CStone Pharmaceuticals (2616) Earnings Call Transcript & Summary

Great. Good morning, global investors. Thank you for joining CStone Pharmaceuticals' ESMO Data Review Conference Call. The call is going to be mainly focused on the CF2009 data readouts at the recent medical conference, ESMO. So today, joined our call, including Jason Yang, CEO of the company; and also Nicky Ni, CFO of the company. This is Ziyi Chen, China health care analyst at Goldman Sachs. I'm going to be moderating the call. Before we kick off the session, I would like to highlight that, this call is strictly for clients of Goldman Sachs and CStone only. And this conversation is not intended for the media and is off the record. Participants will be removed from the call if they cannot be properly identified. And this call is not for the purpose of sharing or receiving nonpublic otherwise confidential information. Attendees are public market participants who may not receive [indiscernible] request on public and otherwise confidential information about issuers or securities or about market securities. So we already see the data coming out, abstract and also earlier today in the day, there has been a presentation. So now I'm going to hand over the call to the management team to get started on the introduction of the data. Then we're going to potentially open the line for potential Q&As if you have any questions. Jason, please.

Yes. So that's great. Thank you guys for calling the data interpretation, also a little bit of update and a little bit of perspective. We have been very happy with the data. I think for Phase I, we just want to look at 4 aspects. I think typically PK/PD safety and the preliminary antitumor activity in the Phase I background and also in reference to similar drugs, right? And I think we see many of this data meet our expectations and quite a few actually exceed my expectation based on the design. So I'm happy to share with you what I think of this data, and what's the next step. Next slide, please. So this just shows you that we -- today, we talk about bispecifics, but we do have quite a few on the right side that will go into clinical next year. So we probably have 3 or 4 INDs next year. So we can talk to you guys later. Next slide. Next slide. Yes. Just to remind you that why we are here today, right? So we designed this molecule 4 years ago. And really the driving force is that we need to have a second-generation IO backbone that not only have [indiscernible], but really, really has OS that can beat the current standard of care, right, PD-1, PD-L1 plus chemo or maybe plus something else. But the bottom line is the IO backbone has to be improved. We have tried all kind of things, right? So recently, we -- I mean, TIGIT is recent example, and they also tried 4-1BB. In early days, people tried [ TIM-3 and X3 ] and all kind of stuff, but that really did not improve upon PD-1. So it's not yet systematically be PD-1 plus chemo, right? So in the end, we realized that CTLA-4 really, really is the only molecule that add to PD-1 can improve OS further, right, anywhere from 5% to 12% over a year of 10 years. And in all those trials, CTLA-4 only dosed 2 to 3 cycles because of toxicity. So now if I can make a molecule that contains CTLA-4 antibody, and it can be dosed together with PD-1 continuously for 1 or 2 years. So I would expect a lot more improvement of OS upon PD-1, right? This is because we know PD-1 does improve OS over standard of care. Now we want to improve OS upon PD-1 by adding CTLA-4. Now the VEGF is very important. It's not only important because it has the anti-angiogenesis function. More importantly, as I will explain further, so in the tumor microenvironment, it can create a framework that's actually tracking PD-1, PD-L1 -- sorry, PD-1, CTLA-4 in the system, so to make it work might be much, much harder, okay? Now the problem, of course, has been how do you make CTLA-4 tolerable, right -- safe and tolerable. So this actually bring us back to [indiscernible] of our price in the past several years. The first one, of course, is the function of CTLA-4, discovery of CTLA-4 as a break immune checkpoint, right, that actually block CD28 from overactivation because that's the so-called immune second signal to stimulate by binding CTLA-4 binding to CD80, right? So that's the first [indiscernible]. The second one is this year's peripheral immune tolerance, really just to say why we can -- the normal human body does not attack our own tissue. So that's the peripheral immune tolerance and that's TREG. Now the -- the TREG, the [indiscernible] function is the same because CTLA-4 is blocking CD80 binding to CD28. So there's no overactivation either. Okay, now the toxicity associated with CTLA-4 is actually coming from this tube. So on the right side, this tube, right? So your CTLA-4 antibody is so potent, it pull CTLA-4 from CD80 so that CD28 can get activated. Okay, so how can we generate or create a molecule that does not actually pull CTLA-4 from active T cell or Treg cell. Now if you look at CTLA-4, it's actually a dimer, right? And CD8 is a dimer. So essentially the binding is like a monoclonal antibody. So that's a [ nanomol ] binding very, very tight, okay? Now so then we make an antibody that monovalent, only one arm of CTLA-4. So the affinity will much weaker around the 20 to 30 nanomol instead of a single or sub-nanomol, right? So in this case, it will not be able to pull CTLA-4 from CD80 in the peripheral. Okay, now in the tumor microenvironment, that's different because tumor microenvironment has T cells that get exhausted as they all express PD-1, right? And then tumor express PD-L1. So interact with PD-1 that actually suppress effector T cell. Now we have the trispecific antibody that also contains PD-1 antibody. So in this case, PD-1 and CTLA-4 will bind to both PD-1 and CTLA-4. So in this case, they could have driven enhancement of affinity, restore the binding affinity of PD-1 -- monoclonal PD-1 antibody and also monoclonal CTLA-4 antibody. So this kind of synergy make antibody works [indiscernible] PD-1 and CTLA-4 actions in the tumor microenvironment. And this action is further enhanced by the VEGF antibody [indiscernible] to the VEGF bind VEGF-A, which is a [indiscernible], so causing cross-linking [indiscernible] work and that will further enhance the PD-1 and CTLA-4 affinity by turning forward to [indiscernible]. So in this, the trispecific antibody working on 3 targets synergistically within the tumor microenvironment, working very, very hard. But in the proper tissue, they are acting pretty flat, so not working so hard. So then we expect a lot of efficacy, but relatively low safety issues there. So that's good to see whether