Arcus Biosciences, Inc. (RCUS) Earnings Call Transcript & Summary

All right. Welcome back, everyone. This is another session of the 2024 Virtual Oncology Leadership Summit. My name is Yigal Nochomovitz, one of the biotech analysts here at Citi. As you know, the drill is if you have questions, just e-mail me, [email protected] and I'll do my best to monitor the e-mail and relay the questions to the management team. So it's my great pleasure to have with me the President and Head of Research, Juan Jaen of Arcus. Juan, welcome. Thank you very much. Obviously, you had some important updates yesterday on the call. And so let's start with those. But before we do that, if you could just give maybe just a 2-, 3-minute overview of the company. What are the key assets. And then as I mentioned, we'll start with the new data on the HIF-2 alpha.

Absolutely. So first of all, thank you, Yigal. Thanks for the opportunity to spend a few minutes here today. Arcus is an oncology R&D company. We presently have about -- not about, 8 different molecules in clinical development. The most advanced of those programs is our Fc Silent TIGIT antibody, which is presently being investigated in 3 Phase III pivotal clinical trials, all 3 of them in major areas of clinical need, an all-comer frontline metastatic non-small cell lung cancer population, Stage III inoperable lung cancer, in frontline upper GI adenocarcinoma. We -- over the next 12 months that we anticipate opening a fourth Phase III study together with our collaborator, our partner, Gilead, in the perioperative non-small cell lung cancer space. We also have a very promising small molecule inhibitor of HIF-2 alpha, a space where we believe that we are primarily competing with Merck's belzutifan. And if we have an opportunity, I'll share some of the specifics for why we think we may be developing a better differentiated molecule. We anticipate that molecule to also be in a Phase III study in clear cell RCC within the next 12 months that will make our fifth Phase III study at that point. And finally, we are pretty much, I think, at this point, the leaders in the adenosine space. We have a small molecule inhibitor of CD73, a molecule, quemli for which we presented data at ASCO GI earlier this year with some really, really promising data in frontline metastatic pancreatic adenocarcinoma. And we're moving as fast as we can towards the initiation of a Phase III study with that molecule. We hope that, that will occur towards the end of this year, early next year at which point that will make it our sixth registrational study with 4 of our own molecules moving forward. We're very well capitalized at the end of January, following Gilead's latest equity investment, we were sitting on $1.2 billion. But we're also very fortunate to have partners like Gilead, Taiho, AstraZeneca, Exelixis, they are helping us operationally in some cases, but also financially, move forward each one of these registrational studies. We also have a very active research engine that continues to feed the pipeline. And if we have an opportunity, we can talk a little bit about some of those earlier-stage assets. So I'll stop there and take your questions.

Okay. Well, the obvious focus today, given the market is HIF-2 alpha so let's begin there. For those that maybe didn't have a chance to listen yesterday and didn't see the slides, if you could just give us a quick recap of what you've shown, casdatifan or cas for short. You showed some initial early efficacy data. So can we go through that? And then as you mentioned, the arguments around why you believe this could be differentiated from the Merck drug -- from belzutifan.

