Stoke Therapeutics, Inc. (STOK) Q3 2026 Earnings Call Transcript

I'll start with some highlights from our phase three study. To date, approximately 110 of 162 enrolled patients have completed their week 28 visit. This is the time point at which data for the primary endpoint of change in major motor seizure frequency is captured. Approximately 20 patients have also completed their week 52 visit, marking the end of the study treatment period and the time point at which key The secondary endpoint data will be captured. All of these patients are continuing into the open-label portion of the study. Notably, there have been no treatment discontinuations to date. Barry will provide some additional metrics and context in a few moments. Next, I will highlight our recent interaction with the FDA that gave us the opportunity to discuss our five years of clinical data for Zareva nursing, including our Phase I, II, and OLE studies. And to align on key aspects of our planned 2027 USNDA submission. We were pleased with the agency's engagement, responsiveness, and interest in zirubinocin. An important outcome of our meeting was the FDA's willingness to accept a multi-component assessment of Vineland-3 data for purposes of evaluating the key secondary endpoint related to changes in cognition and behavior. This is a positive outcome because we believe I believe a multi-component approach acknowledges the complexity of the disease and the fact that each patient's experience with Dravet syndrome is different, especially when it comes to impacts on cognition and behavior. Coming out of the discussion, we are reiterating our timelines for a data readout in Q3 2027, followed by a planned NDA submission in the second half of next year. Clarity on these timelines is important as we scale our organization and capabilities to deliver Zareva nursing to patients following a potential U.S. approval in early 2028. Our medical affairs team has done an incredible job over the past two years, establishing a foundation of awareness and understanding of Dravet syndrome and the data to support the disease-modifying potential of Zareva nursing. We are continuing to expand our educational Thank you. and other genetic epilepsies. We are encouraged by the actions of stakeholder organizations such as American Epilepsy Society's Presidential Task Force on preparing for disease-modifying therapies and education and advocacy related to the need for genetic testing. Jason will share more on these efforts in a few minutes. We see haploinsufficient diseases as a clear and unique opportunity for our platform, which is why we are excited to advance STK002. Autosomal dominant optic atrophy, or simply ADOA, is a rare genetic disease that is primarily caused by a haploinsufficiency that results in 50% of functional protein production. SDK 002 is designed to increase naturally-occurring, fully-functional OPA1 protein production to address the underlying cause of this disease. In addition to our preclinical data, recent data from the field supports the potential of the OPA1 targeting hypothesis. Dosing is now underway in the third cohort of our Phase I Osprey study, and we expect to initiate the fourth-dose cohort in early 2027. Initial safety and efficacy data are anticipated in the first half of next year, which will inform future development. As we expand our research efforts, we will continue to focus on severe genetic diseases driven by haploid sufficiency with an emphasis on the CNS given our experience in Dravet syndrome. With that, I would summarize by saying that Stoke has made great progress Thank you, Ian.

