Amgen Inc. (AMGN) Earnings Call Transcript & Summary

I'm the operator today for Amgen's conference call from ASCO. [Operator Instructions] I would now like to introduce Arvind Sood, Vice President of Investor Relations. Mr. Sood, you may now begin.

Thank you, Erica. Good afternoon, everyone. Welcome, and we very much appreciate your participation in our ASCO event. We have presented important data at this conference and very much looking forward to the discussion. Presenting from Amgen will be our Executive Vice President of Research and Development, Dr. David Reese, who will briefly review our oncology portfolio strategy, followed by our Vice President of Global Development, Dr. Greg Friberg. Greg will review the Lumakras data in non-small cell lung cancer followed by Dr. P.K. Morrow, who is also a Vice President of Global Development. P.K. will then review the bemarituzumab beta in gastric cancer and Tarlatamab data in small cell lung cancer. Tarlatamab, by the way, is our DLL3 targeting BiTE also known as AMG 757. Don't worry about taking pictures of the slides as a PDF version of the presentation will be posted within the Investors section of our website immediately after the conclusion of our prepared comments. So with that, I would like to turn the call over to Dave.

Thanks, Arvind, and good afternoon, everyone. If we could please advance to one more slide. So you'll see data today on 3 programs: Lumakras, bemarituzumab or bema and Tarlatamab or AMG 757, all of which we think have made meaningful progress and really represent in one way, the culmination of our oncology strategy over the past few years which has a combination of a focus on precision oncology and very high-value targets such as KRAS or FGFR receptor 2b with immuno-oncology with a fight focus on our BiTE platform. If we can advance to the next slide, please. Here has provided for your reference and overview of the entire portfolio. Of course, I won't review this, but we have provided this as part of the enduring materials, and we'll be happy to take questions about any of the programs in the oncology portfolio. Next slide, please. Now the focus for today, of course, is on 3 advanced programs, the Phase II updated data, including overall survival and quality of life have been presented for the Lumakras program. Greg will review that, as Arvind noted. In addition, we have updated data from the Phase IIb FIGHT trial with bemarituzumab, specifically overall survival data, which we think gives us great confidence in this molecule as we move forward with regulatory interactions. And then finally, updated first-in-human data from the Tarlatamab or AMG 757 program, again showing, I think, clear efficacy and a path forward for this molecule in small cell lung cancer, and we'll provide some remarks around that by P.K. Morrow. With that, I'm going to turn things over to Greg, who's going to start with a review of the Lumakras data.