Absolutely. So belzutifan, that was -- just a couple of months ago, received its first approval in advanced clear cell renal cell carcinoma is an inhibitor of HIF-2 alpha. HIF-2 alpha is a transcription factor that for many years now, has -- its profile -- its pharmacokinetic profile in humans has been well publicized. And so we set out to identify a molecule that could be as potent and selective, if not more, than belzutifan, but one that might not suffer from the pharmacokinetic limitations of belzutifan. In a nutshell, the limitation stems from the fact that once you reach a certain dose, it happens to be 120 milligrams, it's very difficult to achieve greater absorption, greater amounts of belzutifan in circulation in humans, and that's what's supported by the publicly available data from Merck. So given that the molecule across a number of settings and different studies, appears to provide broad benefit to patients with clear cell RCC as demonstrated by some of the single-agent studies where the large majority of the patients experienced some level of benefit. Yet, very consistently, the objective response rate is in the low 20%. In the most recent study, LITESPARK-005 that supported the recent approval as a single agent in advanced clear cell RCC, the median PFS, which -- just under 6 months. So we believe that there was an opportunity to do better perhaps if you could get more drug on board. If you could hit the target harder in a tumor type where it's generally accepted that the large majority of clear cell RCC tumors are being driven by an overactive HIF-2 alpha mechanism. The question stands out is, well, if that is true, and we are pretty much across the field, that's taken us as a given, then why is that only 22%, 24% of those patients reached the objective response threshold? And could you do better if you could only hit the target in the tumor harder? So we think we've accomplished that. By now, we've evaluated our compound cas in a healthy volunteer study, where we've defined with a lot of precision, the PK and PD effects of the drug. And when I talk about PD in healthy volunteers, obviously, I'm not referring to PD on the tumor. But the PD in the normal -- effects on normal physiology. HIF-2 alpha happens to be present in healthy kidney, where it regulates the production of erythropoietin. And so we -- before we even went into the cancer trial, we already understood a fair amount about the PK and peripheral PD of our molecule. The study that I think caused -- drew so much attention yesterday and today is a study that we refer to as ARC-20. It's a study -- it's our first study in patients. The study is made up of 2 stages, the dose escalation portion and in 3 different expansion cohorts in all cases, the population in all 3 expansion cohorts is the same. The only difference is the dose that's being evaluated in each one of those cohorts. And so the data that we shared yesterday was -- we spent a lot of time talking about the PK/PD advantages of our molecule in the early observations of both -- very promising clinical activity both in some of the patients in the dose escalation cohorts that obviously have been on drug longer as well as some very -- comments about the early signs of clinical benefit that we're seeing in the expansion cohorts.

Okay. Well, that's great. Let's -- you mentioned the PK/PD arguments, and I think that will be a good point just to go through that. I know it's a bit technical, but I think it would be helpful. You made the point around the EPO as a biomarker, not related to the tumor. And if you could just kind of go through that argument because as you've shown in the slides, you see a saturating effect on EPO as you move from 20 to 100 milligrams of your drug, of cas. But I believe your point is that while you're seeing the saturation on EPO because you have much higher doses as compared to what belzutifan achieved on the EPO curve that you would have potentially an outside impact on downstream genes being controlled by HIF-2 alpha that are relevant in the tumor. So maybe I'm just -- if I've characterize that right, tell me, but that's kind of...

