Here for the afternoon session on day two of Citi's Back to School Biotech Summit. I'm Yigal Nochomovitz , senior biotech analyst at Citi. Our next company for the afternoon session is Olema Oncology. I have with me, of course, Sean Bohen, who's the President and CEO of the company. We have a lot to discuss, a lot of developments in breast cancer over the last year or so. Sean, maybe just set the stage in terms of what's in the pipeline and what are the key readouts coming up. There's several things.
Yeah. Thank you, Yigal, and thanks to both of you for attending. Obviously our lead asset is palazestrant. It's a complete estrogen receptor antagonist for the treatment of ER-positive, HER2-negative breast cancer, and we're studying it in phase III in two contexts. One is as a monotherapy, in the second, third-l ine settings. These are patients who've progressed on a prior CDK4/6 plus an AI, now going on to another therapy. The control arm is fulvestrant or exemestane. Experimental arm is single agent palazestrant. That trial is finished enrollment and randomization.
We're awaiting readout, and we forecast that we will have our top-line data in Q1. In that population, as you know, it's very interesting because it's really two separate groups. There are those individuals who progressed on their prior therapy with an ESR1 activating mutation. It's about 40%-50%. The remainder still have estrogen receptor in the wild-type form and is not mutated. There's been prior success in the mutant population, but no one's been able to do better than fulvestrant or exemestane in the wild type. We're testing both of those populations, both of those hypotheses in the trial.
In total, it's about a $5 billion a year market opportunity. For us, very significant. The other trial that's ongoing and enrolling very well is in the first-line setting, and in that case, we are combining with the gold standard CDK4/6 inhibitor, which is KISQALI. That trial is called OPERA-02. The control arm is KISQALI plus letrozole and AI. Experimental arm is KISQALI plus palazestrant. It's enrolling very well. It'll take a couple years to read out just because of the effectiveness of that treatment.
But that's a $10+ billion market opportunity, should we be able to access it. Both of those trials are based on phase II data that suggests that we can do better than the existing therapies and have activity in these different settings we're describing.
Okay. Let's go in order, I guess. Start with the OPERA-01. I guess the first question is, originally the data was the second half of the year, and now it's 1Q 2027. This is like the age-old question of some of these event-driven trials, the significance of moving the endpoint or moving the data out a little bit. Anything you want to say there, or is it just
Yeah, I wouldn't over-interpret it. First of all, the trial obviously is not a blinded trial, right? It's an open label trial. But the reading of the endpoint is blinded.
Right.
It's a blinded, independent radiographic review. Progression-free survival is the primary endpoint of both OPERA-01 and OPERA-02 trials. Yeah, we do blinded event rate, and so that's in there to some extent.
Mostly the slight delay had to do with a slowing toward the end of enrollment in the enrollment of the ESR1 mutated subset of patients. We think, we're not 100% sure, we think from our investigators that that might have been related to greater availability of ORSERDU in Europe.
But again, it's past us. We are completely enrolled in the trial. We did not compromise the number of patients in either population, so it's fully powered for the statistical plan. I think the other important part about this is that overall, we had predicted 40%-50% of patients would be ESR1 mutated. We actually did end up in our predicted range. I think that part isn't different. It's just that we were in the higher percentage earlier on, and it sort of fell a little bit toward the end.
Okay. You mentioned, of course, the two populations, the mutant and the wild type. The mutant obviously is the lower bar. More specifically on the wild type, tell us more about what gives you a confident view that palazestrant can show a difference there versus standard of care.
What evidence do you have previously that would support that? You pointed out that together, I think it's $5 billion, although even with just the mutant, it's probably like $2 billion or something.
Something in that range, yes. I'll go into what molecularly we think might be going on, but I think the strongest evidence is our phase II data, right? The question is obviously can you replicate that in the phase III setting? In the phase II setting, post prior CDK4/6, in that case, a little more heavily pre-treated patients because a significant number had had prior chemo that's not allowed at OPERA-01.
