Read the Full Video Transcript

Zach Klaassen: Hi, my name is Dr. Zach Klaassen. I'm a urologic oncologist at the Georgia Cancer Center in Augusta, Georgia. We are live in San Antonio, Texas, at the AUA 2024 annual meeting. I'm delighted to be joined by Dr. Steve Williams, who's the chief of urology at UTMB Galveston. Steve, thanks so much for joining us.

Stephen Williams: Well, thank you so much for having me today. It's always a joy to come and speak with you, UroToday.

Zach Klaassen: We always have a little bit of fun, don't we, on these?

Stephen Williams: We tend to.

Zach Klaassen: We tend to. We're going to talk today about some really exciting data you presented at this meeting, looking at an AI biomarker for non-muscle invasive bladder cancer. So just tell us, by way of background, why is there a need for biomarkers in this disease space?

Stephen Williams: Sure. Well, like most disease spaces, we focus solely on clinical predictors. That's all we've had for such a long number of years. And then bladder cancer, for whatever reason, even only until recently, we've risk-stratified bladder cancer. In 2016, I believe it was low, intermediate, and high-risk non-muscle invasive bladder cancer, and then really now we're just trying to refine it and better understand particularly patients that also may respond to treatments. The historical treatment, of course, is BCG intravesical, particularly for patients with high-risk, non-muscle invasive bladder cancer, because as you know, those patients have an increased risk of recurrence and progression. And really, we're just trying to better understand the response to those treatments and kind of move the needle forward.

Zach Klaassen: Yeah, absolutely. So tell us, walk us through the data you presented here. It's a histopathology AI biomarker. Tell us about the validation, development, and some of the key findings from that study.

Stephen Williams: Sure. Well first, over 300,000 specimens were used in order to quantify and develop this algorithm. Importantly, a little over a thousand patients with high-risk, non-muscle invasive bladder cancer were included. But really, the development of the algorithm is quite intriguing. An outstanding team, and over 60 million microns of tissue was examined by this model to develop it. And then really what we ended up doing and were presented here, there's several presentations, but one of which that I presented was really trying to understand also the applicability in identifying recurrence but also progression among patients that have been treated by BCG and then really trying to determine the next steps for subsequent therapies. And really, what's really exciting, and I don't have the graph to show, but you really are able to separate the patients into a low and high risk, and these are high-risk non-muscle invasive bladder cancer, but if they have the markers present or not, then determine what subsequent therapy they should proceed with, which separates it much cleaner than the historic EORTC risk strat or any of the other clinical predictors.

And the thing I really love about it is the ease of a urologist being able to interpret it, not a urologic oncologist, an academician, but really your everyday community doctor. And to have that discussion, it really empowers our patients as well to make the decisions. And then hopefully over time, as the model will develop and become more robust, then it's only going to get better.

Zach Klaassen: Yeah, absolutely. I think, and you've mentioned this in your presentation too, there's some real key aspects of this to delve into a little bit further too. Several countries, several institutions scanned across multiple scanners... which is important when you're using a digital test... and then looking at high-grade recurrence-free survival as well as progression-free survival. And so really the generalizability of this right off the bat is quite good, isn't it?

Stephen Williams: Yes, and that's what we wanted to do is really make this generalizable. And then something that I think we need to move forward with is get away from a lot of, I guess, the handwritten annotations, the actual report, and we test it out at our facility. It's digitized... we use a digital scanner, everything's online, you upload it via portal, you download it, and it's really a seamless process to where everything is also embedded into the EHR and protected. So I think this is... it's not the future, but this is where we need to be. And I just love that now we are bringing this live and online and we're actually testing the value stream with our patients, particularly at UTMB. So being a part of really the development, testing, validation, and then now coming up with the actual sheet of paper, if you will, although it's online.

Zach Klaassen: It's a PDF.

Stephen Williams: Yes, it's a PDF that you're able to share with the patients, and that's where we need to hopefully make this so easy, undeniably beneficial so that we can improve the care of these patients.

Zach Klaassen: Let's delve a little more into your experience because you guys are one of a few sites that's now using the test. And what's your experience been with the ease of use? You mentioned it's easy to use, your nurses are using it, they're able to interpret it. The patients, what's the feedback from them and what's the experience been at UTMB with the early access to it?

Stephen Williams: Sure, exactly. So we're one of the early access sites that's testing this. And really, as you mentioned before, the value stream to our customers, the most important stakeholders. But really in regard to the work to upload it, the nursing staff have found this incredibly easy. It's not labor-intensive. Obviously, you have to be computer savvy, but not really, you've just got to be able to use an online portal, upload the system information, but also too from the pathologists. So our pathologists, we just have an exciting team. We have an entire digital lab pathology so they don't use the traditional methods that have been used historically. And they're able to quickly upload the images, the slides if you will, and then we're able to get a report within 24 hours or less. And for me, from a value, I do, say, a procedure on a Tuesday, I get the specimen read by Wednesday with our pathologist team.

