All right, so welcome to this next fireside chat with Compass Therapeutics. With us today, we have Tom Schuetz, CEO. Tom, welcome. Thanks for joining us.
Thanks, Michael. Thanks for inviting us.
All right, I'm really excited about this story. Before we dive into some of the questions, Tom, could you just provide us with some quick background for those that are new to this story?
Sure. Compass, we're located here in Boston. We're a monoclonal antibody discovery and development company in oncology. All of our therapeutics are either monoclonal antibodies or bispecific antibodies. We have got three drugs in the clinic. Our lead program, which I think we'll talk quite a bit about today, is a DLL4 VEGFA bispecific antibody, currently in a randomized trial in patients with advanced biliary tract cancer. We also have a very novel PD-1, PD-L1 bispecific. We recently announced some Phase I data, and we have responses in the post–checkpoint inhibitor patient population. We also have a next-gen CD137 agonist antibody. We also recently announced with our Q3 earnings that we have a PD-1 VEGFA bispecific antibody. We shared the full preclinical characterization of that drug at CITSI last week, and we reiterated our IND filing timeline for this quarter.
Early next year, we'll have four drugs in the clinic with multiple different clinical trial readouts coming up in 2026.
Great. I'd like to start out the discussion with Tuvesimig, no surprise here, and the COMPANION-002 Study that we've spent some time on discussing in the past. Before we go to that, maybe first remind us of the overall commercial opportunity for Tuvesimig in biliary tract cancer, or BTC, where you're evaluating the therapy.
Sure. BTC, biliary tract cancer, is much more common than you might think. We and others have done a claims-based analysis, which is consistent with the published epidemiology data. About 25,000 new patients in the United States are diagnosed with biliary tract cancer each year. There are a couple of different anatomic subtypes, depending on where the actual tumor is located: gallbladder, bile ducts, et cetera, but a total of about 25,000 patients annually. Earlier this year, we conducted some third-party market research, which confirmed that number. Unfortunately, only about 15%-20% of patients have an actionable mutation that makes them eligible for a targeted therapy. In the second-line setting, for the remaining 80%-85% of patients, there's literally no approved therapy. Front-line therapy is chemotherapy plus a checkpoint inhibitor, either Keytruda or Imfinzi. Then second-line regimens, there are multiple.
Our study that we're running is in the second-line setting, and we believe about 70% of patients ultimately will move on to receive a second-line therapy. So 15,000 patients annually, let's call it on drugs, something like six months, and at any reasonable assumption of oncology drug pricing, it's a very significant commercial opportunity, well north of a billion $ annually in second-line BTC in the United States alone.
Great. Obviously, you're running the COMPANION-002 Study here. You announced success on a primary response rate endpoint earlier this year. Yeah, just remind us again, what were the key learnings from this initial data disclosure?
Sure. As you mentioned, we're running a randomized trial, which is Tuvesimig, our DLL4 VEGFA bispecific, plus the chemotherapy drug paclitaxel versus paclitaxel alone in patients with advanced biliary tract cancer who have received one prior line of therapy. As you mentioned, the primary endpoint of that study was overall response rate. That was assessed by blinded independent central review, about as rigorous as you can get. We more than tripled the response rate in the control arm by adding Tuvesimig to paclitaxel, and that was statistically significant with a p-value of 0.031. Interestingly, we learned something else from the best overall response rate analysis, which is there was a very large difference in the fraction of patients who had progressive disease as their best overall response. Remember, best overall response was the endpoint.
If progressive disease was your best response, by definition, that means it occurred at the first time point. Because if you had stable disease or a response at the first time point, you'd be in that category. The first scan in the study was done at week eight. In the paclitaxel arm, there were 42.1% of patients who had progressive disease at week eight, whereas in the Tuvesimig plus paclitaxel arm, that number was only 16.2%. Of course, progression-free survival, it'd be 100 minus those numbers. At week eight, the incidence of progression in the control arm, 100- 42, is 58% at most, and it could be less than that. Similarly, in the combination arm, 100- 16, it's 84% at week eight. It appears that the progression-free survival curves are already separating in a very significant way.
Right. Great. And what about safety tolerability?
