GSK plc (GSK) Earnings Call Transcript & Summary

Good morning, and welcome once again. We are here with the top management of GlaxoSmithKline at Cowen's 42nd Annual Healthcare Conference. We're very, very pleased to have with us Dr. Hal Barron, who is Chief Scientific Officer and President of Research and Development. Dr. Barron has shown over the last 30 or so years that he's clearly one of the industry's most prolific drug development experts. So it's always great to have conversations with him.

So Hal, thanks again for your time this morning. I'd like to start out and ask, what have -- do you believe, have been your greatest accomplishments in R&D since joining GSK a little over 4 years ago?

Yes. Thanks, Steve. Thanks for having me. I think in R&D terms, 4.5 years, that's still a small number of years to really have a huge impact. The things that I'm most proud of when I joined was all about trying to improve, optimize the pipeline. And if you look across companies, it's really hard to determine exactly what metrics one should follow to assess success in this. The one I tend to use I'll describe in a second and set out to optimize on these parameters. So in, let's say, how best to say this, in August or July, when we committed to The Street of a top line 5% CAGR and a bottom line CAGR of at least 10%, which we think will put us in the top quartile in terms of growth, one way to assess R&D's productivity is to ask what percentage of that growth is coming from recent approvals and adjusted value that's ascribed to the pipeline. And over the last 4.5 -- a little less than 4.5 years, there's been 13 medicines or vaccines that we've got approved. Those medicines and vaccines will likely account for around, if not a little higher than, 60% of that CAGR. And that is one strong number. The second, I think, equally strong is that over 40% of that CAGR comes from late-stage pipeline, Phase IIb and Phase III. And during the 4.5 years, we've doubled the number of Phase III programs to enable that. When one looks then at that progress as a function of spend, which is, of course, important, and you look at the number of approvals per billion spent, another metric, and compare that to our peer group, which is done by an external group, so I don't -- all I have is our own data relative to 12 other peers, we were in the top quartile in terms of number of approved per -- well, actually the number of approved period, and then the number of approved per billion was also top quartile. And importantly, it's been a bit of a critique. When we looked at the number -- the peak year revenue anticipated by The Street at the time of approval of the median peak year revenue for those 13, we were also in the top quartile. And I showed that data at the investor meeting in July, and I think those numbers, to a large extent, reflect the progress. The other thing I'd say is we've made a lot of progress on business development, which was a focus. And some of those approvals, in fact, were a result of that. We probably doubled the amount of business development activity we've conducted over the prior decade per year. We, as I mentioned, doubled the number of Phase III assets but also took a very unique strategy in research where we really refocused the entire research organization on 2 key areas and significantly reduced the number of areas in focus on research. So there was 30 or 40 different areas people are pursuing. We really try to simplify it down to 2, really a deep focus on immunology and a deep focus on human genetics, functional genomics and machine learning. And that, of course, takes a longer time for that sort of research strategy to manifest. But through business development deals like 23andMe and FinnGen Open Targets, UK Biobank, et cetera, we've now had, moving into the clinic, validated -- genetically validated targets. Late-stage research now has, for just 23andMe alone, over 40 projects that have resulted from interrogation of human genetic data sets. With the functional genomics expertise that we've developed internally as well as through collaborations with Jennifer Doudna and Jonathan Weissman in an organization called the Laboratory for Genomics Research as well as with the Broad where we've now been using a lot of different techniques to understand how a variant that you can see in the genetics actually translates into a functional impact and, therefore, be able to understand and interrogate the pathways responsible for that. And I think -- and that's resulted in a significant number of late-stage research targets. And I would say that we predominantly focused that on synthetic lethality where we have one in the clinic, MAT2A inhibitor with IDEAYA, 2 really very promising targets behind that, POLQ and Werner Helicase inhibitors that we think we have a handle on as to where they would work synergistically as a synthetic lethal and probably 10 other targets that are a little bit further behind that will come from all this effort. So when you think about pipeline, both late-stage, mid-stage and early-stage as well as late-stage research, I think we've made a lot of progress. And lastly, I'm very proud of the changes we've made in terms of talent. Over 2/3 of the VPs in -- are new to role in GSK since I joined. We have got some incredible talent. Chris Corsico is now running a very large development group. He's an outstanding drug developer and is really optimizing that function. Phil Dormitzer, we hired from Pfizer to run our Vaccines group. He's outstanding. Tony Wood, of course, we hired to run our technology, and he's taking over for me. Hesham Abdullah is running our Oncology group now, from the development perspective, really outstanding. Kevin Sin, we hired to run the business development, and I mentioned the progress. There's plenty other really outstanding talent as well that I could point to. And all of that talent, combined with a cultural transition to really following the science, really being proud of good decisions, some of which are wrong, but they're smart risks, and some hopefully good decisions that really result in this growth, but not fearing failure is something that we tried really hard to insert. As well as this -- a culture of what we call single accountable decision maker, where we've set up governance models where individuals, not groups of people where you can hide behind decisions or where consensus sort of evolves. We've got a strong focus on individuals throughout the organization to make the key decisions and listen carefully and get input. But that was a bit of a shift in culture as well. So I think it's a combination of all those things. Probably, the most important is the pipeline progress.