that we actually needed this goal. All right. So we started Phase I trial March this year, and we finished enrollment of 72 patients really, really, really fast, okay? Because this is a dose escalation with back failure. So it's -- as you guys have been probably just seeing a lot of trials, this is an unbelievable speed. Actually, we also started Phase II because with the data we have, we started Phase II already, and we expect to enroll treatment-naive non-small cell lung cancer by end of this month. And we should have enough data by mid of next year to start Phase III trial. So our plan is to engage U.S. FDA pretty soon for the Phase II expansion. And then by the middle of next year or Q3, we engage with FDA for the end of Phase II and the pre-Phase III negotiation so that we will have a dose and trial design and agree with U.S. FDA. Indication-wise, the most important is the chemo combo in first-line squamous, non-squamous non-small cell lung cancer. And then the other 2, one is in the second-line setting combined with docetaxel or something else against docetaxel. That should be a very easy win trial. That's post-IO, of course, because you will see post IO, the drug is working, our drug is working. So this will be a good indication to win. Of course, there is always the monotherapy in the first-line setting, whether we are going to do that or not, I think we will together make that decision with our partner because that's going to be a pretty challenging trial because enrollment, right, is not going to be easy. Okay, so this is our plan. So now let me just show you the data from the dose escalation. Next slide. As I just mentioned, 72 patients, the median age is a typical Phase I patients. So it's pretty old patients. And the trial enrolled more than 55% of non-Asian, right, non-Chinese, let's say, mostly it's all active from Australia. So 54% are White. On the prior therapy also typical Phase I, so almost 40% of patients have 3 and above 3 prior therapy. And then prior IO, so 51% of patients have prior IO therapy. Some of those have multiple rounds of IO, including bispecific. So essentially, every patient that should be treated with IO has been treated with IO, and there's also off-label treatment with IO. Angiogenesis, same situation. So everyone that should be treated with VEGF has been treated and it is also off-label treatment. So the tumor type, non-small cell lung cancer is the most numerical patients because we focus on that. Even though a lot of those patients have not yet had a first tumor assessment, we only have 17 as you will see, but we enrolled 33 already. Next slide. The first important readout for us is the safety, and that's also what our multinational company has been telling us. So they always said, if you can make this antibody work, we don't worry about efficacy, we worry about safety. So we -- if you look at the safety of the grade 3 and above TRAE, so right now, it's only 30.9% and the grade 3 immune-related TEAE, it's about 4.2%. We know this is a relatively short follow-up, but you also can tell that 20 milligram actually has been follow-up, 21 patients has been followed for quite a while since 3 to 4 months and 20 milligram per kg is not a low dose, it's pretty high dose. And also 1 to 10 milligram have quite a lot of those patients escalated from 1 to 3, from 3 to 10 and 10 to 20. So there has been quite a bit of up intra dose escalation to 20 milligrams. So you can see those data are much more mature. As you know that [ RE ] typically occur in week 4 to week 8, right? So I think for the dose 1 to dose 4, I think we pretty much all pass the day of week of 8, at least, right? So I think this safety looks -- even though it's immature, looks very, very attractive. Next slide. This is just an easy way to visualize the safety, the most frequent AE. So on the right side, you can see very few grade 3 TRAE. Next slide. So obviously, we want to look at relevant comparators, right? So in our case, we have a trispecific. So we definitely look at both PD-1 and CTLA-4 bispecific. PD-1 and PD-L1 plus CTLA-4 combo, CTLA-4 [ low ] and then PD-1, PD-L1, VEGF bispecific and the PD-1 monoclonal antibody. So if you just look at the Grade 3 TRAE and [ TRE ] leading to treatment discontinuation against all these classes, you can see right now, it's very, very favorable. We do expect where it goes up, but we think that it will be similar. At worst, it will be similar as the PD-1 VEGF bispecific. Next slide. So we also look at the RAE. So in this case, we look at any grade RAE as well as grade 3 RAE versus the same classes of molecules. Again, the frequency right now of RAE and the grade 3 RAE are very low. Again, I think even if the data mature, we will still be in a pretty safe places. Next slide. So we're looking into the detail of frequent RAE, typically skin rash and the thyroid disorders and [ STLT, ] right? So there are real ones, but we are not seeing many real ones yet because the sample size is still small. But even if you look at the more frequent ones, right now is much lower than the PD-1, PD-L1, CTLA-4 bispecific or PD-1 plus CTLA-4 monoclonal combination. So it's -- right now, it's very good -- it's very well tolerated. Next slide. This is actually interesting is to look at the VEGF related tox because we have the same VEGF that's from [indiscernible], right, bevacizumab as [ AK112 ]. And also the moderation is the same. AK112 and both molecules are 200 [ kilo ] [indiscernible]. So we have the same moderation. So meaning that our 200 milligrams contain 15 milligrams of bevacizumab as AK112 does. But right now, of course, with a shorter follow-up, if you look at the 4 type of VEGF-related tox, the hypertension, proteinuria, hemorrhage and the symbol of embolism. Right now, it's very low. So we have 2 cases of hypotension, one case of proteinuria and one case of hemorrhage for a patient who have the oral, so that patient should not be enrolled, but it's enrolled into the 3 milligram per kg. So in general, right now, the VEGF-related tox is really, really low. So this is -- we are very happy. So our investigator is also very happy because they think that hypertension they typically can manage, but sometimes proteinuria is a problem, especially grade 3, right, recurrent because you're losing essentially nutrition. So, so far so good for the VEGF-related tox. Next slide. So PK, so typical a textbook like linear PK. So that means we don't have a lot of neutralizing ADA antibody interfering with PK, or there's no target-mediated drug disposition. So it all looks good. Half-life is 6 to 8 days, similar to AK112, probably 1 day longer than AK112. Next slide. The receptor occupancy in the peripheral T cells. So