Read the Full Video Transcript

Sam Chang: Hi. My name is Sam Chang. I am a urologic oncologist at Vanderbilt University Medical Center in Nashville, Tennessee. And one of the joys of my life is being able to teach rising superstars in urology. And one of those is Dr. Bryn Launer, who is joining us today. She'll be presenting her work that was given at the AUA in 2025 in Las Vegas, focusing on a computational histology artificial intelligence biomarker. That's a mouthful. Dr. Launer will actually talk about the use of this marker in evaluating and personalizing risk stratification for patients with non-muscle-invasive bladder cancer. So Dr. Launer, take it away.

Bryn Launer: Thank you, guys, so much for having me. I'm really happy to be here with Dr. Chang in UroToday to discuss this exciting work. This was a multicenter collaboration with the Valar Labs, as well as multiple centers as you can see listed below, looking into this CHAI biomarker, which stands for Computational Histology Artificial Intelligence. And we actually found that it enhances the risk stratification for high-grade TA non-muscle-invasive bladder cancer. And here's what we found.

So for some background, this is an exciting biomarker that's been studied, recently published in JU 2024 in October, a validation cohort for this biomarker that could actually identify patients with high risk non-muscle-invasive bladder cancer and again identified patients at high risk of recurrence, progression, and BCG unresponsiveness after BCG intravesical treatment. So this high risk group was really exciting, and we wanted to know more about the powers of this computational biomarker and how we could use it.

For this study, we selected a subgroup of patients with high grade TA non-muscle invasive bladder cancer. And it's important to emphasize that these were patients with high grade TA on pathology regardless of risk stratification. So some patients were intermediate risk. Some patients were high risk, but they all had high-grade TA on their pathology slides.

And this is an exciting population because it's so heterogeneous. You know you have some patients with small non-concerning lesions that you may watch. You have other patients with multifocal, large volume high grade TA that you treat completely differently. But they're all in this same group on the pathology slide. And so the aim of this study was to report the prognostic performance of the CHAI biomarker to stratify risks of BCG-naive high-grade TA non-muscle-invasive bladder cancer patients and compare this biomarker to the current guideline risk models from the AUA and the EAU.

For this study, we included patients from 12 different institutions between 2004 and 2024. They all had banked pathology slides, and these were all BCG-naive patients with high-grade TA non-muscle-invasive bladder cancer. These patients when we included them in our study, were BCG-naive. And then all of the patients then went on to get BCG.

We used the available slides from these patients from all these different institutions, and we evaluated their slides with the biomarker at the time of their first resection with this institution, where they were found to have high-grade TA. And these slides had not previously been used in development of this biomarker. That's important to note.

We analyzed whole slide images, and we used statistical analysis as you can see here. In total, we analyzed 269 patients, 269 whole slide images for this study. And the patients had a median follow-up of 32 months. So here's our results. This is first analyzed by recurrence-free survival. And you can see here that we categorize the 269 patients in their slides by AUA and EAU guideline definitions for intermediate risk and high risk, including factors like focality and size and things like that.

And then in the graph labeled C, we use the CHAI biomarker. And that's just binary, present or absent. And so you can see here that we found no significant difference between recurrence-free survival rates for EAU intermediate versus high risk. But we did find a significant difference for both the AUA risk categorization and the CHAI biomarker categorization for significantly being able to predict recurrence-free survival.

We also looked at muscle-invasive progression-free survival. So again, by the EAU risk categories, the AUA risk categories, and by the presence or absence of this CHAI AI biomarker. And neither the EAU nor the AUA intermediate or high risk was able to statistically distinguish patients at higher risk for muscle-invasive bladder cancer progression-free survival. However, we did find a statistically significant signal for the CHAI biomarker, as you can see in the column on the right.

So in conclusion, this was a multicenter study, 12 institutions, over a period of 20 years, involving 269 patients and their whole slide images, which were run through a previously validated and locked AI biomarker. And we use this to statistically significantly prognosticate recurrence and progression-free survival for high-grade TA non-muscle-invasive bladder cancer.

Specifically, we found that the presence of these high risk prognostic CHAI biomarkers indicated significantly higher risks of high-grade recurrence and muscle-invasive progression which outperformed both the AUA and the EAU risk categorizations. Certainly, limitations of this study include retrospective design and comparison to an EAU risk stratification model that did look at progression risk from a historical population that was not receiving BCG maintenance therapy.

However, this is an exciting finding and these biomarkers may enable personalized medicine to be applied for our patients to provide more accurate prognostic value for each patient, especially those in heterogeneous populations like high-grade TA, even beyond our current guidelines. Thank you for your time.

Sam Chang: Bryn, what a wonderful presentation. The-- To clarify some points here, the CHAI, the C-H-A-I, the Computation Histology Artificial Intelligence biomarker, I like that CHAI eponym. So the CHAI algorithm or test has already been set based upon hundreds and hundreds of previous slides. Is that correct? So it's either a positive or negative, just so you said?

Bryn Launer: That's exactly right. It's binary positive or negative. And this is an already validated and locked cohort. It was previously published in JU, and it's actually already been CLIA approved as well.

Sam Chang: And so importantly, this is based upon just an H&E slide a hematoxylin eosin. You don't need anything else. You don't need special antibodies. No mutational markers. It's just from a glass slide. Is that correct?

Bryn Launer: That's right. Just from the glass slide. And they actually send it from many institutions as a digital image. So you don't even need to ship the slide itself. We can upload the digital image in our pathology lab and just email that file over.

Sam Chang: So Bryn with this in mind, you can see yourself actually being able to better counsel patients regarding, hey, maybe we need to move sooner to another therapy. Or maybe we need to consider, say, a timely or early cystectomy based upon the risk to muscle progression. Is that at least it generates that hypothesis requiring probably further study?

Bryn Launer: Certainly. I think that's very much on the table based on these results.

Sam Chang: So where do you go next? What's the next research project then?

Bryn Launer: So you know this has been CLIA approved. So we're looking next, I think, into clinical trials for saying, we can see that this biomarker, this tool based on our retrospective data and how we know these patients progress through their treatment course looking back did. Now, can we use it truly to help guide therapy for patients prospectively and going forward? And can we help patients find better treatments, better fits for them based on this biomarker?

Sam Chang: Right. And what is the turnaround time? Bryn, do you have any idea?

Bryn Launer: That's a great question. I think it's pretty darn quick. But I don't know the exact time frame.

Sam Chang: So today in practice as we advise our patients, this is just another piece of information that we could use in conjunction with urinary markers, or with imaging, or with obviously further histopathologic features like lymphovascular invasion or other high risk features. I can see it as being part of that evaluation process.

Bryn Launer: Exactly right. And I think probably similar to how we use some of these biomarkers in prostate cancer. And again, like you mentioned, other urinary biomarkers. I think this may be the most helpful in those borderline patients. You have a slam dunk patient with high volume, high-grade T1. They need their bladder out. We don't need to go down this pathway to futz around with those concerns.

But I think especially for the high-grade TA patients or some very small foci of other concerning disease, this is a great option. These patients can have long-term effects from intravesical therapy. And they're in a situation where they don't know if they're the right choice for that. And so I think a biomarker for these borderline patients can be really, really critical in guiding our practice.

Sam Chang: Bryn, thank you so much. We look forward to your future meteoric rise in urologic oncology as you pursue a urologic oncology fellowship. And we look forward to future presentations as well.

Bryn Launer: Thanks so much for your time.