Eli Lilly and Company (LLY) Earnings Call Transcript & Summary

Thank you guys for joining us. Pleasure to have Dan from Lilly management, an R&D organization, join us. I feel like it's become a bit of an annual tradition and maybe Dan's the only one to raise his hand to agreeing to do this every year, but I apologize in advance, Dan.

Thanks for that. I didn't know I [ cursed you ], Umer. I thought this was a part of my job.

Fantastic. Well, I feel like we should start today with a topic that you guys never get questions on, which is Alzheimer's. And maybe perhaps -- actually, even before that, Dan, let me turn it over to you first. Maybe lay out the top R&D priorities from your perspective. I want to ask that in part because from an external perspective, for whatever reason, it's become this GLP in Alzheimer's and nothing else matters kind of a perception, even though I feel like there's been a very well-diversified growth story among launch products and a very bold diversified pipeline. But I'll let you kick it off.

Yes. Well, I take that head on. I think the 2 big trends that probably define the next couple of decades in the pharmaceutical business are likely to involve the treatment of cardiometabolic disease, diabetes, obesity, that whole spectrum of disease, which now we think we have a handle on. Lilly can be a leader there with GLPs in our next-generation incretin technologies here. And the second theme is probably the ability to treat neurodegenerative diseases, which is still earlier. We don't have as many proof points there. But halting Alzheimer's is huge. So I understand the focus there. But I always sort of think, if we didn't have either of those areas, we'd still have a great portfolio. We'd be spending a lot more time talking about drugs like mirikizumab for inflammatory bowel disease, which I think is a big opportunity; lebrikizumab for atopic derm. Both of those are Phase III. They've had positive Phase III readouts. We have a pipeline of Phase II molecules behind them in immunology, some of which we'll be able to highlight next month at our R&D Day. In oncology, we've just, over the last 3 years, completely rebooted our oncology R&D strategy. We've broadened Loxo Oncology, which now represents our entire oncology effort, leads our entire oncology effort. We've highly differentiated Verzenio, launching the adjuvant opportunity now and behind it, prostate for Verzenio. Lots of opportunities for growth with Verzenio. And then pirtobrutinib, which I think would be the star of any company's portfolio. If we didn't have donanemab and tirzepatide, probably we would be getting exclusively pirtobrutinib questions, which is a big opportunity. And we started to talk about some of the molecules behind that in oncology, which things like our PI3-kinase mutant-selective inhibitor, which could be another mega-sized oncology opportunity. So it's really exciting to be at Lilly at this time when all 4 of our therapeutic areas, oncology, immunology, diabetes, neuroscience, are finally working. I think we've had times when we had output from 1 or 2 of them, but never before have we seen all 4 in good shape and delivering important innovations for patients.

Outstanding. Okay. So maybe let's jump right into it. I know there's so much focus on Alzheimer's. I know specifically perhaps, and I'm going to drill down into some specific aspects of the data we've seen. When I look at the iADRS endpoint or the CDR Sum of Boxes at endpoint, when looked at on a mixed-model approach, there was this period of week 12 to week 36 that was remarkably flat on the active arm, which wasn't quite the case on some of the other stats approaches you guys showed. And I was very confused why that was. I'd almost never seen something like that.

Yes. Your super smart question here. So MMRM, which is...

Mike wrote it, by the way.

Okay. It's a good observation. It's a mixed model for repeated measures that is commonly used as a primary endpoint -- primary statistical method, rather. And if you look at, for example, iADRS, it's not even just flat. The treated group actually like slightly improves over those early time points. CDR Sum of Boxes is essentially flat on drug and declining on placebo. So one thing that happened is Lilly has been arguing for a while that MMRM is not the best way to look at data. We should probably use a Bayesian approach like disease progression model. But in our discussions with the FDA, we were asked to look at different methods as well as just look at the raw data, so not the model data but the real data we observed to see which model best fits the observed data. And lo and behold, it was not MMRM. And that initial blip where it actually gets a bit better on drug for iADRS or stays flat for CDR Some of Boxes is an artifact of the MMRM model. So I think that's unlikely to be real. And we have that slide that we showed at a scientific meeting where we looked -- actually, I showed it, I think, on the Q2 call as well. We looked at the cubic spline and quadratic mix model as well as MMRM. I think those other models in DPM reflect the observed data more accurately. So I don't expect that anomaly in the early time points of MMRM to reproduce it in a bigger trial. And I don't think we should put too much validity on that very rapid effect.

