Evolution of Active Surveillance - YouTube

ASPI - Evolution of Active Surveillance Aug 26, 2024

Introduction

“Evolution of Active Surveillance" features LeonardMarks, MD, professor of urology at David Geffen School of Medicine at UCLA, whose database of 5,000 patients was used to develop Avenda Health’s Unfold AIsystem. Marks said the system may be able to predict the durability of active surveillance using an AI map of prostate cancer to determine how big a lesion is. Marks says, “smaller is better” in terms of lesion size and durability of Active Surveillance as a management technique.

Also featured is Scientific Director, Meghan Tierney from Artera, the developer of cancer diagnostics. Meghan is a Scientific Director with Artera. She is an RN and holds a PhD in nursing from the University of Minnesota. Her background spans oncology clinical practice and clinical research with a focus on improving outcomes and care experience for patients with cancer and their care partners.

Her research experience includes hospital-based clinical trial operations, patient outcomes research, real-world data research, and biomarker development. She is part of Artera’s clinical development team, focused on research collaborations and the development and validation of digital histopathology-based multimodal AI biomarkers.

Also from Artera will be Medical Science Liaison, MiaLi-Burton. Earlier this year Artera released its Artera AI Prostate Test, included in the National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines for Prostate Cancer, that helps men going on radiation therapy determine whether adding hormone therapy will be beneficial. This month, Artera released an updated version of its test to “include personalized insights on a patient’s relative risk of their prostate cancer showing more aggressive features, helping to inform the decision between active surveillance and active treatment in patients with lower risk prostate cancer.”

Summary of Chapters, Key Moments

14:07 Dr. Marks

· Discussed the evolution of active surveillance (AS) since the 1990s

· Highlighted changes in diagnostic criteria and follow-up methods

· Presented data from UCLA's ASCAP program, showing improved outcomes with MRI-guided biopsies

· Introduced the concept of using AI to map prostate cancers

· Discussed the potential of focal therapy to extend AS

· Emphasized the need for long-term follow-up and clinical trials

33:31 Meghan Tierney, Mia Li-Burton from Artera

Mia Lee Burton (Artera):

Introduced Artera's AI prostate test
Explained the concept of multimodal AI (MMEI) algorithm
Described how the test incorporates clinical parameters and histopathology imaging
Outlined the prognostic and predictive biomarkers provided by the test
Explained the process of analyzing biopsy images using AI

Megan Tierney (Artera):

Detailed the clinical application of Artera's AI prostate test
Explained how the test compares to NCCN risk groups
Presented validation studies for the test's performance
Introduced new active surveillance insights provided by the test
Discussed how the test results are presented and interpreted
Explained the test's ability to predict benefit from ADT in intermediate-risk patients

1:01:09 Q & A

Here's a summary of the Q&A period, with each question and a summary answer:

1. Q: Where do we stand now with AI in prostate cancer management?

A: Dr. Marks praised the advancements but cautioned that it's still early. The Artera team emphasized their rigorous approach and ongoing research.

2. Q: Is focal therapy considered part of active surveillance?

A: Dr. Marks explained that while not pure AS, patients treated with focal therapy are followed similarly to AS patients.

3. Q: Who is a candidate for focal therapy?

A: Candidates typically have less than half a lobe involved, usually with Gleason 3+4 or 4+3, and are not Gleason 6 patients.

4. Q: How does a doctor order the Artera test?

A: Doctors can contact Artera directly. The company coordinates with pathology labs to obtain and process slides.

5. Q: Does the Artera algorithm differentiate between targeted and systematic biopsies?

A: Currently, it does not, but this is a potential area for future research.

6. Q: What happens if a clinician's judgment differs from the AI test results?

A: The Artera team emphasized that their test is one tool among many and should be considered alongside other clinical factors.

7. Q: How does the Artera test compare to other biomarkers like Decipher or Prolaris?

A: There's currently no head-to-head comparison data available.

8. Q: Can the test use PSMA PET scan information instead of a biopsy?

A: Currently, Artera's test requires histopathology from a biopsy.

9. Q: How does tumor location (transitional vs. peripheral zone) affect prognosis?

A: While location can be a factor, Dr. Marks emphasized that Gleason score and other predictors are more important.

10. Q: How reliable are MRIs for patients with metal hips?

A: Newer hip implants are often MRI-compatible, and adjustments can be made to the MRI strength for good results.

11. Q: How is PSA density interpreted in active surveillance?

A: Dr. Marks explained the relationship between PSA, prostate volume, and PSA density, emphasizing the importance of considering multiple factors.

12. Q: How long do biopsy slides remain viable for testing?

A: The Artera team reported success with slides from the early 1990s, but noted that slide retention policies vary by institution.

13. Q: How does Artera ensure long-term validity of their 10-year predictions?

A: They use historical data and continuously update their models with more recent data to maintain accuracy.

14. Q: Is the Artera test FDA approved?

A: It's not FDA approved, as it's a laboratory-developed test, which is common for this type of diagnostic.

15. Q: Can individuals volunteer their slides to help improve the AI?

A: While there's no direct process for this, patients can inquire with their doctors about participating in research.

Edited Transcript

Bill Manning: Welcome everyone, and thank you for joining us today. My name is Bill Manning, and I'm the executive director of ASPI. A quick footnote: I've been on active surveillance since 2009. Our purpose here is to educate individuals about options regarding AS, which is not a treatment but a management plan. I'd like to remind everyone that we are not anti-treatment, but anti-unnecessary treatment.

A couple of quick housekeeping measures: We're going to hold off all questions until the last speaker. You can submit questions in the chat box, which we'll try to address after the last speaker. At that time, you can also click the raise hand button from the bottom of the Zoom window toolbar, and if there's time, we'll try to call on you. Please keep yourself muted during the meeting, and if you unmute yourself by mistake, we'll be happy to mute you.

The speakers presenting here today are generously donating their time, so we really want to be respectful of that. As a reminder, we are not here to dispense medical advice. Nothing said during these meetings by our speakers, ASPI, or the audience should take the place of consultations with your medical professionals.

You don't need to take notes. This meeting is being recorded and will be available in approximately a week from now on our website. You can also subscribe to our YouTube channel and automatically get notified when videos appear.

The last item is watching the meeting. If you'll notice in the upper right corner of your Zoom window, there's a button that says "View." When you click that, you can switch between gallery view, which is everyone, or focus on the speaker. We highly recommend selecting the speaker view.

Without further ado, we'll have a few quick words from Mark Lichty, our chairman. We'll then turn the meeting over to Howard Wolinsky, the organizer of this meeting. So relax, enjoy the meeting, and I hope you leave a little bit smarter. Take it away, Mark.

Mark Lichty: Thank you, Bill. I have really been looking forward to this meeting on AI and active surveillance. This is on the leading edge of future treatment, and it has the potential to really fold into what we're about, which is trying to help you all to avoid overtreatment. I'm just very excited to have this program, and Howard, I really want to thank you for having uncovered this and arranging to have these folks show up today. This was done on very short notice, and yet we've got over 350 people registered.

This whole AI arena is really going to not just disrupt the active surveillance world, but it's going to disrupt the medical science world. We have a big voice, and we need all of you. Let me just give you an example: Over a year ago, the NCCN made radical prostatectomies and radiation the primary focus for dealing with prostate cancer. They did not recognize active surveillance. We wrote to them, and they changed their guidelines. They listened, and we appreciated that. That's how powerful your voice is. It was not our voice alone that was responsible for the NCCN changing its guidelines, but I can assure you your voice was very important.

I too wanted to mention that PCRI has a conference coming up in LA in about two weeks. We'll be there, we'll have a table there. I've been there many times. I won't be there this time, but I strongly recommend that conference. It was at PCRI where ASPI was born. It was Mark Scholz's leadership on the issue of AS that really prompted us to create this organization, and we were the first active surveillance organization in the world at this point in time.