Yes, I think it's generally correct. I mean it's a well-accepted principle in pharmacology or in biology that the amount of drug present in the blood of a subject is what feeds whatever amount of drug is going to make it to different tissues, but that extent of penetration into different places doesn't have to be identical. In fact, quite often it's not. So you're used to talk -- to hearing people talk about brain penetration of the drug or concentration in the liver. The tumor happens to be the one place where drugs in general tend to have the hardest time getting in. And in the case -- in the situation that we're discussing, primarily, we have tumors that are highly dependent on this one transcription factor for growth and survival. So we talk about peripheral PD as measured by the amount of erythropoietin, EPO, that the healthy kidney can produce. And we've known from the Merck data that about 2/3 of the EPO produced in healthy kidney is controlled by HIF-2 alpha, right? Once you reach that level of -- but -- a certain level of inhibition and you reduce it by [ 65, 70, 75, ] something in that range, going higher in dose, having more HIF-2 alpha inhibitors is not going to move the needle any further because just empirically describing the observation is that the rest of the EPO seems to be controlled by other mechanisms that are not HIF-2 alpha, okay? So that's a measure of peripheral pharmacology. It takes Merck or belzutifan a 120 milligrams and a certain amount of drug in the blood. It takes us -- we've now demonstrated this in the healthy volunteers as well as in our dose installation patients, it takes us only about 20 milligrams of cas to achieve the same level of peripheral -- by peripheral, I mean in the blood, in the healthy kidney, the same level of EPO reduction that belzutifan achieves at 120. We do it with significantly lower plasma levels because our compound -- just empirically, I'm just restating the empirical observation, our compound appears to have better partitioning from the blood into tissues than the belzutifan does. So that's sort of a normalizing point, 20 milligrams drug, 120 of belz. The second part of our premise is that -- and we've shown this in the updated corporate deck is that unlike belz, we can continue to go up in dose of cas and continue to see dose proportional increases in the amount of drug that's in circulation, right? So the question on the table -- which belz is unable to do. Question on the table is the dose of belz, which happens -- was selected primarily based on the pharmacokinetic limitations of the molecule, does that also happen to give you a maximal inhibition of the target in the tumor? Or is there room for more, right? We think about that dose of belz almost as not being random, but it wasn't selected based on any biological endpoint, it was just selected on -- can't go any higher. So -- and we're seeing some tumors shrinking, so it cannot be a bad dose. And so that's what we're trying to do is say, "Okay, is there more room for more biology, clinical benefit, if we can go to a higher level of a very effective HIF-2 alpha inhibitor. And what we've shown now in our PK profile, as I said, is that we can dose proportionately, so at 100 milligrams of our drug we have 5x as much as what is required to completely shut down all of the HIF-2 alpha dependent EPO in the periphery, right? And now we're really just at the beginning. In some of the data that we shared yesterday, and I think you could sense Terry's excitement, is the very early signs that start to suggest that our hypothesis may in fact be correct. We're starting to see with still relatively mature very early looks, particularly at the expansion cohorts, we're starting to see a profile that suggests to us that, in fact, we may be onto something. So belz tends to be a slow onset of benefit. The median time to response in the most recent study is just under 4 months. The majority of our patients in our expansion cohort at 100 milligrams daily have only had 1 or 2 scans for -- in our study is 6 to 12 weeks. And in spite of that, we're already at a place where we're seeing response rates that look pretty much in line with what belz recorded after a couple of years of follow-up. So our -- so the question on the table is, are we bringing in the responses earlier? Are we going to continue to see more and more responses? Or that's, I think, what we will continue to accumulate data on -- over the coming months and intend to present that more mature data in the second half of this year. The other thing that we talked about yesterday is that if you look at the LITESPARK-005 data that led to be a recent approval of belz, about 1/3 of the patients progressed by the time of their first scan. So literally, 33 -- almost 34% of the patients in that study, their first -- by 2 months have progressed. And the early indications from our expansion cohorts is that our, let's call it, primary progression rate is significantly -- and our CMO would argue clinically meaningfully lower. That's a number that's not going to change. Everybody that we have in the expansion cohort has already crossed that threshold. They've already had at least 1 or 2 scans. So we pretty much know what the ultimate PD rate will be in that expansion cohort, and it's going to be meaningfully different than what LITESPARK-005 reported. That number isn't going to change in this study.

So you're saying it's around the 20%, 22% from this trial, now the some of these -- you said you had -- some had 1 scans, some had 2. So you're still waiting for some of these early responses to be confirmed. Is that correct?

Correct. So we're seeing some responses even in the early scans but the majority of the patients in the expansion cohort have only been on study. The majority of those patients are still on study. But a good half of the cohort hasn't made it past the 3-month mark yet. And so even with a very early look at the expansion cohort, we are already -- we can already see response rates that are very much in line with what belz has reported very, very consistently. There are 4 separate cohorts that you can look at, at belz, the original Phase I expansion cohort. A recent study was also described at the most recent ESMO, comparing 2 different doses of belz. And then the fourth data is LITESPARK-005. And the objective response rate is uncannily consistent across those 4 groups, right around 22% to 24%. So it's a very solid objective response benchmark against which we can compare what we're seeing in our expansion cohort.

Okay. And then on the -- so that's -- just so we know when -- you said the next update, are we going to get the more detailed data on those 30 in the expansion in the second half of the year? Are we going to see the typical data plots like the waterfall plot and the swimmers plot, and things like that? Should we expect that....

That's our intent. That's our intent. Definitely, objective response rates that will probably almost certainly not be the final number because I anticipate many of the patients will still be ongoing and both for belz and even some of our -- one of our patients in the dose escalation, we've seen that those delayed responses do occur, don't get me wrong. It just -- it may very well be that we continue to see responses much later, but we definitely will be presenting fairly mature response rate in the second half of this year. Our intent is to share some aspect of PFS. We should be so lucky as to not have mature PFS data. The median PFS in LITESPARK-005 was just under 6 months. You can do the math. We may be able to infer that we're doing better than that. But our intent is to be fairly generous with the data that's available from ARC-20 in the second part of this year, and we haven't communicated yet the venue for that.