Okay. Got it. So -- but for Phase III, is it MMRM as the primary endpoint?

So we focused on disease progression model, although that's still a controversial statement. So I think these things have to be worked out in some detail with regulators. But certainly, the data that we have from the Phase II trial and our explanation of different statistical methods would give one pause to the assumption that MMRM will always best reflect the true data here. I think there's room for other analysis methods, which could include disease progression model, which is a Bayesian method or for people who want to be a frequentist, probably a cubic spline or quadratic model is better than MMRM for a disease like Alzheimer's.

And from the FDA reviews I've seen, it looks like MMRM ends up being the default in a lot of the primary stat analyses?

Yes.

So has FDA signed off on the DPM? Or is that something you guys are hoping to get signed off?

Yes. Yes, maybe I don't get into too much detail and certainly don't put words in the FDA's mouth here. But I think your observation is right that MMRM is almost always, maybe always, the FDA preferred primary endpoint for clinical trials. Sponsors can take risk with other methods, and then you get into the question of what happens if you hit on your prespecified [ statistical ] method but miss on MMRM? That's an interesting situation. If you're in that situation, I personally would rather have specified the one that hit as my primary and argue why that's better than have missed on my primary analysis and say, " Wait, before you throw out this drug, look at this other analysis." That's never a good position to be in, although it's worked for some.

Got it. Okay. Got it. And then also the Phase III design change, the 500-patient changed to 1,800 and now iADRS is the endpoint. Can you maybe elaborate on that? It sounds like there's been some level of regulatory discussion on how the Phase III is being conducted.

Yes. There, of course, has been regulatory discussion. But I point out that when we designed this trial initially, we designed it as a Phase II trial. And we said at that time, like, we don't have the TRAILBLAZER-1 results. And if TRAILBLAZER-1 looks good, we'll need to expand this trial, but we want to just get it started so that we're ahead of schedule. So we started as a Phase II with 500 patients, knowing that there was almost no chance it stays as a 500-patient trial. Either we would stop the trial if TRAILBLAZER-1 failed or expand it to Phase III if TRAILBLAZER-1 looked good. So it was a bit of a placeholder. So we expanded it to 1,500 to make it an adequately sized Phase III trial as soon as we got the TRAILBLAZER-1 data. Remarkably, we achieved that 1,500-patient enrollment. At the same time, we didn't -- it didn't cost us any extra time. We were able to accelerate enrollment fast enough that we enrolled it just as quickly as we were going to enroll 500 patients. And then in fact, we overenrolled it beyond the 1,500 because there's so much enthusiasm for this trial in the last few days of enrollment, which maybe cost us a few days on data -- on last patient [ entrant into the ] trial. But in the end, it will be closer to 1,700 or 1,800 patients in this trial. So that's how we got to sample size. And then so it's overpowered for its primary endpoint design.

Got it. And then also, Dan, the inclusion criteria around tau, where you guys basically exclude the high and the low ends of the bell curve, if I may, has that ever been applied to an aducanumab dataset to see if efficacy signal could look more prominent?

No. Not to my knowledge. I don't think they have the tau scans to be able to do that, unfortunately. I think the very low tau patients that we excluded at the left end of the curve, those patients don't progress anyways. So like the placebo group will be flat and so I'm not exactly sure what you'd hope for in the drug group. So excluding those, it's easy to think about. The high tau patients that we exclude, and that's about 1/3 of the population, they do progress. I just worry that the drug won't have much effect on them. So in TRAILBLAZER-2, we're actually letting them in the trial. They're just not in the main cohort. So just the 1,500 patients that we intended would have been 1,000 patients who have low to medium tau, plus 500 of high tau for a separate analysis. And the primary analysis is really on the low, intermediate tau, although we put some alpha on the whole population as well.

Got it. Okay. Mike, anything on Alzheimer's you want to touch up on?