I'll start by expanding on our Phase III progress that Ian mentioned. EMFEROR is a global, double-blind, sham-controlled Phase III study of Zareba Nursin in patients with Dravet syndrome. The study is designed as a 52-week treatment period in which patients receive two loading The primary endpoint is percent change from baseline in log transformed major motor seizure frequency at week 28. All secondary endpoints are measured at week 52. To protect the integrity of the study, we will not review or analyze any data until all patients are measured at week 52. The study is progressing very well. All 162 patients enrolled in the U.S., U.K., and Japan are through week eight of the study, which means that they have received either the two loading doses of 70 milligrams of zarevanurcin or sham. Approximately 130 patients are through week 24, and therefore have only one of the four doses left In the 52-week treatment period, approximately 110 patients have also completed their week 28 visit, the time point at which the primary endpoint of change in major motor seizure frequency is measured. Approximately 65 patients are through week 40 and have therefore received all four doses of zarevonersin or sham. Approximately 20 patients have reached the end of the 52-week treatment period. All of these patients are continuing into the open-label extension period. No patients have discontinued treatment in EMPERER. When we originally designed EMPERER, we set an enrollment target of 150 patients. This was based on a goal of 128 evaluable patients at the end of the study with an assumed 15% discontinuation rate. We ultimately enrolled 162 patients, and as I mentioned, there have been no discontinuations to date. These metrics reflect a well-designed trial, the need for better treatments, and a deeply committed community of investigators, site teams, and caregivers. The primary endpoint of the emperor is change in major motor seizure frequency at week 28. There's no doubt that anti-seizure medicines are providing benefit for patients. Even still, seizure control remains the number one goal for new treatment. Demonstrating a disease-modifying impact on cognition and behavior is expected to take longer, which accounts for the year-long treatment period and evaluation of the Vineland results at week 52. As Ian mentioned, for the purposes of our US NDA, we plan to use a multi-component assessment that combines four pre-specified 5 and 3 subdomains. Expressive communication, receptive communication, interpersonal relationships, and personal skills. We believe the multi-component assessment is the most appropriate way to evaluate changes in cognition and behavior due to the variability Let's take a look at how the disease affects patients. To generate a multicomponent score, we first calculate the individual treatment effect for each subdomain using a multivariate mixed model for repeat measures, or NMRM. We then average all four scores to produce a single number that reflects the overall treatment effect. I'll show you what that looks like in a minute. Before I do that, I want to remind you of an analysis that we performed back in 2024 to inform the design of the Phase III study, and specifically to help us prioritize the selection of Vineland subdomains. On the left side, you see data from a subset of patients treated in their Phase I, II, and OLE studies. These patients received a total cumulative dose of zerebinersin that was consistent with our plans for phase three. On the right, you see data from patients followed in our natural history study whose baseline characteristics were matched to the treated patient population. Week 68 was used for these analyses because it was the most proximal to a 52-week study design when accounting for the pause and dosing Between the last phase 1-2 dose and the first OLE dose. Selection of these subdomains was further supported by feedback from clinicians and caregivers. Now, with an additional two years of data, and in light of our recent discussion with the FDA, we generated a multi-component analysis using propensity score weighting to calculate a treatment effect. Use of propensity score weighting mimics randomization by adjusting for baseline differences between treated and natural history groups. This is the closest we can come to replicating a randomized trial. This analysis demonstrates the overall treatment effect across the four subdomains using a multicomponent average, as well as the contributions of each of the four selected subdomains. Even with small numbers of patients, a highly statistically significant multicomponent score was demonstrated. With five years of clinical data, an assumption of 150 evaluable patients, and the change to a multicomponent assessment of the Vineland 3 for our US NDA, we have run a calculation of the minimum treatment effect to achieve a p-value of 0.05 to 0.01. The study has 90% power to detect a treatment difference between treatment arms of 46 to 53% relative reduction in seizure frequency at a two-sided significance level of 0.05 to 0.01, respectively, at weeks 28 and week 52. For the second key, secondary endpoint, measured by volume 3, the study has 90% power to detect a treatment difference of approximately 1.8 to 2.1 raw score points in the multicomponent assessment at week 52, using the same two-sided significant levels of 0.05 to 0.01, respectively. We are often asked to explain the real world meaning of these violent scores. What I can say is that even small changes can have meaningful impacts for patients and their families. This has been documented in the literature and is evident in the video case study that was included as supplemental to our data published in the New England Journal of Medicine earlier this year. We will continue to keep you updated as we progress the study, and we look forward to our data readout in Q3 of next year. Let me turn to our second clinical stage program, STK002. Momentum and awareness of this program are increasing. Earlier this month at the American Academy of Ophthalmology Annual Meeting, our teams delivered two presentations of data that support the potential of SDK 002 as a disease-modifying treatment for ADOA. ADOA is a rare genetic disease that leads to progressive degeneration of the optic nerve and vision loss starting in the first decade of life. It is the most common inherited optic nerve disorder, affecting approximately 13,000 in the U.S., U.K., EU4, and Denmark. The majority of cases are caused by mutations in one copy of the OPA1 gene, resulting in haploid insufficiency and reduced functional OPA1 protein. Approximately half of people living with ADOA will progress to legal blindness. There are currently no approved treatments for ADOA. STK002 is designed to increase OPA1 protein expression with the aim of improving vision in people with ADOA. We are encouraged by our preclinical data, as well as emerging data from the field, demonstrating that upregulation of OPA1 protein has disease-modifying potential. In ADOA patient-derived cells with different Opa-1 variants, STK002 significantly increased Opa-1 protein expression and improved mitochondrial function compared with healthy controls. On the next slide, you see that after a single intervitreal dose in nonhuman primates, STK002 was also shown to reach retinal ganglial cells and increase retinal OPA1 protein. On the left side, staining shows STK002 in red in retinal ganglial cells with increased signal correlating with increased dose. In the center, immunofluorescence staining with an Opa-1 antibody, again in red, shows an increase in Opa-1 protein in the retinal ganglion cell layer. Opa-1 protein levels increased at all doses by week four by 44 to 47 percent at the 0.3 milligram and 1 milligram doses, respectively. The Phase I OSPRI study in people with ADOA is ongoing in the UK and Europe. The study aims primarily to evaluate the safety, tolerability, and exposure of single ascending doses of SDK 002, following a standard dose escalation design. OSPRI will also assess clinical activity, including changes in visual function and patient quality of life, As well as retinal mitochondrial dysfunction, data from our natural history study show higher levels of mitochondrial stress strongly correlate with decreased visual acuity. OSPRI is designed to provide clinical proof of concept by testing whether increasing OPA1 protein expression with SDK 002 translates into measurable effects on visual function in patients with ADOA. The study is progressing well. We recently completed dosing of the first two cohorts of patients, and dosing in the third cohort is underway, with completion expected by year end. Subject to ongoing safety assessment, we expect to begin dosing of the fourth cohort in early 2027. We expect cohorts three and four to reach dose levels that may produce measurable effects on visual function. Initial safety and efficacy data from these higher-dose cohorts are expected in the first half of 2027. With that, I will turn the call over to Jason.