Mark Lichty: I did want to mention too that, well, you probably recall in my case, I've been on active surveillance for 20 years. I barely even think about prostate cancer at all. It does not affect the quality of my life at all. Part of what I speak of here is that all of you can reach a place of peaceful vigilance where you're not worried, where you're not getting a PSA and worried about a PSA or worried about an MRI. You can get there. We've gotten there, I've gotten there, others of us have gotten there.

The other thing I wanted to mention too is that we had a founder, George, whose doctor on call is Mark Scholz. The neat thing is that Mark and George, they're willing to push the envelope. I don't think George would mind my saying, he has a PSA of 40. He's not getting intervention because he's 79. These decisions about treatment and whether or not you get treatment are very related to age. If you're 79, you might not be making the same decision that you made at 59 or whatever.

I did want to mention as well our Awards program. We have a wonderful Awards program, all online now. This was a special Awards program because it really presented a history of prostate cancer with Mike Scott speaking, with Freddie Hamdy and Jenny speaking on the ProtecT trial. You really need to watch these Awards because there's so much information there. The ProtecT trial was really what opened the door to establishing the validity. It said you're going to basically live just as long if you're on active surveillance as you are if you've been treated, and not have the side effects that you might otherwise have. So that was a huge trial, and we gave the award for that. The folks that were involved with that were very appreciative of receiving that award.

Anyway, let me introduce my good friend Howard Wolinsky, and he'll introduce the speakers today.

Howard Wolinsky: Well, thanks Mark. I'm one of the co-founders of ASPI and I'm also the editor of ActiveSurveillance.blog. ASPI's first medical adviser was Dr. Jerry Chodak. Back in 1994 in the New England Journal of Medicine, he suggested the idea of taking a more conservative approach to low-risk prostate cancer. Because of his efforts, other doctors picked up the challenge later in the 90s and developed what became active surveillance.

Mark a moment ago mentioned Awards, but he didn't mention that one of the awards is named after the late Dr. Chodak, and this year we gave it to Dr. Carroll of the University of California San Francisco. You might catch that online as well; it was very moving.

Back when Dr. Chodak was speaking up, there was a huge increase in the incidence of prostate cancer as some new technologies were catching on - PSA testing or screening, the biopsy gun, ultrasound. Chodak warned that prostate cancer was being overdiagnosed and overtreated, leading to many patients experiencing side effects that impacted the quality of their lives.

By the end of the century, Ballentine Carter and Carroll had developed active surveillance. Uptake was slow in the United States. In 2010, when I was diagnosed, only about 6% of us opted for surveillance. It's been growing; it's over half now, but we lag far behind Sweden and the United Kingdom.

Meanwhile, in my time on AS, technology has continued to change. I was among the first to have a prostate MRI. Then came genomic and genetic testing, biomarkers, fusion biopsies, and now AI.

Dr. Marks, who will be our first speaker today, will be taking on the evolution of AS. "It's not your father's AS" is his title. When Dr. Marks arrived at UCLA 16 years ago, he had two goals. The first was to start an active surveillance program, and we thank you, Dr. Marks, as do over a thousand patients who you have on active surveillance. Second, Dr. Marks wanted to improve the quality of biopsies. He has succeeded at both.

Howard Wolinsky: He has seen 5,000 patients since he joined UCLA. He also built a huge database that has resulted in the development of an AI program to map prostate cancers. This can guide not only surgeons but also radiation oncologists in offering their various treatments, including focal therapy, prostatectomies, and radiation treatment. His mapping software keeps these doctors on track, keeps them on the right trail, but he'll also explain today, I hope, how AS patients will benefit from this new technology, and there are a couple of applications that are coming. Dr. Marks is co-founder of Artera Health, which developed the Unfold AI mapping software.

Dr. Marks will be followed by Artera's Medical Science Liaison Mia Lee Burton and Megan Tierney, who is Scientific Director at Artera. Earlier this year, Artera released its AI prostate test, which is sparing 40 to 60% of men from the side effects of Androgen Deprivation Therapy. I was taking a glance as people were coming in, and they let in a guy I know from the Detroit area who actually benefited from it, so it might be nice to have him speak up if he's up to it.

Will men on AS benefit from undergoing AI testing? We'll see. I'll turn it over to Dr. Marks, and we'll follow with the speakers from Artera. So Dr. Marks, the floor is yours.

LeonardMarks, MD, professor of urology at David Geffen School of Medicine at UCLA

Dr. Marks: Thank you very much, Mark and Bill and Howard, for what you have done with this organization. I think it's extremely commendable. I think you fill a void. I congratulate you and thank you for asking me to speak today.

What I hope is not disappointing to you is that I think artificial intelligence is a rather small part of the evolution of active surveillance. It may be a big part in the future, but those words get going in the popular parlance and sometimes get ahead of the curve in terms of the evidence. So I'm eagerly looking forward to seeing what the Artera people have done. It looks like we don't really compete; we kind of have different missions here. I think there's going to be a lot more AI companies that come along, but this is very, very early. I think for any of your listeners to go to their doctor and say, "Are you using AI?" is way premature.

I titled this talk "Evolution of Active Surveillance" because a lot has changed since the beginning days. The subtitle is "It's Not Your Father's AS." This slogan, "It's not your father's," really came out of the automobile industry. This gentleman who's the founder of the Oldsmobile company, his company subsequently languished, and a bunch of Madison Avenue executives got together in the late '80s and dreamed up a new form of this car which would appeal more to younger people. They called it "It's not your father's Oldsmobile" and they introduced the Cutlass Supreme. That worked and it turned things around. Incidentally, the slogan has stuck around much longer than the car, but the concept is right because active surveillance is already more than a generation old. So this is not your father's active surveillance anymore.

Just to pay tribute again, Howard scooped me on this a little bit. I've followed this story since the very beginning, and to give tribute to these three pioneers: Jerry Chodak at the University of Chicago, Jonathan Epstein and Ballentine Carter at Hopkins, and Laurie Klotz up in Toronto. All got the ideas about the same time, and it's really in the '90s that active surveillance became formalized as a management strategy for prostate cancer.

Dr. Marks: Just a few words about the differences between then and now. In the 1990s, the buzzword was "We're looking for insignificant prostate cancers." Diagnosis was made by blind biopsy, ultrasound-guided only. We were only interested in 3+3s. These were the Epstein criteria: less than 50% of a core, less than two cores involved, generally PSA less than 10, and follow-up was at yearly blind biopsy. The endpoint was not very well defined; apparently, anything more than this was the endpoint.

Now, it's a different ball game. The term "insignificant cancers" has more or less disappeared. We now use risk. The Gleason grading system has replaced the Gleason score, and we're talking more about GG1s and GG2s now instead of 3+3 and 3+4. Cancer core length is not well defined, and the number of cores is not well defined because with targeted biopsies, we get longer cores and we get more of them. So this is not as important an issue as it used to be. PSA is now acceptable up to 20, especially when the prostate is large, and PSA density has largely replaced PSA. Follow-up is with MRI and MRI-guided biopsy, and the endpoint, which is pretty well agreed upon I think, is that the appearance of any Gleason score 4+3, now known as GG3, prostate cancer should not be part of active surveillance because of the metastatic potential of these cancers.

My interest in active surveillance began around 2009 when I first came back to UCLA full-time. Since that time, we called our program ASCAP. We developed our own little logo to try to popularize this. We formed this program with some government funding and private philanthropy. We have dedicated personnel maintaining it. It's a prospective IRB-approved project. We have a full-time database manager who keeps us up to date, and our focus from the very beginning has been on prostate imaging.

To date, around 1,200 men have been enrolled in this program, and the information I'm going to share with you this morning deals with the 869 men who have been followed up for as long as 12 years with MRI-guided biopsy at baseline and at least a year of follow-up with MRI. So I believe this may be the largest collection of men with MRI-guided biopsy at baseline who have been followed for any length of time. You can correct me if I'm wrong, but we were not the first to this game.

