Amgen Inc. (AMGN) Earnings Call Transcript & Summary

My name is Cindy, and I'll be your conference facilitator today for Amgen's conference call from the World Conference on Lung Cancer. [Operator Instructions] And I would now like to introduce Arvind Sood, Vice President of Investor Relations. Mr. Sood, you may now begin.

Okay. Thank you, Cindy, and good evening, everybody. So let me begin by thanking you for taking the time on a Friday evening to participate in our session. My initial logic was to conduct this session as close as possible to the presentation of our sotorasib Phase II data in the scientific session at the World Lung Conference. But as you know, we ended up putting out a press release yesterday containing most of the data as there was a media leak of embargoed material. Regardless, we are delighted that you can join us. We are pleased with the progress we have made with this program on several dimensions: the speed with which we have gone from first-in-human to NDA submission; quality of the data having undertaken an intent-to-treat analysis and central adjudication; and breadth of the program considering the many combinations we are exploring. So very much looking forward to having a discussion with you today on this important data and what sotorasib can mean for treating lung cancer patients. To lead this discussion, I'll turn it over to Dr. David Reese, our Executive Vice President of Research and Development, who will make some opening comments and introduce other presenters and investigators who are joining us today. Dave?

Thanks, Arvind, and thanks again, everyone, for joining us on a Friday evening, and then we know in some cases, early Saturday morning. If you are following on along with the associated presentation and turn to Page 4, you can see the agenda that we put together for this session. I will make some brief introductory remarks and give you a status update of the program, including some important regulatory milestones. And then turn things over to Greg Friberg, our Head of Global Development for Oncology, who will briefly review the presentation that was just shown and discussed at the World Conference on Lung Cancer. For those of you who may not have been able to participate in that session, Greg will also provide some additional color on the data. And then we will open it up and should have plenty of time for questions and answers. We're very pleased this evening to be joined by Professor Hoss Borghaei from Fox Chase Cancer Center in Philadelphia, one of our primary investigators. As well as another primary investigator, Professor Jürgen Wolf from the University Hospital of Cologne in Germany. Professor Wolf is joining us in the middle of the night from Germany, and we are grateful for his participation. If you turn to Page 5 in the presentation deck. As you're all aware, sotorasib is a first-in-class KRAS G12C inhibitor with many first milestones, first in the clinic, first to complete a pivotal study and the first to show comprehensive efficacy benefits. We've now treated over 700 patients across 13 tumor types and across 5 continents. There are 10 Phase Ib combination cohorts currently enrolling. And I know that's a focus for many of you. We expect the initial data in the first half of this year from some of those cohorts at meetings like ASCO and then as we get into the summer at the ESMO meeting. We remain quite pleased with the overall efficacy and tolerability profile of the drug. In fact, I would say we're thrilled with it. And the -- one of the remarkable things that I think we've witnessed is the absolute consistency of the behavior of this drug since we first introduced it into the clinic to now. We're the only once-daily oral dosing regimen, and we are quite confident of our dose and schedule moving forward. If you turn to Page 6 in the presentation, you'll see that we continue to advance rapidly through global clinical development and regulatory submissions. As we announced in December, we had filed with the U.S. FDA and the EMA in Europe. More recently, we have completed regulatory submissions in Canada, the U.K., Brazil and Australia, marking 6 major regulatory submissions within about 29 months of the first patient being dosed in this program. In the United States, sotorasib has been granted Real-Time Oncology Review and Breakthrough Therapy Designation. The former has been important because it allows, as the name implies, real-time submission of data as opposed to aggregation of data packages and lump submission to the FDA. This is designed for programs that are promising and to expedite regulatory review. In terms of clinical updates, we would like to point out that we do plan a first-line non-small cell lung cancer study that we'll be launching in the coming months. We will be biomarker selecting these patients. They -- and many of these patients, such as those with STK11 mutations, have significant unmet medical need as we will discuss in a moment. As we have previously discussed, we have a 10-arm master protocol open designed for rapid progression of promising combinations. And I think a nice example of that is the fact that we've cleared the safety hurdle for full dosing of sotorasib in combination with a MEK inhibitor. That is now open to an expansion cohort to examine efficacy, and we expect to have those efficacy data over the course of this year. We've also enrolled a triplet cohort with sotorasib, a MEK inhibitor and an EGFR antibody, another example of the flexibility of this protocol that allows us to add and expand arms based on emerging clinical data. You can see a summary of the global development program on Slide 7. That is provided for your reference. And now I'd like to turn things over to Greg Friberg, who will take you through the clinical data that we presented today. I would remind everyone that this is the culmination of a 40-year quest in the field to drug KRAS, and we are simply thrilled with the results that we've seen. Greg?

