All right. Good afternoon, everyone. I'm Stephen Willey , one of the Senior Biotech Analysts here at Stifel, and glad to have with us presenting in the next session Tom Schuetz who is the CEO of Compass Therapeutics. Fresh off a pretty important data release last week. Tom and I are going to have a discussion. This is intended to be somewhat informal. There is a Q&A function for whoever is listening to send questions in, and we'll get those answered as deemed appropriate. Tom, thanks for the time.
Sure. Steve, thank you. Thank you to Stifel for having us here today.
All right. Now, I'm probably going to pepper you with a lot of nitty-gritty questions around maybe what was shown and what was not shown last week. I mean, look, we're obviously getting a lot of questions around all the stuff that's happening on the periphery here on the macro tariff side. You're a late-stage company, right? You obviously have the potential of securing an approval here before the term of this administration comes to an end. Can you just start off talking about your current manufacturing supply chain, how you may or may not be impacted by these policies, and then where is your IP domiciled
Sure. Thanks again, Steve. Unfortunately, a very important question. Let's just answer both of those. All of our manufacturing is done by a contract manufacturing organization that is based in the United States. Full stop. All of our intellectual property is domiciled in the United States. Period.
All right. We like, immune.
Hopefully, we are in fact immune to many of these things.
Okay. Let's get into it. You disclosed top-line data from the COMPANION-002 trial in second-line biliary last week, obviously achieved stats on response rate, which was the primary endpoint. You're now guiding to having this event-driven secondary endpoint data in the early fourth quarter. How do you think what we learned about response rate and some of the associated metrics that were provided now de-risks PFS?
Sure. Thanks. As you introduced today, last week, we had a very important disclosure. Happy to answer additional questions about that. First of all, as you mentioned, by definition, the study is positive. Just period. One of the questions that we have gotten over the past week is, "Oh, can this response rate change as patients go through the study?" The answer to that is no. We get to do this analysis one time and one time only as pre-specified in the statistical analysis plan. This is the final analysis of ORR. It was statistically significant, about three times greater than the response rate in the control arm, and about three times greater than what you see with regimens that are used in patients with this disease. You alluded to other metrics that we reported last week.
We clearly have a difference in disease control rate. Our disease control rate is approximately 70% in the combination arm versus the disease control rate of approximately 42% in the control arm. I think more importantly is the number of patients whose best overall response was progressive disease. Best overall response can be determined at any time during the study. If your best overall response is actually radiographic progression, that means it occurred at the first scan, just by definition. The first scan, we're doing scans every eight weeks. The first scan occurred at week eight. We know that at week eight, we have 42% radiographic progression in the control arm compared to 16% in the combination arm with Teva-semag. There is a clear difference in the fraction of patients that have progressive disease.
Thinking about sort of early mortality in this disease, I think we have some confidence here that we're going to hit the progression-free survival endpoint. I think very important observations that came out of the best overall response analysis.
Okay. Yeah. I mean, look, I certainly kind of look at that PD number and kind of struggle to see how a median control arm PFS is somehow greater than three months, right? That just seems kind of a bit of a heavy lift.
I agree.
Can you maybe just talk about median duration of follow-up time when the response rate analysis was conducted, how you were modeling PFS event accrual in the Teva-semag containing arm, and then how these two things kind of now govern your thoughts with respect to what a PFS number might look like later in this year and if that gives you confidence around both stats and clinical meaningfulness?
A couple of questions in there. First of all, we currently are north of 13 months of median follow-up. We had announced that the study was fully enrolled in August of 2024. The study more or less enrolled linearly from August of 2023 to August of 2024. You can just get the median from that. You asked a somewhat subtly different question. The data cutoff based on the statistical analysis plan for the study was 28 weeks after the last patient was randomized. That was March 12th. At the data cutoff time, we had 12 and a half months median follow-up. The second question you asked in there, how are we modeling PFS events? I just want to give a super clear answer to that. We're not. We are monitoring pooled OS events.
The time-to-event analyses are triggered by 80% pooled OS events in the study. That is what we are monitoring. We cannot do any modeling, right, because we have pooled events. What we announced is that just based on where we are at, we think that we will hit 80% OS events sometime toward the end of Q3 of this year when, I think, interestingly, we will have something like 19 months of median follow-up. I think we are outside of where we thought we would be at the beginning of the study based on how we thought about these medians.
Yeah. I guess my poorly asked question in terms of modeling had to do with your powering of progression-free survival and how you were thinking about those events occurring over time.