Thank you, Barry. Shifting back to Dravet, I'll start by sharing an overview of the patient population. While we've shared these data previously, our latest market research continues to validate our assumptions and strengthen our confidence in the numbers. There are an estimated 38,000 patients with Dravet syndrome across the seven major markets where we're running the EMPROR study, the U.S., U.K., EU4, and Japan, including approximately 16,000 in the U.S. alone. Of those 16,000 in the U.S., we estimate 6,000 are under the age of 25 and likely under the care of a pediatric provider. These are the patients who would be immediately addressable at the time of a potential U.S. launch. Pediatric neurologists and epileptologists tend to follow the field closely and are typically the most well-informed about the disease and current treatment options. An ICD-10 code was established in 2020 and is used to track confirmed Dravet diagnoses. With the increasing awareness of Dravet syndrome and the importance of genetic testing, we're seeing 40 to 50 new patients entering these claims databases every month, equating to 500 to 600 every year, Including newly identified younger patients and older patients finally arriving at a diagnosis. As a result, we anticipate the total number of immediately addressable patients to increase between now and the time of launch. We're confident that we know where 80% of the 6,000 immediately addressable patients are being cared for today. In the U.S., the Dravet population is highly concentrated around Centers of Excellence and other key treatment centers. Approximately half of identified U.S. patients are cared for by the top 50 sites. All 50 of those sites are experienced in administering intrathecal medicine, and over half of these sites have participated in a Zareva Nursen study. There's a high level of awareness of Zareva Nursen in anticipation for the Phase III readout and potential approval. The Dravet community has been involved in this program since its earliest days, and sites are anticipating strong demand upon a potential U.S. approval. Several of the top treatment centers have proactively reached out to us, asking what they can do to be ready to deliver Zareva nursing. Our teams are beginning to engage with sites to help them prepare the infrastructure protocols and staff expertise necessary to administer Zareva nursing with appropriate capacity. The Dravet community has been waiting years for a medicine like Zareva nursing. Our job is to do everything we can to ensure they don't have to wait a day longer than they need to. To that end, we're adding resources and capabilities with an emphasis on people who have launched high science new medicines, particularly for rare genetic diseases, and are committed to delivering excellence in patient care and overall experience. Our head of patient services recently joined Stoke, and she's in the initial stages of designing a patient support program that reflects the unique needs of this patient community and enables timely access. Our goal is to make the process between prescription and treatment as simple and easy as possible. Co-creation sessions with caregivers, clinicians, and office administrators are underway to design a program that meets their needs. On the payer side, we recently deployed our payer account team, who will start their educational efforts with medical directors at insurers across the country. These educational efforts build upon the years-long effort of our medical affairs team, who have supported our clinical teams and have done an exceptional job of education and awareness with healthcare providers and treating physicians around the world. Our latest HCP market research shows a high conviction in the data and in the disease-modifying potential of Zareva Nursin. We're also making progress among the broader neurology community, who may be less familiar with the latest advances in the treatment of Dravet syndrome. We are very encouraged by the growing awareness and interest we experience at the recent Child Neurology Society meeting. As recently as three years ago, very few people at this meeting knew about Dravet, Zareva Nursin, or Stoke. Today, there's a far greater appreciation among general neurologists and epileptologists for the genetic drivers of Dravet syndrome and the reclassification of genetic epilepsies as DEEs. The ability to genetically diagnose these diseases is also prompting physicians and medical societies to rethink the treatment paradigm. One such example is the American Epilepsy Society's new Presidential Task Force on Disease Modifying Therapies. The intent behind this new initiative is to prepare those who treat these diseases for the arrival of genetically targeted treatments. We're encouraged by initiatives like this one, which complement the work we're doing, signaling a shared commitment to improve outcomes for patients and recognition of the potential for a new class of medicines. Beyond our educational efforts, we're preparing to initiate two additional clinical studies of Zareva Nursin by the end of this year. This includes a study in adult patients, which is intended to support access and reimbursement, as well as a study in infants and toddlers. The primary objective of these studies is to evaluate the safety of Zareva Nursin in these populations. These are not registrational studies. Consistent with our breakthrough therapy designation, we believe we can achieve a broad label without an upper age limit. Taken together, the strength of the clinical package, the continued positive feedback from payers and providers, Thank you, Jason.

