Matthew Galsky: Thanks, Sam. Thanks for having me. I'm going to focus my few comments here on ctDNA. utDNA is emerging rapidly in terms of the potential clinical impact, but it's not quite at the point that we are at with ctDNA. So, I think what we know about ctDNA from a clinical practice standpoint, so for those of you who see patients every day, what do you need to know about ctDNA in bladder cancer? So I would say, one, ctDNA tells us who needs adjuvant treatment. I think we have sufficient evidence to make that a true statement right now, and I'm going to briefly show you some of the data that supports that. I would say, I would give a yield sign to the statement that ctDNA tells us who doesn't need adjuvant treatment, and I'll explain what I mean by that. And then, I would give the stop sign or the caution sign to the statement that ctDNA tells us who doesn't need neoadjuvant treatment. I don't think we're quite there yet, and I'll show you why. So, the adjuvant setting in bladder cancer, really the perfect setting to study ctDNA, right?
This is the holy grail of adjuvant treatment is that we know by definition that we overtreat patients in the adjuvant setting. We know that many patients are cured with surgery alone or neoadjuvant chemotherapy followed by surgery. So, giving patients additional treatment who don't need additional treatment is certainly something that we want to avoid. And we can simply look to the phase three data in the Kaplan-Meier curves from the CheckMate 274 study as shown here as evidence for this. You can see the placebo line. A large subset of those patients never recur. Those patients are cured already. They don't need adjuvant treatment. We need to know who those patients are. And so, I describe this as this double biomarker dilemma. In the metastatic setting, we have a single biomarker dilemma. We have a bunch of drugs. We know that they don't all work in all patients. We need to define which drugs work in which patients. In the adjuvant setting, we have a double problem. Not only don't we know who benefits from certain drugs, we don't actually even know who needs certain drugs. And those are often conflated, but they're actually separate considerations and might require separate biomarkers.
I would say that the most promising technology for determining who needs treatment is certainly circulating tumor DNA. We know this from several studies, including the retrospective analysis of IMvigor010, which I'm showing you here. And based on that analysis, you can see that patients who had detectable ctDNA at a single time point after radical surgery seemed to have a benefit from adjuvant therapy with Atezolizumab in this retrospective analysis, whereas patients who did not have detectable ctDNA on that first test after surgery did not seem to have a benefit. I think this is a distinct consideration from patients with undetectable ctDNA being "out of the woods". And I make that point because I'm going to contextualize this when I show you in Figure 11 results. And what I mean by that, of course, here is that you can see that in patients in those top curves, the ctDNA undetectable group, while in a retrospective study, it was hard to show a benefit to adjuvant therapy in that group. It's not that all of those patients are cured, and you can see that based on the DFS curves and the OS curves. So, those are related but distinct considerations. I'm going to come back to that point. CheckMate 274, for those of you who missed this, we have now shown very similar findings to IMvigor010 in CheckMate 274 and a retrospective analysis, again, using the same ctDNA technology, Signatera. If you look at patients who have detectable ctDNA and after surgery, that's where the benefit of adjuvant therapy seems to be enriched. And it's more difficult to show that in patients with undetectable ctDNA.
Again, I'll make the point that if you look at these DFS curves and look at the patients who were randomized to placebo, if you have 90% of patients on placebo developing metastatic disease with detectable ctDNA and, say, 30 to 40% in the ctDNA negative group, one could imagine that you need a much, much larger sample size in the ctDNA negative group to demonstrate a benefit adjuvant therapy versus a group where 90 plus percent of patients "needs treatment". So, that takes us to IMvigor011. So IMvigor011, the first prospective study seeking to establish clinical utility for ctDNA testing, designed very similarly to IMvigor010, except one, restricted to patients with bladder cancer. So, patients with upper tract disease were not included in this study. And two, after enrollment based on pathological eligibility criteria outlined here, patients had ctDNA testing with Signatera. And this is where the study gets a little bit interesting and somewhat difficult to follow the results. And so, I'm going to break this down. Patients with detectable ctDNA on that immediate post-op test were randomized to Atezolizumab versus placebo. Patients with undetectable ctDNA on that baseline postoperative test were followed every six weeks with serial Signatera testing.