Yes. In terms of the powering of the analysis for progression-free survival, we used the lower bound of the 95% confidence interval from our phase two study for Teva-semag. The median progression-free survival in that study was 9.4 months. We used 5.4 months for the powering calculations for the randomized study, which was the lower end of the 95% confidence interval. We used three months for an estimate for—you mentioned three months earlier. We used 3.0 months for an estimate of the median PFS in the paclitaxel control arm. Using those assumptions, we have approximately 80% power to detect a hazard ratio of 0.6. I think you also weaved into your question was clinical meaningfulness. I think a hazard ratio on PFS of 0.6 would be unarguably meaningful.
I think the other thing I'll maybe add to that that I think is important in this disease. In patients with biliary tract cancer, maybe unlike some other solid tumors, patients have real deal local anatomic consequences of their disease and anatomic complications. For example, obstruction of the biliary tree. That is a catastrophic anatomic complication of this disease. One of the things that we saw in the waterfalls was an enormous number of patients with tumor declines that did not reach PR threshold. I think any decline in linear tumor burden in this disease because of the anatomic complications is probably very meaningful.
Yeah. That's actually a really good point. When you modeled out PFS, I mean, I'm sure you incorporated some consideration of patient sensoring.
Sure.
I know you talked about kind of utilizing a very standardized approach here to how you were censoring patients. When this Kaplan-Meier curve gets disclosed, how heavily censored do you think that curve will be, right? I know that you had about 15% of the patients in the trial who were non-available, right? It doesn't mean that they didn't have a death event at some point, but not available. You also have some undisclosed proportion of patients who are going to discontinue therapy for reasons related to an AE, to whatever.
If you think about patient censoring, you think about the need to trigger 80% progression events, do you kind of reach a point on the curve itself where death almost becomes the progression event, and then you get an OS number and a PFS number that are actually really proximal to each other? Is that the way that we should maybe think about this, or..?
It's interesting. I'm not sure. I think in terms of the 15% of patients that didn't make it to week eight, let's start with those patients. Unfortunately, if you look at the OS curves from the FOLFOX study, the two-month mortality in the FOLFOX arm was almost 20%. I mean, think about that number. I mean, that's just staggering. We had 15% of patients come off study before week eight. Some of those patients, unfortunately, will have died, but those patients will then be captured as an OS event. I think the way you phrased the question, I think, was quite interesting because in our phase two study, the median PFS was 9.4 months. The median OS was something like 12 months. Call it 2.5 months approximately.
The way you asked the question, I think, is good because I think it's important to keep in mind that if you come off the study for any reason other than radiographic progression, okay, which that could be many different reasons, clinical progression, right? You have some patient is just much, much sicker. They have to come off study. They don't get to a scan. Some patients withdrawal of consent, right? Patients, it's very, very hard to do these clinical trials. Death, an adverse event, right? Those are just four examples. If you come off study and then you die within a month, say, which is not uncommon, unfortunately, those patients will be captured as a progression event.
I think that I don't think that the way you phrased it might not be correct that progression will approximate OS because we already know, as you pointed out from the data that we have in the control arm, it looks like the median PFS is going to be two months and change, certainly less than three months, right? We don't have 50% mortality at month three, right? I'm not, of course, every time you do a Kaplan-Meier curve, you have to be worried about censoring. I think unless there's a ton of patients who are going on to receive third-line therapy, which I think is highly unlikely, I don't think we're going to have a problem with excess censoring.
Okay. We talked about the proportion of PD events in the paclitaxel arm. I think this was a little bit higher than what we've seen in other second-line chemotherapy trials. I think if you look at three-month landmark PFS in ABC06 with FOLFOX, it's like 67%, right? You've got two-month PFS landmark of 58%. Again, I know it's hard to make kind of cross-trial comparisons without knowing patient baselines, which we haven't seen. Do you think that this maybe triggers some kind of conversation around the sufficiency of paclitaxel as a control arm?
You're right. We can't control for baseline differences across two studies. I would also highlight for you some very big differences in terms of trial conduct between our study and ABC06. First of all, scans in ABC06 were every three months rather than every two months in our study. That's one big difference. The most important difference is all of ABC06 was investigator-assessed responses and progression.
Everything that we showed last week was blinded independent central review. Those are some very big differences that I don't think we can really compare those numbers directly to each other. I've been asked about the response rate in the control arm, which was 5.3%. I actually said, "I'm quite happy to see that because the response rate in the FOLFOX arm in ABC06 was 4.9%." FOLFOX is 4.9%. Paclitaxel is 5.3%. We're right there. I think that number alone suggests that the control arm is fine.
Okay. I know a question that we were getting from a lot of investors. Again, I'm not sure it's necessarily relevant in the grand scheme of things, but was this notion of confirmed versus unconfirmed responses, right? I think maybe the question was amplified a little bit in light of the fact that you were doing scans every eight weeks, right?