Okay. And then on the -- can we touch on the safety side, are there -- what are you seeing in the early safety that would suggest a different profile versus belzutifan, specifically on the anemia, for example, and hypoxia?

So we're seeing an incidence of anemia hypoxia that appears very, very similar to what's been reported for belz. Now -- so remember, our premise is not that we're going to be safer, it's that we're going to be equally safe and more effective. The peripheral safety signals is the direct result of pharmacology on the healthy kidney. And so if you inhibit 2/3 of the production of EPO in the kidney, human physiology dictates that, that's going to translate into x percent incidence of anemia. And so we don't anticipate. Now what we're seeing so far is that as we predicted that once you reach that point of peripheral saturation, we're not seeing a big difference in safety profile from 20 milligrams to 100 milligrams, which is, again, consistent with the fact that, okay, you turn the spigot down to -- by 2/3 and it stays -- you go up in dose in the periphery, it's the same number. So you would expect a very similar safety profile pretty much across the dose range. It's the tumor where we are anticipating more and more, greater and greater inhibition. And the fundamental hypothesis that at that low dose where you can max out the effects in the periphery in those tissues that are easily vascularized or well vascularized that the tumor requires a higher level. We used an analogy yesterday and some of our calls of everybody knows that -- and every -- I'm sure everybody can relate to this, everybody knows that pembro saturates PD-1 in the blood, probably at a dose of 20, 30 milligrams. It's saturated or oversaturated. Yes, you -- the current thinking in our field is that in the tumor may require -- this is probably -- it's a rule of thumb that we all apply to antibodies. It probably is going to take 10x higher to get to a complete inhibition in the tumor because it's harder for the drug to reach. And so we don't dose people at 20 milligrams every 3 weeks, you give pembro at 200 milligrams every 3 weeks. And so you can go down the list of many, many different kinds of class of -- types of drugs where the stuff that you measure in the lungs or the kidney or in the blood itself where drugs have an easier time reaching, you still go up in dose and associated plasma levels of the drug in order to get complete pharmacology in the organs that are harder to reach. And the tumors, a metastatic lesion is the extreme of difficult to reach by any drug. So that's pretty much what we're trying to squeeze out of this mechanism of action.

Okay. And then given you mentioned that example and then, of course, you mentioned EPO as the biomarker in healthy tissue. Are there tumor specific downstream genes that are controlled by HIF-2 alpha that at some point, you could interrogate and show, obviously, in addition to the tumor control data, but just show that you're -- like your thesis is that you're hitting the target harder and having a differential impact at the level of the renal cell tumor?

Absolutely. That is an excellent question, one that we're working very hard on. And I hope if we're lucky with some of the experiments that we're doing, that we'll have some numerical information at that point later this year, I suspect as the clinical program is moving that we will actually have the clinical -- that we'll be talking about the clinical, the numerical data that supports our hypothesis possibly before I can show you the biomarker data. So I suspect it is going to be a case where it's going to be after the fact that I'm showing you how HIF-2 alpha regulated gene X, Y and Z, went up or down as a function of dose but already with full knowledge of what the clinical activity is.

Sure, sure...

It's a program that's moving very, very, very quickly. It's enrolling like gangbusters.

Well, we know drug development isn't necessarily a linear process, so we understand that. And then speaking of the plans for the Phase III, can you provide any thoughts as to how you would structure that Phase III? What would be the potential design? Or what are the different scenarios for how you could combine belzutifan with probably other molecules that are already approved?