I will remind you that full financial results can be found in today's press release and our 10-Q that will be filed next week. Today, I will focus on the strength of our balance sheet position. We ended the quarter with $358.8 million in cash, cash equivalents, and marketable securities. This excludes restricted cash of $3.2 million. Our net loss for the three months ended September 30 with $73.6 million, which include $11 million of non-cash stock-based compensation expense. Our priorities at Stoke have not changed. Advancing our Phase 3 study of Zareeva Nursen, preparing the organization for potential U.S. commercialization, and investing in our pipeline while maintaining a strong financial position. Along with the reimbursement we received from Biogen for Zareva nursing-related expenses and a potential development milestone payment, we expect our cash position to fund our operations into early 2028. Thank you. I will now ask for the line to be open for Q&A.

Thank you, ladies and gentlemen. To ask a question at this time, you will need to press star 1-1 on your telephone and wait for your name to be announced. To withdraw your question, simply press star 1-1 again. Please We stand by while we compile the Q&A roster. And our first question coming from the lineup, Kevin Strang with Goldman Sachs. Elon is now open.

Great, thanks. Just a question on the design for Emperor in terms of data quality monitoring. Can you walk us through, you know, how you designed the trial with that in mind in terms of the Violent 3 subdomains cores, how you could consistently collect that across different time points, across different sites? That would be helpful, and then as just a brief follow-up, on the multi-component analysis plan that you have in place now, are there any meaningful differences between this plan with the FDA and what you have planned in Europe? Thanks.

Thanks, Kevin, for the question. This is Barry. So, the collection of the data for the Vineland in the study is vitally important, and so what we have put into place are several measures. There's one important one is that there is a trained central rater who collects all the data and puts together the scores so that that is uniformity across all the different sites. Second is that we encourage the same family member to be the reporter on the questionnaire. So at every time that the questionnaire is asked, it's the same person giving the response. And then we also do have an internal spokesperson who is reviewing on nearly a daily basis to make sure that the Vineland tests are being collected and done appropriately at every site.

Kevin, it's Ian here. I'll take the U.S. versus Europe question. Correct, and I'm sure we'll talk about it more on this call, the MC4, as we call it now. You know, that's for the U.S., but in Europe, they still have what's called a net treatment benefit for individual violence subdomains, and there's five of them, in a similar way that we had before we got to the MC4. Biogen is in control of that filing in Europe, and we'll wait to see how they proceed with those authorities overseas. Great. Thank you.

Thank you. Our next question in queue, coming from the line of Laura Chikwu with Bush Ilan is now open.

Good afternoon. Thank you very much for taking the question. Just two for me. I guess following the interactions with the FDA, I'm wondering if you could elaborate if there's been any change on the potential for an advisory committee meeting. How have the expectations there changed? Just a point of clarification, what remaining areas need to be resolved prior to the rolling submission? submission. And I'm sorry, one last one for Barry, perhaps. You recently presented some data at the CNS meeting on the OLE data. Could you just remind us on how the protocol there handles discontinuations of pre-existing ASMs? I guess maybe said it differently. I'm trying to remember whether patients are coming off some of their baseline ASMs when they were, what they were on at study entry. Thanks.

So, Laura, I'll take the first two, and then Barry can take the The question regarding maybe a change in medicines or ASM medicines through to week 28. And so firstly, ADCOM, we don't anticipate an ADCOM. There was absolutely no discussion relating to an ADCOM when we had our FDA interaction a month or so ago. So as far as submitting data for the NDA, we anticipate submitting data throughout the first half of 2027 with a completion of the submission in the second half of 2027. And what we want to do is we want to submit that data to give the FDA the option to review early. There's no guarantees they will review early until they actually confirm that they're doing a rolling review. And so we will await that, but we will be submitting the data to the FDA to give them the option. And we'll be continuing to work with the FDA to talk about the importance of what our medicine has shown, basically through our Phase 1, 2, and our OLE data. And we've had multiple meetings outside just this meeting with the FDA to show the importance of our data and the urgent need to move fast through our submission and potentially through our approval. And Barry, do you want to talk about how we treat change in ASF medicines through week 28?

Sure. Thanks, Laura, for the question. So, the background anti-seizure medications, the patients are monitored for the first six weeks before treatment with the background anti-seizure medications kept stable, and then they're supposed to stay stable through at least the first six months to the week 28 endpoint. And, again, the background medications are continued, so any effects that we're going to see with Zareva Nursin will be on top of the baseline anti-seizure medications that they're already on. This is the way we saw in Phase 1-2, that the improvement that we saw in terms of, or the reductions that we saw in seizures, seizure frequency, were on top of background standard of care anti-seizure medications.

Laura, could I just also add, it's a great question, one actually Barry and I frequently talk about, and that I want to remind everyone that there is an eight-week screen period before you come into the EMFRA study. That eight-week screen period, there's a number of things, but one of the things it does is it allows you to assess that these patients are already on a stable background of medicines, and so when they enter the study, whether they're going to sham Our active drug arm, they're on a stable background of anti-seizure medicine. So we would anticipate there to be small changes in terms of anti-seizure medicines, but there probably will be changes, and we will assess week 28, and we'll look at the data with and without the changes of those background seizure medicines.