They upload the image, we get the results, the patient comes back on Thursday and I'm able to review that and delineate not only their diagnosis but what's their response to a treatment, BCG, and then their recurrence and progression. That's where we need to move medicine, to where it's seamless and interactive. And then the patients love it because we've also provided some nuances to the report for the providers, but then we have a great just one box that just tells you, yes, you have an increased risk of progression and recurrence with BCG, or proceed with BCG. And that part is really beneficial.

Zach Klaassen: When you've seen these AI high-risk patients and maybe BCG's not the ideal treatment, do you still give them BCG with the caveat that you're being really careful that they're already sort of designated as high-risk or are you completely switching to something else?

Stephen Williams: That's a great question, right, because it's only with BCG and we haven't tested particularly with Gem/Doce or any of the other novel agents, but we're having exciting discussions and are hopefully going to be able to provide that information. So I mean, it's taken right now that if patients, and we leave it up to them and the providers, but for myself for instance though, if someone did have... if they are going to respond to BCG of course you proceed, and then if they don't, it depends on the shared decision-making. Also too, as you know, I have a number of clinical trials, including the BRIDGE trial at UTMB where we're comparing BCG naive patients to BCG versus Gem/Doce.

Combining all these together I think it does provide some value-added, but then at the same time you leave it to interpretation. But most patients when they're able to see that data and that information either they have increased confidence or it gives a moment of pause. And then describing the toxicity and if they want to lend, although we don't have level one evidence at present but will be coming soon, then you could go to Gem/Doce, is kind of our go-to, if you will, now. Although we have a number of other agents coming online shortly.

Zach Klaassen: Yeah, absolutely. Let's put our crystal ball on display for a minute and let's look at the next 12 to 24 months for this biomarker. Where do you see it going over the next one to two years?

Stephen Williams: Sure. Well, for one, we have some exciting publications that are coming forward. So I think really providing the evidence out there first, and hopefully also over time commercialization. So then bringing this readily available to also our payers. And then in addition, just getting the word out to particularly the community. And the one thing I love about this is we all don't have the academic medical center, high-volume GU pathologists, just across the country. Right? And now what we have is this algorithm that really has an AUC of the curve up in the high 90s, so high fidelity, reliability, we're able to get those, I guess, pathologists that we've all dreamed that we wanted to have everywhere in the world, real live. And that's really the most important factor.

Zach Klaassen: That's a great point.

Stephen Williams: And value-added, cost-effective. You no longer need to send your slides to wherever, X, Y, and Z. You have it right at your fingertips.

Zach Klaassen: That's great.

Stephen Williams: And that's where I see this, hopefully in the next 12, 24 months. Also too with our guidelines, I'm a member of a number of the groups that help formulate these guidelines, so I imagine this is going to take suit and it's really an exciting space. And I just have the fortunate opportunity to work with an outstanding team. Our patients are highly appreciative that we're investigating, pushing the needle forward, and then leveraging technology as it should be. Right? Value-added but still have that human component.

Zach Klaassen: And as you know, we have a ton of agents in this space and biomarkers like this one are going to really help us sort of delineate maybe where somebody goes with a treatment versus something else.

Stephen Williams: Correct, and in a way that our patients can interpret it.

Zach Klaassen: Yes.

Stephen Williams: So we have to speak their language. And sometimes it varies, right? But you need it. This is you're going to respond, this is one you won't. But also having the confidence to make that determination, and then hopefully get our patients to the right treatments and then testing that prospectively to support all of this work and knowledge. It's really an exciting time.

Zach Klaassen: It is, it is. Always a great discussion. Anything we haven't hit on you want to talk about? Maybe a couple of take-home points for our listeners?

Stephen Williams: I think just be rest assured that hopefully we're going to provide some very useful tools in bladder cancer and then really holistically across the disease space, whether it be in bladder cancer or prostate cancer. And I think from what I see in the sense of energy is that there's no longer fear, I guess, of artificial intelligence and other things in regards to how it's going to impact healthcare. But now we're using it as value-added tools. I think hopefully this will further exemplify that effort and then provide our patients really the necessary tools and resources so that they can make the most educated decisions as we embark upon their journey together.

Zach Klaassen: That's great. Well, congratulations to you and the Value team. It's exciting data. We look forward to more work coming from that group.

Stephen Williams: All right. Thank you very much.

Zach Klaassen: Thanks, Steve.