Read the Full Video Transcript

Pedro Barata: Hello, everyone. I'm Pedro Barata, a GU oncologist out of Case Western Reserve University, University Hospitals in Cleveland, Ohio. Today, I have the true pleasure to be joined by Doctor Laurence Albiges. She's a world-renowned medical oncologist, also the head of the Department of Oncology at Gustave Roussy in Villejuif, Paris, France. Welcome, Laurence.

Laurence Albiges: Thank you, Pedro.

Pedro Barata: And I'm super excited to talk to you about one of another-- one of the great data sets that you presented recently at ASCO GU around COSMIC-313. And you did a great job looking at the overall survival data. And it's also important-- and very important, in my opinion-- to look at the data sets that we have, not just when the trials are positive, but also what can we learn from other trials and the big effort put into it, by the way, from you and your team, and what can we learn moving forward, in this case, for triplets. So I'm hoping we can talk a little bit about it.

Laurence Albiges: Yeah, sure, with pleasure. And thank you for the highlight. So as you stressed, we came to the final results of COSMIC-313 that I'm going to walk you through. So we will be presenting and discussing the data that have been presented at this ASCO GU '25. And these data are important because takeaways, as ASCO asked us now to highlight, show that we now have 45 months of follow-up for first-line treatment triplet strategy that still demonstrates a benefit in terms of progression-free survival, but not in overall survival, as I'm going to highlight.

And the safety profile is in line with what was reported before, but clearly highlighting the challenges of triplet strategy. And another item in this presentation that is important is that it was the first biomarker analysis built on this triplet trial. So I'm going to walk you through all those data. And what is important to understand is that COSMIC-313 is the first of its kind because it's the first triplet phase III trial that has reported, which assessed cabozantinib plus nivolumab plus ipilimumab—that I'm going to refer to as the triplet—versus the doublet nivolumab plus ipilimumab, one of our standards of care in patients with advanced RCC, clear-cell RCC, IMDC intermediate and poor-risk.

And in the past, the primary analysis had demonstrated a PFS benefit favoring the triplet. At that time, the median follow-up was about 15 months, and the PFS benefit had a hazard ratio of 0.73. And this benefit was observed, actually, in the IMDC intermediate risk patient population. So at this ASCO GU meeting, we presented the secondary endpoint of overall survival, the updated PFS, the safety, and also the biomarker analysis.

And I believe that the UroToday audience is very familiar with COSMIC-313. And this is the study design—clear-cell RCC, not previously treated, randomized 1:1 to the triplet versus the doublet. And so the primary endpoint was PFS, but a secondary endpoint was overall survival, the one we are discussing today.

So that's the patient disposition. Overall, 1,200 patients have been screened. Over 800 have been randomized, 420 each in both arms. We now have a median follow-up of 45 months. The vast majority of patients have discontinued treatment, mostly due to disease progression. And these are the baseline characteristics, well-balanced between the two arms. The spread between IMDC intermediate and poor risk is 3:1, so 75% were IMDC intermediate.

And otherwise, these are very standard baseline characteristics. I'm going to draw your attention to the fact that only 65% of patients had prior nephrectomy in this study, meaning that many patients still had their primary tumor in place.

This is the updated PFS in the ITT population. So with this 45-month follow-up, we still see a benefit favoring the triplet—hazard ratio is 0.82. And the median PFS moves from 11.2 to 16.6 months. So we have the same results, same finding as the one we had in the initial PFS report. What is very important is this overall survival final result. As you can see, the hazard ratio is 1.02. The two curves are clearly one on top of the other.

And with regard to the median, both of them are in the range of 42 months. Bear in mind that these are IMDC intermediate and poor-risk patients. There are no good-risk patients in this study. So no benefit in the ITT.

What about the subgroup? We're going to look at both PFS and OS by IMDC risk group. And it was established that the benefit in PFS was mostly within the IMDC intermediate group. And this is still true with more follow-up. However, this is not converting into a signal for OS, even in the IMDC intermediate patient population.

This is the forest plot of the different subgroups. Unfortunately, none of those subgroups really stand out favoring the triplet. OS was comparable between arms across most of the subgroups that have been investigated.

What we can discuss is: What about subsequent therapy? We have more follow-up. How were the patients treated in case they had further disease progression?

So the first information is that about every other patient received at least one line of systemic therapy. Subsequently, in both arms, the time to this subsequent treatment was longer in the triplet, in line with the PFS benefit. And the nature, the type of therapy, was mostly single-agent TKI, as we are doing in routine practice.

Interestingly, about 10% of patients still had sustained PD-1/PD-L1 inhibition. We now know that, based on CONTACT-03 and CheckMate 214, this is not something we are routinely doing anymore. And in this dataset, almost no patient received subsequent HIF-2α inhibition.

What about safety? I think this trial is important in how we are able to provide a triplet strategy to our patients in terms of safety. And it is interesting to highlight the fact that the median average daily dose of cabozantinib was actually quite low—22.4 mg in the triplet versus more than 35 mg in the placebo plus IO/IO arm.