We saw over seven months, median PFS in the mutant setting, 5.5 in the wild-type setting. That is better than anyone has been able to see in either of those settings, but certainly in the wild type. The bar in wild type is to extend PFS by about two months, right?
Eli Lilly got approved with the ORSERDU at 1.7, which is probably a little on the edge, but it did lead to approval, but you'd say about two months. Recognizing that the control arm really should be two to three months, if we can replicate that five-month range, we should be able to access that population, and that's what's being tested in OPERA-01.
Why might that happen when others haven't been able to do that? Well, in addition to being a complete antagonist, we have very high exposure, 14-day half-life, and a very high steady-state exposure compared to other drugs. We think that that is important in the context of maximizing the benefit you would get.
Okay. Let's also go through just the statistical. You mentioned the statistical plan because you have the two populations. Remind everyone how it works in terms of the process of analyzing these two populations, and where you can win, and the order of which things are analyzed.
Yeah. We haven't disclosed in detail the statistical plan, but I can tell you basically the performance characteristics. And the performance characteristics are the trial can be positive in one of three ways, right? There's the obvious one, which is it shows a significant benefit over the control arm in both mutant and wild type.
Mutant and wild type subsets are tested separately. The way the trial is designed, you can actually have two other types of positive trial. One is ESR1 mutant only shows a significant difference, wild type does not. You then get a positive trial in that mutant subset. Not trying to explain why there would be a biological rationale, but from a statistical standpoint, you can do the opposite as well.
Although it would be unlikely.
It doesn't make a whole lot. Yeah. If that happened, we'd all be wringing our hands and wondering what went on.
Okay.
But you could statistically have ESR1 wild type be positive, ESR1 mutant not be positive, and still end up with a positive trial.
Right. Okay. All right. So that makes sense. And then as far as the commercial opportunity, you kind of alluded to it a little bit in terms of the numbers up to $5 billion. Anything further to add on that in terms of how that works geographically? Is that U.S.? Is that global? What are we talking about there?
Yeah. It would be global.
In that situation. Usually, if you look at markets in oncology, the U.S., in revenue terms, not patient number terms, is usually about 2/3, 70%. So it is the majority United States.
Okay.
Two ways, really, to differentiate in this space. One is to do better than the two months prolongation of PFS in the mutant setting. Again, we saw seven, so instead of two to four , two to three , four to five , we could see seven. That would be significant. The other is to get a benefit in the wild type, which at this point is unmet. There is nothing that has been successful in that subset.
Right.
It is probably, as you said, you kind of alluded, probably split up around $2 billion in change per year in the mutant, $3 billion.
Yeah. ORSERDU got. They worked in mutant with 3.9 versus 1.8 or something.
3.8 versus, 3.9.
They worked with the two-month delta.
Right.
Meaning that's a good argument for potentially while you could work. It's a different setting, but in wild type, if you saw two months, you would be working.
You would be. Yeah. I think that's the basis for approval, would be the basis for use.
Yeah.
That's the question. Can we replicate this phase II data, and can we show that delta? Obviously, the market dynamics are very different in these two places because wild type, there's no endocrine agent competition, really. Whereas there are two molecules certainly launched, maybe three coming in the mutant subset.
Yeah. I think earlier you mentioned their phase II data as anchoring your conviction. I think in that data, I believe some of the patient characteristics were less favorable than in OPERA-01.
That's accurate.
Right? Can you speak to that? Because that's sort of like a headwind on your 5.5 number. How does that figure into your math?
Right. We think it is. The biggest difference in that respect was that in the phase II experience, we allowed prior chemotherapy.
Yeah.
We know from a variety, it has been known for a while, but I think the EMERALD data's retrospective analysis really showed that that is a pretty strong negative prognostic. We eliminated that from OPERA-01.
