Amgen Inc. (AMGN) Earnings Call Transcript & Summary

Good afternoon. My name is Annie, and I will be your conference facilitator today for Amgen's conference call in conjunction with the ESMO 2020 Virtual Congress. [Operator Instructions] I would now like to introduce Arvind Sood, Vice President of Investor Relations. Mr. Sood, you may now begin.

Okay. Great. Thank you, Annie, and good afternoon, everybody. It's still morning for those of us on the West Coast. My apologies for the slightly delayed start this morning. Delighted you could join us today, and hopefully, you can join us again tomorrow. We presented some important data this morning in the virtual scientific session of ESMO on sotorasib, also known as AMG 510, which, as you know, is a first-in-class inhibitor of KRAS G12C, and this is advancing rapidly through clinical development. In addition, we are also presenting data tomorrow on our extended half-life BiTE AMG 160 that targets PSMA or prostate-specific membrane antigen. And in conjunction with that presentation, we'll conduct another call tomorrow at 4:00 p.m. Eastern time to review that data. To further review the sotorasib data presented this morning, I'm joined today by Dr. David Reese, our Executive Vice President of Research and Development; and Dr. Greg Friberg, our Vice President of Global Oncology Development. Also joining us for the Q&A session is Dr. David Hong, who presented the data from the Code 100 (sic) [ CodeBreaK 100 ] study this morning. Dr. Hong is the Deputy Chair in the Department of Investigational Cancer Therapeutics at MD Anderson in Houston. So with that, I would like to turn the call over to Dave Reese. Dave?

Well, hello, everyone. Thanks, Arvind, and thank you, everyone, for joining us here on a Sunday. If you'll turn to Slide 3 in the deck that we have posted. As Arvind mentioned, what we'll do today in terms of agenda is that I'll make a few brief introductory remarks. Greg will then review the sotorasib data that were presented earlier today and published more fulsomely in the New England Journal of Medicine simultaneously. And then, we will have a Q&A, for which we'll be also joined by Dr. David Hong from MD Anderson. If you turn to Slide 5, I wanted to remind you that in addition to today's presentation, there was also a poster. The third abstract listed on this slide, that is an extensive real-world data set on the prognosis of patients with non-small cell lung cancer and the G12C mutation. That data set, I think, is quite important in putting everything that you'll hear today into context. It shows that these patients have a prognosis that appears to be the same as the entire non-small cell lung cancer population. It is certainly no better. Importantly, we also had data on duration of response and progression-free survival today as well as our first look at biomarker data. If you advance to Slide 6, I would like to say that we remain very enthusiastic about the sotorasib program, both in terms of the efficacy and tolerability that we've seen so far. And as I mentioned, we now have promising data in terms of duration of response and progression-free survival. We have now several hundred patients who have been enrolled in all clinical studies and are moving forward rapidly with the broad-based development programs. On Slide 7, we list for you that development program, including both monotherapy studies in lung cancer, colorectal cancer and other solid tumors bearing the G12C mutation as well as a suite of combination studies. There are now 7 different arms of a master protocol for combination therapies that have been opened. In addition, we anticipate that 3 or 4 more arms with new combinations will be opened in the near future. So we're moving forward quite rapidly. Very pleased with what we've seen so far. And we'd now like to review the data presented this morning and published in the New England Journal of Medicine. And for that, I'll turn things over to Greg. Greg?