Thanks, Dave. I'm just going to skip ahead a couple of slides and point you all towards Slide 9, the CodeBreaK 100. And we're very fortunate to be able to review some of this data. Of course, Dr. Bob Li just presented this in the Presidential Symposium about 1 hour ago at the World Lung Congress. And we're very fortunate to have 2 other authors, Dr. Borghaei and Dr. Wolf here with us today. So if we move forward to the introduction slide, you can see Slide 10. You're all familiar with this pathway, so I won't review it. But I will take us back in time a brief 4 months ago when we were at ESMO reviewing our data from Phase I. We broke out 59 lung cancer patients in that presentation. We also had a simultaneous publication in the New England Journal at that time, describing what we saw in the Phase I study, and of course, that is listed here. Today, we're going to be focused on 126 lung cancer patients who were treated in the Phase II portion of that study. So moving on to the trial design, which is Slide 11. I just want to point out a few factors about who these patients were and how they perhaps differed from the Phase I population. First and foremost, they were all prospectively selective prior to their treatment with sotorasib using a central and similar assay for G12C. These patients had progressed on prior therapies, anywhere between 2 and 3 -- I'm sorry, they were anywhere between their second and their fourth line of prior therapy, so 1 to 3 priors. And they were to have no active brain metastasis. About 20% of the patients did have brain mets, but they had been previously treated. So now these patients were all treated with 960 milligrams once daily. And of course, one of the more significant differences in this population, a very high-quality analysis using an independent, blinded central review for the cat scans of these patients. And we're all familiar with that often representing a higher bar than the heterogeneous reads of individual contributors. If we move on to Slide 12, we can see the characteristics of these patients. As I mentioned, 126 patients. This was quite the global population from 11 different countries. These patients were fairly heavily pretreated. 90% of them had prior PD-1 inhibitors, about 90% of them had a prior platinum and 81% had, had both. This data cutoff, again, represents a December data cut, and there was a quite significant 12-month median of follow-up time. Other factors, just to point out again, median of third line for these patients, 2 priors. And of course, as is consistent with KRAS mutants, there was a prevalence of over 90% former smokers. Moving on to Slide 13. This is the first efficacy slide. And really the most important number here is the centrally reviewed independent -- by the independent review committee. 37% confirmed objective response rate and that, of course, includes 3 complete responses confirmed by this central adjudication committee. An overall disease control rate of 80%. If we move on to Slide 14, you can see the waterfall, again, giving you some indications of the depth of tumor response here. One of the factors, again, that you'll see here is the shape of this waterfall. As Dave had alluded to, it's quite consistent with what we've seen in other cohorts of lung cancer across the study. About 81% of patients had some degree of tumor shrinkage. And you can see the RECIST responders here in blue and green. And again, over towards the right, you see those legitimate complete responses. If we move on to Slide 15, you have a set of swim lanes depicting the durability of tumor response. So this isn't just durability of treatment. When we're looking at durability of response, of course, we want those patients not only to be on drug, but not have progressed, as is standard. And what you see here, again, are the 46 responders across the study. First and foremost, I want to point out just some of the summary statistics. Median duration of response of 10 months. And if you look at the green diamonds, which is the time at which the first scan indicated the response, you'll see that most of the responses happened early. Median was 1.4 months, but that didn't preclude that patients could have responses farther along. And you'll see down at the bottom there, a patient who actually had their first centrally confirmed RECIST response as a PR, fairly late, 10 months into their therapy. I'll also point out that 20 of the 46 patients here are not only responders remaining on treatment, but they also have no progression as of the data cutoff. And it's quite nice to see. All of those arrows there, again, highlight the patients that remain progression-free. If we move on to Slide 16. Of course, we have been quite interested in describing the responders, but when you have 126 patients, we can look at time-based end points for the whole population. This is both responders as well as stable disease patients. And you can see a quite impressive median progression-free survival of 6.8 months here with the confidence intervals listed there. Just to put that into perspective. Of course, this compares quite favorably to the available therapies for these patients right now. Docetaxel, for example, has anywhere between a 2- and a 4-month progression-free survival. If we move on to Slide 17, we reach the treatment-related adverse events. So for a single agent -- a single-arm study, this is an appropriate way to look at the adverse events. And really, what's remarkable is the drug was quite well tolerated. Most of the treatment-related AEs are grade 1 or 2. There was a 20% rate of treatment-related grade 3 adverse events, and that compares quite favorably to other drugs in the field. The toxicity profile was fairly mild. With regard to GI toxicities, we really did not see significant cardiovascular toxicities. We looked very closely for QTc abnormalities. Did not note any in the study and of course, no treatment-related deaths, something that we're quite pleased to see. With regard to dose modifications, less than 1/4 of the patient required any dose modification, and 7% of patients required discontinuation. Again, if you look at drugs like osimertinib, this is less than half of what you see even with those other targeted therapies. And again, compares quite favorably. Again, as a whole, for a study that was done in 11 different countries, we were very pleased to see this. And again, it compares nicely to other available therapies. So moving on to Slide 18. I just want to take a few minutes and go through the biomarker slide. This is fairly dense with regard to the data here. As I think Dr. Li stated very nicely and Dave alluded to, we wanted to present data, in particular, related to known prognostic factors, factors that perhaps describe patients who are poorly served by the therapies that are out there today. If we start on the left, looking at PD-L1, of course, the protein that's expressed at varying levels of poor man's barometer of inflammation in the tumor, and of course, is also likely predictive of outcomes with immunotherapies. These were 86 tissue samples that we collected on patients. We used the pharmDx IHC assay, the 22C3 antibody and really looked at the question of were their responses across the different types, and the answer was yes. But before we dive into those details, I just want to point out a couple of factors here. One was Dr. Li nicely noted that the samples were taken from patients that they were taking from what we were able to get. The median time before treatment for the samples was about 90 days, so not too far before, though there were some as far as 3 years. So we need to keep that in perspective. But you do see here something that was repeated from our Phase I experience. Those patients that are coming into this study that are in their second, third and fourth line, patients who have had PD-1 previously, tend not to be the high PD-L1 expressors. We saw relatively few of those. Interestingly, and it was nice to see regardless of this PD-L1 status, we were able to see responses. And that's something, again, that we're going to follow very closely moving forward. If we look to the right of this slide, we actually now are looking inside the cell rather than protein expression. We're looking at DNA mutations, somatic mutations. And these 2 mutations, in particular, STK11 and KEAP1, are known to be quite poor prognostic factors for patients who are treated with frontline therapies. There may be a predictive element of that, that relates to immunotherapies, though that's hotly debated. Now these are 2 tumor suppressors. When they become mutated, they have a loss of function. So if you lose function of tumor suppressor, you have more aggressive disease. And we looked at these quite closely. Most of these 104 patients came from tissue, though some came from plasma as well. We used the Tempus xT panel. And in that regard, we saw a couple of things. First and foremost, about 35% of these patients had STK11 mutation and about 20% of them had KEAP1 mutations. Interestingly, there was 10% overlap in patients who had both. Now the takeaway from this, of course, is that sotorasib was able to generate responses in all these groups. Of course, when you get into small subsets, we have to recognize that the error bars get quite large, so we don't want to overinterpret this. But certainly, seeing that there were responses in each category was nice. I think there's another factor that I want to point out, though, which is, if you look just at the orange, you see the STK11 patients. Again, numerically higher response rate than the all evaluable patients. And similarly, if you look at the isolated KEAP1 mutations in gray, you see numerically something that's lower than the all [ patients ]. Again, interesting observations, we're going to need to keep following those. But you should also look at the green bar here. Again, if you have both STK11 mutation and KEAP1, something in between happens. And the story here is clearly one that this is not a 1-gene, 1-outcome situation. It shouldn't be surprising to any of us that there is a complex web of circuitry in these tumors. And if we're going to look at small numbers, if we're going to try to interpret them, we have to truly understand what the totality of the biology of these given patients are. And whether that's intercomparing data sets or trying to interpret our own, we're very aware of that. We're fortunate to have other biomarker data that we're diligently looking through at this point. And the goal, of course, is to identify patient prognostic groups where there might be an opportunity for sotorasib to offer them something better than other available therapies. So moving on to the Slide 19 and our conclusions. Again, we're quite excited to have seen that the overall response rate using a very high-quality, centrally independent blinded review committee was 37%. We can look at durability of response in PFS with 10 months and 6.8 months, again, comparing quite favorably to other available therapies. And of course, from a safety profile, the drug was quite well tolerated. There were no treatment-related deaths. That 20% Grade 3 AE level in this population of pretreated sick lung cancer patients is quite nice to see. We're continuing to develop the drug. We have an ongoing Phase III study, the CodeBreaK 200 study, comparing the drug against docetaxel. And of course, we're -- we've been very fortunate to be given the breakthrough designation in the regulatory filings in 6 nations around the globe. We're looking forward to, again, discussing more with regulators. So with that, again, I want to thank you for the time to frame some of this data. And Dave, I'm going to hand it back to you to start the Q&A.