Just quickly, Dan, the head-to-head TRAILBLAZER trial versus aducanumab, it's -- just curious that -- it's a 6-month trial in which by 6 months, Aduhelm is only dosed at 6 milligrams per kilogram versus donanemab. There'll be 700 milligrams for the first 3 months, I believe, and then 1,400 milligrams after that. So just trying to -- how should we think about that -- the fact that adu won't be at target dose at 6 months?

Yes. Probably we should think that adu works more slowly. So as a result of titration and other factors, you're probably getting less efficacy early on from aducanumab. So the trial is -- will run long longer than 6 months, but 6 months will be the first primary endpoint. Look, I think if you believe that Alzheimer's drugs work because they lower plaques, then lowering more plaque faster is probably the way to go. It's always been the case with other surrogate endpoints, oncology response rate or whatever, that if one drug can beat another at the surrogate endpoint and the surrogate endpoint is correct, that translates to better efficacy. So that's -- the focus here is on the surrogate. I think speed does matter, but we'll also have 18-month data as well on plaque-lowering, and I expect that we have a good opportunity to be superior there as well.

Got it. And then Dan, I have to admit, I love all the work you guys show, but there's one piece of work you guys showed where I tended to disagree a little bit, which was the -- let me get -- the Conrado model. I think you know what I'm talking about, which was the -- remember how the New England Journal paper said, there's -- on a patient level, there's no relationship between A-beta lowering and efficacy, but then this Conrado model came out and it said there is. What I really noticed was you can't establish causality using a Conrado model because it's non-linear. I guess how -- to what extent can we or can we not interpret that?

Yes. I think you're right. We can't prove causality at all. But I think the perplexing thing is why in the New England Journal article, in our first analysis, there was not a correlation. And I was a bit surprised. But then as we dug into the data subsequent to the publication and you'll note how quickly we published that after we got the data, something emerged. And maybe I do a quick analogy. So we were looking at delta amyloid versus delta cognition, so delta-delta change in change -- change-to-change correlation in that paper. But imagine if you had a therapy that lowered tumor burden, maybe it's radiation therapy, focused radiation, and you measured how much the tumor shrank, and then you correlated that with outcomes, you wouldn't find the people who had really massive tumors, but they shrunk to still be big tumors, did better than people who had really small tumors and just shrunk away to nothing. You would see the opposite. And so that's exactly what happened here is that it turned out that the people who we got rid of the most amyloid, who had the biggest delta amyloid, were the ones who had the worst amyloid to start with. And so even after you remove their amyloid to the greatest extent, you were still better off being a low amyloid patient at the beginning who then became a 0 amyloid patient at the end. So delta to delta wasn't as good. The second thing that we noticed is that using the delta over 18 months for plaque lowering isn't that helpful. Almost everyone got to really, really clear levels at 18 months. So again, at 18 months, you're basically just measuring how much amyloid they had at the beginning in terms of the clearance. And there's no opportunity for that level of plaque to affect cognitive endpoints. So we looked at, well, how much amyloid we got rid of at 6 months? Did that predict subsequent decline? And that did, and we subsequently showed that data. So it's more complicated than the simple delta-to-delta correlation. The Conrado model is definitely not perfect and doesn't tell us everything. But I think the way to the data is [ balancing ] that suggests that early and deep removal amyloid is more helpful than a little bit of [ clearance there ].

Got it. Okay. Makes sense. And maybe one last on Alzheimer's. I'm going to move on from this topic, Dan. I feel like nobody is talking about -- just like nobody was talking about Alzheimer's a year ago, currently, nobody is talking about the A4 trial for solanezumab. Can you remind us what's the median dose? Is it 400, 800, 1,600? And also, I know it's a trial with older patients at risk for memory loss. So I guess what I struggle to understand is the endpoint that's being used in that trial because it's not the traditional CDR, et cetera. Maybe just anything along those lines.