Thank you very much.

Thank you. Our next question in queue coming from the line of Pizza's Prepolis with Cancer Fitzgerald.

And team, congrats on the progress, and thanks for taking our questions. A couple from me. For the Vineland 3 subdomains, you showed a very impressive forest plot, you know, compared to natural history. You mentioned on the call that even small changes can be meaningful. Just, you know, help us understand, you know, what those treatment effects sort of translate to, you know, with examples. Also, you know, that was also at 18 months. You know, what gives you confidence that the one-year time point is sufficient to capture differences between the two different arms under the current multi-component assessment of the four?

Pete, this is Barry. Thanks for that question. Very important. So, when we look at the treatment of the Vineland changes. We know that even small changes can be reflected impactful and impact on the families. And we've heard that from many families. The measures of the Vineland are scores of zero, one and two, for instance. And the so that means that a score of zero would mean for a particular activity, that child would never do that. And a score of one would be usually. So if you ask the question, for instance, does the child respond to their own name or be able to say their own name? Those are questions. If you can imagine a child never being able to say their own name as opposed to sometimes being able to say their own name, that's a change of one point, which, as you can imagine, is quite meaningful for a family. Another would be following a simple instruction of one action. Again, zero being they never can do that, or one meaning they can't. So as you can imagine, these are effects that are having a huge impact both on the quality of life of the child as well as the family.

Pete, I'll just add to that. I'm not sure how familiar people are with Vinyl 3. To be honest, it sits on my desk. I look at it most days of the week, and I look at the questions, and Barry gives some good examples, but it also includes things like, do you recognize your name? Never, you get a zero. Sometimes you get a one. Usually you get a two. And that's how it works. And there's a long list of these type of questions that you add up in terms of total score at baseline. You then do it at week 52 in our study, and same questions, and you add it all up. Let's say there's a total score of 150 at baseline, but then you'd get a total score of 154. That would be four violent points, four extra points. That would be what, you know, maybe the drug is contributing towards. So that's how the scoring works. And so to Barry's point, you shift a somebody that never does something to a patient that does something sometimes, that's a one point score and therefore meaningful. And I would advise people to take a look at the Vineland scoring or the questions, and you'll get a good feel for the type of questions that are asked. As far as your question regarding the 12-month, and so, yes, we did put up the forest plot for the 18-month data. You know, interestingly, what we're trying to show here is that you, I would start with saying, when you take the basket of endpoints approach, you have multiple ways to measure disease modification. And contrast that to a hierarchical assessment of individual violent domains. You know, there is sometimes a little bit of a lottery there, if you can hit the first one, so you can go to the second one and hit the second one to go to the third one. With the multi-component assessment, we remove that hierarchical assessment, the need for hierarchical assessment. And what you're doing is you're assessing four domains equally as a basket to get an average score. And that's a much better way in a disease that manifests itself with multiple comorbidities that, frankly, do manifest at different variations within each patient. What it also does, which is a really nice outcome, it's a statistical outcome, but because you've got more valuable data, you're measuring an endpoint four times, you actually get more powering. And you asked about the 12-month data and scoring if we ran the same assessment. We did. And interestingly, in those four domains, we missed p-values on three out of the four, but we hit one out of the four. Overall, our MC4 score, so our multi-component score, achieved an overall p-value of less than 0.001. Just to give you an idea of how much it enhances the powering, because you've got more valuable data, and you're assessing four endpoints equally. Now, those three endpoints that missed, they did have trends, and they were close to 0.05, so understand that, and there was one that was below 0.05, but overall, the MC4 p-value was 0.001, 0.01, sorry. So, less than. And so, you know, what we're trying to show here is that we've enhanced the powering of the study, but as a principle, it's enhanced and it's more appropriate of how to measure the cognition and behavioral benefits our medicine may be providing to these patients because it is a basket approach.

Congratulations again on the progress.

Thanks, Pete.

Thank you. Our next question coming from the line of Andrew, side with Jeffrey. See you on this now.

Hey, guys. Congrats on all the great execution. This is John on for Andrew. Thanks so much for taking our question. So earlier this week, your partner, Biogen, made a comment about how they think differentiation and differential pricing, especially for XUS, as far as revenues tend to have an impact on cognition. So just curious how you guys think about their comments and whether there is an alignment on the underlying value proposition. Thanks.

Well, I think Biogen can give you, you know, give you their best answer, and so I don't mean to speak for Biogen. But I've been involved with rare disease, genetic diseases, and readouts in the single-payer markets within Europe, for example, Which is what I think you're asking about, John. I will tell you that it is more challenging in those markets than it is in the US. And in the US, I would say the approach, and we're very comfortable talking about the US because those are our territories. But in the US, it is about the totality of the data. And we've been sharing four-year OLE data, New England Journal of Medicine, various peer-reviewed presentations with the payers, and they understand the value of the medicine, and Jason will probably comment on that later. Whereas in the EU, for example, they do hold you closer to the label in terms of valuing a medicine. And they have a whole scoring system in certain countries. But overall, it is more challenging over there. But Biogen should be able to provide you more understanding of that. But it is more challenging in Europe.