Yeah.
You could see a greater than 30% RECIST-defined reduction in tumor burden. And then in less than eight weeks, that patient could have progressed. It kind of brings in this question around what's the quality of the response that you're giving to the patient. I know from a regulatory perspective, it doesn't matter, right? Because both unconfirmed and confirmed are defined within the parameters of RECIST for a randomized trial. Do you think that this is a regulatory conversation if, for whatever reason, the confirmed response rate, there's somehow a discrepancy, a significant discrepancy between what is reported as a headline ORR and then what is subsequently confirmed?
Yeah. I can refer anybody who's listening to section 4.6.1 of—sorry. Thanks, Steve.
That's my favorite section.
T he RECIST 1.1 paper. The concept of a confirmed scan only applies to single-arm studies. RECIST is incredibly clear about this, that that's not required for randomized studies. One thing I think I might disagree with you on, just the way you phrased it. You used the phrase "quality of response," okay? That's not what confirmation is about. Confirmation is about measurement error. RECIST 1.1 is super clear about this. In single-arm studies, you need a second scan in order to control for measurement error. Whereas in randomized trials, that's taken care of by the randomization. We followed RECIST 1.1 to the letter. We sent that into FDA, and there was no comment on that.
Okay. Maybe last response-related question. I think one of the debates that seemed to be ongoing ahead of the disclosure, right, was really kind of related to the depth of response or maybe lack thereof that was observed in the phase two, right? Where most of your responders, I think, had less than 60% reduction in tumor burden. Most were kind of within that 30-45% range. In this trial, you had four patients who achieved 100% reductions in measurable tumor burden, which I guess kind of seems interesting to me, just given the history of anti-angiogenics, really in any tumor type, has not necessarily been correlated with a strong depth of response. What do you think is happening here in terms of this response?
This is a great question. For people who are following along, I just happened to pull up the waterfalls for each arm just to give a visual to the question that Steve is asking. You can see in the waterfalls, there's an obvious difference here between the two treatment arms with the majority of patients who received Teva-semag plus paclitaxel having declines in their linear tumor burden. Steve, you're correct. We had four patients who had 100% declines in their linear tumor burden, which is an extraordinarily unusual observation. One of those patients was officially declared a RECIST CR, which is based on non-target lesions. Of course, to be a RECIST CR, your non-target lesions have to disappear as well. The other three were PRs. You're absolutely correct. This is a very unusual thing to see with an angiogenesis-targeted agent.
It suggests that blocking the interaction of DLL4 and NOTCH1 might have some anti-tumor effect that maybe we did not anticipate. I think that's an incredibly important observation and question that you just asked.
Okay. I know you've talked about wanting to get this response rate data in front of FDA at some point this quarter, I think you had mentioned. Are you hoping that that conversation that you have with FDA will give you some kind of clarity in terms of what's going to be needed from secondary endpoint data to either support a full or an accelerated approval? I know I've probably asked you this question 100 different times over the course of the last two years, but do you think the agency will provide you with that kind of clarity in a preliminary meeting without necessarily having the secondary endpoint data in hand?
Sure. Let me just say our intention is to submit the data for Teva-semag and seek full approval. That is our intention, just period. You ask a very difficult question, of course, here. We're still working through exactly how we want to engage with FDA. We were so focused on delivering the response rate readout. In the past week, I would say we've sort of begun to think about the structure of that interaction. What do I expect from that interaction? I'm not sure. I think it's unlikely to get some epiphany clarity here at this stage in the development program. I mean, our base case assumption has always been that we would need the whole data from the study in order to be able to submit a license application. I suspect we'll have a more robust interaction with FDA later this year.
Okay. Maybe the last question. Does this data now somehow influence what you want to do with this drug in other solid tumor types?
Yeah, maybe. I mean, we clearly have anti-tumor activity, clearly. I think we're now sort of looking at other DLL4 positive malignancies, colorectal, gastric, renal, ovarian, and thinking about how we might expand the potential indications for this drug. I think we're going through that thought process now as we're continuing to evaluate these data. I think maybe later this year, we'll make some determinations about some of these other indications. Yes, I think we clearly have a drug here, and we're thinking about other potential commercial indications.
Okay. I know we didn't get to any of the pipeline stuff.
All good.
CTX-8371 data later this year, CTX-10726 into the clinic. I'm sure we'll have more opportunity to discuss these things here, but just wanted to get through some of these nuanced questions. Thank you for obliging me. Thanks for your time today.
Sure. Super thorough. Thank you.
All right. Thanks, Tom.
Take care.