Great. And you mentioned Tony Wood. What changes do you think he will bring to GSK research?

Well, Tony is an excellent scientist. He's been -- he's an excellent scientist. He's an excellent leader. He's an excellent manager. He's a very good person, too. He was really instrumental with me in building the technology piece of this strategy. And I think functional genomics, the machine learning component were built by him through hires, Chris Miller in functional genomics and Kim Branson, who is really a leading machine learning scientist. And through building these groups, I do think that that's given us this competitive advantage in research, which will translate into the future. And I think that Tony's approach will be to continue this journey of using human genetics, functional genomics and machine learning but take it to the next level. We've been focusing it almost entirely on discovery. And there's a lot of opportunity beyond discovery that I think Tony is excited about. We've talked a lot about whether this could be someday applied to the clinic. And in particular, in the clinic, there's other opportunities to do deep phenotyping of patients using image-based deconstruction. There's a massive amount of high-dimensional space included in images. And as data sets get larger and larger, they start to approximate the size needed for machine learning to deconstruct the semantic representation behind all these different features. So I would think it's going to be a double click on machine learning, functional genomics and human genetics 2.0 as well as continuing to explore the intersection between that and immunology and how immunology can have so many incredible impacts on not just autoimmune disease, which has been well characterized, but of course, oncology, where we've doubled down on immunology in the CD226 axis, get example of work so highly overlapping between human genetics and immunology. And I think that's an area that Tony will want to pursue as well.

Okay. Why don't we move to one of the most exciting readouts that GSK will have in 2022, that being otilimab? And I have to admit, I was quite struck by your level of confidence and enthusiasm on the Q4 call. It seems justified. It's just -- it's a new target and one that is unclear. So -- and I think at that point, you said that the study was basically over. You were following patients. That was a little over a month ago. How close are we now to that final readout? And is your confidence still quite high?