You saw the pioneers, but fortuitously, when we got started with this, it was about the same time that prostate MRI, and especially MRI-guided biopsy, became readily available around the year 2010. We assembled a team at that point, at the early onset, to include - and it was all focused around MRI and MRI-guided biopsy - a dedicated pathologist, a biomedical engineer to help me understand the technology (there's a lot of technology in MRI-guided biopsies), a research coordinator to keep our books, and a radiologist who became a world-class expert in MRI interpretation, which is very important for any program. This team stayed together a long time. Now all of them have moved on to better jobs except me, and I'm still in the trenches here.

The advances that I want to emphasize to you in 2023 compared to previously is that we are dealing with more and more intermediate-risk men with prostate cancer. There are some provisos with this. MRI-guided biopsy is very important upfront because it helps stratify patients and also can be used to predict the durability of active surveillance - how long somebody can actually stay in there. You can tell that to a guy up front; very helpful for counseling.

Dr. Marks: We've also learned by doing lots of follow-up biopsies with MRI guidance that MRI can replace biopsy in follow-up in many cases. Not everyone needs a follow-up biopsy.

We've learned that anxiety-driven events have decreased over time. There are several reasons for this. There was a time when people dropped out of active surveillance very quickly because of anxiety, went under radiation or surgery. That momentum has diminished.

Finally, the point that I want to share with you is that focal therapy can actually extend the life of active surveillance. We've been interested in this from early on, and I will show you some data on this.

First, to look at the effect of Gleason grade upon initiation of active surveillance. These are men who've had MRI-guided biopsies at baseline. Here's years of follow-up, and this shows their progression-free survival on the vertical axis. This is known as a Kaplan-Meier survival curve. We have introduced one new term. We have men with Gleason grade 3+4, that is the GG2s. We have men with GG1s, and we have a lot of men who've come to us with outside biopsies done by the ultrasound-guided method only, without MRI, who had documented small amounts of prostate cancer. When we enroll them with MRI-guided biopsy, we can't find it again. So these are very low-risk patients.

This is a risk stratification from very low to low to low-intermediate risk, and you can see that using this, we actually are extending active surveillance out pretty well compared to previous years. If you look just at the five-year point, you can see that about - even the highest risk patients, the low GG2s, almost 50% of them or so are still in active surveillance, have not progressed. About 70% of GG1s are still in active surveillance, and almost 90% of those with the low-risk patients are in active surveillance.

I want to use this same chart to show you how an MRI-guided biopsy upfront allows you to predict the durability or the longevity of that person in active surveillance. This has not been well emphasized in the past, but a man who comes into active surveillance with very low-risk prostate cancer has little to worry about going forward, assuming the biopsy was done properly. He does need follow-up. We do follow up with PSAs and occasional MRIs depending, but we can predict how long that man will be in active surveillance. Look at five or six years, notice the difference between these lines. Those are statistically significant, and it's helpful for a man to know his chances going forward based on his initial biopsy.

I told you that MRI can replace follow-up biopsy in many cases. This is something we've discovered by doing follow-up biopsies in everybody, but when you correlate those with the MRI, what we found - here are three groups: GG0s, GG1s, and GG2s. On the vertical axis is the negative predictive value, that is the ability of MRI to correctly rule out prostate cancer. The MRI for the low-risk and very low-risk groups approaches 90% or more in follow-up biopsy. For men entering with GG2, we're more concerned, and we do want to do follow-up biopsies on them, but in the majority of men, MRI can replace biopsy in follow-up.

Anxiety-driven events have decreased over time. This is our experience beginning in year 2010 on the horizontal axis out to 2022. In each bar is the number of surgeries or radiation therapies encountered by men who decided to do that despite being stable in active surveillance, that is, not having progressed. This is a trend line showing that that trend has decreased over time. Part of the reason is the accuracy of the MRI-guided biopsy and the safety that people understand that they have with very accurate biopsy.

Dr. Marks: Finally, I want to show you that active surveillance can be extended by judicious use of focal therapy in these patients. This has not been well established, but I do believe this is a very important factor. Here on the horizontal axis is years from focal therapy eligibility, that is the time when GG2 or, if they're in surveillance, GG3 is recognized. Instead of kicking them out of GG3, enter them into a focal therapy program. On the vertical axis is the freedom from surgery or radiation therapy charted over time. You can see the men who are undergoing focal therapy stay in active surveillance twice as long as men who do not have focal therapy. So we view the advent of GG3 in a patient in active surveillance to be a crossroad and not an endpoint necessarily. Some of these men can be salvaged for active surveillance by judicious use of focal therapy - cryotherapy or HIFU at our place.

The artificial intelligence tool that we have evolved over time - and there now is a company, Artera Health, established to promote this product - is based around our biopsy data. Each of our patients at baseline undergoes a biopsy where there is targeting of any visible MRI lesion and systematic sampling throughout the rest of the prostate. We've done this since 2009. Anybody who has a biopsy like this can generate a lot of data, which goes into a - which when put together with clinical data, can be used to generate an artificial intelligence program.

Keep in mind we've done around 5,000 of these targeted biopsies over time, and each of these biopsy cores is tracked in the software of the image fusion device. So putting all these together, we can not only identify the cancer, but we can actually use the locations of those biopsies in relationship to the MRI visible lesion to create a map, a cancer map.

This is an example of such a map where the red spot is the MRI visible lesion, and going out beyond the red spot shows you how far out the cancer may actually extend. We've learned that the MRI visible lesion is just the tip of an iceberg, and that almost always that cancer extends far beyond the visible area. It may extend in fingers and not in a smooth concentric pattern.

What you can generate from there is a treatment margin. We did this primarily for focal therapy guidance to begin with, but it also is appropriate for trying to decide where to give a radiation therapy boost if that's appropriate, or where to extend a surgical margin. It also may have a role in active surveillance.

The current challenges as I see them in active surveillance are increasing efforts to stratify GG2s because they're not all the same. We're looking at cancer core lengths, percent pattern four. The advent of the PSMA scan has changed urology dramatically. We are actually doing biopsy guidance in many patients using PSMA identification of lesions within the prostate. Current indications for PSMA are to stage prostate cancer pre-treatment or to detect recurrence, but it also has a role in looking at cancer within the prostate. This is new and going to be very important in the future.

Molecular markers - there's already a fair amount of data out about the Decipher score and a number of other molecular markers which can be used to try to predict the likelihood of a cancer getting out of control in the future. Very appropriate for men in active surveillance, need more data.

And then I think further evaluation of the role artificial intelligence will have is also important. I think it's an exciting time to be in this work. I think that this all began around 2010 when the MRI-guided biopsy became available, that is, the evolution of this became available. And the use of artificial intelligence to map cancer has many roles in active surveillance.

Dr. Marks: The size of this lesion may help predict propensity to become more aggressive in the future, and that's one thing we're measuring. There's no other way to measure the size of the cancer except by the use of this software right now. So we're very enthusiastic about that. Lots of new challenges in the future. I'll be happy to take questions, and I thank you very much for the invitation to be here today.

Howard: So, doctors Mia and Lee Burton, the floor is yours.

Mia Lee Burton. Medical Science Liaison at Artera

Mia Lee Burton: Thank you so much Bill, Howard, and Mark. My name is Mia Lee Burton. I'm a Medical Science Liaison with the Medical Affairs team at Artera. As an MSL, one of my key functions is to engage with clinicians, mainly urologists and radiation oncologists, to provide medical and scientific education on Artera's multimodal or MMEI biomarkers. Hearing the stories from the clinicians about how the Artera AI prostate test has given them and their patients more confidence in their joint decision-making is really humbling to me. Thank you so much for inviting us. It's a pleasure to have this opportunity to introduce this test to the ASPI members today.

We have come a long way in terms of risk stratification tools in prostate cancer since the 1960s when the Gleason score system was established. Following the FDA approval of the PSA testing, the nomograms that integrate multiple clinical parameters, to the rapid adoption of multi-parametric MRI. Also, the genomic classifiers have been around for the past decade. Finally, in the past few years, artificial intelligence-enabled tools have emerged. For example, Artera has developed our multimodal AI or MMEI algorithm that provides AI-enabled biomarkers to inform prognosis and, one step further, to predict treatment response.