Right. So as you know, Merck is now talking about belzutifan as a blockbuster or a drug with blockbuster potential. And so they're already very active in a number of settings, including the one that -- the only clear cell RCC setting in which they just obtained approval. And so we're trying to be, obviously, as you would expect us, very careful about selecting combinations and settings where we can either be best or first. And when we announce -- later this year, I anticipate when we announce the specifics of what our first Phase III study will be, hopefully, you will agree with me that you can check both of those marks. But at this point, we're not -- we're definitely looking at a combination with a standard of care. It's public knowledge that we started a study in collaboration with Exelixis called STELLAR-009. There, we are evaluating primarily the safety and PK profiling of our drug in combination with their second-generation VEGF TKI, zanza. But we're also -- within our own ARC-20 study, we are in the process of enrolling or evaluating the combination of our drug with cabo. So we're keeping open the options. But you can see where that is going. It's going to be -- we're evaluating the safety and initial profiling of combinations with 2 VGF TKIs that are generally accepted, I think, as being probably your best choice when it comes to an agent in that class. As you know, Merck is, for better or worse, stuck with lenvatinib, an agent that in this setting is, I think, a lot of practitioners have a problem with that it's harder to manage the toxicity associated with it. So that's one of the areas where one might conceive that a better HIF-2 alpha inhibitor combined with an equally effective but safer and better to manage TKI might -- it might be a combination that has the potential to be best.

Well, so -- okay. So STELLAR-009 is with zanza the company you mentioned, but why is the cabo combo with 521 not part of that trial. You said you're just doing that yourself, right?

Yes. So -- at this point, we're trying to generate foundational safety and early evidence of activity data with both combinations. Over the coming months, we'll settle on which of those 2 doublets we're taking forward into a Phase III study. Clearly, the -- zanza is, at this point, an experimental drug. Cabo has been -- a well-accepted standard of care in frontline and later lines in clear cell RCC. So there's some pros and cons of going one way or another.

But -- so you'll tell us the designs, I guess, later this year for the Phase III, and that's starting -- I have in my notes that's starting early 2025. Is that right? For the Phase III?

Correct. And that's about as much resolution -- a level of time resolution as we should bring to that. We're still a year away from the most likely initiation of that Phase III study. And I included in my 12-month forward look of how many Phase III studies we're going to have but that one is right -- I would say it's probably right at the very end of that 12-month look.

Okay. Fair enough. All right. Well, we spent half an hour on HIF-2 alpha. So let's try to hit some of the other assets. You talked -- you mentioned quemli at the beginning. So there was some good data in pancreatic, the ARC-8 trial. So can you walk us through that? And of course, I mean, I noticed and others noticed that zim didn't appear to have much of a contribution there. So I just wanted to walk through that aspect of the data, too.

Absolutely. So the study in question is ARC-8. It's a study in which we evaluated the benefit of quemli, our small-molecule CD73 inhibitor in frontline pancreatic in combination -- on top of gemcitabine. Abraxane backbone, chemo backbone. And some of the some of the patients in the study received that combination plus our anti-PD-1 zim and another portion of the study did not. The data that we presented at ASCO GI last month clearly demonstrates that any differences in clinical benefit as a function of having zim or not having zim is entirely the result of the percentage of patients with liver mets in each particular group. So we did a very detailed analysis of that aspect of it. And when you look at patients with liver mets as a subgroup in patients with no liver metastases subgroup, the survival parameters for both of those cohorts is -- it's within days of each other. So our conclusion is that there's no benefit associated with the inhibition of PD-1. That's consistent with what we all know that in pancreatic, anti-PD-1s or PD-L1s really haven't had much activity when combined with standards of care. So that's very consistent with our observation. And we were seeing -- the remarkable thing, I think the most remarkable thing from that study is that -- and it's very consistent with an immune-linked mechanism of action is that numerically, the objective response rate was not indistinguishable from historical chemo numbers. In fact, we conducted several synthetic control-arm analysis, and we should talk about that for a couple of seconds, and I think it's a really new direction, not just for our study, but for clinical trial execution in general. We saw that a numerically objective response rate, that's not worth the benefit of inhibiting adenosine formation is going to manifest itself. An interesting benefit on PFS and an unbelievable level of benefit when you look at OS. So we ended up with a median OS in our study of just under 16 months, compare that in the synthetic control analysis that we performed, the synthetic control arm came in just under 10 months, which is smack in the middle of recent large Phase III study sort of benchmarks, which range from 9 to 11. And like I said, it has a ratio of 0.67 for the specific synthetic control analysis that we presented at ASCO GI. We conducted a total of 4 of those analysis, looking at different subgroups, what percentage...

Can you explain to everyone what -- how you did the synthetic because you use an external company that has expertise in this whole setup of external control -- I mean, synthetic controls and then how they pick the patients from other studies, but in a very specific way to propensity match. So can you...