You could not have prior chemotherapy in the metastatic setting. The other thing that we did is we said, look, whatever your prior endocrine regimen was before going on this trial, you had to be able to be on it for at least six months.
Without progression.
Without progression.
Yeah.
That's an effort to avoid primary endocrine resistance. These aren't strong selectors, but they do seem to have some power. So the 5.5 was achieved in that, and obviously in the phase II we didn't make that criteria.
So the 5.5 was achieved with the chemo, with the enrollment of patients who progressed relatively quickly, and then in OPERA-01 we don't do it. So the hope is that those truly refractory patients, we somewhat select against their enrollment.
Okay. All right, so then data, as you said, 1Q 2027, unless there's a further change, but that sounds like where it's headed.
I think that's where we're going to be.
Okay. Where are you with regard to thinking about the commercial build-out?
Yeah.
If this is a tractable market, the second, third line setting that you would be running the launch yourself. Is that right?
In the U.S. Yes. A couple of things, right? I'll do our strategy, and then I'll do the commercial build. I'm going to reverse your questions a little bit. Our plan is to launch this initial indication, second, third line monotherapy in the U.S., promote it ourselves. That is a doable, that's probably less than 100 field sales that is needed to do that effectively.
We are enabled to file outside the U.S., to progress the regulatory packages. But we do not plan to launch and promote, so we will have to seek a collaborator. Obviously, we're six months maybe and change from the data, so our assumption is that a collaborator will want to see the data before they enter an agreement.
That's where we are with that part of the strategy. With regard to that U.S. launch, we've already started the commercial build. We've hired people, we've done market research on it. We have a full plan to ramp up. Some of it will continue ahead of the data. Obviously, the data being supportive of launch, there will be an inflection point, and we'll ramp it up quite quickly at that point. Probably early next year.
Okay. One other question I forgot to ask, but maybe you haven't disclosed this yet. You're counting the PFS events. Is it across both mutant and wild type, or do you need a certain number of PFS events in mutant and a certain number in wild type? Or how does that work?
Yeah, it's interesting. The way the trial is designed, if you look at the trial from the standpoint of execution or a patient coming into it's one trial. We have one inclusion/ exclusion criteria. There's one randomization. It's the same. You're stratified, right? So you get tested and you're stratified, so you make sure it's even between the arms.
When you go to the analysis plan, it really looks like two trials, actually. It looks like an ESR1 mutant and ESR1 wild-type.
Okay.
There are actually statistical design that are different between the two. What happens is you have to achieve a certain number of events in each of the populations. The thing is, there is one unblinding. There is one closing of the database, there is one unblinding.
Right. It is not like you can.
Whichever one comes later will be the one that triggers that unblinding.
Right.
It is not like you can read them out separately.
Right. Even if one meets the threshold on the events.
We just wait. It will probably have more. We will get a little more power in that particular population than we had planned in the statistical design, but that is fine.
Okay. Cool. All right, so that's second, third line, then you mentioned front line and OPERA-02.
Yeah.
There's a lot of interesting things that have happened with persevERA, and we have the SERENA-4 obviously coming up soon.
Coming. Yep.
Watelle, first of all, we learned a lot of things a little bit from persevERA. You learned something about the control arm performance, which maybe helped you in thinking about your frontline trial.
Right. Yes.
Is there an opportunity to revamp the powering now that you have a cleaner sense of the control, the modern control of arm performance? Maybe just speak to that first.
Sure. Right. Just to set the context here, one of the things that happens in oncology fairly broadly, not necessarily specific to this, but seems to be applying to this, is that a given regimen often performs better over time. And what that probably is it's probably learning curve from the oncologist. They are very good at learning how to manage, particularly the AEs maintain dose intensity. You will see in many different tumor types that the initial approval of a given regimen has a certain PFS benefit, and then as you look at what happens over time, as more trials are done with that, it improves. We suspected that that might be the case here, because what we're using for the design of OPERA-02, which is a 1,000-patient trial now, 500 per arm.