Yes. Thanks, Steve. I think I would just back up a little bit and let me explain my enthusiasm on the program. And I'll explain what I even mean by the program in a second. But when you think about rheumatoid arthritis, I think the number is something like 1%, I think, of people over 17 will develop rheumatoid arthritis. And the current treatment is although completely transformed from 15 years ago, and to watch that and participate to that to some extent at Genentech, it's been amazing to see where we've moved in the field. But despite all these terrific medicines, we're still at a place where, probably, I don't know, it's probably over half of the patients will continue. I think it's about 40% of patients will still continue to have daily pain. The vast majority, of course, not remission. And these patients are in need of better treatments, treatments in particular that will address this pain signal because that's oftentimes missed in some of these endpoints where you're looking at cytokine levels, you're looking at bone remodeling and disease modification, which are all important. But we've kind of lost sight, I think, a little bit of the opportunity to help with pain. When the Phase IIb data unblinded, which was a little bit before I arrived or right around the time I arrived, it was -- importantly, it is -- it was negative for primary endpoint. So the Phase IIb study had a very high bar endpoint disease remission, I think. And on a simplistic view, your Phase IIb data means you don't move forward. What was intriguing though is the endpoint was a very high bar. And I think this -- the philosophy at the time was that RA is extremely crowded. There's not a huge unmet medical need left. If we're going to move forward with RA, maybe we should only move forward if it meets this enormously high bar. And for the reasons I just described, it seemed like maybe there was an opportunity to think about subgroups of patients, those that continue to experience pain. So when we looked at the data, there was a number of endpoints that were positive. It's clearly biologically active. And the signal, again, always dangerous to be interpreting retrospectively, but we were able to see that there looked like a trend towards a greater symptom improvement than we would have expected for the traditional endpoints like ACR 20, et cetera. So with -- in addition, another point I should mention was, when we looked at the data, it came very clear partly because of some excellent PK modeling that was done by the team, that we're still not sure exactly why this was the case, but it looked like the dosing was -- that the patients have been slightly underdosed. And the decision to move to Phase III was based on 3 things. The first was -- and we really needed to be convinced of all 3 of these to feel like it was a smart risk to take. The first was, would the FDA agree with our PK modeling and allow us to go to Phase III with a higher dose number studied in Phase II? That's usually not the case. But we felt that for us to have the best possible shot, we'd need to convince them that that's a smart thing to do, and they worked. The FDA found the modeling very appropriate and allowed moving to Phase III with a higher dose. The second thing is we needed to understand this pain signal a little better because if a negative Phase IIb with a trend in a nonspecified endpoint is, I can tell you from 30 years, you pointed out 30 years of doing this, that's a dangerous place to be. But what we did was look very carefully preclinically at why this might have done so and looked to see what different proteins, GM-CSF, when it activates a monocyte leads to what proteins and whether any of those proteins might be involved in pain. And that most overexpressed protein when its monocytes are stimulated by GM-CSF is CCL17, and we looked preclinically at whether CCL17 inhibition would reduce pain in sort of an osteoporotic model, and it did. And so that gave us more confidence that, that signal might have some relevance because of GM-CSF's ability to reduce CCL17. Now in fairness, when we looked at CCL17 reduction, it wasn't reduced anywhere near the magnitude that we expect for antibodies, and that was -- gave us some pause. But directionally, it definitely lowered it. So we have put together this program that is really going to be interpreted at the -- in the second half of this year where we'll put all the data together to determine whether this is likely to be a novel mechanism with, hopefully, attributes on pain. But I do want to also highlight that this is part of a program. And we put CCL17 antibody in -- we rushed quickly to find an antibody to CCL17 and get it in the clinic. And I think part of the story is going to also be -- so part of this program is really going to be looking at CCL17's impact on pain. If all this biology is correct, the pain signal might be more prominent with CCL17 inhibition, where you can get 100% or close to it reduction. And so that's why I look at this as an overall program rather than any one study. But again, I think that there's such an unmet need and, particularly, in pain because if we were able to uncover a mechanism, maybe an important mechanism of sort of inflammatory pain where the treatment might be an analgesic that's not addictive, this would be quite transformative. So it's a high risk that -- but we thought given that the Phase III costs were what would be assessed as to whether that was a smart risk, we thought this was reasonable thing to do given we get up the dose, given that we're going to advance the CCL17 antibody into the clinic. And I should say Phase I data for that CCL17 antibody should be available around the same time lines as the overall program for otilimab. So again, putting this all together, we'll have to wait and see, but it could be a potential opportunity for RA patients with a lot of pain.

Great. And let me ask one more question, then I'm going to turn it to Mike to ask some additional questions. But this is on the RSV maternal vaccine program. So I realize it's just been, I think, maybe 1 week, 1.5 weeks since the most recent news on the trial stoppage. As GSK does its analysis of the likely causes of these preterm births, what is kind of rising to the top as the most likely culprit? And what might be less likely, if that's a question which can be answered based on available data?

Yes. Thanks, Steve. It's disappointing when any trial was stopped due to lack of efficacy but particularly so when there's a safety signal for such a significant amount of medical need. We all -- we had an IDMC in place that gave us the recommendations. And we quickly halted the trial based on it, confirming that the signal was real. But we're still puzzled as to exactly why this occurred. And we're clearly going to have to do further analytics to better understand the safety data and why the signal occurred. It's important for a comment, and I just discussed a few minutes ago to highlight that this was not something that we're expecting to see in RSV OA given the different population, obviously. Those are 60-plus-year-old patients and rarely, if ever, are even considering getting pregnant. But also importantly, that this is not due to some speculation that this might have been due to the adjuvant, which is not the case, and maternal RSV vaccine was just a protein. And so again, surprising, not something we expected, and additional analyses are going to have to be undertaken to see -- to confirm the signal and assess risk factors, maybe therefore pointing to some biologic rationale. But right now -- but we just don't have that kind of data to be able to speculate with any kind of confidence.