you can see that at 20 milligram, the blue line, we saturated the receptor occupancy and that situation also can be observed even before the second dose, so-called dose interval, right? So it's a cross dose interval. We still have more samples coming. So the 30 milligrams will have more longer follow-up. But overall, you can see from the blue line, which is 20 milligram is active, the receptor occupancy is constant. So that help us determine dose going forward. But this certainly is an underestimate of the receptor occupancy within the tumor microenvironment because this is a peripheral because in the [indiscernible], there's no synergy. And in the tumor microenvironment, there's a lot of synergy. So that binding will be much tighter. Next slide. The pharmacodynamic biomarker, so the T cell proliferation as indicated by upregulation of Ki67 and also the CTLA-4 specific activation as suppression -- as indicated by ICOS. If you look at the top, the proliferation on CTLA-4 -- CTLA-4 proliferation essentially is driven by CTLA-4 blockade. So you can see a plateau at 10 milligram per kg, okay, against in the peripheral. And for Ki67 in CD8, that's contributed by both CTLA-4 blockade and PD-1 blockade, and you see also a plateau around 10 milligram going forward. ICOS also behaved similar at 10 milligram. So essentially 10 milligram it should work, and we see it works, right? But going forward, you always want to have a little bit of margin going forward, 20 or 30 milligram probably will be the right dose taken in Phase II. The next slide. This is interesting. So this is the serum-free VEGF-A. You can see even at 1 milligram per kg, immediately post the dose, the serum-free VEGF-A dropped from about 500 to undetectable level before 10 pg/ml, and it stayed that way. And that's true for 3 milligram per kg, 10 milligram per kg, 20, 30 milligram per kg. So this is actually very different from either AK112 or bevacizumab. In those case, the serum-free level actually jump up and down after dosing, right? So a lot of people are asking why this one behave differently, why the suppression is so much tighter. I think it's hard to explain the only way that I can explain this is because we've got 3 targets in the tumor microenvironment that cross-linking causes the VEGF actually [ trap ] inside the tumor microenvironment. Your environment, say, does not come up that much. So that's the -- would be a reasonable explanation. But of course, there's no evidence to say that's true. But right now, there's no need to really know the mechanism. But as long as that we know the trispecific has definitely neutralized VEGF-A. So that's, again, is another way of proof of mechanism. So next slide. Okay. So the [indiscernible] data, this is the cut for the ESMO poster. And just 1 week later. Next slide. Just one patient converted from SD at third tumor assessment to PR. So at 20 milligram per kg, right? So then dose-wise, it looks just -- it looks better. Essentially every dose, you got some response, right? But because the denominator is different, so the response rate actually as you go off the dose, it looks like there will be more response rate, but that will wait to see, right, data will mature further. I think important to see is that the disease control rate. Remember, all these patients come in with tumor progression. So if there's no drug -- if drug has no effect, the tumor will keep growing. So you will not see a lot of stable disease. And that's reflected in the disease control rate. So we have a 71.4% of disease control rate in this Phase I patient population. So that means the drug is actually working. Next slide. Okay, right, so then we look into which tumors actually have reached a partial response level. So we have 17 patients from post-IO non-small cell lung cancer. So everyone has IO treatment and 41% has VEGF treatment. So the response rate is 17.6% and the disease control rate is 42.4%. Now within these 17 patients, there are 5 patients who are EGFR or something else mutation. As you know, EGFR mutant non-small cell lung cancer does not respond to monotherapy of IO, so if we remove those 5 patients, and they are actually 3 out of 12, 25% and DCR have reached 83.3%. Now this is a pure post-IO AGA active -- AGA negative non-small cell lung cancer. So this really support this drug will probably work better in both frontline and second line, right, because you are overcoming IO resistancy. If you remember EGFR TKI, the third-generation EGFR TKI are much better than first generation because it can overcome some of the resistant mechanism of first-generation EGFR TKI, okay? So this is important. You need to have this type of data to be confident you will be able to beat KEYTRUDA, for example, right, because those are PD-1. So right now, we know that our drug works post PD-1 like KEYTRUDA, and we actually have a lot of patients who have KEYTRUDA from Australia. Then you look at ovarian cancer, relatively cold cancer, TNBC, relatively cold and NCC RCC, non-clear cell or renal cell carcinoma, relatively cold and the soft tissue sarcoma very cold. So the drug is working across this type of tumors look at disease control rate as well as response rate. Next slide. So for the 17 non-small cell lung cancer, we look at the duration of treatment and also the timing of response. So if you look at the left side, the green one is 10 milligram. The 10 milligram, you have 2 patients that still are on treatment and another one is PR and the third tumor assessment, right? So it's coming a little bit slow, but it's still coming. That's also a typical of later-stage clinical trial. You don't see all response in the first tumor assessment, right? In first line, you see probably 75% of response in the first tumor assessment, remaining typically in the second tumor assessment. But in later line, the sensitivity is lower. So you see quite a few response coming later, okay? And also 10 milligram apparently not enough. So you have 3 patients progressed at 10 milligram. Now for 20 milligram, the more patients stay on treatment right now and also you have 1 patient becoming a partial response at second tumor assessment. Now for 30mg right now, we have 4 patients. So 1 patient progressed. So 3 patients, one of those patients had a tumor response at first tumor assessment. So again, we have a lot of other patients at 30 milligrams that have not had tumor assessment. So when those happens, we expect tumor response rate will go up more than probably 35%, okay? On the right side, we're also showing you the each one has IO and then tumor type, squamous, non-squamous, both response, right, because there's quite a few non-squamous. So the safety is not an issue. And then the VEGF