Thanks, Dave. It's really my privilege today to be able to present to you the Lumakras data. We were quite pleased 1 week ago today when we received approval from the FDA, a little under 3 months in advance of the PDUFA date. And just as a quick reminder, we've got additional reviews ongoing in 7 countries and regions around the globe. The program itself remains a broad clinical program, over 800 patients treated. You see a listing of the different studies that are ongoing here. I'm not going to list them all off, they'll be available for your reference. I do want to highlight at the bottom though, that of the over Phase Ib combinations that are underway, we will be reporting initial data for the MEK inhibitor, an oral EGFR inhibitor and an EGFR antibody combination in the second half of this year. Next slide. This is a menu of the pictorial representation of all the studies that are ongoing. Again, it's here for your reference. Next slide. I'm going to focus now on the updated data that was presented this morning by Dr. Skoulidis at ASCO. And what this represents is an updated data set from the World Lung Congress data set that you saw all in January. This is up -- 3 additional months of follow-up from that data set and about 6 additional months from the label that the FDA has. Next slide. The study design is the same. This is the CodeBreaK 100 study, single agent, 960-milligram once-a-day, non-small cell lung cancer patients only and, of course, high-quality, blinded, independent central review of CAT scans. Next slide. The characteristics on these patients, as you would imagine, are exactly the same from World Lung Congress and the median number of prior therapies is 2. So this is patients on their median of third-line therapy. Next slide. Of these patients, you see again the primary response elements here. And of course, the overall response rate remains stable at 37%. Of note, there is 1 additional complete responder since the last time we reported. That was a patient who was previously a partial response, who has now gone on to a complete response. The disease control rate is stable. And with additional follow-up, we're about at 15 months of median follow up now. You'll see that the durability of response, duration of response has ticked upwards and is now 11.1 months. Next slide. The progression-free survival data is exactly the same as we reported previously, again, a testament to the fact that for this endpoint, the data at World Lung Congress was quite mature at 6.8 months median. Next slide. Here's the new data, and we're just going to focus on this moving forward. This is the overall survival. So again, with the additional months of follow up now at a median of 15, we're reporting that the median overall survival is 12.5 months for these patients. You see the shape of the curve here. There are several -- actually many patients were still alive and censored out on the right side of the curve. We hope they remain so. And with this regard, putting this data into context, we're quite pleased to see 12.5 months here. We think this compares quite favorably to what would have been expected of these patients. If we take an optimistic view of it, somewhere between 8 and 10 months would have been -- what would have been expected of these G12C positive third line and later non-small cell lung cancer patients. Next slide. I want to take a few moments and drill down on safety. We received some questions. So I want to, again, frame what this is and what this isn't. We're reporting our safety update that has the 3 additional months. And you'll see in the box on the left that it's actually -- almost essentially the numbers are exactly the same as we've reported at World Lung Congress. Now just to ground this again, this is treatment-related adverse events. Again, Amgen doesn't determine that. That's determined by the physicians. This is different from treatment-emergent adverse events, of course, which are reported on the label, and this data is just lung cancer patients. So again, just the patients in the Phase II portion. And the differences again versus the label are that the label involves other disease types as well. I'll also just highlight that the terms that you see here are the standard med returns. This is our industry standard for reporting. And for that, as we dive into some of these, you'll get some additional detail here. As I have noted, in World Lung Congress, it remains true that there have been no fatal treatment-related adverse events on the study. For that matter, there are no treatment-related adverse events across the whole program as of the data cut that we used for the label, which is our last formal beta test. Similarly, the discontinuation rate based on treatment-related adverse events is stable at 7%. It's 9 patients. It's about 13 different terms here. And you can see that most of these terms, a majority of them are related to liver dysfunction, which again is a known side effect that we've well established and described on our label. For example, drug-induced liver injury, again, this is the term. This doesn't refer to his cases. There, in fact, have been no ILO cases in this study nor have there any been again at the same data cut that I mentioned across the whole program. Similarly, you see that there are 2 cases noted here of pneumonitis. We've received some questions on this. There was one additional case that was noted on the label. Of note, these 2 cases did have some compounding factors, all of them at either prior radiation or current radiation, one of them even had prior ground glass capacities and of course, prior PD-1 therapy was common in these folks. So with that, I'm going to move forward. I would just add before we move