Thanks, Dave. I want to point you to Slide 9, just to begin, and remind you that the data that we're going to be reviewing today was presented by Dr. Hong in the early morning hours here in the U.S. this morning. And this represents an update on our non-small cell lung cancer patients from the Phase I study, both the escalation and the expansion part of the Phase I. We were fortunate to have a simultaneous publication in the New England Journal of Medicine, and that publication involved the entire cohort, including these patients, but also patients who had colorectal cancer and other tumors. And the total number of patients there was 129. We move on to Slide 10. A quick review of the pathway. This audience certainly doesn't need much details here, but I will remind you that the non-small cell lung cancer patients that we're going to review today represent the largest tranche of patients who have this G12C KRAS mutation. About 1 in 8 non-small cell lung cancer patients have that at 13%. If we move on to Slide 11, we see the design of the Phase I study that Dr. Hong presented. We're calling this the CodeBreaK 100 study. As a reminder, this data is from the Phase I portion of the study and not the Phase II portion that we'll be revealing data before the end of the year. This is a fairly standard Phase I study. One of the commonalities, of course, across the different tumor types that were enrolled were that they all had the G12C mutation, as determined in their tumor biopsies. Next slide, Slide 12, jumps into the disposition of the patients who were treated, who have non-small cell lung cancer on the study. And the take-home message here is that there's more patients, we have 59 total patients now compared to the last time that we were speaking. And we have more time of follow-up, almost a year of follow-up, 11.7 months. And of course, with that additional follow-up, that allows us to, in a credible way, start describing some time-based endpoints that we'll dive into a little bit later on. From a demographic standpoint, I just want to highlight that all of these patients had, had prior chemotherapy that had failed them. They have progressed through that, and about 80% to 90% of them had, had prior PDL-1, PD-1 therapy. So the take-home message, again, these patients were under median of third- or fourth-line therapy. And of course, as with the demographics of this disease, most of them were prior smokers. If we move on to Slide 13, you see a summary of the treatment-emergent adverse events. There were no dose-limiting toxicities that were seen in these 59 patients. There were no fatal adverse events that were seen. And of note, there was only one treatment-related adverse event that led to discontinuation. It was 1 out of the 59. It was actually 1 out of 129 if you look at the New England Journal article. So certainly, the monotherapy has demonstrated a relatively safe safety profile. Moving on to Slide 14. And we're very fortunate today, we'll dive into some of the treatment-related adverse event terms. We have Dr. Hong on the line as well who can provide additional color. And really, the terms that you see on this slide are quite similar to the presentations we've seen previously, diarrhea, LFT abnormalities, a little bit of anemia. I think Dr. LoRusso summarized it quite nicely in her editorial in the New England Journal. This is quite a favorable safety profile and has the potential for minimal overlapping toxicity with other agents. If we move on to Slide 15, we start diving into the efficacy that's been seen in the non-small cell lung cancer patients. And what you see here is that there are 19 total partial responses confirmed as according to RECIST across the 59 patients here. If you only look at the cohort that was treated with the highest dose level, 960 milligrams, it's a little over half of those patients, 34 of them, and we clocked in with a confirmed partial response rate of 35%. And of course, as we've seen previously, that doesn't seem to describe the entirety of the activity of the drug. There was a disease control rate above 90% in this cohort as well, with a significant number of patients demonstrating stable disease. If we move on to Slide 16, we're presenting an update to the waterfall plot that shows the best RECIST response across the patients. 57 of the 59 patients had the appropriate scans, both pre and post, to be able to put them on this grid. And one of the factors that really jumps off the page here is that this looks remarkably similar to the prior presentations in terms of the shape of the waterfall. Again, we see that 32%, 35% confirmed response rate, but you also see 71% of patients who have some degree of shrinkage, and again, suggesting that perhaps the overall response rate number isn't describing the totality of the benefit. If we move on to Slide 17, we have an update of the swim lanes for you, including all those patients, those 19 responders that are in there. And looking at those 19 responders, there's a couple of trends I just want to point out. One is that 10 of 19 of these patients still remain in response at the time of the data analysis. Second would be that the responses are usually occurring fairly early in a patient's treatment course. Though that's not universal, there are some patients who had responses later. And of course, you see now we have a significant amount of follow-up, and that allows us to have enough data to start looking at Kaplan-Meier estimates. And below, you see again exactly that. You see a durability of response for those 19 patients that's estimated to be 10.9 months. So something that we feel quite encouraged by. On the right side of the slide, we now have the first Kaplan-Meier curve for progression-free survival. And as we noted, responses are an important benchmark to see in patients, but progression-free survival is measuring the effect for all the patients who come on to the study. So if you take all of those 59 patients and you estimate out what their time to be progression-free and alive is, you see a median of 6.3 months, again a number that we're quite encouraged by considering that these patients are on their third and fourth lines of therapy. Just to put that in context, standard docetaxel chemotherapy, depending on the patient population, probably has a progression-free survival in the order of 2 to 3 months across a similarly matched group, and that's not even accounting for the KRAS status. So moving on to Slide 18, we presented a patient case. This is an example of one of the best responders on the study. And while it's anecdotal, I think it has a few important points that are worth mentioning. First and foremost, this was a heavily pretreated patient again. Had a nice response, even though this patient did have an STK11 co-mutation, a very poor prognosis marker that was identified in their plasma. So that mutation doesn't seem to preclude the possibility of response. And this was also a patient with brain metastases that did shrink on scans on the study. This is obviously something that we're keenly interested in, and we're enrolling a cohort of similar patients and want to be able to describe this more broadly as time goes on. Moving on to Slide 19. We dig into a portion of the presentation that was not covered in the New England Journal article, and this is some of the first biomarker data that we've looked at. Clearly, we want to understand the factors that might predict responsiveness or sensitivity in these patients. So we took the patients for whom we have that data, and we did just that. We looked at responders and nonresponders, for that matter, and frankly, progressors, and ask the question, do we see differences in these biologic factors across the patients? And the answer was no. It does appear that sotorasib has clinical activity across a range of the G12C patients regardless of mutational allele frequency in tumor tissue, regardless of PDL-1 status as determined by the pharmDx assay in subset of patients and irregardless of tumor mutational burden as determined by a serum assay. Of course, these are early days, but it's a pertinent negative that's giving us insights that these are not eureka moments in terms of pointing out factors for sensitivity and resistance. We'll continue to look at these as time goes on. If we move to Slide 20, we are showing data for co-mutations on the patients. This is looking at their tissue, again, with the QIAGEN comprehensive cancer panel. So over 250 hotspot mutations that are looked at here, and we presented some of those that we feel are the most important here. Of note, 33 patients had tissue available for this analysis. And what you see here are a couple of factors. The take-home message really is that the presence of actionable co-mutations were quite rare. For example, you see 4 patients with EGFR mutations there. Details are important. None of these were T790M. None of them are actually even actionable EGFR mutations. The co-mutation level, whether it's EGFR or looking at epidemiologic factors over studies of ALK and other potentially actionable lesions, are rare in these patients. We also looked at co-mutations that can predict poor prognosis. And you see KEAP1 listed here. It does not appear to preclude responsiveness, the presence of KEAP1. The question came up with Dr. Lindsay earlier today, why are there no STK11 data reported here? The answer there is simple. We looked at it, and in the tissue, we actually did not find any cases of STK11, probably reflecting the poor prognosis of those patients and the fact that this was a Phase I study. They would have had to survive long enough to make it onto the study. Again, I think the take-home message here is that none of these co-mutations appear to predict sensitivity nor do they seem to predict inability to respond. We'll continue to look quite deeply at this, but this biology appears to be one that is not as simple as one might hope. So moving on to the final slide, Slide 21. We have a summary. Again, the New England Journal paper is out as well looking at the totality of the patients. And the safety profile was reported there as having a very encouraging profile. That, in addition to the activity that was seen, the 35% durable objective response rate in the 960-milligram cohort, the median progression-free survival in all patients of over 6 months and of course, the durability in responders estimated to be over 10 months. We see these as quite encouraging. And the 960-milligram dose, of course, that was identified was brought into the single-arm Phase II study in non-small cell lung cancer that we completed enrollment to back in December. So with that, I'm going to wrap up this presentation and hand it back to Dave, and we'll move on to the question and answer.