Thanks so much. And then maybe just one quick follow-up, if I may. I was curious, like, besides the Vineland 3 adjustment in the recent FTA meeting, what other topics may have come up in the meeting with the FTA? Or was Vineland 3 kind of the main focus? Thanks.

There was a few other things. I think Vineland 3, getting to the multi-component 4 assessment, was the priority focus. The submission of data to the NDA was important, and there was also, we did ask the question regarding the OLE data, the four-year OLE data. data, which will be five-year OLE data by the time that we submit our data. We asked them whether that can be included in the label, and we got a clear answer from them, which is it will be a review issue, which is a good answer for us, and that this is a chronically dosed medicine, and by the time we file, we're going to have five years' worth of data to Show how these children may have compounded benefits to year one, to year two, to year three, to year four, to year five. And we believe that's important, and the regulations for, you know, labeling is that you put clinical study data on the label that is supportive of an understanding of the benefits that a medicine may provide. Overall, the interaction with the FDA is always a great opportunity to continue to educate, And that's where we started with the OLE data, because it is a chronic disease, and we're dosing a chronic medicine. And we've come a long way with the FDA over the last two years, and I think that was one of the reasons it drove the outcome to the multicomponent assessment, where we're assessing

cognition and behavior on a basket of endpoints now.

Understood. Thanks so much, and congrats again on all the progress. Thank you.

Thank you. Our next question in queue, coming from the line of Sumon Kulkani with Ken Court, Yelanis Nelvin.

Good afternoon. Thanks for taking our questions. This is again related to the violence scale and the multi-component assessment used as a key secondary endpoint. Does a relative contribution of a subdomain to the average score assume importance, and would high deviations from the mean on subdomains make a difference to being able to support a disease-modifying claim on the label? I guess asked another way, would an average that meets the definition of clinical relevance Can the difference still be viewed that way if one or two subdomains drive most of the difference, even if the others may take away from the effect?

Thanks for the question, Suman. You know, the way it does work, and I'll keep this simple, which is you do assess equally each subdomain. So there's four subdomains you assess equally. Let's say you score a three, a three, two, and a two. Make the numbers easy. Divide that by four, and that gives you a multiple MC4 score of two and a half. That's what you then assess the p-value. One of the things I want to be very clear on what we're trying to help you with these forest plots that we've showed today, that two and a half, if you score it, because you're assessing Vineland four different ways, and you've got more valuable data, you get more power to that score. And that's why when I referred to the 12-month or even the 18 And so what it really translates to is you could miss on one of the domains in terms of a p-value, but still achieve a significantly low p-value, potentially 0.01, as we saw on the 12-month data. And I'll just take this opportunity. When I say we missed one on the 12-month data, I want to be clear, and Barry was in his remarks earlier on this call, the 12-month data was suboptimal dose. There was a six to seven-month gap in dosing, and they did not get the level of dosing that we're dosing in our phase three. So I do want to remind people of that, but I'm giving you the example. We still hit a very low p-value, less than 0.01, even on the 12-month data. So it is a simple kind of average, but because you've got more valuable data, you get greater power.

And I'll just add to that. This is Barry. So what Ian was talking about also, the 12-month data, those were a relatively small number of patients, quite small number of patients. So obviously we're going to have more. Addressing your question a little bit more specifically, there is no, of the four subdomains, there is no one that is more important than the other when we do this calculation. And the other thing, the advantage of doing this multi-component is that it smooths out the variability and the differences between different subdomains. So even if there is one that is extreme, extremely high, for instance, or extremely low, the multi-component will take that into account and average it out.

Got it. Thank you.

Thank you. Our next question, coming from the line-off, you're on Werber with T.D. Cowan.

Great. Thanks so much, and really nice progress. So, Barry, on the last point, the MMRM is a big deal on the composite, right, because essentially it uses the entire curve in every visit. You're not just doing baseline against the end. And as you mentioned, it really helps reduce heterogeneity and ultimately really boost the power. And also, I believe that when you use MMRN, the way you handle dropouts is you don't have to impute. Can you comment on that? And also, the second question, are you doing the MMRN, the multiplicity at the same time, or is each score still hierarchical of the four domains? Thank you.