By definition, artificial intelligence is a technology that enables computers to simulate human intelligence and problem-solving capacities. A couple of examples would be GPS guidance or autonomous vehicles. Machine learning is a subdiscipline of artificial intelligence. It involves the development of artificial intelligence algorithms that use neural networks to extract features from the data and make predictions about what the data presents, just mimicking the decision-making process of the human brain.

When we think of AI, the examples that come to mind first are likely to be generative AI, which is designed to create some new content that mimics human-like creativity. On the slides are some examples. On the other hand, medical AI, instead of focusing on creativity, is designed to support clinicians in diagnosing, staging, and treating disease by utilizing machine learning models to analyze high-quality data, for example, clinical trial data, through a rigorous testing and validation process, with the goal to improve health outcomes and patient experience.

By utilizing machine learning methods, many AI-enabled tools have further advanced the diagnosis, prognosis, and prediction of treatment response. The Artera AI prostate test is intended to be used after a diagnosis of localized prostate cancer has been made. It provides both prognostic biomarker and predictive biomarker results. The prognostic biomarker results inform the disease aggressiveness of the tumor - how likely is it to grow and spread, and how likely it is to cause mortality in the future. The Artera prostate test also provides predictive biomarker results, which inform whether a particular patient will benefit from a certain treatment.

Here is a brief overview of how the test works: After a localized prostate cancer diagnosis has been made, the Artera AI prostate test incorporates two streams of data sources into its multimodal AI algorithm. One being patients' traditional clinical parameters including age, PSA, and clinical tumor stage; the other being patients' digitized histopathology imaging data. So the MMEI algorithm deploys the machine learning algorithms to analyze the data input and provide both prognostic and predictive biomarkers.

Mia Lee Burton: The prognostic biomarkers we have are the endpoints of 10-year risk of distant metastasis and 10-year risk of prostate cancer-specific mortality. Those two endpoints are provided to all patients with localized prostate cancer across NCCN risk groups. Starting from August 5th, we also started providing additional active surveillance insights for patients with NCCN very low, low, and favorable intermediate risk, which Megan will discuss in more detail in the upcoming slides.

We also provide predictive biomarker results. Bill and Howard have mentioned that we can predict whether a patient with NCCN intermediate risk disease will likely benefit from adding short-term ADT to their radiation therapy.

This slide shows how many clinicians have seen are using the AI prostate test to inform treatment intensification decisions. First, the prognostic biomarker results inform the prognosis, the disease aggressiveness. If this patient has a good prognosis, then maybe treatment intensification is not necessary. However, if a patient has a poor prognosis, the predictive biomarker becomes crucial, particularly for patients with NCCN intermediate risk disease, including favorable and unfavorable, and it can guide Androgen Deprivation Therapy intensification decisions.

In the next slides, my goal is to walk you through the Artera multimodal AI algorithm and its architecture. This slide shows the architecture of the MMEI algorithm. First, patients' traditional clinical parameters are converted into a feature vector, which is a numerical representation of the data that is easier for the downstream machine learning algorithms to analyze. We then digitize patients' biopsy slide to generate digital pathology images. I wanted to point out there is no tissue consumption during this process. Everything is performed using the images of the biopsy slide.

After we digitize the biopsy slide, we then segment out all the images of the biopsy cores and paste them together to form this image quilt. Then a digital grid was laid over on the image quilt to divide it into small patches, or the small squares you see here. The patches were then fed through this pre-self-supervised or unsupervised machine learning algorithm which converts image data into numerical feature vectors. Finally, those two streams of data were combined to generate an AI score which will be used to generate prognostic and predictive biomarker results.

I wanted to point out here that the self-supervised machine learning algorithm we used - there was no pathologist labeling before we fed the data through it. It allows the AI to learn from vast amounts of raw data to discover novel patterns beyond current human understanding, and this is where the power of AI lies.

Here is just an example of a biopsy core on a digitized prostate biopsy slide, and this slide shows you how the MMEI algorithm analyzes images. First, the MMEI algorithm analyzes each patch of image to determine whether there are important tissue patterns within each patch related to the outcome of interest. As you can see, there are some patches that are highlighted, indicating that the AI identified them as containing important patterns. The darker the patches are, the more important they are to the outcome of interest. You can see the AI deemed different areas, different patches as important for the prognostic biomarker and for the predictive biomarker for short-term ADT.

This is a brief overview of how the MMEI algorithm works, and this brings me to the end of my part of the presentation. I will now hand it over to Megan, who will delve deeper into the test and the test report and the biomarkers it provides.

Megan Tierney

Megan Tierney: Great, thanks Mia. I think that's a really great background into the AI technology that is the core of what we do here at Artera AI. My portion of this presentation is going to take us a little bit more towards what is offered in our current clinical test and what we have come to the point of our newest edition that provides a bit more insight into active surveillance itself.

Megan Tierney: For those of you who are not as familiar with the Artera AI prostate test, it has two primary elements, as Mia described earlier. One being our prognostic element that provides risk stratification for certain outcomes down the road, the other being our predictive model that provides information about whether or not a patient is likely to benefit from the addition of short-term ADT to their radiation therapy.

I think the predictive element was described in quite a bit more detail back in a webinar in February, I believe, when we were last here to talk with this group. The ASPI team does a fantastic job of making these resources available on their website, so anyone who's interested in more detail on the predictive model can go back and reference that webinar. But today, I'll focus primarily on the prognostic risk stratification element of our test because that's really what generated the new addition around active surveillance.

This slide shows a summary of what is used to generate our prognostic risk scores: the combination of patient information in addition to digital images from a slide of the patient's biopsy. That is fed into our AI algorithm, which analyzes that data and outputs a personalized risk score that estimates long-term outcomes such as distant metastasis or cancer spreading to other parts of the body, or prostate cancer-specific mortality, which is the risk of death specifically due to the prostate cancer itself.

This particular model was built and validated using data from five large clinical trials, about 7,000 patients' data, really representing a wide variety of patients. We do the initial development of the model, we lock the model, we validate the model, and that first effort gives us information about how well the model itself is able to predict these outcomes - whether it is doing what we expect it to do and performing the way that we expect it to perform.

But that's not necessarily where our research ends. We continue to do additional research to evaluate how the biomarker compares to other prognostic tools that exist clinically, and also to do more research to explore the generalizability of the model itself. I'll touch on a couple of those additional research efforts that have led towards the creation of these insights for surveillance.

This slide shows an example of how we have compared our test to other existing tools that are clinically available. One common tool that gives an idea of a patient's prognosis is the NCCN risk groups. When we have compared our Artera AI prostate test to the NCCN risk groupings in terms of predicting distant metastasis and prostate cancer-specific mortality, we actually see that the Artera test does a better job of predicting those outcomes. This is great because now we know how the model itself works, we know how it compares to some other existing tools. Here we're seeing that it does add value beyond just the NCCN risk group. It confirms that the AI is able to pull valuable and relevant information from those histopathology images to contribute to its ability to predict these outcomes.

But as I said, this model was originally developed using data from five clinical trials. These clinical trials were primarily patients with localized prostate cancer who were using radiation therapy as part of their primary treatment for their disease. So then that begs the question: Does the model perform similarly in patients who are using prostatectomy or patients who choose active surveillance? That has led to further research.

Here on the left side of the slide describes a recent validation study that we did with an expanded cohort of patients. Here, a little over 3,300 patients who have been managed with radiation therapy or prostatectomy or active surveillance. We see here that in this expanded cohort of more diverse patients, our MMEI prognostic score was still significantly associated with distant metastasis, which is great because it increases our confidence in being able to use this tool across different choices for disease management or treatment.