treatment, prior treatment and also the AGA mutation. And look, the AGA positive ones do not have partial response. So they do have some SD and also PDs for the patients. So the drug is still working in this population, just not as sensitive as AGA [indiscernible]. Next slide. So we look at all the 49 patients who have had at least 1 tumor assessment and looking for duration of disease control, right? So the first patient is a soft tissue sarcoma patients. The patient has a partial response at the second CT scan. The first CT scan has a 24% of tumor reduction. So we dose escalate patients from 1 milligram to 3 milligram and the patient has been on 3 milligram per kg as of today. So patients are still receiving treatment with a tumor reduction of more than 40%. These patients failed 3 lines of chemo before coming to the clinical trial. So I think this is really lucky for this patient because otherwise, there's no treatment, right, soft tissue sarcoma. So in the early doses is longer follow-up, you can see the disease control rate is pretty good, right? We're talking about 7, 8 months already for lower dose patients. Now for higher dose patients, there's a lot of patients still on treatment, right? The follow-up time is still short. But I think it's very promising for us to expect to see a pretty long PFS for this Phase I cohort patients when the data mature. So next slide. Correct, So then we put all this data in perspective of relevant drugs with the Phase I trial data, right? So you can see you're all familiar with the Phase I molecules, right? They are relevant bispecific molecules, either bispecific VEGF or bispecific CTLA-4. So immediately, you can see that our trial has much higher prior IO therapy, and we have much shorter follow-up because we need to report data early for a lot of other purpose. And then disease control rate is much higher. And overall, it is very similar, right? But it's hard to compare because each trial enrolled different patients. Like AK112 Phase I enrolled quite a bit of ovarian cancer. So you got a 25.5% in that trial results. But for others, it's enrolled different patients. So it's hard to compare. But the DCR, it tells us a lot, okay, meaning that the drug our drug is working across different tumor type. So next slide. So we are trying to do apple-to-apple comparison. So we pick up all the non-small cell lung cancer post IO patients from various trials from this molecule. So in our case, we already show you that out of 17, so we have partial response. If we get rid of AGA neck positive, the response to 35%. And the second column, that's a relatively new trial for AK104+VEGF biosimilar. It's a relevant trial because that's on a second line, pure second-line setting, 47 patients, and they are all AGA negative. I think this is a good trial design. So they definitely want to focus on AGA negative second line and trying to generate data potentially trying to head-to-head against docetaxel, right? Because you can only add to docetaxel, you already have 2 drugs that's going to have a lot of problem -- toxicity problem. I think with this 12.8% response, I think to trying to head against docetaxel has some challenge there. Whereas in our case, we can add to docetaxel, right, because we have a single drug that's much more and also much safer, so just talking about the going forward. And then you look at some other trials. So for example, the AK104 Phase Ib/II trial that has 23 patients. So those, again, are AGA negative or unknown. And in this case, it's line 2 and above. In this case, 23 patients have no response. And then the PM8002 trial, so that has 8 patients, that 8 patients, they are all AGA negative and the second line post IO, they have 1 response, 62.5% disease control rate. So if you look at across this, you can probably appreciate that the trispecific antibody is probably going to do better in posted IO non-small cell lung cancer with or without an AGA mutation, okay? So this is actually good for us because we want this drug to work in first-line non-small cell lung cancer and also second-line non-small cell lung cancer in combo with chemo and KEYTRUDA with chemo. And ceritinib can also do a monotherapy, but I said that before, that's going to be a tough decision to make. Next slide. Correct. Next slide is correct. So we see the Phase I data from how our development molecule. When the Phase I data was wrong, people are probably not impressed with that first Phase I data, right? But when those molecules come into Phase II treatment naive either as monotherapy or come with chemo, the response rate is suddenly jump is getting much higher than a typical PD-1 or PD-1 plus chemo. Each one is slightly different because sample size and also baseline. But border line is the bispecific plus chemo or bispecific alone has a better response, better PFS than PD-1 plus chemo or PD-L1 alone, that's a given, that's a fact, okay, just like HARMONi-6 reported yesterday. And of course, the problem is that whether you will have OS, right? And we -- here, we fully expect that we will see similar or slightly higher response rate, longer PFS and much, much longer OS. And we truly expect that the CTLA-4 will add 10% of OS on top of PD-1, right? And then with the bispecific PD-1 VEGF that adds roughly 20% together, you can add about 30% of OS on top of PD-1 -- currently PD-1. So then you will be pretty safe to run Phase III trial. So that's the expectation going forward, okay? So next slide. Right. So this is a summary of what I just said. I think the safety tolerability with immature data is very encouraging, okay? And also, we have very low infusion reaction. So that's really a good thing to have. And the fact that we dosed the 45 milligram per kg, it continued dosing multiple cycles already. The highest cycle we have is probably 12 cycles already, right, really saying that we manage CTLA-4 toxicity really well and also VEGF toxicity. And the PK/PD really showing the drug is behaving well inside the body and the PD showing that the drug has bind to its target and it's doing what it's supposed to do on PD-1, CTLA-4 and VEGF and antitumor activity, you definitely see good response in post-IO, post standard of care of tumor in a Phase I setting, and you see very encouraging efficacy in post-IO non-small cell lung cancer that support our Phase II focus on non-small cell lung cancer. Certainly, we have 15 cohorts, there's a lot of other cohorts. But because non-small cell lung cancer is the most compatible one market-wise, we want to get into that start Phase III before we box start -- get a box start. That's the future development, okay? So I think that's it. I'll be happy to discuss further if you have any questions.