Yes. So this is quite a large trial of Alzheimer's prevention, secondary prevention, I guess, people who have pathology in the brain, so they're amyloid plaque positive but don't yet have symptoms. And as you know, we started this study with the same dose of solanezumab we used in the Phase III trials. But upon a careful analysis of our Phase III trials, we felt like we underdosed solanezumab. So we increased the dose to 1,600 milligrams. And I think most of the doses administered in this trial will be 1,600 milligrams. So when we get down to trial, I hope we support that dosage if it works. But you can't -- and maybe I also point out that this is a different mechanism than donanemab. This is not a plaque-clearing antibody. Even at these doses of solanezumab, I do not expect plaque clearance. It's finding some soluble species. So if it works, we have to believe that soluble amyloid is also important in disease progression. But because this is patients who don't have symptoms yet, the normal Alzheimer's things that we're used to, like ADAS-Cog or CDR Sum of Boxes or activities of daily living, they don't really make any sense here because these patients won't show progression on those scales. So different scales are used in this population. And in this case, we're using something called the PACC, the Preclinical Alzheimer Composite of Cognition -- or Cognitive Composite. And PACC is a combination of 4 different measures really heavily weighted towards memory testing, sort of subtle changes in memory that are quite sensitive for decline. Even so, I think it takes a couple of years to see change in this population. And that's why it's a long study. And I hope we'll see some decline in the placebo group and less decline in the solanezumab group. Those results will come out, I think, early '23.

Early '23?

So really another important readout [ for the field ].

Got it. Okay. Got it. So it sounds like lower expectations what you're sort of hinting at, just given the plaque clearance? Or it's harder to tell?

Yes, I think it's reasonable to have lower expectations for solanezumab and donanemab, given that solanezumab didn't hit in its Phase III trials and donanemab hit in a Phase II trial when there was a larger effect size. On the other hand, I wouldn't say solanezumab was completely inactive in its Phase III trials. In each of the trials, there were trends in subgroups and a flavor of efficacy here, which if you believe is correct, you might believe that a higher dose would give you more efficacy, then you might also believe going earlier would give you more efficiency. If either one of those is correct or certainly if both of those is correct, this could be a positive trial. At the same time, I think plaque clearance could be a more powerful mechanism. And we have a trial in a similar population now running with donanemab, which we'll try and do a bit faster here and get data in that prevention paradigm for donanemab as well.

Got it. And Dan, just before we move on, not a commercial question but just in talking to physicians and your former colleagues, et cetera, do you get a sense there's level of interest in A-beta antibodies? I feel like judging off of how aducanumab has been doing, and again, there's a whole list of issues that surround that launch, it's not clear if there's even an end user demand that's coming in so hot. It's not like everybody is putting in orders and they're not getting them approved. A few of them are. Most of them are not. And I wonder like -- I wonder what you hear from folks you speak to, like, actual prescribers and Alzheimer's specialists, not forgetting all the commercial.

So a couple of things. I think one is obvious. A lot of skepticism about the FDA decision. And so then people are waiting for another source of authority to come. And the next source of authority here is CMS. So if CMS gives doctors some reasonable guardrails here that need to be tighter than what they updated, but if they can define either in terms of the patient population or prescribers or mechanism of reimbursement, that I think will stimulate use. Still, I think we should have modest expectations for use of amyloid lowering drugs until some of the Phase IIIs read out. And there's 4 Phase IIIs coming in the next 18 months. If they're all flat negative, probably not using these drugs is a wise decision. If there's a trend towards benefit or a meta-analysis of benefit or one of the drugs is successful, then I think that would push us the other way. So right now, we're in this strange period of time. I think it's great to have accelerated approval, but I think it's right to expect modest use of these drugs until we have the full data.

Okay. Excellent. Okay. So let's go to the -- let's switch to oncology, perhaps. And I know you mentioned pirtobrutinib. There's obviously a lot of interest in the BTK space, et cetera. And I guess one of the questions I've had is, why didn't you guys prioritize a head-to-head versus BTK inhibitor in CLL? Why wasn't that the first study?

Yes. Well, I think we have to remember where this drug came from. So first of all, this was a molecule that was designed to take care of a mechanism of resistance to ibrutinib and other covalent BTK inhibitors, the C481S mutation, and this was designed to take care of that. And what we found in ibrutinib study was that not only, yes, do we have responses in that, but surprisingly, we were seeing responses in patients who failed ibrutinib for other reasons. So we don't know why they failed and why their tumors came back, but it probably was inadequate inhibition of BTK because now with a cleaner BTK inhibitor that, yes, takes care of this mutant but also can inhibit BTK more fully without dose-limiting side effects of ibrutinib, we could now get responses in patients who were previous refractory. So already, that data paints a picture of what the key initial opportunity is. It's patients who failed ibrutinib or other BTK inhibitor and now have relapsed/refractory disease. So I think that's a growing unmet medical need, given the number of patients that have been on these drugs. Those patients don't have other great options. So providing a drug to rescue them just makes a lot of sense. And we should do that as quickly as possible. Those patients need us. For the first line, it's a question of offering them something better. There already is a good therapy in the approved BTK inhibitors, but I think we can offer something better. That is going to take a long time to get the readout from because response rates are already high for ibrutinib. We'll start that study next year, and it will take a while to read out. So I don't think it's a question of not being excited about it. It's just going to take longer to do that and get that data. The quicker path is to the second line as well as we have a first-line study that's not head-to-head against ibrutinib, but against CIT, chemoimmunotherapy, which should read out faster.