We did not release any interim safety data. We did disclose the study was overseen by a data safety monitoring committee, and that committee met four times during the study and compared arm-specific safety data and did not come back to us with any findings of any new safety signals or recommendation to modify the study.
Right. And so the COMPANION-002 Study, so despite the fact that it hit positive outcomes on response rate, is still continuing to look at PFS and OS, right, which you guided to disclosing likely early next year. And so how important are these time-to-event readouts to support potential regulatory filings, in your opinion?
Sure. The secondary endpoints, by the way, we're using hierarchical testing to control for alpha statistical spend in the study. Overall response rate was the primary endpoint. Because that was statistically positive, the full 0.05 alpha now rolls over to the first secondary endpoint to be tested, which is progression-free survival. The next endpoint will be overall survival. The time-to-event endpoints are triggered both by OS events. When we hit 80% OS events, the secondary endpoint analyses will be triggered. We recently guided that in August, with our Q2 earnings, we announced that there are fewer deaths in the study than we had originally projected with our powering assumptions, suggesting that we could be seeing an effect of the drug as measured by overall survival. Last week, with our Q3 earnings, we refined that guidance a little bit to late Q1.
In late Q1, we expect to report the analyses of progression-free survival and overall survival.
Yeah, so how important are those two outcomes to potential regulatory submissions?
FDA never said that to us. I think we do have some important regulatory precedent here, which is the study of the IDH1 inhibitor, Tibsovo. That study was called CLARIDY, C-L-A-R-I-D-H-Y, CLARIDY. The CLARIDY study was a second-line BTC study, randomized study of Tibsovo versus placebo. That study hit on progression-free survival. That study allowed a crossover, which our study also is allowing. They did not hit on the intent to treat analysis of overall survival, but did hit on the crossover-corrected statistical analysis. That data package got Tibsovo approved in the US. We also prospectively declared that the crossover-adjusted analysis of overall survival is the primary analysis of overall survival in the study.
All right. People sometimes ask, what if you succeed on PFS, but there's only an OS trend, but no statistically significant improvement?
Yeah, I don't, as you know, FDA doesn't engage in any of these kinds of speculations. Of course, because FDA doesn't speculate, they never said any of these things to us. Just based on what the Tibsovo data, I think if we hit on PFS and we have a strong trend, I think that's a very, very powerful package.
Right. And so obviously, I think the study was initially intended to read out in summer 2025, is that correct?
So yes.
Right. It has been sort of overdue. The events are occurring at a slower rate. When that happens, there is always the question, I think, hey, is this due to higher dropouts than expected? Is it due to patients just generally doing better? Or is it because your drug actually works? How do you think about those options? Is there a way to perhaps look at that more closely?
Sure. I'll take those in order, finishing with the last option, which is the most important, of course. Are there a lot of dropouts? In other words, are we missing a lot of data? I know the team has worked incredibly hard to maintain tremendous follow-up, and the sites have worked very hard. I know that we have incredibly few patients that have been lost to follow-up, something like 5%. We're not going to have the reason we're not seeing fewer deaths is because we're not counting them properly. There's a very, very small number for a study like this of patients who are lost to follow-up. The second one, could the control arm somehow be performing so much better than we thought it would be?
I think we have a little hint in the overall response rate data that we have, again, 42% progressive disease at week eight. The median PFS in the control arm is probably going to be something like 2-2.5 months, something like that. The response rate in the control arm was 5.3%. We really have no evidence that paclitaxel is doing something crazy in the study. That leaves us with the third possibility that what we're seeing here is an effect of the drug on survival in real time. Earlier this year, we moved to monthly survival sweeps. These data are incredibly up to date. We get pooled mortality data weekly. I think it's almost impossible to look at the data and not see a tail forming on the Kaplan-Meier curve in real time.
Right. All right. So that's definitely a very promising, unusual circumstance in a way, just given that we have so much information already. The other question that sometimes comes up, and when we talk to docs, they really suggest the use of FOLFOX in BTC, which is in NCCN guidelines. The ones that we spoke with, they do think paclitaxel is fine too. Have you had any regulatory feedback on the use of paclitaxel in the control arm and any thoughts on the choice of that particular chemotherapy?