on that cases of interstitial lung disease pneumonitis, we are seeing across the entire industry that for lung cancer, this is commonly becoming class labeling, and we've even seen it with some of the large molecules that have been approved recently like [indiscernible]. Next slide. Focusing back on efficacy again. We can look at both response rate and now we can look at median overall survival. Now this is a single-arm study, but we can look at different subsets. So we looked at age, we looked at performance status, prior lines of therapy and then what those therapies were. And the long and short of it was that the response to Sotorasib have actually appeared consistent across all these groups. There was an interesting early signal potentially with patients who had prior PD-L1 but no platinum chemotherapy, where a slightly higher response rate and survival. But again, we want to caution you 13 patients, it's fairly small, and we're going to continue to follow this up. Next slide. Similarly, we can look at a variety of biomarkers now, and we're reporting some updates with regard to response rate and survival and biomarker subsets. The story that's being presented on the slide in front of you looks at mutation allele frequency, so allelic burden, and that does not predict responsiveness or nonresponsiveness. Similarly, we looked at super mutation burden, again, did not predict responsiveness or nonresponsiveness. Similar response rates in both high and low groups. And finally, we looked at 3 key tumor suppressors that are often mutated where there's loss of function in lung cancer. And you can see that with TP53 and with STK11, there didn't appear to be decreased responsiveness in patients with mutation in those genes. However, keep one, a well-known poor prognostic factor did appear to have a slightly lower response rate as again, we've reported previously. Next slide. Now we can take this to the next level and actually look at response rate and overall survival. And what you see, as I noted on the prior chart, that if a patient was green with a KEAP1 mutation or gray with a KEAP1 mutation, those patients not only had lower response rates, but they have had a lower progression-free survival and median survival on the order of 5 to 8 months. And yet, if you look as compared to the all-evaluable patients, the patients in orange, these are patients with STK11 mutations but with wild-type KEAP1. Those patients actually look like they did a little bit better, slightly higher response rate at 50% and the median overall survival of 15. You see the confidence intervals there. So again, we're going to watch this very closely, but certainly, interesting hypothesis generating data. Next slide. In summary, the new data, median overall survival of 12.5 months, which we're quite excited by. Efficacy was seen across all subgroups. And of course, this STK11 mutation signal is something that we're hopeful we can follow up on because these patients tend to be poorly served by available therapies. I'll just add that we do have a poster available as well by spirit all, and this is presenting patient-reported outcomes. The long and short of it again is that we saw either maintenance or improvement in the single-arm study on quality of life measures, including physical function. We saw some improvement in cough and in dyspnea, and then patients also reported in a great majority that they were not at all bothered by the treatment side effects of the drug. Next slide. We're thrilled to also report that there was a concurrent publication in the New England Journal data today. And this is -- I would refer you to this. There's a supplemental index that has a lot of the methods that were used for the biomakers example, if there are any questions. Next slide. The final focus that I want to go to today is with regard to our colorectal cancer data. Now we've had a preliminary look at the data on this study. We'll have additional follow up, of course, but we wanted to present the top line data to you today. This was a Phase II single-arm cohort from our CodeBreaK 100 study. There were 62 patients with advanced colorectal cancer that were treated in the study. Again, they were all treated with 960 milligrams once daily. It's a single arm and we used blinded independent central review, just like we did with lung cancer patients. And for the top line data, we saw an overall response rate near 10% using those stringent criteria. Again, the -- from this preliminary look, the drug is clearly active. Since ESMO 2 years ago, we've been discussing that the circuitry in colorectal cancer looks slightly different than what we've seen in lung cancer. It's analogous probably to BRAF. And so what we're hoping, again, as we dig into this data is that we're going to be able to provide additional insights sometime in the second half for publication and/or presentation. We looked at this, and we actually determined that it doesn't meet our criteria from the monotherapy bar that we were looking for in order to do a single agent accelerated approval filing. And again, that's probably related to the underlying biology. We do, however, feel that again, this drug is active, and we're very focused on bringing this forward where we think there would be most value for colorectal cancer patients. And that's clearly in the combination studies that are ongoing. We're hoping that with the continued focus on combinations, we'll be able to present some of that data to you shortly. And initial data should be coming out in colorectal specifically for our EGFR antibody combination in the second half of this year. So with that, I'm going to hand off to Dr. P.K. Morrow and let her cover some of the additional data from ASCO.