Okay. Great. And I'll turn it to Mike.

Thanks, Steve. Thanks, Hal. I'm wondering if we could turn to some of the oncology programs you have ongoing and maybe we'll start with BLENREP. There's now more than one BCMA-targeted cell therapy approved. During BLENREP's launch, how has physician decision-making come down around BLENREP versus a BCMA CAR-T? And how might that change in the next several months?

Well, thanks, Mike. BLENREP is a really exciting medicine, and it really is emblematic of probably one of the most important targets that hasn't been fully exploited to help patients with myeloma, and that is BCMA. And as you point out, there's a number of different ways one could target BCMA. There's what we have is an antibody-drug conjugate. One could have imagined a monoclonal without the drug conjugate because BCMA not only signals but because you can optimize the FC receptor, which is what we do with BLENREP, you can induce ADCC and have some cell killing. We've decided that we would have 3 mechanisms, not only the signaling inhibition through the monoclonal antibody-binding BCMA and the A2 cosylation of the FC to induce ADCC but also to tag it with the antibody-drug conjugate warhead, which interestingly, preclinically induces an immunogenic cell death, which we think might actually be in part why this is so active. If you look at the DREAMM-2 data, the data that resulted in approval of 12-0, both from the ODAC, it's very, very active. We're seeing above 30% responses in even the most refractory population, including even many of the patients with [ data ] refractory. And what's nice about BLENREP is it's easy to give. It's off the shelf. It's a very quick infusion. And relative to some of the other opportunities to target BCMA, say, with CAR-Ts where you plasma freeze and then the whole process by which you reinfuse the cells, that's another approach. It's probably going to be more limited to centers that can plasma freeze and maybe academic centers that have experience with that. And then there's the BiTEs, which are BCMA CD3-targeted approach. When we look at the landscape in terms of what clinicians are looking for, it really depends on the stage of the patient. The therapies are so effective these days that you really are having a different clinical scenario when you're in the fourth line versus the third, second, first. In the fourth line and beyond, you're looking for anything that might work, and this is a very potent drug. As you move into -- and then we think that it's a very, very significant opportunity for patients in the setting and clinicians are very excited about its use there. When you move into the second and third line, the safety signal that we see with BLENREP, which is this ocular toxicity that is basically these corneal deposits of basically keratopathy, which resolves on its own usually with turnover of the corneal epithelium, but reducing the risk of that is important in the second, third line. And we've reduced the dose because in second and third line, you're going to be giving them with [ single care ] pomalidomide or Darzalex or other second-line drugs. And by reducing the dose and, to some extent, reducing -- changing the schedule from Q3, which is what it was given Q3 weeks, given in the single-agent most refractory population to exploring Q4, Q6, even Q8, which has been done with a couple of studies now, we're finding that efficacy is reasonably well maintained, and ocular toxicity seems to be coming down. In addition, we're also monitoring the patients and holding doses differently in the fourth line, where you're, again, looking for efficacy. We've -- we held the dose after observing a grade 3 ocular tox. As you move more proximally in the second and third line, holding the dose at a grade 2 ocular tox can obviously prevent grade 3, which is what the clinicians in this stage are looking to do from, say, the 60-plus percent that you see in the most refractory to somewhere around 25% to 35% in the second and third level, which would make them very comfortable. And then the last lever we have, which is interesting from a biology perspective and really borne out in the functional genomic assays that have been conducted, is using a gamma secretase. Gamma secretase inhibitors can prevent the clipping of BCMA off the plasma membrane and, therefore, expose the -- actually increase the density, if you will, of the BCMA protein to enable more binding and, more importantly, more toxin to get into the plasma cell at an effectively lower dose. So we're studying using gamma secretase inhibitor from SpringWorks. We're using that drug in combination with the dose of BLENREP that we are pretty confident has no activity whatsoever, 0.9 mgs per kgs, about a little less than -- a little more than 1/3 the dose of what was used in the pivotal trials, 2.5. And again, expecting no response from that alone, we're looking to see if we can potentially use 1/3 the dose and get -- maintain maybe the efficacy that was seen with that higher dose. And if we can do that through this combination of gamma secretase, combinations with other drugs like pomalidomide and VELCADE, Darzalex, et cetera, as well as this dosing frequency, Q3 moving to Q4, 6 or 8 and, importantly, changing the holding patterns, we think we can get this ocular tox down to a very reasonable rate where this off-the-shelf option will be very effective and safe enough for clinicians to use in early lines. Now as we think about the front line, the profile even needs to be further optimized because safety in the front line is the most important component. Activity is important. But given how long these patients are going to live and how well they're going to do, you're really looking for something that has very limited ocular tox. And so for us to move into the front line, which we're expecting to be able to do, all sort of 4 of these modules, if you will, need to be combined. And they each need to be additive, if not synergistic. So gamma secretase needs to work. We need to have the dose optimization through the schedule, lowering the dose just because of concomitants and the other aspects that I mentioned. So if we can put all 4 of those levers together, which we think we can, each individually contributing, we think we can maintain the significant activity to drug test and get it to be safer. And if that's the case, this will be, I think, a leader in the field of BCMA. But the different constraints, efficacy being the most important and fourth and beyond and front line being safety and second and third line in between. But we think we have a good shot with the data that will be emerging.

