2. UroToday Home
  3. Video Lectures
  4. Non-Muscle Invasive Bladder Cancer

Identifying an Aggressive Subgroup in Low-Grade Bladder Cancer with a Molecular Classifier - Ewan Gibb

November 25, 2025

Ewan Gibb discusses the development of a molecular classifier for identifying aggressive TA low-grade bladder cancer tumors. Dr. Gibb describes using long non-coding RNA clustering on the UROMOL cohort, identifying an LC2 cluster with significantly worse recurrence-free and progression-free survival. The team developed a supervised machine learning classifier with 179 gene features and validated it on the Knowles cohort, identifying seven LC2 patients with six experiencing recurrence within 24 months. The classifier reveals distinct biological signatures including higher proliferation and FGFR3 activity but depleted immune signatures. Dr. Kamat discusses how this tool might complement existing clinical risk stratification systems like the IBCG scoring system, with Dr. Gibb explaining the classifier could reclassify low-risk and intermediate-risk patients to higher-risk categories based on genomic aggressiveness. 

Biographies:

Ewan Gibb, PhD, Assistant Professor, Department of Urologic Sciences, UBC Senior Scientist, VPC, M.H. Mohseni Institute of Urologic Sciences, Vancouver, Canada

Ashish Kamat, MD, MBBS, Professor of Urology and Wayne B. Duddleston Professor of Cancer Research, University of Texas, MD Anderson Cancer Center, Houston, TX

Related Content:

Low-Grade Ta Bladder Tumors Show Unexpected Metastatic Potential in Long-Term Study - Alexandre Zlotta

Genomic Landscape of Non-Muscle Invasive Bladder Cancer from UROMOL Consortium - Lars Dyrskjøt

AUA 2025: Active Surveillance for Low-Grade Intermediate Risk Bladder

SUO 2024: Low-grade and High-grade Intermediate-risk Non-muscle Invasive Bladder Cancers Respond Differently to Bacillus Calmette-Guerin and Gemcitabine-Docetaxal
Read the Full Video Transcript

Ashish Kamat: Hello, everybody and welcome to UroToday's Bladder Cancer Center of Excellence. I'm Ashish Kamat, Urologic Oncologist in Houston, Texas. And it's a pleasure to be joined in the forum by Ewan Gibb, who's assistant professor at the Department of Urological Sciences at UBC, and a senior scientist at the Vancouver Prostate Cancer Center. Ewan, welcome and tell us about the reclassification and the molecular classifier for aggressive tumors in an area that's quite problematic clinically for us, the low-grade bladder cancer patients who sometimes go on active surveillance sometimes go on aggressive therapy. So really looking forward to hearing what you have to share about your presentation that was done at IBCN recently.

Ewan Gibb: Absolutely. Well, first of all, thank you for the opportunity to be here. It's great to be able to talk about some of the work we've been doing. So I'll just jump right in. Me and my group, we're really focused on clinical biomarkers, and so the way we approach this problem is to start asking what the clinicians are facing in their workflows for stratifying patients better. So in this, we started to think about TA low-grade because this is, as you said, a group of patients with largely indolent type tumors, but there's a number of them that have high recurrence rates and this cost of quality of life burden to the patient, but also cost in the system.

So the clinical question we're asking this particular study is, could we identify patients with TA low-grade tumors that actually have higher rates of recurrence, and can we then watch them more closely or treat them more aggressively? And this sort of falls into the biomarker challenge. Can we develop a supervised machine learning classifier to actually achieve this? And so what we did in initial work is we took the UROMOL cohort, which about 270-some TA low-grade tumors, and we did long non-coding RNA based clustering and identified a four cluster solution. And interestingly, one of these clusters, this LC2 cluster, actually had unique biological and clinical characteristics. And you can see here on the panel on the left that the recurrence free survival was much worse for these patients, but also the progression free survival was significantly worse, and this held up on MVA as well.

So we took these labels and we decided to develop a supervised classifier to identify these patients using the UROMOL cohort and the Knowles cohort for validation. And so the basic workflow for this is we use the LC2 as labels for the classifier and we take UROMOL and Knowles and we do what's called quantile normalization, so we balance out the expression profiles across the two cohorts and then using a machine learning approach, we basically reduce the gene features down to features that predict best the LC2 characteristics. And we ended up with the final model of 179 features and we deployed this on the Knowles cohort for validation.

