, and is using the Artera post-prostatectomy test to help stratify patients who are interested in that trial to high versus low-risk groups.

And then there are randomizations for intensification for the high-risk group and for deintensification for the low-risk group. So I think that's a real area where we'll see strong impact over time.

Leslie Ballas: So I think the question that I have as a clinician is, when I have a patient with intact prostate cancer, and I have multiple different tools that I can order to personalize their care, whether that's one of multiple genomic assays or Artera, I'm faced with, which should I order? What is best, if you want to say that? Which is giving me the most important and useful information for the patient sitting in front of me? How do you answer that?

Timothy Showalter: Well, I think that's a hard question, because there are several options available. And there's been a lot of work in this space. And I think as clinicians, we're lucky to have so many tools. And it's not just these tissue-based tools. We also have great imaging studies available for us now. And so the question is how do you use this information.

I think that any of the—there are three tissue-based tests available, and there's the Artera test as the only digital-pathology AI test recommended on guidelines. And so you've really got four options for this sort of thing. I think in general, these tests each provide effective risk stratification. I'm very confident in the level of clinical validation for the Artera test, just based on the number of clinical-trial data sets that we've been validated in, and we take that very seriously.

I think one point of distinction to think about is the experience of getting the test result for your patient. I think one advantage of an AI-based approach like Artera is that the turnaround time can be really fast. And so, generally speaking, from the time we receive the specimen in our laboratory, we're turning around results within a day.

And so what that can mean for your clinical practice is that when you're seeing a patient—I have a big clinic day on Mondays, and if I were to order this test for my patient, I know by the following Monday I'm going to have results back. And that's just not been something that we've experienced with the genomic test. And it can work with a small amount of tissue.

So generally, we're receiving clinical specimens from pathology labs. They just send us one slide that represents the highest-grade tumor. And so it's easier for them. We get the results quickly, and we have a high confidence that we'll get an actual analyzable result. So I think that's one difference between the AI test versus the tissue-based test if you're thinking about how that fits in with your practice.

The other thing I would say is that for the specific question of which patients benefit from hormone therapy, or pretty soon, which patients benefit most from abiraterone, I think that's a real point of distinction for the Artera test. And then otherwise, more broadly, I think it's sort of like, is your practice, and what are you comfortable with, and how is the information presented? And so certainly, people are going to see value in one test versus the other and be more used to it. So yeah. That's a non-answer to your question, but—

Leslie Ballas: It's a hard question. Are you guys being made aware of or are people looking at discordance between these tests?

Timothy Showalter: Yeah. So we're very aware of that, and we receive questions about that all the time. And it is true that none of these tests were co-developed to have any concordance with each other. And I've had the most experience, and we've even had some research published with Decipher, where some investigators—again, this is the Phil Sutera and Phuoc Tran group—have looked at a patient population where our test has been validated as well as the genomic classifier.

And what's interesting is that both tests appear to work and both these tests are well-validated tools. But we see a relatively high rate of discordance between the two. And so I think it's really challenging to use these tests together. And I think it's also important for clinicians to always interpret these results in the context of what the clinical risk factors are as well.

So these tests don't remove what is this patient's Gleason score and PSA, and the test doesn't remove what are the imaging findings or other considerations. So always think about those results—these are meant to inform or supplement your risk assessment. But yeah. I don't have an answer right now for how to use the two-test results side by side.

But I would caution listeners to lean towards one or the other rather than to try to use them both together, because I've certainly heard from my clinician colleagues that it can lead to some challenging dilemmas when you've got one test showing one thing and one showing the other. I think usually, people choose the answer they prefer for that particular patient.

Leslie Ballas: Of course. Well, it sounds like you guys have an incredible product and a lot of exciting things on the horizon. Thank you so much for taking the time to educate me and our listeners about Artera and its usefulness. I just really want to say thank you.

Timothy Showalter: Well, thanks so much, Leslie. It's been great to talk through this with you.