Dan, do you expect efficacy to differentiate beyond C481S?

Yes.

Okay. Okay. And then also what about afib and the long-term dosing?

Yes. So we now have a lot of patients who've been on this -- on pirtobrutinib for quite some time from the BRUIN study. So one of the things that we've wondered about and people have asked us is some of the BTK side effects that they've seen so far, are they target-dependent side effects that we should see them also? Or are they molecule-dependent side effects, in which case maybe we'll avoid them because of a different method of inhibition. So you have to wait to get the answer to that question. But at ASH, I think we'll have some updated data from BRUIN that should be able to address the side effect profile, including rates of afib.

Got it. Got it. Got it. Okay. And then the first data on your BCL-2, when should we expect that and the differentiation?

Yes. So the BCL-2 inhibitor that we in-licensed is not -- we're not pursuing a differentiation hypothesis. Maybe it will differentiate, which would be great, but that's not the rationale for pursuing it. I think you can think of this more as a life cycle strategy with respect to pirtobrutinib. Clearly, it needs to be competitive with venetoclax, meaning it has similar efficacy and safety. But I think patients are going to -- a lot of patients are going to need to be on a common [ recommendation ] of BTK inhibitor and BCL-2 inhibitor. And so there -- I think owning both parts of that regimen could be important parts of the adoption. So that's how we see it.

Okay. Makes sense. And then I feel like, Dan, some of the emerging commentary on CDK4/6, like, I remember I was speaking to Mike after your last earnings call, and I kind of missed some of those comments around you guys having put out a letter and then I finally found the follow-up to the CDK4/6. The follow-up 2 weeks later, there was this sort of letter where the most important disclosure, the OS curve, was finally posted. And I was very puzzled by sort of the progression of disclosures on that. I was curious if you could just walk us through what happened and why it wasn't that published in the first place? And I have to admit, the first time I got a question on it, I was like, no, the OS has never been shown. They're like, no, no, no, it was published. And then I was like, I thought I read the paper, but I didn't read the paper.

Yes. Okay. So it's kind of an unusual situation. So we said, I think, early on that the FDA was looking at OS and that would be part of their determination here on approval. And I think when we initially said it, we said the overall OS wasn't -- it needed to be lower than one. And so we would take some pause during the review cycle that there [ was some risk ]. I think we said that. When we published the data, journals definitely do not publish, and we usually don't publish either their highly immature OS data. And that's the key point is that this OS data is so, so early, something like 90 events or something. And so because of that, and I think if you look at the OS curves that we finally did publish, it's a bit hard to interpret. I don't get too excited about the 0.77 or whatever it was in the one population, [ I think ] it was 7 or the OS in the overall population because the number of events is so small. Now -- so why did we publish it if these OS curves are uninterpretable because of their immaturity? I think a lot of people, when they saw the publication, they saw a hazard ratio on -- I think we showed IDFS and basic disease-free survival. The hazard ratio for the high Ki67 population and the low Ki67 population was almost the same. We had a similar strong benefit in both groups. So we showed that data. And then the next day or the same day or day before, I don't remember the sequence, the drug was approved just for the high Ki67 population. People said, why is that -- how did the FDA come to this decision, given that the data you published all looks like it works equally well in both groups? And so then we dropped the immature OS data with the caveats that it's hard to interpret this, but this is the reason why the FDA went to the high Ki67. Now what do I expect? Why is it looking good in high Ki67? Is it because IDFS is more likely to translate to OS in that population? No. I can't see any reason for that. It's just that the data are more mature, and you've had more time because it's more aggressive disease for differences in IDFS to translate to differences in OS, whereas in the rest of the population, those take longer. But I'm confident the rest of the population will eventually translate and that will be our opportunity to go back to the FDA.