P.K., before you start, let me just provide a quick update. Apparently, folks are having a difficult time accessing the slides on the webcast. So we have posted the slides under the Investors section of our website, and we are also sending you a link separately. So please follow along the audio on the webcast. And like I said, you should have the slides accessible now to our website. P.K., please go ahead. P.K., I think you have us on mute.

Perfect. Thank you so much, Arvind, and it's really a delight to meet with you all today. So if you can advance to the next slide. So I'm delighted to present the updated results from the FIGHT trial, a Phase II randomized study of bemarituzumab or bema, as we affectionately call it, combined with modified FOLFOX 6 versus FOLFOX 6 alone. And this data was presented by Dr. Dan Catenacci as part of a late-breaking abstract today. So with continued follow up, we are very encouraged by the confirmed efficacy of bemarituzumab in this population. Next slide. So first, a little background. Bema is an IgG1 monoclonal antibody that is specific for FGFR2b. And it has, as depicted on this slide, 2 mechanisms of action. It blocks fiberglass growth factor signaling and causes internalization and degradation of the receptor as you can see in the top portion of the slide. In addition to that, it also enhances antibody-dependent cellular cytotoxicity by recruiting the natural killer cells or macrophages in the tumor microenvironment, which is depicted on the bottom portion of the slide. Importantly, as you know, Bema's selectivity for FGFR2b means that it doesn't affect FGF23 ligand binding, which means it doesn't cause the hyperphosphatemia, which is seen with FGFR tyrosine kinase inhibitors. Our scientific rationale for Bema's activity was first validated, as you know, in a late-line study of gastric cancer in which Bema demonstrated an 18% overall response rate in FGFR2b overexpressing gastric cancer. Next slide. So these lendings led us to the FIGHT trial, which was a Phase II trial that randomized previously untreated patients with unresectable locally advanced or metastatic FGFR2b overexpressing or amplified gastric cancer or gastroesophageal junction adenocarcinoma to treat with either Bema and FOLFOX or placebo plus FOLFOX. And the primary endpoint was progression-free survival with key secondary endpoints, including overall survival as well as response rate. Next slide. It was previously reported that all outcomes, including progression-free survival, overall survival, overall response rate, improved with increasing expression of FGFR2b in tumor cells. Notably at the time of the data cut off, which was September 23, 2020, on this slide, the median overall survival had not yet been reached. Next slide. This slide depicts the fact that both PFS and OS were improved in the Bema arm across all the subgroups including age, gender, region, prior therapy and whether or not the patient had received a prior dose of modified FOLFOX prior to randomization. Next slide. And this slide emphasizes the fact that Bema provides a benefit in gastric cancer patients who are found to have FGFR2b overexpression by IHC, amplification by ctDNA or both with the greatest benefit in those with both IHC positivity and ctDNA positivity. Next slide. So we're now really happy to present the updated overall survival results seen with a longer follow up of 12.5 months and a data cutoff of February 28, 2021. In this updated analysis, median overall survival has now been reached in the intent-to-treat population with a median overall survival in this population of 19.2 months and a hazard ratio of 0.6. As you move across the graphs from left to right, you see the outcomes continue to improve with increasing levels of FGFR2b expression. And in the middle graph, you can see at a higher cutoff median over survival has not been reached and the hazard ratio remains 0.52. And in the rightmost graph, at the highest cutoff of more than 10%, median over survival is now 25.4 months versus 11 months in the placebo arm with a hazard ratio of 0.41. Next slide. Bema's toxicity profile continues to be consistent with previous experience with an increased frequency of both dermatitis and dry eye in the Bema containing arm. I'm going to double-click on the corneal events on the following slide. Next slide. Notably at the median time to onset of the ocular event, which was approximately [ 24 weeks ], and the fact that no prophylactic regimen was recommended in the FIGHT trial, this truth provides us with a clinical window of opportunity to add a prophylactic regimen such as a simple ocular lubricant into the Phase III trial to reduce the risk of these adverse events. Next slide. So in conclusion, you can see and we've reviewed the fact that for both progression-free survival as well as overall survival, favored Bema over placebo across all prespecified subgroups. And at this now, longer median follow up of 12.5 months, we've reached a median overall survival and an improvement of more than 5.7 months over placebo in patients with FGFR2b expressing tumors. In general, Bema improved outcomes in patients with tumors expressing FGFR2b regardless of ctDNA versus IHC status, and the delayed onset of coronal adverse events may really help us to support the utility proactive ocular prophylaxis. We are now planning a Phase III study and program underway for the first-line treatment of gastric cancer patients. Next slide. So I'm now delighted to present updated results of the Phase I study of Tarlatamab in small cell lung cancer, and this data is presented by Dr. Owonikoko at ASCO. Next slide. One more slide. So Tarlatamab is a half-life extended bispecific T-cell engager, which is typically administered every 2 weeks. It targets DLL3, which is highly expressed in small cell lung cancer and engagement of Tarlatamab with a DLL3 expressing small cell lung cancer tumor cells leads to activation and proliferation of T cells and ultimately, tumor cell apoptosis. Next slide. This slide depicts the key inclusion and exclusion criteria for Tarlatamab as well as the baseline characteristics. And of note, more than 70% of patients had received 2 or more lines of therapy, and 1/4 of patients had experienced progression through at least 3 lines of therapy and almost half had received a prior PD-1 or PD-L1 inhibitor. Next slide. This slide demonstrates very encouraging preliminary efficacy data with an overall response rate of 20% in this heavily pre-treated population and tumor shrinkage seen across a range of doses. At the 4 highest doses, overall response rate improved to approximately 28%. Next slide. Notably, Tarlatamab achieved not only encouraging partial response rates but also an impressive duration of response of 8.7 months. In addition, both Tarlatamab's efficacy and tolerability are supported by the fact that approximately 7 out of 13 or more than half of patients with confirmed PRs are still receiving therapy and having ongoing response. Next slide. This slide talks about the toxicity profile of this program. And while all clean CRS occurred at a rate of 44% and the updated analysis, the majority were grade 1 and did not recur in subsequent cycles, emphasizing tolerability of Tarlatamab in this setting. You can see here, there was 1 episode of grade 5 pneumonitis that occurred in 2019, which was concurrent with disease progression approximately 6 days after the second dose of this drug. Based upon our own internal analysis, we amended the protocol to implement an additional dose of dexamethasone prior to the second dose of Tarlatamab and have not experienced any further events since then. Next slide. So in conclusion, Tarlatamab has demonstrated a confirmed overall response rate of 20% across all doses with a duration of response of almost 9 months. Majority of events were low grade and did not recur after cycle 1 and a small proportion of patients, 14%, developed anti-drug antibodies. These demonstrated no effect on either exposure nor adverse events. We are now planning and actively working to move into a potentially pivotal dose expansion with 1 or more doses, and we'll be discussing these plans with regulators. And I think that was my last slide.