Megan Tierney: But we know that there is a lot of interest in being able to develop more prognostic tools that can really inform the active surveillance space in particular. We run into the challenge that distant metastasis as an outcome isn't necessarily as useful for the surveillance space because so few patients with lower risk disease actually develop distant metastasis. That information is not necessarily as useful.

So we wanted to look at an outcome that was a bit more relevant to the active surveillance space. If you go to the next slide, it will refocus us to the right side of the slide here where we highlight the outcome of adverse pathology at radical prostatectomy, or we refer to it as AP at RP for short.

What this means is that patients who eventually go on to have a prostatectomy, we look at the tissue from that surgery. Does the patient have more aggressive features in that tissue than were present at the time of their original diagnostic biopsy, suggesting that their tissue has sort of progressed and evolved over that time? These more aggressive features that we're looking at here are those features that would make them no longer good candidates for active surveillance. For example, progressing to a grade group three or higher.

The way that we addressed this with our prognostic model was to take the patients who were part of that active surveillance group in the expanded cohort that we see on the left over there. Of those patients who were managed with active surveillance, 292 of them eventually came off of surveillance and had a prostatectomy. We know we have the follow-up data for these patients, so we know of those patients which ones had adverse pathology and which ones didn't at the time of their prostatectomy.

But we were able to feed their diagnostic information, so their clinical information at around the time of their diagnosis and their image data from their original biopsy, into our prognostic algorithm. The algorithm generated a risk score for those patients, and then we looked at whether that risk score was associated with whether or not a patient had adverse pathology when they had their prostatectomy.

What we found was that yes, there was some statistically significant correlation between their MMEI risk score and whether or not they had adverse pathology. So we're hoping that this endpoint is a little bit more relevant for the active surveillance space, can provide a little bit more information about how likely a patient's tumor is to evolve to develop more aggressive features in a shorter time window than distant metastasis.

This really provides the backbone for what has become our latest test report update, providing more insights around active surveillance. The next few slides we'll walk through what that actually looks like in translation to our test report.

For people who haven't seen our test report before, this is a snippet of our results page where in the box highlighted here, we see a patient's 10-year estimated risk of distant metastasis. This has been present on our test report historically. For this patient, their MMEI risk score translates to a 1.8% risk of distant metastasis in 10 years, and that places them in the low-risk bucket for our test. You have potential to be placed into a low, intermediate, or high risk of distant metastasis based on how patients' risk scores come out with our test results.

Then on the next slide, you see below that our 10-year risk of estimated prostate cancer-specific mortality. So again, for this patient, their MMEI risk score translates to a 1.1% risk of prostate cancer-specific mortality within 10 years.

On the next slide, it'll highlight the right side of this snippet here, and this is what is new as of earlier this month. This provides additional insights around active surveillance that stem from that research that I just described.

Megan Tierney: Here we see we don't provide a specific percent risk of adverse pathology at prostatectomy. We are still doing work and further development to get to the point of being able to provide a sort of absolute risk of an outcome that's relevant for active surveillance. But here, our goal is to be able to provide as much information as we can as soon as possible in terms of being able to help support clinical decision-making.

So here we are, first of all, highlighting active surveillance as a potential option for patients on this first page of our test report. Hopefully, this will trigger the question of whether this patient should consider active surveillance. Then we provide some more contextualization of this patient's risk compared to other patients who have chosen to be managed with active surveillance. That comparison is based on the subgroup of patients from our validation study looking at risk of adverse pathology.

We see for this patient, their individual risk score falls within the 35th percentile of those 292 patients from our study. So their risk is higher than about 35% of patients and lower than about 65% of patients from this larger group who were managed with active surveillance.

We also see along this blue scale a little white bar that shows the distinction between the number of patients in that larger cohort who would fall within the low-risk category on the distant metastasis on the left-hand side of the snippet, and the other 21% of patients would fall into our intermediate risk.

So it also provides a little bit more of a comparison for how this patient compares to this broader cohort of patients who have been managed with active surveillance. Again, hoping that this can provide a little bit more information that can support decision-making around active surveillance for patients who are newly diagnosed and thinking about what is the best way forward for them. They can use this information in combination with all of the other information that their doctor has available and can bring into that discussion, but this can be one piece of that that can hopefully be helpful.

Then I think we have one more slide. I was going to skip over these predictive biomarker pieces. What we wrapped up with the active surveillance insights is all part of our prognostic element of our test. It's been brought up a couple of times, as I said, that we do have this alternate predictive biomarker as well that provides information about whether or not a patient is likely to receive benefit from adding Androgen Deprivation Therapy to their radiation therapy.

This particular result is only provided to patients who fall into the NCCN intermediate risk disease group because those are the patients who are most likely to be making these types of decisions. This has been really well received. I think Howard mentioned that there are some patients in this group who have used this test and potentially benefited from this test.

In our validation study, our model identified that really only like 32% of patients were classified as biomarker positive and likely to benefit from using ADT. So there's quite a few patients who could be spared the side effects of ADT if their tumor is not as likely to respond to that particular treatment.

Then one more slide. This helps to give a visualization of how our results are provided based on patients' NCCN risk groups. Here in the US, NCCN really provides the core of our standard of care, and so we want to provide our tests in a way that aligns with that standard of care.

Patients across all NCCN risk groups will receive our prognostic information about risk of distant metastasis and risk of prostate cancer-specific mortality. Patients who are candidates for active surveillance according to NCCN - so very low, low, or favorable intermediate patients - will receive those active surveillance insights. And then as I mentioned, the predictive results around short-term ADT are provided for patients who have NCCN intermediate risk disease.

So I think that wraps up our presentation for today, and we can probably hand it back over to open up for questions.

Q&A

Bill Manning: We appreciate your presentation and we will open up for questions for all the doctors above. I had a question for Mia and Megan. Dr. Marks expressed some caution about AI in general, and I think we all would want to stay away from overhyping. So maybe this is a question for the three speakers or two of them or one of them: Where do we stand now with AI? Is this still a period of caution? Is some of it okay? Is none of it okay? Dr. Marks, do you have any thoughts on that?

Dr. Marks: Well, I think it's great. I commend the Artera people. I think it's really interesting what they've done. I think they've taken older concepts and really modernized them tremendously. I have to think back to the Partin tables and the CAPRA scoring system. What they've done is to take those scoring systems and put them on steroids with a lot more data. I think it's great. I think we just have to wait and see how the data shake out.

This is a very interesting time in medical history for the incorporation of AI. This is like the beginning of automobiles in the early part of the 20th century. We're going from horse and buggy to horseless carriages. It's all about the data. The problem with AI in general is it's a black box that's very hard for others to duplicate. That's true with our system, that's true with their system. They have their own algorithms that make this work. So I think we just have to see how it plays out. Looks interesting.

Bill Manning: Mia and Megan, do you have any response specifically or generally to that?

Megan Tierney: I think Dr. Marks makes really good points that AI is really growing rapidly in the medical space and more broadly as well. I will say that our Artera test is part of the NCCN guidelines, so there is some kind of confirmation there in terms of the level of evidence that is supporting our AI and the kind of amount of continual research that we do.

We try to be very rigorous in our approach to developing AI. We work with a lot of institutions, a lot of physicians as we are developing our products to try and come up with things that are scientifically sound as well as clinically useful. I think we're also seeing a lot of changes in the regulatory space around AI, and so I think that will also be something that all of us who are working in the space will be navigating in the years to come.

Howard Wolinsky: Now Dr. Marks, we've gotten a lot of questions about focal therapy. Maybe Jeff in the background is looking through the questions and can bring some up, but I have a question for you. To be a purist about active surveillance, I take it to mean you haven't been treated. Focal therapy, and it comes in a variety of flavors, is a treatment. So I'm wondering, technically speaking, to be a stickler, would you still, after you treat somebody with focal therapy, consider them on active surveillance? What do other doctors think about that? Or am I just being too much of a stickler?