And then remind us again, Dan, what's the population where it's 0.77?

Sorry, that's the high Ki67.

High Ki67.

Yes, which is the -- Ki67 is a marker for replication of cells. So that's where the more aggressive...

So I guess that would imply, Dan, that Ki67 lows are tracking at a hazard ratio of 1.3 or so. Like how should we interpret that?

Yes. I don't know the math. I'm not sure you can just average hazard ratios like that. And frankly, I haven't seen or paid any attention to these numbers because now you're talking about hazard ratio from 40 events per arm or something. It's hard to get excited about that. I think even when we closed the curve, it made me laugh a little bit because we're looking at lines that are essentially on top of each other and the second -- couple are driving these hazard ratios. So one thing that I look at when I look at hazard ratios are the confidence intervals, and these confidence intervals go from like 0.5 to 1-point-something. I mean, it's hard to get too excited about these until we get more events.

Okay. And maybe my last one on oncology before I turn it to Mike on diabetes. For SERD, I guess, what's the level of excitement for you guys? And also, what type of response rates should we expect in ESR1 mutants versus non-mutants? I feel like you guys were dropping in a very prominent hint.

Yes. I'm not sure I remember what hint we said. But look, I think these agents, they generally have a pretty low ORRs as a single...

Yes, I've noticed.

So I wouldn't read too much into this. And again, it's like another small number problem, when you're trying to compare across the different SERDs. I don't really care what the exact numerical numbers are for -- or for different members of the class or even I think with our agent with different doses, people try to read that. I do think the mutations are important here that, that's the mechanism of -- or one of the important mechanisms of a 0.5 resistance in later lines of therapy to endocrine therapy. And clearly, SERD should be effective. I expect our SERD and others to be effective in ESR1 mutations, whereas a drug like fulvestrant isn't. So that has implications, I think, for the second-line study but I would say is less relevant for earlier lines of therapy like first-line because the patients don't have ESR mutations yet. I think that's important because what do I see as the biggest -- what do we see as the biggest opportunity for the SERD classes? It's got to be in the adjuvant setting. That's where SERDs are going to have the biggest opportunity as a more powerful oral agent. And I don't think ESR1 is -- has much implication there. So let's see the efficacy outside of that mutation. And then based on that, let's see if we can do something in combination with Verzenio in adjuvant therapy. That's the really big opportunity for all patients.

Got it. I mean, I guess I almost wonder, in metastatic setting, is there a chance SERDs end up not showing much at all? Like aren't they tracking [ below where ] fulvestrant was expected to track in terms of single-agent activity?

Yes. I think there is risk in the first-line setting. I agree with that comment, and we haven't [ prioritized anything ] like that. But it will be important to see those readouts, and I don't think we have much longer to wait, I think, to see some actual data at least in the second line from the SERDs.

Got it. But would you agree with my characterization there, Dan, that SERDs are not showing the monotherapy activity that fulvestrant did?

Maybe I'm not close enough with what fulvestrant showed at that point in time, but I do think you're correct in pointing out the risk in the first-line setting for sure.

Got it. Okay. Got it. I mean I see -- and I should go back and look out into this data in more detail, but I see ORRs approaching the 40s, although I will -- I need to confirm this if the FALCON study was on a background or not. Okay, that's fine. We can follow up on that. Mike, do you want to head off into diabetes?

Sure. Thanks, Umer. So yes, pivoting to diabetes, I just kind of want to focus on the tirzepatide Phase III obesity trials. Just kind of wanted to get clarity on the preferred dose in that study. Is it 10 milligrams or 15 milligrams? And just recognizing that although the entire dose range is available, I think, in the study design, but any comments on that would be great.