Thank you, P.K. If we could just move on to the final slide here, and then we'll open it up to question-and-answer. So as you heard today, we've got, I think, real progress across all 3 of these programs. You're very familiar with Lumakras with bemarituzumab. We're delighted with the updated overall survival data and are looking forward to upcoming regulatory interactions in the near future to define the Phase III program. We'll provide guidance on that as those discussions occur. And then finally, based on the cumulative data in the Tarlatamab program and discussions with our investigators, we are planning regulatory interactions to discuss a potential pivotal path there. So I think real forward momentum in that program. It's worth pointing out that roughly 70% of the patients in the first-in-human study have 3 or more prior lines of therapy. There's a very heavily pretreated group of patients with a quite poor prognosis. So with that, Erica, why don't we go ahead and compile questions and open it up for Q&A.

[Operator Instructions] And your first question is from Carter Gould with Barclays.

Great. Congrats on all the data. I guess one for Dr. Morrow and David. You're probably going to get inundated with Lumakras questions, so maybe I'll go a different direction to start off. On 757, I guess, first to clarify, have you defined 100 meg as the go-forward dose? And specifically then when you're thinking about future development here, should we think about that Phase II in a similar population as this? Or will you have to focus on a little bit more on maybe a PD-1 experience population?

Yes, Carter, thanks for those questions. I'll take them and then ask P.K. if she wants to add any perspective. She's very closely involved with this program. In terms of the dose, that's one of the things we'll discuss with regulators. It's feasible that we may take a couple of these doses into a potentially pivotal trial with an adaptive sort of design. Given -- when you look at the efficacy and safety numbers across, say, the last 4 cohorts or so, it's not clear you've got a true differentiation there. So I think that's one of the key questions that we want to discuss. And in addition, we're continuing week-on-week to accumulate additional data. But we're quite encouraged with what we're seeing so far. We would expect in a potentially pivotal trial, probably a patient population roughly similar to the one you're looking at here. P.K., perhaps you'd like to provide briefly just a little more perspective on that.

That's perfect. Thanks, Dave. So the only thing I would add is that what we've seen with the -- in terms of the toxicity profile across the ranges of doses is a fairly consistent toxicity profile. So that is the fact that increasing dose isn't driving increasing toxicity right now. So that's -- for that reason, we really would like to explore perhaps more than 1 dose to identify the optimum of those to go into our ultimately registrational plan.

Your next question is from Alethia Young with Cantor.

Congrats on all the progress. I'm going to ask a Lumakras question. I just wondered if you talk a little bit more about the binding with STK11 and KEAP1 and what that might mean in combination with other agents and in earlier lines of therapy.

Yes. Thanks, Alethia. Great question. And obviously, there's not a simple story here. I think it's important for everyone to understand. This is not like a tyrosine kinase inhibitor, where we can predict now with relative accuracy the presence of gatekeeper mutations that may confer resistance. This is a very different biologic setting. Let me ask Greg to comment here. He and the team have been digging through this, and we're continuing to do multi-gene profiling to try to understand and potentially develop signatures of response and resistance. Greg, and then you can comment on implications for combination therapy as well.