It's a one-to-one randomization. Again, that's with KISQALI, letrozole is the control arm, palazestrant is the treatment arm. We had always suspected that the control arm might outperform the 24, 25 months that was historical.
persevERA suggests that that is the case. In the overall population, the median PFS in the control arm, which this, again, that's not KISQALI, it's IBRANCE, but for PFS they were similar, is 28 months. That's three months longer.
But actually, remember, they made a mistake. They enrolled endocrine-resistant patients who had progressed within one year of completing adjuvant. Those patients did very poorly. If you exclude those patients, it's actually 33 months. So, it's almost five months longer than, I'm sorry, 31 months. So, it's almost five months longer than the historical.
We may have to revise our statistical assumptions. What can you do about that? What it does is it, obviously, you're doing a hazard rate. The way you design a trial is you have, "Hey, what's my control arm going to do?
How much delta do I need to show to be clinically significant?" You calculate out a hazard ratio, and you decide how many events you need, and then how many patients you need at risk, and what will the timing be.
The only variable we really can control right now is patients, number of patients. We may increase the size of the trial. What would be great, persevERA 's informative tells us two things. One, maybe the control arm's doing better than it did historically. Not a big surprise. It's not a ridiculous amount, but it is significant. Two, don't enroll the endocrine-resistant patient.
That's it.
You're not.
We don't.
Right.
Nobody does. SERENA-4 doesn't either, and AstraZeneca rightly has made this a significant point about one of the differences in their trial. OPERA-02 never enrolled them.
But for persevERA , they erroneously enrolled some of these endocrine resistant, or they slipped into the study, or they weren't careful enough, or?
They did it on purpose.
They did this on purpose.
Initially.
Right? Because if you go back and look at the history of it, they allowed these patients on. It is not done, none of the MONALEESA, MONARCH, if you look at the pivotal trials for the CDK4/6, they didn't do that. To your point, SERENA-4 does not do that. OPERA-02 does not do that. They did this. On their first amendment, they stopped.
Clearly, at some point, somebody decided, "Wait a minute, we don't want to be doing this through this trial." They then removed the eligibility of those patients progressing who were endocrine resistant and went to the more traditional. At this point, they already had 10% of the total. Those patients did very poorly on the trial.
The overall data that you saw there was 28 months in the control arm letrozole, 33 months in the giredestrant arm. So a clear signal, though it did not achieve statistical significance. Interestingly, in that group of patients, about 100, who were treatment-free interval less than a year, they did poorly on the control arm. It was 19 months instead of the 28. But even more poorly on the giredestrant arm, 14 months versus what was 33 in the overall population. Even though it's only 10%, it can hurt you. We didn't do that.
Right.
I think the other thing we learn is maybe there's this outperformance of the control arm. So why haven't we changed the trial and told everyone what we're doing? SERENA-4 has the same control arm, basically. It's palbo plus AI.
It would be great to have two things. One, two data sets, right? It just helps you refine what's really happening here. The second thing is, as we've discussed, the SERENA-4 population is the OPERA-02 population. So, it's a more relevant trial to learn from.
Okay. We're going to get that relatively soon so you'll see that then you'll assess situation.
With both data sets, we'll be able to take a look.
Okay.
This is part of the reason we aren't communicating. I mean, OPERA-02 is enrolling brilliantly, but we aren't communicating a timeline for readout yet because you don't communicate it if you think there's a reasonably high likelihood you're going to change the trial.
Yeah. Okay. Speaking about probabilities of success, I mean, persevERA, 0.89, as you say, it had an effect, but it didn't make it statistically.
Exactly.
SERENA-4, we'll see what happens. We've made the argument in our research that palbo's got some potential advantages, obviously, on PK/PD in terms of exposure and mechanisms. If SERENA-4 hits, then talk about how you think about the probabilities of OPERA-02. If you need to up raise the power, you raise the power. If not, you don't.
Right.