All right. I'd like to ask too about the anti-TIGIT program you have going at GSK. I suppose there's an agent that was recently in-licensed. Some of your competitors have moved relatively quickly into broad Phase III programs with their respective anti-TIGIT molecules. To some extent, TIGIT is validated, but the jury is still out to some extent as well. How do you think about TIGIT as a target, the molecule you're developing and the pace at which you might expand or...

Yes. It's a great question. And validated mechanisms in immuno-oncology, one could even argue, is an oxymoron until you see the completed Phase III data because of the sort of massive complexity in it and the clear challenge in developing IO drugs. That said, when they do work, they're, as you know, completely transformational for patients in a way that can change the trajectory of a company in a pretty profound way as you can see with pembro and Merck. We liked the CD226 access for 2 separate reasons. One, when we started interacting with 23andMe, we were able to show that, that pathway is genetically validated for oncology. And that gave us a lot of confidence based on our belief that genetics tell you a lot about the future of drug development. In addition and in parallel, before we knew that from 23andMe, we have found preclinically that the 226 axis was very compelling. So based on that, we decided to go to take a deep dive into 226 and advance both our CD96, which is our first and best-in-class CD antibody, to part of that pathway but predicted actually that the Genentech TIGIT program would be successful, and we're acting as if that was the case. Now as you said, there were some challenges in interpreting that data. But my interpretation is that it's very active, and the Phase III repeat study will probably regress a little bit to the mean but be very active. I'm not as confident in the small cell study. It was a little bit confusing by that. But the non-small cell study, I think, will be very positive and possibly another opportunity to be in basically IO version 2.0 with the TIGIT PD-1. And having -- GSK having missed the whole PD-1 sort of era, if you will, having what might be one of the best-in-class PD-1 inhibitors, Jemperli, with what we think is one of the best in class if they are differentiated TIGIT antibodies where you can see actually pretty nicely in the clinic a depletion of the T-reg cells that not sure every antibody against TIGIT does. And so we're expecting that, that combination will be broadly active. We're pursuing multiple indications. There's always a misconception that if you're not first, you're behind by a lot. But I think there's first in an indication. And pursuing additional indications where you could be first or second has proven to be very, very successful for many companies who have come to the party like with PD-1. And so we think that this combination of Jemperli and our TIGIT will be very successful. Where we really get excited, and this is, of course, much higher risk and more speculative, but there's the possibility that clamping down on the entire 226 axis will be even more effective, and that might even usher in IO 3.0, if you will. And the CD96 inhibitor or antibody CD96, plus the antibody we in-licensed from Surface Oncology, the PVRIG antibody, have an opportunity to be triplets, which if the doublet works, and again, I don't want to get too far ahead of myself because CD -- SKYSCRAPER study hasn't went out. But assuming that's positive and we're launching into 2.0, our vision is that maybe we could become the triplet that is superior to the doublet and a very high bar but not uncommon in medicine. So we're clamping down on the entire access. If you could do it safely, it's more effective than clamping down a part of it, particularly immunology where there's system oftentimes works around you. So we're excited to have probably the most robust component of 4 different drugs and the only CD96, I think, that's in the clinic. So we should be getting -- we should be starting our Phase III program in one with TIGIT this year ideally. And we'll be having combination data with the triplet, hopefully, later in the year as well.