Thanks, Greg, and we have plenty of time for Q&A. So Cindy, why don't you go ahead and get a few folks in the queue, and let's start with questions?

Okay. So your first question, from Chris Raymond of Piper Sandler.

Just a question on the companion diagnostic effort. And I know you guys have announced collaborations and efforts to get both liquid and tissue biopsy companion diagnostics. But -- and I think you guys have said previously that this would not be a rate-limiting step for approval, but maybe just talk about the effort there. Will there be one available at the time of approval? And maybe just talk about the commercial implications, if not. And then, Greg, I had a question on the swimmer plot. You gave a lot of great detail on the time to response, and the median of 1.4 months is great. But just noticing there's a few that seemed to sort of cluster around that 3-month time frame and a couple beyond. Is there any hypothesis as to why you see some of these later responders?

Great. Thanks, Chris. Perhaps I'll take the first part of the question around the diagnostics, and then Greg can talk about the time-to-response kinetics. In terms of diagnostics, with both the tissue-based and the plasma assay with our partners, those submissions are going forward. The goal is to have available when the drug is approved. Those companion diagnostics, again, we don't feel that, that should be rate-limiting at this point. It's also worth pointing out that many centers, certainly many major academic centers through their CLIA-certified labs, have large gene panels and are already doing KRAS sequencing. So is there work to do in terms of widespread use and reporting of KRAS G12C mutations? Sure, there always is. But these are, we believe, are solvable problems, and we're moving very quickly on many fronts there. Again, do not believe that should be rate limiting. Greg, do you want to talk again about the time to respond kinetics?

Yes, sure. And so on that slide, I think it's Slide 15, just for those who are interested the order of the patients is actually by their durability of response. So it looks lumpy bumpy because some of the patients start the clock of when they have the response later than others. That's why it's stack-ranked. You ask a very fair question, and we're always in the field for these molecular pathways, we want to have an easy answer. There doesn't seem to be a single answer here. We've looked both at the flavors of the progressions in terms of compartments and whether it's target lesions or nontarget lesions, and there's no one pattern that pops out there. We've also looked at, at least in the patients that we have data both from the Phase I as well as this, we've looked at some of the baseline molecular characteristic. And all that is clear is that there is no one clear pattern here. It shouldn't be surprising to us that there are many ways to -- that a cell can reengineer itself and reset its signaling. But again, all our efforts thus far have not popped out a single answer on the resistance nor have we seen that any one resistance factor can preclude responsiveness. So clearly, we're going to look at this more. The panels that we used have a great depth of data in them, and we'll be interrogating that additionally as time goes on.

Before we lose this topic, and I know it's one we've received many questions from you about, perhaps it would be good to get, first, Dr. Borghaei and then Dr. Wolf to comment on diagnostic testing. Their views about what you're doing in your institution and then your view of the community testing. Dr. Borghaei?