Yes, absolutely. I'm happy to do that. Before I do that, let me just sort of walk through some real basic math, okay? If you have a place or a setting like frontline metastatic pancreatic, where there are at least 2, if not 3, very well-known, hugely prognostic features, the presence of liver mets, the geography of your -- the ethnicity of your patients, whether the disease is de novo or recurrent after prior locally advanced surgery, if you try and run a randomized study where you theoretically try to stratify for each one, even just a couple of those variables, you would easily end up with -- you would need well over 100 patients in your experimental arm and a similar number in your control arm. To try and run a randomized study with 50 patients versus 50 patients, yes, you randomize them, but you really didn't control for these prognostic variables. And whatever conclusion you might draw from that comparison is subject to the most likely unevenness of these elements. The beauty of a synthetic control arm is that you can go into a much larger pool, and you need to be careful about where that data is coming from. The data set that we use, that Medidata has accumulated contains, as I understand it, over 100,000 individual patient records. These are patients that participated in recent Phase II or Phase III multinational studies and received the standard of care, not the experimental treatment. So there's single patient data accumulated from those studies that allow you to do what we did in collaboration with them. We went in and said, "How many patients do you have in your database that were -- that received gem/Abraxane in frontline pancreatic in -- but studies that read out in the last x number of years so that you're not reaching back to the beginning of time." And the answer was over 500 patients fulfilled those conditions out of 100,000 patient database. Then they apply all kinds of statistical models that honestly, I don't understand but they went in and they did this for us 3 different times. Looking at the totality of our cohort, they did the patients that received only the no-zim combination and then they did it for the patients that received the zim, 3 completely independent analysis. And then when looking into the pool of patients for the ones that best matched all the prognostic features present in our study, but they did it without -- while being blinded to the results. They didn't know what the -- that there was a protocol for how the study was going to be done. They didn't know what the right answer was. And they did that 3 times. And so what we ended up with is in our -- the totality of the patients that received the 100 milligrams of quemli was 122 patients. They ended up with 122 patients that looked like the parallel universe alter egos of our 122. That mean down to the age, performance status, presence of liver mets, ethnicity. I mean it's like you can go down the list of every one of these things. And at the end of the day, it felt like a much better way to sort of say what -- in that parallel universe where these patients would have received only the chemotherapy, what would have happened? And that's the comparison that we ended up placing so much value on. We clearly beat out all the cross-trial comparisons, but there's always that question of are your patients different or whatever. So we were very happy with the results of those synthetic control analyses.

But your argument is that, I mean, it's obviously -- it's not a prospective randomized trial. But your point is that in some sense, the control is maybe even a better control in a certain sense because of the way you can select from a much larger database and find the more completely matched patients to what you actually enrolled in your active arm.

Correct. And I think that from a decision-making perspective, I think the regulatory agencies are sort of gradually becoming more and more open to the possibility of maybe a mixed real-world randomization, supplemented with some synthetic control analysis. But as a company -- and believe me, we try very hard to talk ourselves out of doing anything further. I think until we dug deeper into the last 10 years' worth of clinical trial development in pancreatic, we were -- we suffered from the same superficial, often repeated notion that pancreatic is so hard. So we tried to talk ourselves out of going forward, and we tried to challenge the observations that I just described every which -- way. At the end of the day, the data just kept coming back to, no, you're looking at a very, very strong signal. And so part of that process of getting comfortable with the meaning of the data also required us to dig -- spent a lot of time looking at the last 10 years' worth of failed Phase III studies in pancreatic. And our conclusion was that there were very few, if any of those failed studies that were preceded by any kind of proper Phase II data. In many cases, there were single-arm Phase I data, 12 patients, looks good, launch a Phase III study. Okay, fails. Okay. Well, so I think we're probably the first -- this -- our study is going to be the first registrational study in the last 10 years that's really preceded by what we think is a very robust and strong signal.

So then getting back to the clinical development, I think you've said you're going to be starting the Phase III early next year. What's the plan then as far as the regimen? I mean I have my guesses, but tell us is it going to be the triplet or the quad for the Phase III?