Hi, everyone. Thanks. This is Barry. So as far as the MMRM, yes, we're using the MMRM for the Vineland, and the advantage of that is that it takes into account all of the data collected up to that time point. So as you mentioned, it's not just at week 52. So we do have a collection of the Vineland, for instance, at six months in the study. So that would be combined with the data that we have at the 52-week endpoint, and it does allow then for any missing data and could then impute for missing data or take that into account. As far as the hierarchy, the multi-component endpoint will be assessed as all four of the subdomains together. And so, there is no need to put together any hierarchy, and that was part of what we had in agreement with FDA, that we can measure just that as the key – as the second key endpoint, we will just use the multi-component and not have to do any ranking of the subdomains. So, that obviously is an advantage in terms of having

power for that endpoint.

Thank you. Super useful.

Thank you. Our next question in queue coming from the line of Joseph Stringer with NITAM Milan is now open.

Hi, thanks for taking our question. Just a follow-up question on the statistical analysis. Just given the change to the MC4R analysis from the individual component, can you speak to how this affects the alpha allocation? Does this change the split across the primary endpoint, the week 28 So, Joe, it's Ian.

I think Barry referred to these endpoints and how they're measured, the primary, given this change with MC4. In the prepared remarks, we gave you the updated powering, and so let me just remind you of that. Firstly, we do have more patients that are being retained in this study than we initially anticipated. Just to remind you of the initial expectations, we thought this would be a 150-patient study with potentially a 15% dropout. That gave us an estimate of needing 128 patients to power the study. If we bring it all the way forward, as of today, we've got 162 patients in the study, of which zero dropouts, and 110 of those are already through the primary endpoint. So we believe there is going to be more valuable patients, and what we showed you today is we're making an assumption, that's all it is, but maybe there's going to be 150 valuable patients. Assuming there's 150 of valuable patients, and with this new assessment, we anticipate that we lead a 46% to 53% treatment effect for seizure reductions just to achieve p-values of 0.05 or 0.01. And for that Vineland endpoint, it's now 1.8 to 2.1 to achieve a p-value of 0.05 to 0.01. So the valuable patients that are still in the study and this change has really enhanced the overall powering of the study. But I do want to continue to kind of echo, I'll say, or repeat. The real benefit here is this is a disease with many comorbidities, and they evolve at different rates within different patients. And so being able to measure these endpoints in a basket approach is the real win here, because that's the way you should be measuring disease modification.

And I'll just add, Joe, to your question. In terms of split, there is no alpha split. So, the seizure reduction at 28 weeks is the primary endpoint. 52-week seizure reduction is the first secondary, and the Vineland is the second secondary. And so, alpha is not split between those.

Thank you. Our next question in queue coming from the lineup. Jessica Pye with J.P. Morgan.

Hey, guys. Good afternoon. Thanks for taking my question. I was curious, what do experts tell you is a clinically meaningful drop in seizures?

Clinically meaningful drop in seizures? Actually, Jason, why don't you take that? You just spend a lot of time in the field.

Yeah, it's a great question, Jess. You know, when we talk to healthcare providers in market research, I think, you know, the first thing they talk about when you talk about what the unmet need is in Dravet, it is the persistent seizure burden that exists today with the anti-seizure medicines and the desire, the strong desire to ultimately get as close as possible, if not to seizure freedom itself. And so any additional reduction in seizures on top of the best standard of care medicines is considered an advance, but I think most are expecting something in the neighborhood of 40% to 50% additional reduction over what you would see in the control arm.

And to be clear, is that an expectation for the trial, or is the threshold for clinical meaningfulness really 40%?

And it's more what their expectation is of a disease-modifying treatment versus clinically meaningful. I think if you ask them what's clinically meaningful, fewer seizures is always clinically meaningful relative to more seizures.

And just last one for me, is success in EMPROR defined by achieving STAT-SIG with a P-value less than 0.05 or less than 0.01?

Yes, this is Barry. Yeah, the answer is .05. So, this is a standard assessment that we have put into the protocol. So, the .05 for the primary as well as the key secondaries would be statistically significant.

Great. Thank you. Thank you.

Our next question coming from the line of Mark Goodman with Lyric, you're on this now open.

Yeah. The comment around the five-year data being a review issue, did that refer to if the study just misses, then the FDA would consider the longer-term five-year data a review issue? Is that what you were saying?

When it comes to misses, it always depends on how close you are on the miss, and it's always a discussion. I mean, there's many drugs that have p-values that miss on the 0, 5, but have been close. And when you take into the totality of the primary and secondary endpoints and tertiary endpoints, you can absolutely have a discussion with the FDA. I do feel as though we're in a good position with the OLE data, which will be five years of OLE data or effectively six years of OLE data by the time we file. And so it does put you in a good position. But we didn't discuss, Mark, we didn't discuss with them about missing. It was more of an education of the disease. And we started there, just as we did this time last year, when we asked if they would support an NDA submission on the basis of the OLE and Phase I data. And this continuous education, certainly over the last two years, has really brought the FDA forward to be in a really good spot of understanding the disease and how our medicine affects that disease. And so it will be the totality of the data that will drive the label.