Sure. So thanks for the invitation and intro. So I think molecular testing has been part of what we do with patients at the time of diagnosis but metastatic non-small cell lung cancer. I think the -- most of the academic centers that I am aware of in the U.S. have access to a next-gen sequencing platform where that would have access anywhere from 50 to upwards of 300 mutations, amplifications, fusions and whatnot as needed for the management of these patients. In the community setting, more and more centers have access to either larger facilities that conduct this or central laboratories where next-gen sequencing is occurring. So when it comes to something like KRAS, at least from what I see in my practice and for the patients to come in to me as -- for a second opinion, most of the platforms that are being used, at least in the Northeast part of the U.S. where I live, include KRAS as part of the analysis. So I think that information is available to majority of physicians and patients who have molecular testing done. I think the liquid panels have the potential of becoming the great equalizers. A problem, but lung cancer has always been inadequate biopsies. The disconnect is between pathologists, molecular pathologists and clinicians and causing some delays in getting tissue. But I think liquid biopsy panels can be very helpful because they're controlled by the physician who are seeing the patients. And again, majority of the panels that are widely used in the U.S. do contain a KRAS as a marker. So I think we can always do better for molecular testing with our patients with lung cancer. But I think the information is already there.

Great. And Professor Wolf, your perhaps point of view from Europe.

Yes. In my country, in Germany, and this is very similar, at least in the Western European countries, there's a growing tendency also to do molecular diagnostics in large centers with -- and all these centers use next-generation sequencing platforms. So for instance, in Germany, about 70% of patients with advanced lung cancer are covered now by these NGS platforms. And all these NGS panels, of course, include the KRAS mutations.

Your next question from Yaron Werber of Cowen.

Congratulations on the data. So first, just you mentioned that about 20% of patients had brain mets that were all previously treated. Can you provide any insight on to whether you followed those mets over time and how they might have changed? And then my other question is just about the Phase II study you mentioned that's planned for first half in the highest unmet need, including STK11. Are you planning to include the KEAP1 patients in that, especially the ones that are KEAP1 mutant and STK11 wild type, just given that, that group kind of showed the lowest response? I know, obviously, since the small sample size, it would make sense to see what happens. But just curious if you are planning to not include them or include them.

Great. Thanks, Yaron. Both important questions. Greg, you've looked at all of this in detail. So perhaps if you can address the question regarding brain metastases. And regarding the first-line study, we're finalizing that as we have a deeper view of the biomarker data. But it will proceed along the lines that you've indicated, Yaron. Greg, you can add additional color.

Yes. So with regard to the brain mets, I mentioned about 20% of the patients had previously treated brain mets on the study. And in order to be eligible, those had to be not only treated but stable. They, of course, can count as measurable lesions in the RECIST measurement. But typically, when we're thinking about response criteria, again, there are different criteria used for brain mets. I bring that up because we actually are doing a dedicated study looking at patients in whom brain mets exist, and we'll be using the RANO criteria, for example, to look at those more closely. But what I can tell you is absolutely, those -- that represents measurable lesions in some patients. What we also did not see was a single pattern of progression in patients where, for example, the brain mets represented a majority of the progression events. So beyond that, can't say much more. And to be fair, when we measure brain lesions, we should be using a slightly different criteria. Good news is we're doing that moving forward. And we're, again, hopeful that this is something that we'll have more data on in the near future.

And our next question from Evan Seigerman of Crédit Suisse.

Congrats on the data. I know it's late for all of us. But maybe 2 questions. One, do you have any data to the durability of response in the patients who experienced the complete responses, how long were they in complete response? And then when looking at the landscape, how important do you believe the cardiotoxicity profile, or really lack thereof, is with sotorasib when physicians ultimately make decisions on how to treat patients? I guess, in other words, do you think the lack of any QT prolongation will be a significant competitive advantage for sotorasib?

Great. Thanks, Evan. And perhaps, I'll take the second half of that question and then ask Greg to comment on the first half. And we can also get our clinicians' point of view on QT prolongation. As Greg said, we've looked very intently across the more than 700 patients treated in the program now. QT prolongation is just not something we see. And of course, it's something -- if you have a preference, you don't want to see it and to deal with potential cardiac complications or drug-drug interactions if patients have underlying cardiac disease or on other medications. So is that a potential point of differentiation? Yes, I would say absolutely. Greg, perhaps I can ask you to address the other part of the question. And then we'll have -- ask our clinicians to -- and investigators to comment here as well.

Yes. So our median amount of follow-up for all the patients is 12 months. So again, the durability can -- has to be some fraction of some number near that. And you can see in the swim lanes, we have some just responders in general that are out at 11 months. My recollection is that all 3 of those CRs still are on study. I'll have to confirm that for you, though. Honestly, off the top of my head, I [ just don't remember them ].

Great. And Dr. Borghaei, perhaps just a clinical perspective on these questions.

I mean the first question I really can answer, as you just heard, patients who are having ongoing responses, they're having ongoing responses. I'm sure we're going to have additional updates later on at other meetings and sort of see how long these last. So for that, that's a clinical phenomena that we're going to have to follow. As far as QT is concerned, obviously, from a clinical point of view, the less side effect I have to deal with or the less problems I have to worry about, the better I like the drug and the more I'm likely to use it. QT prolongation has become a hot topic. Almost every drug on the planet can cause a little bit of a QT prolongation. So if you have drugs that are competing with each other or causing problems, then it makes it difficult to manage a patient. For instance, we use drugs that are known to cause QT prolongation, and every time I want to prescribe a simple antibiotic for my patients during the fall and winter here in the Northeast, I have to worry about the potential of exacerbating issues and be careful and do additional EKGs, monitoring and things like that. So obviously, not having to deal with QT prolongation would be a really good advantage as far as I'm concerned. So again, less side effects and less issues for us to worry about and the patients to worry about, the better we like the drug.

Great. Thanks. Professor Wolf, anything you'd like to add?