But also, the number of cycles of ipilimumab. If you click, you will have the animation highlighting those two aspects.

The number of ipilimumab cycles was reduced in the triplet arm because of safety. And I think this is very important. It captures the fact that we were underexposing our patients to ipilimumab in the triplet arm.

And so, with regard to side effects that led to treatment discontinuation of at least one of the components, that was extremely frequent in the triplet—every other patient had to discontinue at least one of the components of the triplet, versus 25% in the doublet.

The type of side effect is listed below. Nothing is really standing out. It's mostly liver toxicity, as assessed by AST and ALT increases.

So I'm going to switch gears to the second part, which is about the biomarker analysis. This is really more food for thought, as it is an exploratory biomarker analysis. What was performed was RNA-Seq analysis with two different topics—one looking at molecular clustering and one looking at what we call immune cell deconvolution. And both of these aspects have been correlated with the clinical outcome in the trial.

With regard to molecular clusters, this is something we are familiar with based on the data generated in the IMmotion program in the past, as well as in JAVELIN (phase III). We are seeing the same distribution of clusters within our trial.

And when we correlate that with outcome, well, it seems that none of the molecular clusters are clearly associated with a specific benefit of the triplet.

So this is a bit disappointing—both in PFS and OS. We cannot use molecular clustering to define our strategy, triplet versus doublet, as of now—at least in this dataset, where each cluster was quite small.

And the second and last part is what we call deconvolution. Here, the hypothesis is that multiple targets of cabozantinib are expressed on immune cells, and we are able to use RNA-Seq data to assess the relative proportion of the different immune cell types.

And once we have this information, we associate those different relative cell proportions with PFS and OS. Interestingly, one type of cell stands out—M2 macrophages, or M2-like macrophages. These seem to be the best predictors of both OS and PFS with the triplet strategy.

So I'm going to move on and highlight here that M2-like macrophages are well-known to contribute to tumor growth, invasion, and metastasis because it seems that these cells are able to suppress the immune response. And it has been associated with a dismal prognosis in many cancer types. And indeed, in our dataset, it is associated with dismal prognosis, poor IMDC risk, and more visceral metastases.

So what about the role of the triplet? Well, interestingly, if we look carefully to the left at PFS and to the right at overall survival, and if we divide the population between M2-like low and M2-like high, it is interesting to see that the addition of cabozantinib seems to reverse the dismal prognosis of M2-like high phenotypes.

And here, clearly, we are seeing that those patients in red, who have a worse prognosis in both PFS and OS, are reversed back to the other group of patients when cabozantinib is added, as shown in the dark blue curves.

To summarize, this study is very important because this is the first report of overall survival for a triplet randomized phase III study. Unfortunately, despite the benefit in PFS, no benefit in overall survival was observed. The safety profile is consistent with the prior report. We could not see, with regard to biomarkers, a relationship between the molecular cluster and outcome.

But this exploratory immune cell deconvolution model helps us to identify, potentially, the role of M2-like macrophages in terms of dismal prognosis or poor prognosis features, and the fact that adding cabozantinib may revert M2-like macrophage-mediated immune suppression, though this will require further validation.

It takes a village to do this study. It's really great that both Exelixis and BMS could partner to test the triplet strategy. And hopefully, we will be able, in the future, to define the patient population for which a triplet could play a role. Thank you very much, Pedro.

Pedro Barata: Wonderful overview, Laurence. I mean, you did, again, a fantastic job—a big, big effort. And so, to me, this is such interesting data. And I'm really happy that you were able to show it and highlight some of the points. So maybe, since I have you here, let me pick your brain on one or two things. Obviously, the first is the safety, right?

I think when we look at the data, we are all commenting on how many people were not able to complete the treatment induction, which you highlighted in the EP. What's interesting, though, is that PFS is almost the same as an IO-TKI, right? Around 16 months—shy of a year and a half. So it appears to me that the addition of a TKI did something, maybe not on the response rate part, but maybe, yes, on the PFS part. However, we don't see a survival advantage.

So is that kind of your interpretation as well? Or do you have a different way—when you look at the data, what are your insights on that?

Laurence Albiges: No, I'm in line with your assessment. I think the fact that it was a blinded study induced some degree of confusion. Because when you have toxicity, then you have to dose-reduce, potentially, your TKI, either cabozantinib or placebo here. But also, if your patient was receiving IO/IO plus placebo, you could not rule out potential immune-mediated toxicity. And then, thus, reduce your ipilimumab when you're still in the induction phase.

And to me, that led to an underexposed position to both cabozantinib and ipilimumab. And so your point is well taken on the fact that—with regard to response rate, which I did not present here, but we had that in the first presentation—it wasn't as high as one could anticipate, maybe because we were underexposing our patients to the cabozantinib component.

And as discussed, we don't see a benefit in overall survival, potentially, from my perspective, because we did not use ipilimumab at its full potential. So the fact that many patients could not receive three or four cycles of ipilimumab may have jeopardized the long-term benefit we could anticipate in the study.