Sure. Thank you, Jason. This is very comprehensive and really congrats on the data. I think the safety data looks pretty promising for us to moving forward. So since we talk about safety, there's a couple of questions we try to understand a bit more. Number one is when we look at the AEE across different doses, right, how should we understand the reverse dose dependency? Grade 3 plus TRAE actually is 24%, 9.5%, 7.4% when the dose goes up. And particularly grade 3 plus IRAE only observed in the lower doses, right, while in the fifth DL5 and DL6, we didn't see any of the IRA. How should we understand that?

Right. The number one is that the early dose is not just lower dose. The early dose has been followed much longer, right? So with the first patient dose in March. So we're talking about 7, 8 months. And as well as then the 45 milligram dose has been followed probably 1 to 2 -- 1 month, 1 month, so it's much shorter. So even though the RAE typically come out 4 to 8 weeks, right? So there's not much come out yet. So as I said, the AEE rate will go up. But even if the AEE rate goes up by 50%, it will be very much within the acceptable range for trispecific antibody, right? So that's one obvious explanation. The second is the 1 milligram, 3 milligram, 10 milligram -- 1 to 10 milligram is not really just 1milligram, 10 milligram. There are quite a few patients escalated already to 20 milligram per kg, right? But it still belong to the initial 1 milligram, 10 milligram cohort.