I guess, Ian, one of the other analysts was asking about at the FDA meeting, and you mentioned three things, and the third thing was they asked about the four-year data, and you talked about the five-year data, and you said it was a review issue. What did you mean exactly when you said it would be a review issue? referring to the five-year data serving as what?

Oh, okay. Yes. Thanks for clarifying, Mark. So, when you go to the FDA, the process is you submit a briefing book. That briefing book has questions in it. One of the questions that we put in that briefing book specifically was, would you support putting the OLE data within our label? Section 14 is known as, but would you support putting that data in our label if we were We had to have approval. And, you know, as you probably know, the FDA doesn't say yes at this point in time, but they did say it would be a review issue, which is a very clear answer, which is it's the totality of data. And so that's what I meant by it'd be a review issue.

Got it. And then just secondly, just to be clear on this MC4, it can be positive even if the benefit is concentrated in one or two subdomains, the way you've described it, right? So that discussion is, you've had that with FDA. Like, if you hit your Vinland secondary endpoint with the p-value, you know, you're going to be good, right, even if it's driven by one or two of these subdomains. Is that fair to say?

That is correct. They've actually gone as far with us in the discussion to advise us that they would like to see it as a forest park, because they want to understand what drives the MC4 score. So the example you just gave, Mark, is a good one, where you may, you know, the example where you say you miss or you get a zero in one of the domains, but if you hit on the others, yes, as long as you hit your p-value on MC4, yes, acceptable.

Thank you.

Thank you. Our next question in queue coming from the line of Rudy Lee with Move Research. Your line is now open.

Hey, thanks for taking my question. Maybe a question for Barry.

For upcoming ADOA readout in first half, so what is the bar for visual retinal benefits to support your dosing selection? I mean, making sure you're reaching the therapeutic levels. Are you measuring OPA1

protein levels, and do you need further dose reading study before pivotal? Thanks.

Yeah, Rudy, thanks for the question. So, the ADOA program is a very important program for us, and we do have quite a bit of preclinical data and data from the field that supports the premise that increasing OPIL-1 protein levels in the cell can have a beneficial effect and potentially even improve vision in ADOA patients. So the phase one study is designed primarily as a safety study, which will be an important readout, but we are also looking at visual acuity, as well as a biomarker measure of mitochondrial function. So those will be important endpoints for us to analyze in terms of the decision to make to go forward. But, again, improvement in vision would be something that we would be looking for as well as the safety effects.

So it sounds like you need another, you know, dose-ranging study to determine the right dose moving forward. Is that fair?

I would say that we'll wait and see what the data are. We are considering different ways to potentially accelerate the program because this is a disease where the patient – there is no approved medicine for this patient. And patients are progressively losing vision. So there's a tremendous amount of interest both on the part of the patient advocates as well as the clinicians to see this sort of medicine. So we're going to look for the way that we can most quickly get the medicine to the patient. And to your other question, in terms of the OPA1 protein level, that's not something we're measuring, but as I mentioned, we have this biomarker for mitochondrial function. So that would be an important way that we could look at the basis for the mechanism of action of the medicine.

Very helpful. Thank you.

Thank you. Our next question coming from the line of Yatin Suneja with Guggenheim, Yulani Snellwin.

Hey, guys. Thank you for taking my question. Maybe just one more on ADOA. Could you maybe articulate for us a little bit more on what What the learnings were from the first two cohorts, and then how would you characterize sort of like the concept later this year? So that's first part. And the second is on the spend side. Could you maybe also comment on how should I model the spend in 4Q in 2027, any sort of color you guys could provide would be great.

Yatin, thanks for that important question, this is Barry. In terms of progressing the study, as I mentioned before, the primary endpoint is safety. We do have a safety monitoring committee that reviews the results, and they did review the results of the first two cohorts, and so that recommendation to move forward to the third cohort was part of the safety review.

Now, just to add, Yatan, to what Mary is saying, as we move into the third and fourth cohort, the doses of 0.3 milligrams and 0.5 milligrams, we are starting to move into the therapeutic dosing level, and that's why we anticipate seeing safety and efficacy results in the first half of 2027.

And, Yatan? And this is Tommy here. Thanks for the question. Just to help you, just to get you straight there, in terms of how to guide forward a Q4-2026, I think you should look at Q3-2026. You saw the increase we saw quarter to quarter. I think you could assume a similar number going forward.

Thank you. I'm showing up for the question. I'll now turn the call back over to Ian Smith for any closing comments.

Oh, thank you. Just want to thank everybody for the time tonight. You know, we approached this call. Thank you for submitting questions to us beforehand. It gives us good clarity. Fortunately, we were in position to help you understand this technical analysis of the primary and secondary endpoints and the powering of the study. All very important and all very good questions, so we appreciate your time and your patience of listening to us and look forward to catching up with you very soon. Thank you very soon.

Thank you.

Ladies and gentlemen, Delta Sunkar conference for today, thank you for your participation. You may now disconnect.