We are learning a lot from this study. If we had to build the study now, I don't think we would go for a blinded study because we would be better able to fine-tune how we handle toxicity. And by the way, now the triplet trials that are being built are not blinded.

Pedro Barata: That's a very good point. And Laurence, I'm thinking about subsequent therapies—it's interesting because even though patients got to second line later on, when you showed the subsequent therapy part, because they had longer PFS on the triplet versus the combo. It's interesting—the breakdown of the subsequent therapy is not that different, although you see more mTOR inhibition on the triplet than you saw in the doublet.

Did you find a median number of subsequent therapies to be different? Maybe I missed that part. Did you guys look into that or not really?

Laurence Albiges: So I would say that access to subsequent therapy was pretty much the same in both arms of the study. So to me, there was no risk of underexposure. The difference is in the timing to subsequent therapy, which, of course, is a bit longer in the triplet because PFS was longer.

And I believe that, in this situation, we've been using TKIs in sequence—the one that hadn't been used previously. And what was not presented here on the slide is that we had great exposure to cabozantinib in both arms. Even the patients that were in the CABO arm potentially subsequently received some. And those that were in the placebo plus IO/IO arm received subsequent cabozantinib.

So to me, the subsequent therapy is not an issue in this trial at all.

Pedro Barata: Gotcha. Fantastic comment. So I guess what comes to mind immediately after the safety part is, of course, looking at triplets, other triplets being explored, including maybe the same triplet in the PEDIGREE fashion, bringing the CABO later on, right? But then, also, different triplet therapies being explored.

How do you see that overall? Philosophically, we could be asking the question: When are we going to stop adding drugs on top of already existing treatments? We start with one drug. Now, the doublet is a standard. We're trying triplets. There are also quadruplets being explored. Do you see a limit to it? Or do you see a world where, maybe with biomarkers—which maybe will be my last question—we can actually figure out that some people might be OK with the doublet, some people might be OK, maybe, with monotherapy (maybe a minority of them), and some may need more treatment intensification, perhaps a triplet or a quadruplet? How do you see that frontline moving forward in the next couple of years?

Laurence Albiges: Thank you. It's not an easy question. I think we would love to have a biomarker to help us define treatment selection. But really, there are some patients we're seeing in the clinic who have very aggressive disease. You want to have a TKI because the patient is highly symptomatic. You're not sure you’ll be able to get to a second line for some patients. And here, I would love to be able to offer a triplet because we know that TKIs can induce tumor shrinkage. But also, I want the ipilimumab or the IO/IO part to make sure that I get the long-term benefit.

So this is exactly the patient setting—highly symptomatic, very aggressive disease—where I would like to be able to use a triplet. Based on COSMIC-313, we are not yet using this, but I hope that the next triplet generation will help us address those patients with very aggressive clinical features.

That being said, the vast majority of our patients are not in this situation. And so I think we need a trial where we'll have biomarkers to help guide us toward one strategy or another. Now, in Europe, we have the CARE-1 study, which is IO/IO versus IO-TKI, with PD-L1 being assessed under the hypothesis that IO/IO would be better in PD-L1-positive patients.

And so I hope that, in the future, we will be able to do that. The OPTIC trial, led by Brian Rini, is very interesting to me because they are using molecular clustering to define which strategy to pursue. In our dataset, this did not pan out, but maybe it's because of the sample size. And I believe that triplets or quadruplets will not be feasible in all patients.

So maybe the PEDIGREE strategy will be the right one—early rescue. But here, what is interesting to me is: Can we cure more patients? And this is still something we need to demonstrate.

Pedro Barata: That's wonderful. One last question before I let you go. You touched on that very important summary that you made on the biomarker part. We happen to have OPTIC as well here. And I can tell you that, in my own experience, it appears that we might be minimizing the PDs as best response by biomarker-selected approaches—IO/IO or IO-TKI—on that particular study.

You have French experience with BIONIKK. What's interesting is—do you have an explanation for why we come up with gene expression signatures in one specific combo within one specific trial, but we cannot validate them in a different trial with similar mechanisms of action? How do you take that?

Laurence Albiges: Trust me, I wish these findings would have been more positive. I was highly disappointed. I think there are some biases. The sample sizes are quite small. Did we really achieve the full potential of cabozantinib in this trial? I'm not so sure, based on the discussion we just had.

So, to me, the molecular cluster is very valid. The IMmotion program is very strong. JAVELIN confirmed that.

But the real question is: How can we translate this into a routinely applicable biomarker? And this is where we're not mature yet. You cannot do gene expression profiling routinely. We did it in BIONIKK, but that was a randomized prospective study. This is not something you can do immediately when you have to decide on treatment for your patient.

So we'll have to be able to translate those clusters into an immunohistochemistry-based test or an easily applicable surrogate staining method.

Pedro Barata: Well, always a pleasure chatting with you. It’s always super insightful, and I learned a ton. So thank you for taking the time. Thank you for walking us through the data. Again, congrats—important to report trials that are meaningful. We’ll keep learning from you and hope to see you soon.

Laurence Albiges: Thank you, Pedro. Thank you very much.

Pedro Barata: Thanks.