Dr. Marks: Well, you're a purist. You're looking at it from a puristic standpoint. What we do is, when somebody has a focal therapy intervention, we follow them exactly as we would if they were in active surveillance. It's like starting the clock over again with them. We build in follow-up biopsies and follow-up MRIs. All our patients on active surveillance who get focal therapy as part of their active surveillance regimen - patients welcome this. After all, the motivation is to avoid surgery or radiation therapy. If they can do that with something that is not going to change their quality of life and allow this tumor in them to be observed going forward instead of treated radically, they welcome it. Our patients are very enthusiastic about this when it's approached.

Howard Wolinsky: And by the way, there is a lot of debate about focal therapy. Maybe you can define for us what focal therapy is and what it isn't, and who is a candidate for focal therapy. In other words, we had another program earlier where we had a speaker, Dan Spratt, who's a radiation oncologist, and he said that focal therapy is being oversold to some patients with Gleason 6. So do you feel like Gleason 6 patients, maybe high-volume Gleason 6 patients, are candidates? Or what's the reality for this?

Dr. Marks: Those are great questions. Number one, we do not treat Gleason 6 with focal therapy, even high volume Gleason 6. We continue to watch and follow with additional biopsy and MRI going down the road. A candidate for focal therapy is someone who has less than half of a lobe involved, usually with Gleason 3+4, some with 4+3. I think most people would agree with that. 4+3 is a little more prone toward metastatic disease, so 4+3s have to be watched a little more carefully than the GG2s, but we deem both of them to be appropriate candidates.

Now, there's a lot of anatomic information that goes into the selection process. If you've got a 200cc prostate and 100cc's of it is occupied by cancer, probably not a good focal therapy candidate. A lot of these questions have not been answered in a formalistic way yet, but in general, half of a lobe or less. We're using cryotherapy and HIFU. HIFU particularly is very good for targeting posterior lesions in smaller prostates. Cryotherapy is more for anterior lesions and larger prostates. There are several other methods of focal therapy which we don't have as much experience with, but when they get followed in clinical trials - and all of our patients are in what we call open label clinical trials, they've all signed consent, these are NCT registered trials - patients embrace this, and we like it.

Jeff: Howard, if I could jump in here for a second, there's a few questions that are directed directly to Artera, and I'd like to give them a chance to answer some of them.

Megan: I can jump in on a couple that are more related to our research and our background of our model. I can maybe take a stab at those first and then hand over to Mia or Leah for some of the more logistical questions.

I see one that asks about whether we have repeat MMEI determination for AS patients and whether our MMEI is stable over time. Currently, our product has been validated to use the earliest biopsies for patients in terms of predicting outcomes. We have internally looked at whether the scores change if we look at serial biopsies over time versus just the first biopsy, and we have not seen a significant difference in the predictions whether we use just the first biopsy versus all biopsies or their most recent biopsy. But technically, our product is validated for using the original diagnostic biopsy.

Then I think there was a question around Gleason score, Gleason grading. When we were originally developing this model, we initially had included Gleason score as an input to the model in addition to age and PSA and T-stage. We were able to remove that because we believe that the AI imaging is actually able to pick up on some of those Gleason features when it's looking at the images of the tissue sample itself. So we might be able to pull that information about Gleason from the digital images as opposed to using that as a clinical input like from a pathologist reading.

With those, maybe I'll hand over to Mia and/or Michael to talk about general cost.

Michael: I already answered the question in the channel, but basically, the Artera AI prostate test does have a Medicare payment rate. So for any patients that are on Medicare coverage, there should not be any type of out-of-pocket cost for the patients. For commercial payers, we do submit a bill to the insurance companies for the cost of the test. Our list price is $1,420, but we will certainly work with the patient. We do have a financial assistance program. We understand that this can be a financial burden for patients, and our goal is to obviously make this widely available. So we will certainly work with the patients and with the insurance companies on the commercial side and try to figure out a path forward to make this test available.

Jeff: Hi Bill, I'm gonna jump in here. I've seen a lot of questions. Thank you everybody who's participating. There's mention of biomarkers. I think for the panel, it would be good perhaps if Dr. Marks could lead us into a discussion of the definition of what is technically a biomarker. My understanding is it's a scientific fact about your body, something that's been tested and shown to exist in you. For example, PSA. For example, some of the tests - there are several tests. So first with the biomarkers, then secondly, how does the artificial intelligence model that we're seeing today, Artera, compare to some of the other biomarkers, and particularly some of the other tests? Some of the questions have mentioned Decipher, Prolaris, and others.

Jeff: And then perhaps capping that is, going back to my first day at the University of California Berkeley - a shout out to Dr. Marks - a certain amount of information goes into modeling, affecting the information that comes out. I would like the presenters to be fair about describing the limitations perhaps that have in this model versus the results. Thank you.

Mia Lee Burton: I'm happy to take part of the question. Currently, there's no head-to-head comparison data between Artera AI prostate test and other genomic-based tests. The two tests are looking at different perspectives of the disease. Genomic tests, of course, are looking at gene expressions, and Artera AI prostate test, we are looking at the histopathology images and more like morphological changes of the prostate tissue.

Another difference is for Artera AI prostate test, like I was presenting in the presentation, there was no tissue consumption. We're using the slide of the biopsy. So there are some differences between those two types of tests. The goal of the test is to provide additional information in addition to the clinical parameters. But in summary, we do not have head-to-head comparison.

Jeff: Let's ask Dr. Marks. Dr. Marks, can you lead us into this broader discussion?

Dr. Marks: So a biomarker, broadly speaking, is an indirect measure of something that you want information about. The Artera people have identified theirs as a biomarker, and actually, the Artera people have also called their cancer map a biomarker. Fine, that's fair if that's how they want to do it. I think that's a fair way to state things.

My personal feeling is that one day, the genomics will replace everything, but we are not anywhere near there yet. I think Artera has a very interesting platform that will allow for later incorporation of additional information that becomes available, like the genetic materials or perhaps imaging, which I noticed is also not part of their system. But it's all about the black box and the algorithm that the AI scientists have chosen to use for their product.

Could there be a competing product to the cancer map? Sure. And could there be a competing product to the Artera? Sure. It's just a matter of getting the data. I think they've done something very clever in using existing clinical prospective clinical trial material to prove the worth of their product. That was very clever, and I look forward to using it.

Jeff: I'd like to ask, how does a doctor get to order this test?

Michael: I can take that. It's actually quite simple. Doctors can reach out to - and actually, anyone, and I know there's a lot of questions on this call - please feel free to email support@Artera.ai. I'll drop that in the chat as well so you guys can just click on it. All inquiries, if you're a clinician, we'll be happy to set you up. Unfortunately, Dr. Marks, since you're in California, I think we are not available there, so I don't want to get your hopes up there. But we certainly love to chat with you and talk to you as we do hope to get California certification in the coming months. But for all other inquiries, please direct them to support@Artera.ai.

Jeff: Basically, the doctor or the hospital mails you the slides and the clinical information?

Michael: Yeah, basically. We'll reach out to the Pathology Labs to get the samples mailed out to us to our lab, our CLIA lab in Florida. We'll scan the biopsy slides and return the slides back to the Pathology Lab, and that's how we run the test.

Howard Wolinsky: I can interject something here. I had written a couple of stories about this, and the Artera people can tell me if I'm wrong, but the goal with this, Dr. Marks, is eventually you won't have to send slides, you just send the image. Am I right or wrong?

Michael: Yeah, that's what we're working for. Currently, right now, we do need the slides to be sent to our lab in Florida, but we are working - and Megan kind of mentioned this earlier - the regulatory path forward. This is sort of a new thing that we're working on with the FDA to make sure that the slides being sent or images being sent from all sorts of different scanners from different Path Labs, etc., is going to be sufficient quality and all that type of stuff for us to be able to use. So those are ongoing discussions, but that is certainly the direction that we're hoping to be able to achieve.

Jeff: Does your algorithm differentiate between targeted biopsies and systematic biopsies?

Megan: Not currently, no. We accept any sort of biopsy tissue. We request that it represents the highest grade tumor from the particular biopsy. But it is in our research pathway, I believe, projects to kind of look at whether we see differences depending on the ultimate source of that biopsy.