Got it. So is it fair to say that we should be taking, let's say, the dosage cohort 1, 2, 3 safety profile given you mentioned that for those lower doses has been followed for a longer time. So meaning Grade 3 plus TEAE is 38%, [indiscernible] about 24%, RAE about 10%. This is kind of like baseline when we are looking forward for a potential -- the safety profile going to be shown in Phase II study?

That's -- I think it's a reasonable assumption except that in early days when we enroll patients, right, there are some -- enrolling is probably not appropriate. For example, we enrolled one patient with [indiscernible], right, start with [indiscernible]. So those actually are supposed to be enrolled because those patients have a strong bleeding tendency, but we do have patient at 3 milligram that patient does have grade 3 hemorrhage, right? So that's actually increased the hemorrhage case. So I think a small sample size, sometimes 1 or 2 events, they can tear this scale inproportionately, right, disproportionately. So again, I think the frequency is one thing. Second is the type of AEE, whether it's relatively easy to manage it. For example, grade 3 hypertension is probably easier to manage than grade 3 proteinuria, right? And also in this case, we have grade 3 STLT without symptom that's also considered as a grade 3 in this country of 5, right? So I can tell you all the 5. So the STLT is one of that. And then in the grade 3 neutropenia, that's also grade 3. So essentially, when you look at all these 5 grade, 3, it's not -- you're not concerned -- we're not concerned.