Sure. Very important question. I get this question a lot. The answer to the question is incredibly simple. Paclitaxel is the control arm because FDA told us that paclitaxel had to be the control arm. It is that simple. We asked FDA specifically if we needed a FOLFOX control arm in the study, and they said no. You are correct, of course, FOLFOX is in the NCCN guidelines. I will also maybe point out there are 10 other regimens listed in the NCCN guidelines. Earlier this year, we did some third-party market research as we began to think about potentially commercializing this drug ourselves. One of the things that we learned from the third-party market research is the hodgepodge of second-line regimens that docs are using because there is just really nothing out there for these patients.
Right. As we think about the data next year, is there a particular effect size or treatment benefit that you think is needed to be demonstrated at a minimum for this to be broadly incorporated into clinical practice?
Sure. Maybe I'll give you two answers to that question. Maybe I'll give you my answer, and then I'll give you the answer that came out of our market research. I always default to our power calculations for the study. For progression-free survival, we assumed a median of 3.0 months in the control arm, probably going to turn out to be a bit of a high estimate, and 5.4 months in the combination arm. That's a hazard ratio of around 0.6. I think if we had a hazard ratio of 0.6, I'll be thrilled, and nobody could possibly question a hazard ratio of 0.6 and the clinical benefit associated with that. I was a little surprised, to be honest with you. In market research we did earlier this year, where we interviewed and surveyed 70 KOLs. For KOLs, the bar is a heck of a lot lower.
I was really surprised at that. Even de minimis benefits were thought to be meaningful to docs. I think if we hit our power calculations, I think it would be, again, no one would question the clinical benefit.
Right. Okay. Super. Great. At this point, I think all we can do is sit back and wait, right? With that, I think we can switch to another program.
Great.
For example, 8371, which is, as you mentioned, you did disclose some data recently. It's a curious mechanism. This is a PD-1, PD-L1 inhibitor. Maybe first, what's the mechanism of action of this antibody, actually?
Sure. You are fundamentally correct. It's a curious molecule. When we were first going down this pathway, we set out to discover a next-generation checkpoint inhibitor. In order to do that, we had to develop a technology. We developed a screening tool that we called StitchMAMS, which allows us to screen for synergy in bispecific drug candidates. That technology also allows us to build small libraries. Now, let's be clear when I use the word library. This is not like a small molecule library that has a million components. We can build small libraries of biologics. We did a screen, and curiously, the best partner for PD-1 blockade in a bispecific antibody is PD-L1 blockade. We spent about a year investigating the mechanism of action, and it's definitely a novel checkpoint inhibitor. It's clearly a T cell engager.
Fascinatingly, it exposes a metalloproteinase cleavage site on PD-1, leading to the cleavage of PD-1 off the surface of effector T cells. It converts PD-1 positive T cells into PD-1 negative T cells. It is a really fascinating mechanism of action. Last week, we updated some data from a Phase I study that we are running. We are running a standard dose escalation Phase I, three plus three design, five dose levels. One of the things we disclosed last week is we have no dose-limiting toxicities at any dose level. A total of 15 patients, because we saw no dose-limiting toxicity, we never had to go to three plus three. We only had to enroll three. Fifteen total patients enrolled, including the de minimis first in human doses. I think it is really interesting.
We might have a differentiated safety profile, which would be consistent with the mechanism of action. We believe that this drug is anchored in the tumor microenvironment via PD-L1, where it can provide local high-concentration blockade of PD-1. Just really interesting. Already in the Phase I study, which, by the way, is all post-checkpoint inhibitor patients, we have three responses. We also recently announced that our response in a patient with triple- negative breast cancer is durable and deepening. That patient who actually relapsed while receiving Keytruda, actually that patient would be categorized as refractory, not relapsed, that's a refractory patient to PD-1 blockade, with 9 centimeters of tumor burden at baseline, is almost a CR. Just an amazing patient. We also have a very deep PR in a patient with non-small cell lung cancer.
This quarter, we're going to begin cohort expansions, 56 patients, 28 in each disease. Within each disease, two different doses, the two highest dose levels, 3 and 10 mg/ kg, in patients with triple- negative breast cancer and non-small cell lung cancer. Last week, we also disclosed, fascinatingly, at the highest dose level, we have a new response in a third indication. We have not disclosed that indication yet. I'm not trying to be obtuse, just waiting for confirmation of that. We have now got a total of 15 patients. We have three responses in the post-checkpoint inhibitor patient population. The Phase I data that we have, I assure you, I'm not trying to trash talk here.