Yes. I can also add a comment first to the duration of response or disease control. So it's not necessarily the deepness of response will lead to a long disease control. Or for instance, I have a patient in this Phase II trial. He has never reached the formal PR. So tumor shrinkage was 15% to 20% only, but he is now in a very good condition since 1 year and 6 months. So I think we have to keep this in mind. And second, concerning the QT prolongation. Of course, that this is not a problem with sotorasib, it's definitely an advantage. And because we have a population of patients, 60 to 70 years median age and 90% [indiscernible] with a high cardiovascular risk profile, and I think, in particular, it will be an advantage if we start now with the combination therapies with TKIs.

Your next question from Jay Olson of Oppenheimer.

Congratulations on these results, and thank you for hosting this event. Now that you're developing a more definitive picture of sotorasib's clinical profile, can you talk about what level of incremental efficacy you're expecting from combinations? And do you have a target ORR or PFS level you're looking for in the pivotal combination studies? And then with the new biomarker analyses, are there any combination studies that you would prioritize over others?

Thanks, Jay, and perhaps I'll address that. It's hard to speak about targets for incremental efficacy outside of the specific clinical setting. That will vary depending on the indication and the line of therapy. I mean, typically, you're looking for a minimum of a 20% to 30% relative improvement, in general, in oncology. And you can think about that relative to both target overall response rates as well as progression-free survival. Although again, as Professor Wolf just pointed out, and he has a patient, we have a number of patients who actually aren't formal responders by RECIST criteria, but have remained on drug with clearly good disease control for a period of time. And so it's really the combination of all of these end points together that tells you whether you have an incremental advantage. In terms of preferential combinations, these have been selected based on preclinical work by us with our academic colleagues or in some instances, based on pairing with commonly used clinical regimens. In some cases, those are the same. And I think now we've amassed a fair amount of preclinical data, and the real test is running these combinations through the clinic as quickly as we can. And that's one reason we've opened the master protocol with a very flexible design, really intended to discern efficacy signals as quickly as possible and expand. So more on that, as we mentioned, over the course of the year. My favorite will be the ones that work the best. And we'll keep you apprised of that as we move along.

And your next question is from Umer Raffat with Evercore ISI.

There's 2 broad areas I wanted to focus on. First, as we look at the response rates you're reporting for STK11 as well as KEAP1 mutation, I find it -- and I try to -- and when I try to compare that versus the data from your competitor, Mirati, it looks like the point estimate of those response rates sort of all STK11 mutants and all KEAP1 mutants, it looks lower. And I'm trying to understand why that would be. And in that vein, could you also address for us if you expect these response rates to look different depending on where the samples were collected, plasma or tissue? And secondly, just as we come out of a lot of the lung data that you accumulated to date, what's your confidence sitting here today on being able to put up monotherapy activity, perhaps not as strong as lung, but monotherapy activity in colorectal?

Thanks. I'll start there and ask Greg to comment. So first of all, I would greatly caution everyone by making cross-trial comparisons as Greg, I think, pointed out in his presentation. Especially if you're looking at single-mutation response data, it's actually the overall co-mutational burden and specific co-mutations that is likely important in directing response. We're quite confident in the activity of this drug in the range of mutations. As you heard described today, we feel it will be as good as anything in the field. One thing we are looking at is the relation of those mutations in terms of time when tissue was collected. Some of these mutations occur relatively early in tumor genesis and are fixed. Others can occur during tumor evolution, and that will be a focus as we go forward with the biomarker program. Greg, let me ask you if you want to add a little color to that. And then we can talk about the monotherapy colorectal data as well.

Yes. No, I think the question was asked whether or not any factors about location of the sample could also add heterogeneity. The challenge, of course, is interpreting anyone, even if it's just your own single-agent data set, let alone comparing to, is that heterogeneity is the enemy here. And where the samples come from, actually, that has 2 flavors, right? There's the are we looking in the plasma versus the tumor? And the short answer is we don't know. If those would result in different results, ours was a great majority from tissue, not from plasma assays. And the other geographic issue is that we had a very global study. And certainly, the circuitry of the patients may be different in different populations. We saw, of course, in the New England Journal commentary that there are epidemiologic differences across the continents of these mutations. And we could only expect that if you start clicking down into second and third co-mutations that we might be seeing a different mix. So again, just coming back to what Dave said, I think it's very difficult to intercompare small data sets in complex biology. And from that matter, again, we're pleased to see that we have responses across all the groups right now, but we'll be diving into this more deeply as time goes on.

Yes. And just to finish up. In regards to the monotherapy in colorectal cancer, we fully enrolled Phase II study. We expect those data to be presented in late spring or over the summer as the efficacy data mature and are ready. And I think that will really be the determination as to whether monotherapy or combination therapy, or both, are viable options going forward in colorectal cancer.

And your next question from Terence Flynn of Goldman Sachs.

I just had one follow-up on the Phase II frontline lung trial that you mentioned. Was just wondering if there's a possibility that, that could be registration-enabling as well. And then on the opportunity in China, congrats on the breakthrough designation. Can you just remind us of the economics there? And then maybe help frame for us the size of the potential opportunity.

Yes. In terms of the Phase II frontline trial, I don't want to speculate on potential regulatory outcomes. That, of course, is dependent on the data and having the appropriate conversations with regulators. There's intense interest in that study by investigators and us. And many of these patients for various reasons are not candidates or unable to receive other frontline therapies and may drive significant benefit. I'll ask Arvind, if you want to join the conversation in a second here to talk about the question regarding economics in China. But I will say from a regulatory point of view, we were quite happy to receive breakthrough therapy designation, which has a very analogous meaning according to Chinese regulations as it does in the United States. Arvind, do you want to add just a quick word?