Yes. Well, we don't have the data yet, but if I had to predict, my guess is 15 milligrams will be the preferred dose for obesity, given a couple of things. So one is even in the type 2 diabetes population, there's a dose response for weight loss that's pretty good. So the 15 shows better weight loss than the 10. The second is I expect that dose response to be greater in obesity than in diabetes. And likewise, I expect efficacy for each dose in obesity. There'll be more weight loss in obesity than in diabetes. That's been the trend in the field for every other drug. So that's what I would expect here. And then finally, obese people are bigger and typically require higher doses of drugs with less side effects and less pharmacology as well. So that, again, would push towards the higher dose. But it's good to test small and let's see what we get and what patients prefer in the marketplace side. You can't underestimate or can't understate my -- or can't overstate. Anyways, I'm extremely excited to see this data. We're in a good position with obesity. The data coming out of our Phase III diabetes trial were beautiful, and we have an incredible weight loss drug, including even starting at the 5-milligram dose, which was superior to 1-milligram sema, 15-milligram dose is going to be great.

All right. That's helpful. And just correct me if I'm wrong. At least one of the trials in the tirzepatide obesity study will occur in obese diabetic patients, correct? And...

Yes. I think we have both people with diabetes and -- obese people both with diabetes and obese people without diabetes and...

Yes. And where I'm going with this is that because could we expect to see the same kind of muted weight loss in that specific population since they do have diabetes on top of their obesity as we see the Novo STEP 2 trial?

Yes, it's a good question. I don't -- I'm not sure. I think some of the reason that we expect to see more weight loss in our obesity trials is because they have higher baseline BMIs. So even among people with type 2 diabetes, if you stratified weight loss based on baseline BMI, people with type 2 diabetes who have a higher baseline BMI should have more weight loss. So I think that still should be okay.

Okay. Okay. Also to sum it, the HFpEF trial that's ongoing, what are the expectations for that? And can the additional mechanism that tirzepatide does offer add any benefit there, in your opinion?

Yes. So maybe. I mean, I think the HFpEF is -- maybe start off by saying, HFpEF is a pretty significant unmet medical need. Excited that we'll have Jardiance in this population, which is, despite Entresto getting label there, Jardiance is the first drug to have a successful HFpEF trial, which is kind of a big deal. Here -- and I think that could have effects directly on the heart or could have -- probably also has other benefits that -- in the kidney that lead to the heart failure benefit. Here, I think it's a bit different. The mechanism of action is likely more related to the body weight changes that we see with tirzepatide. So I think that in itself, we know that the HFpEF patients, when they lose weight, have better outcomes. That's been shown with a number of different trials, including with bariatric surgery, which has much better outcomes than HFpEF. So we're focusing here on people's BMI is greater than or equal to 30. But I think in this population with HFpEF, just the weight loss should be enough to lead to efficacy, and we don't have to worry about benefits in other ways. There could be other real benefits to metabolism or to cardiac health. But if there's nothing except weight loss, that should still be good enough to have a highly successful trial here, exciting drug. It also sort of fits into a general philosophy here, which is that by treating obesity, you can prevent a lot of other consequences of being overweight, things like heart failure, but there's probably a lot of other illnesses that accrue disproportionately to obese individuals that if you reversed it, would be gone. And we've seen that with bariatric surgery studies, including a recent one where I think things like NASH are almost completely [ cleared ] by weight loss. So it feeds into the idea that treating weight or treating obesity will lead to better health outcomes overall.

Got it. That's helpful. Just pivoting to your old GLP, the one that you got from -- you in-licensed from Chugai, any general takeaways from Phase I? Are there any fasted versus fed PK variability issues? And secondly, just your overall take on the competitive landscape of oral GLPs going forward.

Yes. I like where we are with this molecule. We should have a little more to say and show about it next month at our R&D Day. But it's an exciting molecule for us. The molecule overall, I think, can be [ characterized ] as well-behaved in terms of the PK properties. The efficacy profile is yet to be determined or disclosed, but my expectation for oral GLP-1 agonist is sort of Rybelsus-like efficacy. I don't think it's possible to generate tirzepatide-like efficacy from a GLP-1 monotherapy oral or injectable. So that puts a ceiling on what you can expect, given that we'll soon have tirzepatide available, hopefully. This has got to be seen as earlier-line therapy with less benefit. But the goal is to avoid this sort of complicated administration requirements of an oral peptide like Rybelsus, which does require overnight fasting and taking it without food and a certain amount of water. So I think we want to avoid that, and that's why we're focused here on a small molecule. In terms of competition, we're certainly aware of competitive action on the small molecule front, and Pfizer's working in this area too and I like our profile.