The Phase I data that we have is better than the Phase I data that was seen with any PD-1 inhibitor, which includes some of the greatest drugs ever developed. We need more patients. Obviously, those drugs have been in hundreds of thousands of patients. We've got tremendous early Phase I data.
Of the 15 patients, how many of them were at subtherapeutic doses?
You know, that's interesting. I mean, at least three, which was the patients that were treated in the first dosing cohort, which was 0.1 mg per kg. So 100 times lower dose than our final dose. At least three. I include those patients in the denominator, just to be fair about it.
Yeah. How do you approach dose selection, just given that you haven't seen DLTs? And I don't know, is there some sort of dose response or PD analysis that you've been doing?
Yeah, we have been doing that. And we've also done some PK modeling as well, which has helped us with the dose selection. I think in general, PD-L1 blockers are dosed at higher doses than PD-1 blockers. So I think that the dosing of this looks like it's tracking with the PD-L1 end of the molecule. But with checkpoint inhibitors in general, there's not a clean dose response the way you see with some small molecules. So we're going with 3 and 10, where we had two of our responses so far. And we're going to do those cohort expansions. And in the first half of next year, we'd love to present the dose escalation and some early cohort expansion data at an important medical meeting.
Okay. The histologies for expansion are what? I misheard.
Non-small cell lung cancer and triple- negative breast cancer.
Gotcha. Makes sense. All right. Maybe a couple of quick questions. One on 10726 here, PD-1 retinol inhibitor. In general, maybe remind us again how you're tracking towards IND. How does the antibody compare to Sumitomo's and perhaps the BioNTech Bristol drug as well?
Sure. Last week, we reiterated guidance that we're on track for IND submission this quarter, which would mean Phase I initiation early next year and hopefully in the second half of next year, some early data. Last week at SITSI, we presented the full preclinical characterization of that drug. We compared our drug to Ivenesumab, the Sumitomo drug, not to the Bristol Myers Squibb BioNTech drug. We showed in some models, including a human non-small cell lung cancer xenograft model, that our drug, 10726, was superior to Ivenesumab. Now, look, Ivenesumab has done large randomized trials against Keytruda. This is a mouse model, a xenograft mouse model. Let's just be fair about what we're comparing here. We clearly have an active molecule. We look forward to taking it into the clinic. We spent some time thinking about Phase I indications.
We thought about indications where both PD-1 blockade and VEGF targeting are effective as monotherapies. We are going to go into patients with renal cell cancer, gastric cancer, hepatocellular cancer, and endometrial cancer, as those are going to be our four Phase I indications for that drug.
All right. Maybe just one last question on 471, your fourth pipeline product candidate. Just remind us real quick, what are you doing there?
Sure. 471, differentiated CD137 agonist. We did a Phase I dose escalation study and cohort expansions, again, in the post-checkpoint inhibitor patient population. We did not limit the histologies, which if I could go back in time, we might have done, because we enrolled 60 patients with 17 different tumor types. It's hard to think about that. We had five responses, three in melanoma, one in mesothelioma, and one in small cell lung cancer. We did collect biopsies from patients in this study. Fascinatingly, our small cell lung cancer patient, who actually had a complete response, again, in the post-checkpoint inhibitor setting, that patient's tumor was positive for CD56, also known as NCAM. CD56 is its own ligand and receptor. We believe that NCAM facilitates the recruitment of NK cells into the tumor microenvironment.
CD56 is one of the most important cell surface markers of activated NK cells. NK cells can dock with NCAM-positive tumors, where our drug, 471, can activate those NK cells via CD137. We're going to test that in an NCAM-positive basket study that will begin in the first quarter of the coming year. In 2026, we'll have at least three, possibly four different clinical readouts. It's going to be a very, very important year for us. We're also funded into 2028 to execute on all these programs.
All right. With that, I think it's time to wrap up. Thanks, Tom. Really appreciate your time.
Great. Thanks, Michael.
Thanks.