Yes. Terence, I think you're well aware of the fact that it's a joint development and commercial collaboration. And I think all we can say at this stage is that the collaboration is going very well.

And your next question is from Michael Yee of Jefferies.

Two questions that might be related. You announced in the MEK combination that you had expanded. Can you just remind us how that works? Is it a dose escalation? It's deemed to be safe and you expanded. And then how does -- how big is that expansion versus, say, expanding it to a pivotal? Is it like a go, no go? Maybe just kind of explain how that works. And then related to that, since you have so many combination arms going on, some of this was asked before, but is there 1 or 2 most common mutations that either evolve that drive the resistance or the progression of the tumor after sotorasib? Or maybe just think about -- or explain which 1 or 2 do you think are most important in the tumor oncogenesis.

Thanks, Mike. Yes, regarding the MEK combination. So as we announced, we cleared safety at full doses. We're expanding -- this expands into a robust equivalent to a Phase II basically sample size. So I think it will give us a very good insight into efficacy. We should point out that many -- some of these combination studies will, of course, be targeted at specific malignancies based on the combination partners. And again, the biology will differ in those tumors. Hence, it's likely that mechanisms of resistance will differ as well. This might be good to actually get a perspective from our academic colleagues on mechanisms of resistance and what they're looking for as the field evolves. So Professor Borghaei, perhaps we can start with you.

Jürgen, do you want to start? So the question was again the MEK -- yes, go ahead.

I can? Okay. I think we definitely are in the beginning of understanding resistance to KRAS inhibitors, and we should discriminate primary resistance and secondary resistance after response. And what has been learned in preclinical experiments that one reason for the primary resistance means not getting -- or not seeing a response at all, is obviously that not all KRAS-mutated tumors are addicted to KRAS. I mean there seems to be a genetic heterogeneity, and this might explain the early response rate of the nonresponders. And what has also been seen in preclinical experiments is the downstream of KRAS. There is often a kind of a rebound of the ERK-mediated downstream signaling cascade, a kind of a rebound ERK activation, which is a good rationale for adding a MEK inhibitor here as a combination partner. And also upstream of KRAS, so what has been seen preclinically was an activation of the receptor tyrosine kinase SHIP2 access, which is a resonator for adding a receptor tyrosine kinase inhibitors and SHIP2 inhibitors. All this is done in this Phase II trial. And I think there are still many more preclinical rationales. For instance, influence of KRAS and KRAS inhibition on the tumor microenvironment with regard to susceptibility to immunotherapy because there are still a lot of preclinical rationales and because there might be a discrepancy between the preclinical observations and the clinical observations. I think the only way is to do this kind of trial, where you test a lot of hypothesis in such multi-cohort trials with a restricted number of patients in a relatively short time.

I agree that a trial like this is absolutely needed. I think you need to determine what the real clinical reason for resistant is to this particular pathway with the drugs that we have available. So unless you do a study or you are testing multiple different co-mutations and as a potential resistant pathway, you're not going to find the answer. So the SHIP2 that you heard of and the MEK inhibitors, I think the more -- are the 2 that have a lot of data in support of them. So again, the effort in terms of defining what the resistant pathways are, I think, is what attracts me to these kinds of protocols. And yes, I mean, there's still a lot that we don't know about this pathway. There is again, as you heard a lot of preclinical data, but sometimes don't necessarily translate into the clinical arena. So having access to samples and investigating it in the context of an organized trial is what we need to find the answers to some of the questions you're asking.

And your next question is from Geoff Meacham of Bank of America.

I just had a couple. When you look at your best responders, were there any common features in terms of, for example, line of therapy or prior best responses or even perhaps lower tumor mutation burden? And then the second question is does the lower activity in high PD-1 expressors, how does that change your thinking on the strategy when you think about combo I/O therapies down the road?

Yes, Geoff. Perhaps I'll take those questions and then ask Greg to comment very briefly. First of all, we see responders across all subsets. So it's not an absolute and that tells you that any -- right now, the single best predictor of the potential to respond is having the G12C mutation. The PD-1 expression is something that we continue to look at. There may be self-selection going on there in terms of how these patients progress. This is a question Greg and the team have looked at a fair amount, and so I'll ask him to comment as well.

Yes. Thanks, Dave. I think with regard to the high PD-L1 expressors, you have to ask yourself, who were these high PD-L1 expressors that came on to our study for whom PD-1 had failed them. And from that standpoint, there might be, again, a very specific biology for that group of under 10 patients, again, who was on this particular study. But I'll say that, again, we published back at ESMO a look at some of the baseline characteristics. We've been looking at them now. And with regard to predisposing characteristics for the best responders, there is no one-size-fits-all. We've looked at mutant allele frequency. We've looked at TMB. We've looked at a variety of co-mutations through not only at baseline, but also because we've been looking in the plasma at what can develop. And while we're going to need to interrogate that data much more, we've got 46 responders now as opposed to a smaller number in Phase I. We will continue to look for those answers, but there is no one story here in terms of being able to predict nonresponsiveness. I guess, the good news is that we're also not finding that there are anything that precludes response.

And your next question from Matthew Harrison of Morgan Stanley.

I wanted to ask, you obviously have a lot of combination studies ongoing at this point. And as you've collected more patient samples, you have a lot more data on biomarkers, and you presented, obviously, some of those here around responses in different biomarkers. I'm just wondering, do you have any view on which combinations you think are the most promising at this point? And any insight on to some of the biomarkers that may be driving that thought?