Howard: I need to interject. We're almost at the time we said we would stop, but I'm sure if any of the speakers would stick around, people still would like to have their questions answered.

Jeff: Let me jump in here. Here's a good question from Greg.

Bill: Wait a minute. We have to determine if our speakers can stick around, Jeff, before we jump in.

Dr. Marks: I can hang in for a little while. This is fun.

Bill: I'm going to assume the Artera people feel the same. So why don't we go for a little while, see what happens?

Jeff: Thank you for the clarification. Classic case: What happens if the clinician's gut feeling or his knowledge base differs from the Artera report or any artificial intelligence report? Ethically and clinically, what happens?

Megan: I think that's a really good question, and I can just speak quickly from the Artera side on that. I think we tend to view our test as one tool that's available for clinicians and patients to look at. We provide pieces of information based on what our test is able to offer. That is one piece of a picture. Patients, in general, are more complex than I think anything that one test can pick up on. So we tend to position our test as something that can hopefully provide some additional information but should be looked at within the totality of the picture of a given patient and their particular preferences and goals.

Howard: I'd like to get back to some of the focal therapy stuff. Dr. Marks, we were talking about the puritanical view. Let me ask you this: Presumably, with whatever treatment that you're doing, you are removing any Gleason 7s. So does that sort of set the clock back, and are these patients potentially Gleason 6 patients, and so they qualify for active surveillance? Is that a way to look at it?

Dr. Marks: Yes, that is a fair way to look at it. For example, if a patient has Gleason 7 on one side and a random spot of Gleason 6 on the other side, we tend not to worry about that other side with the Gleason 6, and it would not be treated and it would be followed.

The trouble with prostate cancer is you need long follow-ups to really know anything about it. That's why a new drug for lung cancer is going to have an answer very quickly because the mortality rate is so high. A new treatment for prostate cancer is going to take a lot longer to determine its real value.

Interestingly, the PSA test, for example, now dates back to the early '70s, and it's still extremely valuable. But there aren't many other things that have that degree of history and proven value. So we just have to wait and see and be sure that we have good long-term trials designed. Organizations like yours are so important in getting people to enroll in these long-term follow-up trials.

Active surveillance - we are accumulating lots of long-term follow-up for active surveillance. As I think it was Mark who said he's a 20-year survivor in active surveillance. We don't have any 20-year focal therapy people, but we have some 10 years, and we're getting to learn more and more about it. It's something that fills a niche in our armamentarium between active surveillance alone and radical treatment.

Howard: Now Mark Plow brought up a question. Well, he points out that European guidelines suggest focal therapy should be limited to clinical trials. He asks if that should be adopted in the US. But I'm pretty sure the American Urological Association recommends that any patients going on focal therapy be included in trials. And I think you said all of your patients are enrolled in trials. Am I getting that right?

Dr. Marks: Yes, that's correct.

Howard: We've got a couple people that have had their hands up for a while. I think it'd be good to give them a quick shot at it. Robert... Well, Mike Wi has his arm up; it must be falling off by this point. So maybe Mike?

Mike Wi: Hi, this is Michael Wi in Denver. I'm a former professor of engineering, so I got a couple questions. One for Dr. Marks and the second one for the AI team. The first question, Dr. Marks: Through several informal conversations I have with various urologists, they think that the location of your tumor is fairly important. If your tumor is in the transitional zone, it tends to be less risky in terms of progression compared to if you had a tumor in the peripheral zone. So that's my question for Dr. Marks.

The second part for the AI folks: I work with both machine learning and deep learning. One problem we run into is that there's essentially no real good way to query the system to ask us how it came up with that particular prediction. Can you give us a little more explanation of how you do that?

Dr. Marks: Let me take the second question first. I want to be sure I understand. You're talking about the Unfold AI prostate mapping, is that what you're talking about?

Mike Wi: No, I'm talking pretty much essentially about the black box. You've got Bayesian algorithms, and so for these black boxes for machine learning or deep learning, if you feed data to that, it comes up with some prediction value. You cannot really ask the system how it came up with that particular result. I mean, it's very difficult to do that because it's a black box. I think you mentioned that in one of your conversations.

Dr. Marks: Yeah, I don't know how to answer that question frankly. I think somebody else will come up with a little different algorithm, a different black box, and the answers may be different. That's what makes the world go around.

But with regard to your question about transition zone versus peripheral zone cancers, yes, you see across the board about 70% of cancers are found in the peripheral zone, either posteriorly or in the anterior horns as it sweeps up. But I would be more interested in the Gleason score or the genomic predictors or possibly the Artera predictor going forward. I think it would be hard to ignore a Gleason 8 prostate cancer in the transition zone.

Howard: Okay, Robert W, you've had your hand up for quite a while. How are we with metal hips to be confident in multi-parametric MRIs in terms of getting an accurate PI-RADS score?

Dr. Marks: This comes up in my practice every day, and the answer is it depends how old that metal hip is. If it was put in 20 years ago, it may not be MRI compatible. If it was put in within the past five or 10 years, it probably is titanium and is MRI compatible. Frequently, the radiologists will review the device with the manufacturer to determine what can be done. Turning the magnet down from 3 Tesla to 1.5 Tesla typically gives good results. We have many examples of that. There's a lot more that goes into the value of an MRI than just the Tesla, but we do lots of 1.5 Teslas in people with artificial hips.

Howard: Okay, Mr. Hunter, would you like to ask your question? We should note he's calling us from Scotland, so welcome.

Mr. Hunter: Thanks very much. Yes, thank you for this. It's all a little bit over my head. Dr. Marks, if I could take you back to very basics, you mentioned at one point PSA density was more applicable now than PSA, and yet later on you went on and said PSA is an extremely good marker. I have been on AS since mid-2021. In January 2022, my PSA density was 0.24, and in January 2023, my PSA density is 0.36, which seems way above what is acceptable. And yet my PSA is about 8, and my consultant doesn't seem too worried. Can you just clarify that? My PSA density has never been as low as it should be, I don't think.

Dr. Marks: That's interesting. I would assume you have a rather small prostate. If I were you, I would ask them what your percent free PSA is. That's another modifier to help determine, to shift this toward benign or malignant etiology for the PSA.

Dr. Marks: Across the board, prostate volume is the main determinant of serum PSA levels. It's a direct linear relationship between prostate volume and serum PSA levels. Big prostate, high level; small prostate, low level in general. And when it escapes those boundaries, that's when we get concerned. But a PSA density of 0.36 is remarkable. That's very high.

Mr. Hunter: My prostate volume is 21 in the beginning of 2023.

Dr. Marks: Have you had an MRI?

Mr. Hunter: Yeah, and they don't really see anything when they look at the MRI. Twice now, they said, "Oh, it's really nothing to see."

Dr. Marks: It may not be a malignancy. But then you've got to think of other things. Do you have a family history of prostate cancer?

Mr. Hunter: Yes, yes.

Dr. Marks: If you were my patient, I would biopsy this regardless. Another consideration is a PSMA scan if you can get the healthcare system to pay for it. PSMA finds cancers that MRIs do not find. So that's an idea.

Mr. Hunter: Yeah, yeah.

Dr. Marks: But 0.36 PSA density - you would be concerned about that. 0.15 is the normal cut point.

Mr. Hunter: Thank you. I won't take your time anymore. Much obliged.

Howard: How about David Peterson?

David Peterson: I had a PSMA PET scan and everything. I want to know if Artera can use that information instead of a biopsy. I haven't had a biopsy in order to do their test. And if not, are there any other medical AI firms that anyone knows of that can take PSMA PET scan information and use that to determine whether - what I'm most concerned about is whether I need hormone therapy. They've suggested radiology, but I'm concerned about hormone therapy. I have some metastasis, but very little. Thank you.

Megan: I can say Artera has focused our technology primarily on the use of digital histopathology currently. So we don't have the capability to incorporate PSMA PET scan information into our algorithms at the moment. But I'm guessing Dr. Marks can probably speak better to potential options on the imaging side of things.