Yes. Particularly with STK11 and KEAP1, these are loss of function mutations that tend to be very poor predictors of how these patients are going to do. Another way to say that is these patients might be very poorly served by available therapies out there. Based on what we've seen so far, we think, again, there is a variety of hypotheses that could be tested, but one which is listed on our study list is we will be looking to see whether or not, for example, biomarker selection would be a useful way even in untreated metastatic patients to study this drug and see whether or not again, what could be looked at there would look better than what available therapies are. So STK11 is an obvious choice there, and that will be one of the biomarkers we're using to test that hypothesis in previously untreated patients.

Right. And it's important to just recall that no single gene here is probably sufficient to give one a sense of prognosis and likelihood of response.

Your next question is from Mohit Bansal with Citigroup.

Great. Congrats on the progress. A big picture question for you, David. So when I look at your oncology portfolio, it seems like you have targeted smaller cancers, more precision oncology targets that did put to Bema and even small cell lung cancer as a small indication. So is it by design and this is how do you see -- this is how you see your cancer portfolio going forward? And how it relates to also does it how -- would it be the way you go forward and bringing in more assets in the mix as you think about future deals?

Thanks, Mohit. I think you raised a couple of important points here. In terms of these molecules being directed at targeted populations, absolutely. I don't know that that's specific to our approach. I think that is the natural evolution of the field as we become better and better at understanding the molecular wiring of these tumors and use that to guide therapeutics development. So I fully anticipate in precision oncology that we'll see that sort of approach moving forward. In terms of being open to other molecules in the portfolio, I think as we've said, from a business development perspective, across all our therapeutic areas, the aperture is wide open here. We want to bring in assets where we think we can make a real difference that serve an unmet medical need and that will appropriately reward our stakeholders as well as stakeholders of any partner or target. So this is something that wide open and ready to go.

Your next question is from Umer Raffat with Evercore ISI.

I'm very intrigued by this slide on the colorectal where you guys are not moving forward with the monotherapy in colorectal setting while your competitor is. And I guess my question is, what did the AUC look like in the 6 patients that did have a response in colorectal? Was it over 100 micrograms per milliliter? And perhaps conversely, what was the AUC in the nonresponders? I'm just trying to understand, is this more an AUC issue going on while realizing it's a covalent binder? Or is it truly that monotherapy is not very compelling here?

Thanks, Umer. I'll let Greg comment on some of the specifics of the pharmacokinetics, but my suspicion is that, that is not the answer here and that it relates to the underlying biology. I'll point out that, as Greg noted, this was a robust Phase II trial with independent blinded central review. We think that the biology is telling us an answer here. We think that we've got target coverage. And based on our assessment right now, combination therapy seems to make the most sense for these patients. It's implausible to me that there would be dramatic differences in agents that achieve similar sorts of target coverage. Greg, would you like to speak any further on the latter?

Yes. I'll just add, Dave, remember, we've looked at a variety of factors, including pharmacokinetics that might explain what we've seen. We have no leading hypothesis as to why one patient would be responsive versus unresponsive at this point. Again, response rate is just one endpoint. Again, we saw 2 later shrinkage that isn't fully described by the data that we showed today. And when we present that, I think, again, it looks quite similar to what we've seen previously. What's fair to say is that, that your question that you postulate that doesn't seem to be, at least based on our understanding today, what's going on in these responsive versus nonresponsive patients.

Your next question is from Cory Kasimov with JPMorgan.

This is Gavin on for Cory. We were just wondering if you could discuss CNS activity with Lumakras or when the time lines would be for when we get an update on that?

Yes. Let me just turn that over to Greg.

Yes. So about 20%, 21% of the patients who were on our CodeBreaK 100 study had brain mets. We absolutely have seen tumor shrinkage in these patients. Of course, brain mets are a very poor prognostic indicator. The question ultimately will be answered in our randomized study, but additional biomarker subsets, including brain mets, are something that we do intend to publish in the future. We'll have to give you an update on the timing on that. I don't know off the top of my head when we're going to be publishing our brain met data. I'll just say that, again, we have seen activity in these patients. And from that standpoint, we also have not seen one common mechanism of resistance. It's not a case where we're seeing compartmental escape as a common escape mechanism.

Your next question is from Geoff Meacham with Bank of America.