Yes. I think you heard me speak to this earlier. So maybe we can ask Dr. Borghaei and Dr. Wolf to talk about the combinations and which ones they are most interested in. Dr. Borghaei, I think you had mentioned SHIP2 as well as MEK. Anything that you would like to add to that? And then we'll go to Dr. Wolf after that.

Well, no. I mean, again, I would add that the 2 that I mentioned are the ones that we have some data for. Again, as you heard from Professor Wolf that we had some pretty clinical data and now we have had some clinical data indicating that those pathways could be relevant. I think from a clinical point of view, and again, what I look at is that there is a certain degree of unknown about some of the biology here that we're trying to consider and investigate. And I think again, I can give Amgen a lot of credit for going into a study like that, trying to investigate multiple possible mechanisms where for different subpopulations, a particular combination could be useful and effective. So trying to predict which one of these is going to be exactly the right combination and all that, I think it's going to be a little bit challenging. First of all, the data is limited. Second, we've never had KRAS inhibitors to this extent with this kind of activity that we could interrogate clinical samples with actively and effectively. So we're entering in a new area. And again, I don't think I feel confident to say which one of the combinations I am most hopeful for. But the SHIP, the MEK, to some extent, EGFR pathway, I think, can be relevant when it comes to KRAS mutations. But then what happens if you bring all of these co-mutations such as STK11, KEAP1 into it? So I think it just becomes difficult to predict exactly where we're going to go when it comes to combination. But clearly, this is the direction we need to go to.

Great. Thanks. Anything you would like to add, Professor Wolf?

No. And only very briefly because I have already discussed a little bit extensively previously the variety of preclinical hypothesis. And in the moment, I would not dare to define the most promising approach.

Next question from Geoffrey Porges of SVB.

And Arvind, we'd all be out in crowded bars and nightclubs were it not for your call on a Friday night.

Sorry about that, Geoff.

We don't have anything else to do. So I just wanted to ask about the STK11 population. First thing is, could you comment on the proportion of non-small cell lung cancer or G12C patients that have mutated STK11? Is it consistent with the 35% or so that you see in this population? Secondly, could you talk a little bit about whether there was any difference in the duration of response for the STK11 mutated? So it just looks like there's a trend towards slightly better response. And then lastly, I think you said that the STK11, you were only enrolling STK11s in the pivotal trial. Is that correct? And would that ultimately be your final label?

Yes. Again, Greg, you've done a lot of work here. Let me have you handle this question.

Yes. So with regard to the Phase II study that we remarked on, that was an example of a biologic selector that we're potentially using. As Dave mentioned, we're working through a lot of the details of that protocol as we speak. And so there'll be more to talk about in the future there. With regard to the STK11 population as a subset of the KRAS G12C. We've been very fortunate to work with a variety of partners who have looked again at the epidemiology here. And yes, that 35% number is pretty consistent with the G12C population. The G12C population is refined for these STK11 compared to all-comers. But that number seems to be fairly consistent across broader looks at the molecular epidemiology of the G12C population, about 1/3.

And then the duration of response in those patients, any difference?

Yes, thanks for the reminder. Because of the numbers that we're talking about, there's no clear pattern whether these patients differ at this point. But it's something that we're keenly interested in, and we're going to continue to follow up.

From Mohit Bansal of Citigroup.

So just wondering, if you have any hypothesis at this point on why sotorasib is working better in PD-L1-low patient. And the related question is, maybe if the clinicians can help us understand that, if you see potential for usage as a monotherapy in first-line patients with PD-L1-negative patients because 48% ORR that you are seeing in second-line plus setting, that is pretty good. So maybe it could be used as a monotherapy even in first-line and maybe better responses could be seen in those patients.

Thanks, Mohit. And I'll ask Greg to also address this question as he and the team have taken quite a close look at the notion of PD-L1 expression.

Yes. The PD-L1 expression is a story that's still being written. I think the truth is unlike with PD-1 therapies, it's difficult quite to understand the interaction, and you asked about a hypothesis about what may be going on in these patients. And really, that's where I would just give a moment of caution that particularly those small numbers in the PD-L1 high, and again, a recognition that those are patients who have sadly failed or the therapies have failed them in frontline, they flowed through, that may again be a different biology than just a de novo newly diagnosed PD-L1 high. So we're going to continue to look at that quite closely. But again, I think that the error bars overlap, and we want to be cautious. With regard to monotherapy line, obviously, there are good therapies available to -- we live in a world now with PD-1 therapies, and those have changed the landscape. What we're looking for are groups that are more poorly served by available therapies. And again, as we drill into the opportunity for monotherapy cohorts with biomarker selection, that will be our goal, patients in whom we could offer something again that compares nicely to other available therapies.

Great. And perhaps I can ask, again, Dr. Wolf first and then Dr. Borghaei to comment on what you might be looking for in the first-line setting, for example. Dr. Wolf?

So sorry, you mean as a combination [indiscernible] or as a clinical end to end?

As a monotherapy. I'm sorry, as monotherapy. Thank you for the clarification.

You mean efficacy? I'm really sorry that I did not completely get the question. I think in the first-line setting, of course, our expectations depends on the competitor. And of course, the threshold is around an overall response rate of 50%, which is a little bit lower in the PD-L1 high. With the pembrolizumab monotherapy, a little bit higher. And I think definitely this should be higher with the sotorasib. And I think at the moment, it's hard to say whether this can be reached with monotherapy or whether we need the combination because we do not have our final data, which allow us a comparison of the efficacy in the first-line versus the relapsed situation. And remember, the situation with many other of the new TKIs, for instance, [indiscernible] and the MET exon 14, here there is clearly better, a substantially better efficacy of these TKIs in the first-line setting.

Right, thank you. Dr. Borghaei, anything you would like to add to that?