It's going to be the triplet. We would not be acting on the data if we ignore the fact that PD-1 inhibition seems to be adding nothing. It's not even close. It's nothing. And so -- yes, it will be the triplet. It will be a population very similar to the ones that we enrolled in ARC-8, I mean, if not identical. And yes, something that I should mention before I forget, is that completely independently of our ARC-8 study, our adenosine receptor blocker etruma was being evaluated by Roche in one of their MORPHEUS platform studies, pancreatic. We -- the first readout objective response rate differences was unimpressive. So we have pretty much written that study off until Roche share the results. It's a randomized study. So you're comparing gem/Abraxane to etruma on top of gem/Abraxane. In that study, they also have atezo in the experimental arm. Just at about the time that we were analyzing or ARC-8 with quemli, we saw the results looking at both PFS and OS results with etruma, same mechanism of action, different drugs, run in completely different places. In that case, a controlled study, albeit relatively small numbers in each arm. And I don't want to steal the thunder from Roche. They'll be presenting that in the coming weeks at a major medical conference. But I think that you will see that it's very consistent with what we've been talking about frequently. And we find that very reassuring.

Okay. In terms of activity on the adenosine axis?

Correct, yes.

Okay. We did have one clarifying question from an investor. Just going back to the synthetic control arm. I think you referenced 100,000 in the database and then 500 or 0.5% met the criteria. Can you just clarify, was that 100,000 pancreatic or that was all in the database?

Right. And clearly, I'm not an employee of Medidata. So my 100,000 is -- it's a ballpark, but in my recollection from conversation, that refers to all tumor types.

Okay. That explains it. Okay. Got it. All right. And then in the last bit of time we have here, we should touch on domvanalimab. So let's start with the expanded collaboration with Gilead. Can you walk through that?

Yes. So Gilead and Arcus are co-developing and we have -- we share not only that molecule, but at this point, they've opted into 4 programs, a total of 5 molecules. So we're codeveloping and which also requires us to pay for half of all costs. And then we share the commercial rights, our financial benefit at the back end. Operationally, Gilead is in charge of some studies. We run others. In this case, they are running the PD-L1 all-comer frontline non-small cell lung cancer in lung -- sorry, lung on top of chemo. We are operationalizing the frontline gastroesophageal adenocarcinoma study. AstraZeneca is operationalizing PACIFIC-8, which is a Stage II/III inoperable, it's consolidation after chemoradiation therapy. In that case, it's our TIGIT anti-body dom plus durva is in patients that have not progressed after CRT. The fourth Phase III study that we've talked about will be operationalized by Gilead. It will be in the -- its STAR-131. It will be in the perioperative non-small cell lung cancer setting, a place where we believe we could be not only best but also first, given what our competitors are doing.

Okay. And then I believe you're going to have an update from the EDGE-Gastric trial at, I think, ASCO and...

That is correct. So that is...

And what's the...

Yes. So that will be an update following our presentation late last year in one of the ASCO plenary sessions. So we describe data from our platform study in gastric -- in upper GI cancers, where we demonstrated that single-arm data set, but where we demonstrated that dom plus zim on top of standard of care was looking extremely promising. So this will be an update on that data set, it will have about 5 or 6 months greater maturity. But just to remind you of some of the data that we saw there, some of the data that we presented at the time, the progression-free survival at 6 months was north of about 93% in the PD-L1 high subpopulation, which we were -- we and others, we're extremely excited about. Remember that in the pivotal study, and you can go back and look at a couple of reference studies, but CheckMate 649, similar setting in backbone, the corresponding median PFS was just under 8 months in that PD-L1 high population. And so -- so it will be an update on that. We'll have -- I don't -- it will be primarily an update on response rate and PFS. We haven't done the data cut yet. I don't know if there'll be enough maturity, I suspect not to say anything about survival, but definitely it should be a really interesting data set. And that will happen at ASCO in early June.

And then there's also this other study, EDGE-Lung, which is the quemli, dom, zim in that.

Yes.

So -- but you had done the ARC-7. Everyone knows about that one from last year, etruma, dom, zim. So what's the difference? Why are you doing the EDGE-Lung trial?