If they developed detectable ctDNA during that period of time, then they entered the randomization. And so, you can enter the randomization immediately after surgery or in a delayed fashion. This is somewhat akin in prostate cancer if you did a study of adjuvant therapy in patients with detectable PSA immediately after prostatectomy, or you also allowed patients with biochemical recurrence in that same study in a delayed fashion. It's not typically how we do these types of studies, but of course it's very pragmatic and answers to important questions. So, I reconstructed some of the results of the study because the study is not, the results are not described in this manner in the paper because the study was meant to pool all patients with detectable ctDNA who underwent that randomization, whether or not they were included immediately or in a delayed fashion. But of course, when we meet patients after surgery and they have an undetectable ctDNA test, you need to make a decision about adjuvant treatment or not at that time point. You don't have the benefit of knowing what happens over the next year. And so, if we break down the results in this manner, 459 of the 607 patients had a baseline negative test on that first test. And then, of those patients, about 22% had conversion to detectable ctDNA during that first year.
That sort of rings true to what I showed you in those retrospective data sets with DFS curves in patients with undetectable ctDNA. And then, a very small subset developed metastatic disease before their ctDNA converted. That's a really important finding. I'm going to show you one more slide on that. 148 patients had detectable ctDNA. So, here's the results broken down by patients who had detectable ctDNA at that first test. You can see in the top curve, placebo versus Atezo, DFS and OS. And then in the bottom two curves, DFS and OS for patients who had detectable ctDNA only later, at a later time point during that year surveillance follow-up. If you treat patients immediately after surgery or in a delayed fashion, there is a DFS and OS benefit to immunotherapy versus placebo. Very important finding. This study, really these bottom curves shows that molecular recurrence of urothelial cancer is a clinical disease state that can be intervened on in a meaningful way. This figure from the paper is discussed very frequently, and I want to provide some nuance to this figure. So, this shows patients who had a serial undetectable ctDNA during that year of follow-up and what happens in those patients.
If a patient developed detectable ctDNA, they were taken off this curve and they were entered the randomization. So, those 22% of patients who developed detectable ctDNA are not on this curve because they were taken off this curve and entered the randomization. What this really tells you is if a patient has undetectable ctDNA serially for that year, what's the likelihood they'll develop metastatic disease before converting to ctDNA detectable assays? And you can see that likelihood is very small. That's reassuring. That's somewhat expected. This is the concept of conditional probability. The longer event doesn't happen, the less likely it is to happen. Patients with serial negative CAT scans also do quite well. You don't know that a patient's going to have a serial negative CAT scan when you meet them the first time after surgery. So, very important finding, but I think it needs to be interpreted based on what it's actually showing. So, my take home from IMvigor011 is that a first study to show a significant improvement in overall survival with adjuvant immune checkpoint blockade, markedly important finding.
A first study to show that treating molecular recurrence improves outcomes. Incredibly important finding. I think this notion of patients with undetectable ctDNA, can we observe all of those patients? Showing that you can treat on molecular recurrence is not the same as establishing non-inferiority for treatment on molecular recurrence versus immediately after surgery, and that's where I would argue we need additional data. That really takes us to MODERN. MODERN is currently enrolling and it's asking two questions. One, in patients with detectable ctDNA, do we need to escalate treatment even more than single agent immunotherapy? I would argue we do based on if you look at the curves in patients treated with Atezo with detectable ctDNA from IMvigor011, we got to do better than that. And then in patients with undetectable ctDNA, what I was just describing is treating on molecular recurrence in that subset of patients who will need that, estimated to be about 20 to 30% is treatment on molecular recurrence non-inferior to treating everyone immediately after surgery in the traditional adjuvant setting. Last point I'm going to make before I close is how about the neoadjuvant setting?