No, I would agree with my colleagues. Again, several times it's been brought up. I don't think we should put a lot of emphasis on the PD-L1 story. These are limited samples. PD-L1 staining by itself is extremely heterogeneous. So misinterpreting the data based on 9 to 14 samples here and there is easy to do. So I think we should just wait for additional data to build. But I agree with what Professor Wolf said, to move this into the frontline, we need to see a little bit more of the durability and PFS and all of that before we can say in any one particular subpopulation, the drug can be moving into the frontline.

The next question from Kennen MacKay.

We haven't seen spider plots, so I'm just wondering if you could comment on the time to CR, complete response, and if these responses were something that deepened over time from a PR. And sort of along those lines, it looks like a fourth patient appeared to have 100% tumor decrease that wasn't deemed a CR. Just wondering if you could add some color to that.

Thanks, Kennen. I'll let Greg address that question briefly.

Yes. The CRs, as you could expect, the patients who reached CR did so fairly rapidly. I believe it was within 3 cycles for all the patients. There is a fourth patient that has 100% shrinkage of their measurable disease. But of course, in RECIST, if you have nonmeasurable lesions that are still present, we don't call that a CR even if there's 100% shrinkage. I believe there was a lymph node, which are notoriously difficult to measure in this patient. And so again, we're really pleased to see that kind of decrease in the girth of their tumor. But we want to be very upfront and honest, and we are in our presentations about what we call a CR and what we don't.

And your next question comes from the line of Alethia Young of Cantor.

I guess, you guys have done an incredible job enrolling this trial. And I guess, I'm curious if you think the prevalence is slightly larger than the 13% that you quote. And if you have any kind of thoughts on that as you head into a launch cycle.

Thanks, Alethia. No, we think that's pretty accurate number based on work we've done with fairly large databases and our experience in screening for this trial. I think the rapid accrual relates to the potential promise of the drug, the lack of alternative therapies for these patients. And that's, of course, what makes us really eager to try to make this available for patients as quickly as possible. But I think the epidemiology is -- that's reported is probably square on, at least based on our experience. Thank you.

The next question from Robyn Karnauskas.

And great questions by Alethia, Evan and Mohit, by the way. I have a bigger picture question. I love your slide on PD-1 levels and all different mutations. But the reality is when you launch and Mirati launches, you're going to be compared to response rate and the data that you have. So curious how you're thinking about, number one, [ if you don't ], how you think about uptake and what population you think will be a stronger -- where you think you'll have a strong uptick in. And how you think about, given that you have a very broad platform outside of the population combinations you're having, how do you get the highest uptake in a PD-1-positive population or PD-1-negative population. So I mean, it's really a big picture strategy question, and I love the details, but we all know that in the real world, those are science-y questions, and doctors look beyond that. So help me understand how to model this revenue stream a little bit.

Yes, sure. It's a complicated question. I think the simplest way to conceive of this is that beyond first line for patients with the G12C mutation, standard therapy does not provide a particularly attractive option right now. And so I think regardless of PD-L1 expression levels, our belief is that this should provide an extremely attractive treatment option. Greg and the team, again, you've been doing a lot of work here, perhaps very briefly, if you can add your thoughts, and then we'll go to one final question after that.

Yes. I would just add that the details may be science-y, but doctors are scientists. I think they care about central reviews, they care about high-quality data sets. 11 countries are represented in this program. And so in that regard, they care about the fact that all their patients need to be described. Progression-free survival is a very important number to them. And so again, when you take it as a whole and you balance that with the safety profile that we've seen, we're seeing really low rates of GI toxicity. The nausea, vomiting, fatigue, these aren't things that patients want to have. And so this combination of robust, high-quality data from a data set, again, that raises the bar in terms of the central review, with the tolerability profile, we feel really good about what we have in a once-daily therapy.

And Dave, before you go to the last question, I just wanted to clarify on Terence's question on the commercial rights. That is a part of our broader commercial collaboration. BeiGene is going to commercialize in China. And what we have with them is a 50-50 profit share, and rights to most products will revert back to Amgen in 5 to 7 years. But BeiGene does not have the right to keep sotorasib in China. So I just wanted to clarify then. So with that, why don't we go to the last question?

David, if I could just clarify, we did confirm that those 3 complete responses are still on therapy. So we'll save an e-mail for someone.

From Michael Schmidt of Guggenheim.

Nice to see the data really here in KRAS G12C-positive patients. I'm curious, maybe bigger picture here, what programs Amgen has in place in R&D to potentially addressing other KRAS mutations. And if not, whether Amgen has any plans to expand into the broader market opportunity in this category.

Thanks, Michael. Yes, it's an important question. Of course, there are G12D and other mutations that are important in various malignancies. That's something we're quite interested in, and we've got preclinical work ongoing. And as that progresses, I'll be talking about that over time. And we will certainly keep you updated and provide guidance. With that, I think we were about 10 minutes over here on a Friday night. Let me just make a few closing comments and thank everyone again for joining us. We're absolutely thrilled with the data. We're thrilled with the clinical progress that we've got. I personally, as a medical oncologist, am very proud to be involved in the introduction, we hope, of the first KRAS G12C inhibitor, something that I read about as a fellow 30 years ago and now being able to see come to fruition. We'll have a little more to say on our earnings call in just a few days, and we look forward to talking to many of you then. Between now and then, we wish you all a good weekend. And again, we want to extend our thanks to Professors Borghaei and Wolf for joining us at off-hours. Thanks, everyone, and have a good weekend.

Thank you.

Good night, everybody.

Thank you. Bye.

Ladies and gentlemen, this concludes today's conference. Thank you, everyone. You may now disconnect.