That's an excellent question. So our EDGE-Lung and EDGE-Gastric studies are platform studies that we use for a number of purposes, signal generation, dose optimization, data sets. The main potential use of that combination may be one of the things that we're -- we have been intrigued for a while is the possibility of looking at quemli in the Stage III non-small cell lung cancer setting. There was a study run by AZ, described -- 1.5 years, a couple of years ago called COAST in which they demonstrated -- that was a fairly large-randomized Phase II study that demonstrated that their CD73 antibody provided a very significant benefit over the durva control arm, durva being the PACIFIC regimen, standard of care in that setting. And in fact, they're currently running a Phase III study PACIFIC-9 evaluating their CD73 antibody, oleclumab, in exactly that setting. We believe that oleclumab is a really -- I mean, if we find the results of that Phase II study, COAST, actionable, then by definition, we believe that the drug -- their drug is active in that setting, but we've shown preclinical data that supports the thesis that quemli, our small molecule inhibitor is more effective than oleclumab. So oleclumab is unique amongst all of the CD73 antibodies in clinical development that it's primarily effective on CD73 that's still attached to a living cell. As soon as the CD73 sheds floating around in solutions, there's a lot of it, then the way that antibody binds to soluble CD73 is only about 50% effective. So it's like trying to shut off a firehose but only being able to close the opening by half. That's a well-recognized limitation of the drug. So we have a compound in quemli. I think we didn't touch on this. It's a 5 picomolar inhibitor of CD73 that places a 1 or 2 orders of magnitude, even though it's a small molecule more potent than any antibody that I'm aware of against that target. Small molecule gets into the tumor more deeply. And so we think that we should be able to do better with quemli relative to oleclumab or any other CD73 antibody. And so back to your question, if nothing else, the data that we are generating in EDGE-Lung will generate a baseline of safety in combination with dom and zim that we may choose to -- use to build on and go in some of these directions that I've been talking about.

Just so everyone understands. When you say soluble, you mean extracellular setting? Is that right?

Yes.

Yes, okay.

Yes, so CD73 is literally anchored to the plasma membrane in the cells where it's expressed. And there's a lot of it that just gets shed into...

Your point is that your molecule binds the soluble, but it also binds the bound as well as the -- better than the antibody.

It's equally potent against both forms of enzyme. Oleclumab is primarily effective at inducing internalization. So it binds to the CD73, they're still attached to a living cell and induces removal, let's call it that. And in doing that, it seems to be highly effective. But anything that's already been shed, that continues to make adenosine, there's going to be 50% of that adenosine generation that cannot be impaired, no matter how much antibody you use. It's a -- has to do with how it binds to CD73.

Okay. We're running low on time, but I want to make sure we -- there is another program we don't talk about much. You have AB801, the AXL inhibitor. What is the status on that Phase I? And when are we going to get some data?

Right. So we just completed our dose escalation in healthy volunteers, which, again, we tend to do whenever we can, which provides a lot of information on the -- what I would describe as nothing short of spectacular human PK profile. We're very happy with that. Concurrent with that, we are in the dose escalation part of a study in cancer patients. At this point, we're still in the all-comer typical dose escalation type of population. And as soon as that's completed, we haven't really disclosed the specifics, but we will be expanding into a couple of different cohorts, slightly different patient subsets and slightly different combinations. But a couple of points I want to make. One is that the belief in AXL is that it's a common mechanism that cancer cells utilize in connection with becoming resistant to various forms of therapy to survival vest mechanism. The second one is that no matter what anybody may claim, there are a lot of TKIs that list AXL as part of their many kinases that they hit. I don't think there's anything that's ever been tested in humans as efficiently high levels to completely inhibit the AXL. They're typically -- even we were talking about cabo. Yes, you can quote inhibit AXL, but the doses tend to be defined by the toxicity associated with VEGF receptor inhibition. And so we -- I believe and we believe that we will have the first molecule anywhere that would allow us to test whether AXL inhibition is, in fact, effective at reversing that direct resistance profile of many types of cancer.

All right. I think we're going to have to leave it there, Juan. We could probably keep going, but we have another session in a few minutes. So thank you very much. Really interesting data across the platform, especially the recent data yesterday. So thanks again, and we will...

Thank you for the opportunity. And thanks, everybody, for listening in. Appreciate it.