We actually have an e-mail question. So for BLENREP and neuro combo data, what is a win scenario in terms of efficacy, tolerability balance?

That's a great question. And I think I tried to highlight that it sort of depends on the stage. I think for -- I think a win for me would be that you see efficacy at least around what we would expect for the 2.5 mg when you give something that's inactive. In other words, the 0.9 mg per kg dose would have, in a DREAMM-2-like setting, making sure the patients are the same to have, I don't know, 31% is what we saw in the pivotal trial. So somewhere around 25% to 35% response rate. And ideally, that lower dose translates into lower ocular tox in 65. Maybe half that would be ideal for second and third line. Because one thinks about that data -- if then that combined with the idea of stopping, holding the dose, I should say, at grade 2 in subsequent trials and by going to Q4, maybe even Q6, possibly Q8, with this BMCP test, where we don't even have studies looking at that yet, so just to be clear, then that would enable a frontline approach as well. So second and third line, very viable. I think if you get a similar efficacy and a lower ocular tox, than 65 -- I'm making this up. Half of that is probably good. Our market research kind of suggests that somewhere around 35%, 30%, 35% is something clinicians would be comfortable with should there be positive data in DREAMM-7 and DREAMM-8, and that's, of course, adopting Gamma secretase inhibitor. So that profile is why I come up with that. But even lower, of course, for front line, I think you got to get below 15% in ocular tox, maybe below 20%. But we won't have data for that for a while. But when we'll be able to impute the likelihood of that based on a number of trials that will show the impact of each of these 4 modules, if you want to think of them, 4 strategies to reduce tox relative to efficacy.

And we recently noticed 2 DREAMM trials. The primary completions were pushed out not insignificantly. Is there any systemic issue or any reason that we should be concerned about that?

Well, when a trial gets pushed out, there's usually 2 or 3 reasons why that can happen. Some of it's just enrollment rates can be slower than expected, and that's not ideal. And I think a part of those, to be honest, is slower enrollment but also event rates. And when event rates are lower than expected, one can argue that's good news. So without getting into too much, it's a little of both. And they're not pushed out that much. In fact, DREAMM -- I'm forgetting now. I think DREAMM-7 actually enrolled ahead of schedule. DREAMM-8, I think, is a little bit behind. Then, I think, COVID was particularly challenging to enroll patients. Some of the geopolitical things made it a little bit more challenging. But they're actually -- DREAMM-8 is actually back enrolling faster than expected now. So I think we're -- and of course, the bar is very high. These are superior to the Darzalex and VELCADE studies. So I think the most important thing is to get the data, and we should have that soon.

Okay. And we are over time. But allow me one last question, and we're asking this in every session. But over the next decade, what do you think the biggest surprise will be in GSK research?

Well, I'm hugely biased, but I strongly believe that the focus that we've taken 4 years ago, 4.5 years ago on human genetics, functional genomics and machine learning to find novel targets that have a higher probability of success, it's going to take a decade for an idea like that to mature through the pipeline. But I do think that will work. I have extreme confidence that there'll be many examples of synthetic lethality, even synthetic rescue, the opposite side for onc and the identification of numerous, numerous targets that will emerge from data set that you can only see when they become multimillion patient registries. And when they finally get through the system and approved, I think people are going to realize the probability of success is at least doubled. And the impact of the R&D model when you can have twice the probability of success by focusing on this will transform how we think about it. And then likely, if all that's true, what transforms the industry is to think about machine learning and the models created by all this data has been almost the new language of biology. And should we get to a place for the machine learning models that are generated through massive, massive amounts of data start to drive your discovery engine, it's a realization, I think, that, to a large extent, biology and medicine has become a data science.

Sounds like a very exciting future for GSK research, and we look forward to it. So Hal, thank you so much for your time today. It's always a pleasure. Thank you.

Thanks, Michael. Thanks, Steve.