Dan, it felt like there was perhaps a fed-fasted PK variability off Phase I or no? Maybe that's just how clinical trials read?

Yes. I don't -- [indiscernible] clinical trial to test that, but my understanding is that's probably a routine kind of pharmacology trial.

And also, Dan, maybe just one quick one on this GLP topic before we move on is, did you guys ever consider running an Alzheimer's study for your GLP programs, [ one of them ] at least?

Yes. Of course. Understand the rationale here from real-world evidence as well as from our own REWIND trial and Novo's cardiovascular outcome trials, where there was somewhat less cognitive decline in the patients who are on GLP-1 therapy. So there probably is a benefit here. I don't know how big is the benefit. And then you can probably also worry about too much weight loss in this population. I mean it's one thing in people who are obese or have diabetes. If they also have Alzheimer's, they should definitely correct their obesity and their diabetes for sure. They should correct their blood pressure and cholesterol and everything. But I don't know if we need a trial for that. I think it's obvious that optimal metabolic health will be helpful even in the context of Alzheimer's disease. The bigger question is people who don't have a metabolic abnormality who are lean and nondiabetic, will GLP-1s have a benefit in that population? I just don't know yet. I think Novo is being bold in testing that. We'll have to continue to think about it. But right now, we don't have trials in that space.

Got it. Okay. Sorry, go ahead, Mike.

No, I was saying, I think we're coming up on time, but would you want to just pivot to the last set of questions on autoimmune?

Yes. Maybe just before we do that, RNAi collaboration, Dan, anything to be aware of, especially now that Dicerna ended up with your competitor? Does that change in any way the RNAi programs you guys were going to work on? Or is there nothing making any good progress...

Yes. I don't know. They say imitation is the surest form of flattery. So we started with Dicerna a few years ago, and I think we captured the key programs that we wanted to pursue in cardiovascular and liver as well as really the, I think, broad opportunity to pursue CNS targets with their technology. We're going to show some of our progress in RNAi in December. I'm excited for what we've been able to do, working with Dicerna as well as internal programs as well as other collaborators, which we've announced in the RNAi space. It's an important space for the industry. Probably every pharma company will want to have an RNAi platform comfortable -- maybe more than comfortable. I'm really excited about what we have in the RNAi space. So stay tuned and wait until you see what we show you next...

You guys -- the programs you guys opted in to, they're locked in. It's not like they can somehow go back to Novo?

That's right.

Okay. Just making sure. All right. And then on the autoimmune side, how do you characterize the lebrikizumab efficacy relative to what we're learning about OX40?

Yes. I mean, I think it's still pretty early days for OX40, Phase II data. It looks, I think, at least that Amgen one probably comparable to dupi. That's fine, I think, interesting and those could be drugs. We need more biologics for atopic derm for sure. If you look at the options there versus psoriasis, there's no comparison. But even so, even with dupi and, hopefully, lebri and maybe OX40s, we still don't have likely the kind of game-changer in atopic derm that we have with Taltz in psoriasis. So I think there's still headroom here for more innovation. We need molecules that can offer a significant fraction of patients, 90%, 95% response rates. We don't see that in atopic derm. I don't know if it's around patient heterogeneity, and it may be true that people who don't benefit from one class could go on to another and get more benefit. So that will be an important thing to learn. And that will start to play out as we have multiple biologics in the space. There was probably a time when you could have thought that the JAK inhibitors would have a profound effect on the market dynamics in atopic derm. I expect that to be somewhat muted, probably given the safety concerns and anticipated labeling around the JAK class. So the importance of biologics is higher. I like what we've got with lebri. We top lined the first Phase III data and just waiting for the first 2 trials, I guess, waiting for one more and then [ move this for distribution ].

Okay. Excellent. And then my last one and I know we're out of time. What percentage of your IL-23 trials are TNF-refractory, Phase III?

Yes. So about -- this is for the ulcerative colitis trial, LUCENT trial. It's about -- 40% are TNF-inadequate response or loss of response.

Got it. Okay. I'm going to leave it right there. Dan, thank you, as always, for being with us. Always love our back-and-forth on these topics. Thank you, as always, and looking forward to being in touch.

Great. Thanks, Umer. Thanks, Mike, for [ the questions ]. Have a great day.