Read the Full Video Transcript

Sam Chang: Hi, I'm Sam Chang. I'm a urologist at Vanderbilt University in Nashville, Tennessee, and we are very fortunate to have one of the rising stars in urologic oncology.

I've been fortunate enough to know Betsy Koehne for a long time, and she's just completed her fellowship at one of the University of Ws we were talking about, University of Washington. And now, she is a badger at the University of Wisconsin.

Betsy was actually picked by the AUA to be one of the presenters of the key takeaways for kidney cancer. We asked her to focus on the key takeaways for localized kidney cancer. That was for the AUA meeting in 2025 in Las Vegas.

And Betsy, thank you so much for spending some time with us, and we really look forward to your presentation and your viewpoints. And hopefully, we'll be able to have some time for questions afterwards.

Elizabeth Koehne: Thank you. Thank you so much for having me. It's great to be back and yeah. Had an awesome time at the AUA meeting. So I'll look forward to sharing some of the things that I found.

All right. So these will be the key takeaways from localized kidney cancer at the AUA this year. And I'm going to start off with trial sharing some updates about active surveillance for small renal masses. So I'm going to start with a prospective registry trial. And it's a multi-institutional study. And it's the delayed intervention and surveillance for small renal masses.

And so they have a cohort of over 900 patients from 2009 to 2022. And these are patients who came in with small renal masses, and they either went on active surveillance up front or they right away had a primary intervention for their mass.

And so they did an overall survival analysis using one to one exact propensity score matching to look at survival outcomes and cancer treatment outcomes. And so using their matching, they found there was no difference in overall survival for these patients.

They also found an interesting metric, which is that the median overall growth rate was just over 1 millimeter a year, which is a really quite small. And then they had a couple interesting cohorts. One was about 16% of patients that had more growth. So more than a half a centimeter per year. And then a smaller group, about 11% of patients, that had no growth at all.

And so these will be interesting groups to follow over time. They didn't go into much depth about that at the meeting, but they did say that the group, who had growth more than a half a centimeter a year, that didn't really affect oncologic outcomes. So yeah. So we look forward to hearing more from them for that study.

And then next, there are a number of groups using artificial intelligence. And so I'm going to highlight two groups using artificial intelligence for localized kidney cancer. First, this is a group from Mayo who used AI to calculate the 3D contact surface area of the tumor base in patients who are planning to undergo a partial nephrectomy.

And they found that a higher contact surface area was associated with increased odds of this composite outcome that they made, which essentially was a way of calculating morbidity from the procedure. And so it included 30 day complications, positive surgery margins, and perioperative blood transfusion.

And so they found that an increased contact surface area was associated with increased odds of this outcome after adjusting for other clinical pathologic features. And the AI did all of the calculations itself. And so it was a completely automated system.

And then one other group, this was a group from Cleveland Clinic who used AI to also look at a patient's pre-op CT scan. And so they asked AI to predict a patient's age based on their CT alone. And so they found that AI in general was pretty good at predicting the age, but there was a group of patients in which AI predicted their age was older than their actual age. And so they called this the AI age discrepancy.

And then they looked at that group, and they found that this AI age discrepancy independently predicted a longer length of stay and decreased overall survival after localized kidney cancer treatment. So I thought that was a pretty novel finding and interesting use of AI.

So those were some of the main highlights from the localized kidney cancer presentations at AUA.

Sam Chang: Betsy, that was great. I found all three of those actually quite, quite interesting in different ways and for different viewpoints. So I was just going to ask a question about each one. So let's start with the last one, looking at artificial intelligence attempting to predict age based on a CT scan. And when they predicted actually a patient was older, just as we say, boy, that patient looks older than stated age.

Interestingly enough, a lot of times I think we may be off, but AI actually predicted a higher complication or longer hospitalization and actually, a higher mortality rate is what they found. Do you know what factors the AI was looking at when it came to the CT scan imaging?

Were they looking at percentage body fat? Were they looking at sarcopenia? Or how did it learn from the imaging and convert it to an age?

Elizabeth Koehne: Yeah, that's a great question. And that's something that people asked about in the session too. And the answer was-- and I'm not an AI person, but essentially, they said that they can't tell. They don't know, or it was almost like it wouldn't work if they tried to really make AI break it down that much. So that is pretty interesting. And--

Sam Chang: So I assume what they did was basically here are all the CT scan images, all basically-- I'm assuming with or without contrast but consistent. And then here are the age and that's it.

Elizabeth Koehne: Yeah.

Sam Chang: That's really very interesting. Yes.

Elizabeth Koehne: Yeah. And they did have expert surgeons and medical students, and so they also tried to guess the age. They ended up having several different groups of people, and they did find that the expert surgeons were pretty decent at it. They were at the top. And then the less experienced clinicians or no experience were lower down. But yeah, it was just a fascinating use of AI to make a new frailty metric. And--

Sam Chang: Yeah. I mean, it'll be fascinating to see how that compares to some of these symptom scores or index scores that we're trying to accumulate, especially I mean, you have familiarity with the folks at University of Washington and all the work that Sarah Psutka has done, which has been really, really, I think, important as we better recognize these individuals. But perhaps AI can help us even more taking in different factors.