Dr. Marks: Did you say you had metastases?

David Peterson: Yes, a little. One gland right adjacent to the prostate.

Dr. Marks: So this describes a new area of endeavor for urologists and radiation therapists today. This is called the oligometastatic situation. You definitely should have a biopsy. It will be helpful in directing your therapy, which may include that lymph node. I'm happy you don't have any bone metastases. That's a good finding on this.

You should ask what the SUV maximum score of that PSMA uptake is, because if it's - and that stands for standardized uptake value - if your PSMA score, the SUV score, is more than about 10 or 12, that is certainly a cancer and probably a high-grade malignancy.

But the biopsy can give lots of information. It can also give information about treatment regarding drug therapy. If there's a - you know, that's what the Decipher score is all about and other these tests looking for genetic mutations that may be treatable. There's a whole new class of drugs aimed at this. So don't be afraid of the biopsy. It may help save your life or prolong your life.

Howard: Well, if he gets the biopsy, he can turn it over to Artera also, which would provide some direction on ADT.

Dr. Marks: I have to sign off. Great being with you.

Howard: Okay, well thanks for that and very fascinating webinar today. Any remaining questions for Artera, Jeff?

Jeff: I have one if I may. Classic case - my wife is in the art world and a lot of artists post their images online, and some of these images are being used, you might say, without permission. The question is: We have a number of biomarkers and a number of tests that are obviously patented or otherwise private property in the legal terms. This is a futuristic question, but how will the AI world be able to generate more complete information by getting competitors, if you will, to coalesce so that the AI algorithms are more powerful?

Megan: I think it's a really interesting question, a very good question, and one that is probably on a lot of people's minds these days as we see these different capabilities in the AI space popping up in different areas, even within just medical AI. I think everybody at the moment is aiming to develop in a particular area where we can have a certain amount of expertise and develop really good models that focus on a specific aspect of medical care. I think it will be interesting to see in the future, as we continue to grow our capabilities in the AI space, where there are opportunities that arise for connection between these different technologies.

Howard: We have a hand up. I was gonna make a point to you, Jeff, on your question. A similar kind of thing is involved with the genomic testing and with the biomarkers. A lot of the companies have been reluctant to have bake-offs or cook-offs to compare their products because they could get bad news that way. So I don't know if it's going to be as easy to happen as Megan suggested. There's a lot of proprietary information there and people are protecting their patents. Something like the NCCN perhaps would be in a better position, some with a broader neutrality, if you will.

Now, we're going to call on Hun.

Hun: My question was, can the Artera test be utilized in cases where the MRI does not show a lesion? I've been told, "Oh, it's too small. We found it in the biopsy, but it's evidently too small." And that went on for a bunch of years. So I don't know whether if it can't be a guided biopsy, then is it in play or not? Thank you.

Megan: I would say we don't currently specify parameters for the type of biopsy that's done, but if there is prostate cancer that can be detected on the biopsy slides or in the biopsy specimen, then it should be able to be submitted for the Artera test.

Mia: I will add a little bit to Megan's answer. Once your physician places an order, after your pathologist mails the slide with the highest Gleason score to our CLIA lab, we'll have our in-house pathologist look at the slide again to make sure tumor is present, and we'll proceed with the order to analyze the slide as long as there's tumor on the slide.

Hun: Any idea how many years those slides stay good? Mine were like 10 or 12 years ago.

Mia: A lot of the trials in our validation and development model were from the early 1990s, and we have quite a high success rate of reading the slide and processing the data.

Howard: But then you run into how long does the person who has custody of your slides hold on to it. My experience was that after 5 years, they may get rid of your slide. So I don't know if Megan or Mia know about the rules on that.

Hun: Well, this was Hopkins, so I think that they probably still have them.

Howard: Gotcha. Well, sounds good.

Mike Wi: Mike Wi again. I'd like to go and ask the second question. I know you guys have been around for the last two years, I believe, right? And so I know that toward the end of your presentation, you talk about how you validate the data and so forth, and you talked about predictive value for 10 years. So how do you do currently your process of so-called ground truth verification to make sure that you're predicting a result, a report kind of map into something that 10 years from now could be still valid?

Megan: I think that's a good question and also kind of echoes Jeff's question from earlier about potential limitations to this process. We have worked primarily with large prospective historical clinical trials because they have the amount of follow-up time that we need in order to look at these 10-year outcomes. But that means that sometimes during that time, treatment standards of care may have changed, treatment approaches may have changed. We do advance in technology and care practices over time.

So it is on our minds and as part of our continual development of our products to make sure that we are staying connected and bringing in as current of data as we can. We're not just developing a test and letting it sit. We develop a test and then we continue to try and optimize that test over time using more contemporary data so that we make sure that we are tracking that that biomarker is working sufficiently as we progress forward in the future.

Mike Wi: In one of the presentations of a different group, they mentioned that you guys consider to be, by the FDA, a level one evidence type of platform. So I'm assuming that the Artera AI test is FDA approved. Am I correct?

Mia: The Artera AI prostate test is a laboratory developed test, just as other genomic-based tests on the market. We are not FDA approved, and it is just the nature of this type of test. It is not required to be approved by FDA.

Eric Milson: Can one volunteer their slides to help the AI learn and increase accuracy? So are you looking for volunteers?

Megan: That's a fantastic question. Thank you for that, Eric. We don't have a process currently for individuals to donate their slides to help with our research. But I will say we work with a lot of data from institutions, clinical trials, or just medical institutions who conduct research and develop databases at their sites. We oftentimes will collaborate with those sites to leverage the data that they have developed.

So I would say a great place to start would be reaching out to your doctor and seeing if there are opportunities for you to contribute your information for research purposes if you're interested in that.

Howard: Yeah, thanks. Okay, Jeff or Bill, do you see any other promising questions? Anybody else want to raise their hand?

Bill: Well, there have been a few more questions, but we're getting real close to the two-hour mark here. Our guests have been more than accommodating in that capacity. We may need to wrap it up.

Howard: I think that's a good suggestion, Bill. This has been fascinating, and my cup runneth over. I'll say that this has been one of the best webinars that we've had. For the active surveillance pioneers that are on the call still, we at ASPI are always in need of your support. We're always in need of volunteers to edit, to help us with fundraising, and lots of different opportunities. Bill is our executive director, and please get in touch with us.

Also, if you go to PCRI this year in LA, stop in and see us at the booth. We'd love to meet up with you face to face.

So thank you, Mia and Megan, and all of you who participated. Thanks again for everything you've done and all your time. The video will be available in roughly about a week, and everyone will get a notice to that effect. I'm sure it's going to be very well watched.

Megan: Well, thank you. We really appreciate the opportunity to come and chat with you, and it's been a fantastic discussion.

Howard: Thank you, Paul, for all the work that you're doing in the active surveillance space.

Megan: Well, thanks, Megan, Mia, and let's not forget the third M - Michael.

All: Thank you, Michael.

Search This Blog

Informed Prostate Cancer Patient Support

Evolution of Active Surveillance - AlteraAI in Low Risk Prostate Cancer

ASPI - Evolution of Active Surveillance Aug 26, 2024

Introduction

Summary of Chapters, Key Moments

14:07 Dr. Marks

33:31 Meghan Tierney, Mia Li-Burton from Artera

Mia Lee Burton (Artera):

Megan Tierney (Artera):

1:01:09 Q & A

Edited Transcript

LeonardMarks, MD, professor of urology at David Geffen School of Medicine at UCLA

Mia Lee Burton. Medical Science Liaison at Artera

Megan Tierney

Q&A

Comments

Post a Comment

Popular posts from this blog

Cancer patients and doctors team up to change how cancer drugs are tested | Fox News

Member Roundtable Discussion - YouTube

Extraperitoneal robot assisted laparoscopic prostatectomy with Versius system: single centre experience | Prostate Cancer and Prostatic Diseases