Got it. Got it. And also, how should we be looking at the potential [ RX2D, ] probably 20 milligram or 30 milligram. And also, if we look at the data, right, well, of course, it's a small sample, but 20-milligram looks the ORR a bit lower, right? But I think your PK chart already showing that it's pretty dose-dependent, linear kind of perfect chart for that. So when the sample is getting bigger, we should be expecting higher dose going to give us better response rate.

So we will really be making decision based on data. You can see that in the dose escalation, 20 milligram, we have 21 patients, 30 milligram, we have 27. So those data is going to mature eventually. And also in the Phase II, we are going to enroll homogeneous patients like non-small cell lung cancer and comparing 20 versus 30, right, for each [indiscernible] has 20 patients, for example, that will be giving a pretty good assessment as to efficacy and safety. And based on that, we will make a decision which dose to take forward.

Got it. And this is a question actually today we got from investors is that just from -- when you're explaining the design and the mechanism that you mentioned about the PD-L1 is a single valent design in CS209, right, versus a lot of bispecific antibodies is actually bivalent. So is it fair to expect lower efficacy for your asset at 20 milligram per kilogram compared to a bispecific bivalent design at the same dose level? Or is it actually relevant because you have the synergy, synergistic effects and also the affinity might be also different through different design.

Yes. I think the biggest misconception for IO is that you want to have a very, very high affinity molecule, right? Now look at the PD-1. So who has the highest affinity for PD-1, who has that? right. So, [indiscernible] actually has higher affinity than [KIT], right?

That's right.

So -- and if you look at the literature right now, people realize that for blocking antibody, you do not want to have super high affinity because in the tumor microenvironment, where you have a super high affinity, which has high [indiscernible] but low [indiscernible], it's binding will not come down. So it's actually forming a protective shield around the tumor and then more [drug] cannot pass through that shield. So the activity, the antitumor activity actually decreased, okay? So you really want to have a moderate antibody that goes on and goes off, right? And then once it gets on, there's all kind of synergy. The synergy actually solidify this network. It's because our antibody, remember, we have a lower ceiling, but we got a much higher dose. So it's really saturating the environment, right? So it's just like I got -- it's like you attack a city, you either have very strong giant, 10 or 20 giant, but I could have 100 just medium-sized soldier, right? So they probably want to do the same job or do better. That would be the idea.

Got it. And also, you mentioned about you're going to run a compound randomized controlled trial between the 20-milligram versus 30-milligram to determine the RP3D trial for the Phase III. So could you elaborate a little bit more about the potential design when you're going to start to do that and sample size, what kind of indication you will be targeting. Also when we're going to have the results for that?

Right. It's funny that you mentioned this. Look, if I want to make a drug, I don't want to make a shady drug, right? So I would not [indiscernible], I trial with 1,000 patients because I'm not confident I will have OS. So I put a very, very high OS in the ratio, and that will necessitate a large sample size, just trying to get a statistically positive trial. maybe we set [OS] ratio of 0.8 or 0.2 or 0.78. I think we need to have a hazard ratio, OS hazard ratio of 0.75 or below in order to build out in the market the current standard of care, right, to convince that this is really better drugs around. So for me, for the first-line squamous non-squamo Phase III trial, I think I will target a hazard ratio [OS] of 0.75. I definitely don't want to go higher than that, okay? So if I have 0.75, my sample size is around 700, right? and cover both squamous, non-squamous. And PFS is not even an issue because no matter what you put there, it's going to pass, as you would expect. So really, really the sample size will be driven by your OS assumption. And I don't want to do a shady trial with 1,000 patients just in order to get it [indiscernible] drug approved. Of course, I probably will not be a person to make that decision because I think when we have this Phase III started, I certainly hope that we have partnered already, joined us. Then for the second line, that actually it is going to be a relatively simple design, right? I think 300 patients might be good enough if we get rid of AGA positive patient population, right? You add on dosing and they compete against docetaxel. The OS only -- OS and PFS, you can do a primary [indiscernible]. OS is very necessary for approval for the second-line setting. I think that will be a fast-to-market trial. We're actually very excited about that trial.

Got it. Well, of course, non-small lung cancer is going to be the key indication you will be targeting for sure, given it is very large. And of course, it's very competitive, right? But in this set of the data, we already see that there has been ORR observed in several other indications, right, ovarian cancer, TNBC and RCC, STS. But among all those indications, potential indications, what could potentially be the ones you will be pursuing beyond small cell lung cancer?