Let's look at the second one, looking at AI from Abhinav Khanna at Mayo. I have the opportunity to speak with him years ago as he was clicking on images and trying to actually gather data to determine what we can do next with all these images that we have accumulated for so many patients with renal masses. And looking at that surface area that it computed, were there any other factors that AI could help determine that led to a determination of who's more likely to have complications or transfusions? Or was that by far the best and most predictive?

Elizabeth Koehne: So that was the only one that they presented in this study. I think it'd be interesting to know how it compares to the renal nephrometry score, something like that. They emphasize that this use of AI and same with the age one but also completely automated and so doesn't require any human effort.

Sam Chang: Measurement and evaluation. Yeah. Exactly. Really, I mean-- I mean, as we go further and further and learn more of its capabilities, we're going to be, I think, more and more astounded by what the predictive capabilities of AI is going to be in so many different settings. And these are just two examples of giving us a little better idea to be able to counsel patients what may or may not happen. So I think it's really we're just scratching the surface. I think it's really important. I think, Dr. Pierorazio and going to the first abstract that was presented with the multi-institutional cohort from Penn, from Hopkins, and from Columbia, I remember when he first started this with others in those institutions.

And Phil, I think, is the only one who's done the trifecta. Well, I guess Dr. McKiernan. No, he hasn't been to Penn, but Phil has done the Columbia as a medical student and then spent time obviously at Hopkins, is now at UPenn.

And so I want to focus on the two cohorts that you mentioned within this group of almost 1,000 patients. I think the key message that we should all realize and remember, small renal masses, surveillance versus surgery, there's really no difference in overall survival. I think with long term follow up, I think that's the key.

But just as you pointed out, I think, the two cohorts that will really be interesting are the fast growers and then the slow or no growers. So the fast growers, gosh, you would think that they'd be more likely to have more aggressive either cancers or worse outcomes and whatsoever. But at this point, just as you said, we haven't been able to rule that out, but it will be interesting to see down the line as we get more information, more follow up, does that really predict something worse?

I mean, those individuals that I follow, the ones that do grow more rapidly, for sure, we tend to biopsy honestly. And it may be-- we may for sure pull a trigger into some intervention more quickly than those that are slow growers. So it'll be, I think, interesting. What do you think about that cohort as well?

Elizabeth Koehne: Yeah, I agree. I think this presentation was a great snapshot from their study, which is so really useful in the clinic when we're seeing patients. And I think based on their work and others, for a while, we've known that active surveillance is safe for these small renal masses.

It's always good when seeing patients coming in for the first time with these small renal masses to be able to continue to reassure them that surveillance is safe. And I think that those two cohorts, the fast growers in the no growth people will be, really the most interesting. And I wonder if there are ways that we can go back and predict that from an earlier place and then with more granular, larger readouts of their study, learning more about the outcomes, particularly including biopsy, which many of us use in the clinic, especially in that faster growing group.

And so they didn't include their biopsy results or how that weighed in in this presentation. But certainly, I think was people's number one question and will be as they continue to share their findings.

Sam Chang: Well, I'm glad I'm not an outlier because that way you anticipating. My next question was how did biopsy information either get accumulated or influence decision making or predicted. So we'll look forward to their data that's coming out.

And then that group that doesn't grow. I'm so glad that's been reported because I have patients that clearly don't grow. And then my question is, can I stop following them? They tend to be patients that are older and/or have more comorbidities and you follow them. And honestly, I switched to ultrasound or something less invasive or start stretching it out. But I never feel totally comfortable to stop.

But boy, it'd be great to know if after a few years, they don't grow, is it safe them to stop or really stretch out surveillance? But then it begs the question of if it stops growing, why does it stop growing? It started from something at some point. And so I'm so glad, number one, it was reported because we've seen those patients. I think all of us have. And we're like, what do we do?

And it's great to know at this point that it's obviously safe to continue to surveillance. But obviously, if we knew those patients, like you said beforehand, we for sure wouldn't be operating on these patients or intervening. So it goes back to maybe we use AI.

Elizabeth Koehne: Yeah, I was thinking that they need to get with the AI groups.

Sam Chang: Exactly. Exactly right. It's like they've got all the images. They've obviously got de-identified, but they've got the images because they've been doing surveillance and following and just putting that all in into-- I mean, it's probably not enough to be totally predictive, but gosh, get some information of let's look at the 11% that didn't grow. Let's look at their images over time. Boom. Tell us, can you predict what are the indications?

And anyway, really, I think three very thought provoking abstracts that I think will all help us in practice over time. The small renal mass cohort obviously is really important, just as you said, for counseling. I think the two AI abstracts that you highlighted really give us an idea regarding the capabilities of AI and where we're going to be going in the future.

And I mean, I think really most importantly, it highlights the excitement that I think the AUA brought in terms of different oncologic tumors and what we've done and what we can do. So Betsy, I really appreciate you taking the time to share, and I really understand why the AUA chose you as being one of the key presenters for the key takeaways for kidney cancer. So thanks for spending some time with us.

Elizabeth Koehne: Thank you so much for having me. And thank you so much for talking with me.