How would you think about the likelihood of the frontline trial working?
Yeah. First of all, we have quite great confidence in OPERA-02. That confidence is based on. This is a fairly obvious thing to state, but I must say that people and investors in particular do get a little lost here. The best way to predict the future outcome of a drug or a combination is to look at its past performance in the clinic.
So when we did the ribociclib palazestrant combo in phase II, those were patients who had progressed on prior CDK4/6. So they got CDK4/6 plus AI. They are now getting a second CDK4/6 now plus palazestrant. Actually, about 30% of the patients had had two prior AIs, so they would have had fulvestrant, or two prior CDK4/6s, so they would have had fulvestrant probably as well. And we got a median PFS in that population of over a year, which is pretty extraordinary.
That has not been seen before. CDK4/6 after CDK4/6 does not work very well in general. So that is what gives us confidence in the OPERA-02 trial. Now, that said, if you have an agent which we think is inferior in camizestrant, and yet it is able to beat the AI in this context, I think your probability of success has to go up for OPERA-02.
I think the other aspect that is really important to us is differentiation. Obviously, one great way to differentiate is to prolong stability of disease longer. But the other thing that is really important is that had persevERA been positive, if SERENA-4 is positive, which we hope it is, there is still a challenge, which is doctors and patients do not want to take IBRANCE anymore, right? They get a survival benefit from KISQALI.
We are the only company doing a trial with a next generation endocrine agent, a CERAN, in endocrine-sensitive patients with KISQALI, and that is the standard of care. Should OPERA-02 read out positive, doctors and patients do not have to sit there and try to make this choice. They just give the CDK4/6 they want to give and substitute in the end.
Okay. Then the other big, bigger.
That's, I should say, a $10 billion a year plus.
Okay. Then back when lidERA worked last year, everyone got very excited about you guys in adjuvant, right? Stock was up crazy, 200% or something.
It's not crazy. It was appropriately bad.
Appropriately.
Yes.
Where does that. That's obviously a big and very expensive study.
That's our biggest, to be perfectly honest, those attributes, size, and cost. Size less so. Once you're doing a trial, once you're at 1,000 patients, 1,400, going to six or seven isn't that big a deal, actually. The cost.
Yeah.
Right? If you think about this, in order to get the events, you are going to do this in at least a moderate to high risk adjuvant setting.
Well, the standard of care in that patient population right now is really a CDK4/6 plus AI. It is Verzenio or it is KISQALI, both of which have labels in that adjuvant setting. What that means is you are giving that CDK4/6 for two or three years, depending upon which one you use, to every patient on that 4,000 or 5,000 patient trial.
You need well over $1 billion to do this, a large proportion of which is drug cost, and we do not have it. We anticipate that when we get to the collaborator discussion thing, that that will probably be a topic people want to talk about. The lidERA data is impressive. It really is an excellent treatment benefit. It looks like it is better tolerated than AI too, which is really attractive in the adjuvant setting.
The challenge with lidERA, and it was I did my little bashing of persevERA . The challenge with lidERA is it was designed appropriately when it was designed. Doing monotherapy AI versus monotherapy giredestrant was appropriate at that time. The problem is that for most of the population in the trial, the standard of care has moved
to CDK4/6. What do you do? How do you decide that? Roche, to their credit, has said, "We are going to go run a CDK4/6 adjuvant trial," which is absolutely the right thing to do. It is just going to take a long time.
Okay. All right. Well, we got to speed up here and cover some other territory, because you have other programs.
Yeah.
Can we try to rapidly summarize where you are with the OP-3136, the KAT6 inhibitor? You had some data, I was at the poster at ASCO.
In June, yeah.
Right? So yeah, so just sort of summarize where we are with that one.
Yeah. That molecule is, we haven't selected a dose yet, so it's kind of continuing in phase I. We don't have a maximum tolerated dose. We don't have dose-limiting toxicities. But we hope to get that resolved in the not-too-distant future. We have already started the combinations in breast cancer, so fulvestrant and palazestrant, both at full dose of the endocrine agents, are currently dose escalating OP-3136 in there.