So when we look at the results of this validation run, you can see here the results for Knowles, the model LC2 patients, we identified seven of these, and of these seven, we had six recurrences within about 24 months. And this was significant on MVA as well. Interestingly, this Knowles cohort is actually quite a bit less aggressive than UROMOL, but you still find we could capture these patients quite accurately. And rather than doing the standard ROC or AUC validation, we opted for a biological validation since this is the biological classifier. And you can see here that the pattern gene expression signatures between Knowles and UROMOL are very, very similar. These patients have higher proliferation signatures, higher FGFR3 activity, but are depleted for immune signatures and for Sonic Hedgehog. And that is the fast and short version of my IBCN talk.

Ashish Kamat: Great. Thank you so much, Ewan. Again, really appreciate all the work that you do in this field. One of the things that we look at clinically and when we're looking at patients with, say, TA-low-grade cancer in front of us, the predominant group of patients that we see is that intermediate-risk bladder cancer patients. And as I'm sure you're aware to try to address this, the IBCG International Bladder Cancer Group has developed a risk-scoring system back in 2014, which we then refined in 2022, and that's been validated in multiple different cohorts of thousand-plus patients. So a question to you, is there an incremental benefit in using this RNA classifier over clinical parameters? For example, did you test and see if it improved the performance of the IBCG risk-scoring system or any other scoring system? How would you recommend we clinicians use this information in our clinical counseling of patients?

Ewan Gibb: Actually, that's the question. So in the actual, the original UROMOL study we published in EUO that included TA-low-grade high-risk patients, but really what the model here is doing is taking low-risk and intermediate-risk patients and restratifying them to the high-risk category based on the data we have so far. Of course, requires some validation work, but so far so good. It looks very, very promising. So clinically it would be an additional parameter that you could use in stratifying your patients to particular risk group in that the genomics are actually aggressive, whereas a clinical presentation may be less aggressive. So it could add a little bit of more dimension to what you're looking at, what you're working with.

Ashish Kamat: No, absolutely. So just to clarify, this essentially is still the low-grade patients that you're saying is a higher risk category, right?

Ewan Gibb: Right.

Ashish Kamat: Exactly. And the IBCG risk criteria, essentially multiplicity of tumors, size of tumors, number of tumors, et cetera, kind of categorizes these patients into three different tiers. It'd be very interesting to sort of pull all this data together and see whether we can... And I think we can, because biology always trumps other things. I think we can further refine which patients, for example, we may not be quite as comfortable putting on active surveillance because recurrence is one thing, but progression is another. Help us understand a little bit the progression definition. Did you mean progression to muscle invasive disease or even include grade progression?

Ewan Gibb: So this was progression to muscle invasive disease based on the UROMOL cohort. Unfortunately, the Knowles cohort is less aggressive, so didn't have any progression events at all, so we didn't really validate that particular aspect of the classifier. Future cohorts we're recruiting should include some progression events, so we'll be able to confirm that that actually works for this classifier as well.

Ashish Kamat: Okay, great. Great. And again, I presume with the relatively small numbers, as you mentioned, you didn't really have a chance to see whether this is a predictive prognostic of what treatment was offered to the patient, or did you?

Ewan Gibb: No, unfortunately not. The treatment is confounding variable in the UROMOL cohort and in Knowles similarly. So we need better clinical data for that aspect, definitely.

Ashish Kamat: Absolutely, and that's one of the reasons we're doing this, right?

Ewan Gibb: Absolutely.

Ashish Kamat: Because the viewers of UroToday, I'm sure there'll be many that will be interested in collaborating and we'll of course put your contact information at the bottom, and I would encourage anybody that's listening to this to reach out, and I'm sure you're looking for collaborators, so that'd be great.

Ewan Gibb: Absolutely. Yeah, I'd love to work with other people. Definitely.

Ashish Kamat: Right. Thanks so much Dr. Gibb for taking the time sharing with us. Right to the point.

Ewan Gibb: Okay. Thank you very much.