We have data from NIAGARA with ctDNA using the same platform, Signatera. If you look at the frequency of detectable ctDNA in that relatively large data set, about 57% of patients had baseline pre-neoadjuvant chemotherapy detectable ctDNA. And if you look at the EFS curves, here's patients with detectable ctDNA at that time point versus patients with undetectable ctDNA. So, it's definitely prognostic, definitely provides important prognostic information. So, why did I say that we are not quite ready to not give neoadjuvant therapy in patients with the baseline undetectable ctDNA? Is if you look at this slide, all of these patients got neoadjuvant therapy and the vast majority went on to surgery as per protocol. And you can see that even with neoadjuvant therapy and surgery, the EFS is not 100% in patients with undetectable ctDNA at baseline. Plus you can see in the curves on the right that if you add Durva, you seem to improve outcomes even further in that ctDNA negative group beyond chemotherapy alone. So, if you can still boost outcomes further in a ctDNA negative group, that tells me that a baseline ctDNA negative test is not sufficient right now based on current technology to say that patients don't need neoadjuvant therapy. I'm going to stop there. These are my conclusions that I started with. Hopefully I convinced you of these statements. Thank you.
Sam Chang: Matt, that was great. A combination of evidence and a little provocativeness at the end. So, let's start with, all right, next test, what do we do? You ended with the current tests, we're not able to actually determine, make those changes. As we improve the test, as you know, really, the proof is that initial evaluation of the tumor that you have, the DNA that's actually gathered and then evaluated serially, are we going to get to the sensitivity that, say, the breast cancer colleagues have, to be really able to then determine who needs neoadjuvant versus who doesn't? You think that's just more data, more time, etc., or do you think... Tell me what you think.
Matthew Galsky: I think there are two considerations. One is whether or not this is a technology issue that just needs to be solved? And I think yes, we are going to get closer with better technology, better sensitivity. And of course, what we didn't talk about at all is the pragmatic issues in the neoadjuvant setting. Current tumor-informed testing can take four to six weeks to come back. You don't necessarily have that time to make your neoadjuvant decision. So, I think that's in part going to be solved. I think this notion of whether or not there's biology here, that's not necessarily just a technology issue. So, what about tumor dormancy in those patients with undetectable ctDNA in the adjuvant setting who later develop recurrence? Would better technology have identified all of those patients? Or maybe some of them have dormant tumor cells in some deposits in the body, are not shedding ctDNA, something changes and that leads that cancer to grow at a later time. Maybe we can never detect those with our current approach to ctDNA.
Sam Chang: Right. So, that leads me to the second question, because that adjuvant group that are in fact ctDNA negative, that's the group that kind of interests me the most in terms of identifying those who eventually have issues and are we going to be able to then, again, prevent this? But is there also a safe point? So, if you look at your curves, you said overall things look quite good, that you can definitely see that, but obviously concerning that you have this metastatic group that's ctDNA negative. But over time, and I don't know this in terms of the study makeup, are they going to continue to follow those patients? Because it seemed to have a tail after the two-year mark. Obviously, we don't have longer term data, but do you think we'll get to the point where, hey, after two years, we look really good? After two and a half years, as you stated, and this is what I tell patients, the longer you go, the longer you'll go. And so, tell me your thoughts about that.
Matthew Galsky: So, I think from IMvigor011, I don't of course know the plans for long-term follow-up, but I assume there will be long-term follow-up. The serial ctDNA testing in that study, importantly, that was for a year, every six weeks for a year. And so, what ctDNA undetectable assay serially for two years means versus one year, I don't know. That we'll get from that study. And then I think what I had started to allude to, but maybe I didn't complete the point when I went through the slides is that even though ctDNA with current technology is not perfect and a baseline postoperative test for determining 100% of patients who will never relapse, it doesn't tell us that based on all the data. That's not necessarily the same thing as saying that it's not good for selecting patients who need adjuvant treatment. It might still be good because, again, there's this difference between needing treatment and benefiting from treatment, and maybe it's those patients with later recurrence who just don't benefit from immune checkpoint blockade, and that's what we're seeing.
Sam Chang: Yeah. And then, finding the other treatment or another treatment or another series of next step therapies that will actually be then beneficial for that group of patients. Matt, thank you. As always, people don't know that you've actually gotten up early for this, and I gave medical oncologists a hard time, but know just the opposite. We have Dr. Galsky here who has donated his time. So, thanks again, Matt, and look forward to catching up with you again.
Matthew Galsky: Thank you.