Preclinical data suggests that palazestrant should differentiate there, so we're very excited about that data. Fulvestrant's a little ahead because obviously it's a well-established, approved agent, so we had to do a little more safety running with palazestrant, but that's progressing nicely. The other thing that we saw that was really interesting in the monotherapy data we presented at ASCO was monotherapy activity in castration-resistant prostate cancer.
This was also tested by Pfizer with their molecule, and they didn't see a signal. In Q4, around ASCO, we announced a collaboration with Bayer. They're providing darolutamide, NUBEQA. We're going to start our prostate cancer cohort with darolutamide in Q4. So we're now doing breast and prostate.
We will have an update on the combination data certainly in the first half of next year. There is some possibility, depending upon what we see, that we might be able to get one in there by the end of this year, but we don't know yet.
Remind us, you picked daro versus apa?
Yeah. So there are two factors that went into that. We did talk to both parties. We did two things. So our internal team did an assessment. One was talking to the prostate cancer experts, and patients prefer NUBEQA. It's better tolerated.
It has less potential serious side effects. The other thing that we noted in looking through the profiles through drugs is that pharmacologically, NUBEQA is cleaner. We didn't identify any real liability we were worried about, but we thought, look, hey, better tolerability and cleaner pharmacological profile, this is the one to go with. And obviously, Bayer was interested in the data we were generating, so that is helpful as well.
Okay. One last question on AI, and I don't mean aromatase inhibitor, I mean artificial intelligence.
Yeah.
The one company where we have to make that clarification. Can you briefly, we're asking all the companies this: To what extent is AI featuring in your workflow internally in terms of data analysis, looking through the literature, preparing materials for the FDA, et cetera?
Yeah. You're hitting on where it really is useful. We have a collaboration with a Bay Area company called Collate, and what we've found is that the part of AI that really works well for us is not trying to find some biological association or discover a drug, but actually analyzing these complex data sets and synthesizing, simplifying.
There are two things. One is, what we used to do when we were doing a global trial is we would, say, have to make an amendment or do the trial, and we would have the thing translated. We would spend hundreds of thousands of dollars and wait a month, and we would get back all the verified translations, then we'd be able to send it to the sites. This is done automatically now. It takes hours, and it's verified, and it's fantastic, and it doesn't cost us anything.
The other thing is we do anticipate in the data analysis, not so much the analysis, but the assembly of the package for the regulatory filing, that we will use it to help us put those things together.
You still have to review it, and by the way, the regulatory agencies want to know exactly what you did, because they don't want a filing that's just done by AI. They want the sponsor to really review the data. I think it's going to be a great assistant, but we still have to review it.
Are there specific platforms or modules that you use? There's Copilot, there's Gemini, there's a whole bunch of different ones. Or that's
No. We have a licensed version of ChatGPT that the company knows how to use and can use for their
everyday work. But really, when you are dealing with highly regulated, controlled documents like clinical trial protocols, consents, data that comes from the trial, we have to have a verifying system, and that is the one we use, is from this company.
Okay. Very good. Last question. Just at key catalysts, I know, of course, the readout in 1Q on OPERA-01, but anything.
SERENA-4 is a catalyst.
And SERENA-4, of course.
We've got to say that. Love it to be positive, make my life easier. Then OPERA-01 in Q1. I do think KAT6 as we start to generate more combo data is definitely a catalyst. That, again, is a $5+ billion market opportunity just in breast cancer.
Prostate cancers, we haven't forecast it yet. We're still working on it. We know what we want to do in the trial, but we haven't really done a commercial case. We think it'll be quite compelling. I think those are really our major catalysts for the near term.
Okay. Awesome. Well, great. Thank you very, very much.
Thank you, Yigal.
All right.
Pleasure. Thank you all.