I just had 2. So for Lumakras, I also have a question in colorectal. What are some of the options when you think about down the road potential registration? I'm just thinking of combinations that could be more synergistic in CRC versus lung outside of EGFR, which is what you guys have in development now? And then in Bema, just talk a little bit about the cornea AEs in terms of how you can manage that in future studies? And if that could down the road be problematic?

Okay. Thanks, Jeff. And I'll turn this over to Greg and P.K. sequentially to talk about the colorectal Lumakras question in the Bema ocular toxicity prophylaxis question. We've got a number of combinations that could be a way forward in colorectal cancer. And one of the things that's potential of interest there is also combinations ultimately with backbone chemotherapy. Let me ask Greg to provide additional color, and then we'll hand off to P.K. to talk about the keratitis prophylaxis with bemarituzumab.

Yes. I would just add, Dave, that we're going to take forward the combination that we think has the best chance of helping the patients. And so we've got multiple combinations we're studying right now, as you mentioned, including chemotherapy backbones. Now this is a population that's about 3% to 5% of colorectal cancers. So clearly, what we'd like to do is be able to find a combination that we could insert earlier than these third, fourth, fifth line metastatic colorectal cancer patients. And so we're actively working on trying to pick which of those combinations is the right one. And I think we'll be able to show more data as the year progresses.

Yes. So thanks so much for the question on the corneal events. So as I mentioned, given the fact that there was a delay in terms of the onset of these events, approximately 6 months, we do have this particular window of opportunity in order to intervene. And when we've engaged with ophthalmologists behind the scenes to review and for them to take a look at our patient cases, the recommendation has been: one, monitoring; two, to ensure that there's no significant effect on visual acuity; and third, that there is no -- for the most part, there may not be a requirement for aggressive treatment of prophylactic treatment. And what I mean by that is not steroid drops actually, but simply just eye drops that are lubricating the eye may help to reduce the risk and the intensity of these ocular adverse events. So we're planning to implement this in Phase II.

Your next question is from Ronny Gal with Bernstein.

Just first, a housekeeping question. How many patients are still on the CodeBreaK 100? And what is the duration of therapy? When we model things, it's always the question of the tail and how many patients stay there for a really long time if you can just share that with us? And second, can you discuss a little bit the mechanism of escape from KRAS that you're seeing in both lung and colorectal just as kind of helping us understand what future therapy combination might be?

Thanks, Ronny. I'll let Greg talk to both of these questions. The second one, I think we've described that there is no I think single escape mechanism, which tells us that we need to dig a little deeper to understand are there particular patterns here that will guide us towards combination therapy. Greg, do you want to give an update on a number of patients still on therapy?

Yes. I can't give you the exact number off the top of my head, but I'll tell you, in the New England Journal Paper, we actually lay out the latest data cut very nicely and show that further responders that, again, we haven't -- we'll reach a median duration of response that we can report. But of course, the arrow bars look -- include a nonreportable number. So there's a good number of patients who are in response who are continuing on therapy. My recollection is a shade over half, but I would have to look at that chart again. With regard to the question of resistance mechanisms, we looked at this. We haven't seen one single mechanism of resistance. We haven't seen again compartmental escape as an obvious one. I haven't seen any sort of rapid resistance versus delay that are clustering into one effect. And so in that sense, the curves that you see, all the data is there, it is reported, but there is no one obvious mechanism of resistance that's shown itself up till now. We're going to continue to follow the patients very closely. Obviously, we've got biospecimens particularly from plasma over time. And if some sort of signal emerges for a molecular mechanism that we think is appropriate to, again, adjust our strategy to, we will. That has yet to reveal itself. It's been elusive as to finding one single common mechanism of resistance at this point.

Are you seeing multiple escape mechanisms per patient?

Yes. I think, Ronny, the likelihood is that there is more than one escape mechanism between patients, whether within patients, intrapatient escape mechanisms differ. I don't know that we have enough data. As Greg mentioned, we're following these patients longitudinally, collecting plasma for circulating tumor DNA, and we'll continue characterizing that. It's probably worth reminding everyone of the BRAF inhibitor story in melanoma, response rates are 50% to 60%. In colorectal cancer, they are 0% when the tumors bear the same V600E mutation. So the contextual wiring and co-mutational burden is clearly important here. And I suspect we're seeing a similar story emerge with KRAS G12C.