Well, thanks, Greg. And again, thank you, everyone. So Annie, why don't we go ahead and open up the line for questions.

[Operator Instructions] Our first question comes from the line of Evan Seigerman from Crédit Suisse.

Congrats on the update. So two questions. One, can you help us better understand why some patients progress so quickly after initiation of treatment, especially at the high dose? Is anything unique in these patients? And also, do you have any color on the median duration of therapy or duration of response in the 960-milligram cohort? Is it similar to that 10.9 months in the overall population?

Yes. Thanks, Evan. Those are good questions. We have looked at that. It's the real -- the 960-milligrams in terms of follow-up suggest that Kaplan-Meier estimates. But we certainly would have no reason to believe that, that would be inferior to the overall population. Let me ask Greg if he wants to provide a little more color there, but I think the answer is fairly simple.

Yes. The data just hasn't reached maturity yet, but we have no inclinations or reason to believe that there's something dramatically different in that group. With regard to the other question and potential for progression, we have more patients treated at the 960 dose. If you look at the waterfall plot that's not only in this presentation or the New England Journal, you'll see that there were a handful of patients who had tumor shrinkage who then didn't go on to be confirmed. There is no one story to describe that. One answer is sometimes it's a math issue, and patients have a very -- close to 30%, and the next will be right above or below the line. Second would be that we have seen a couple of patients with some heterogeneous tumor responses. And we have seen a small number, but a couple of patients who had tumor shrinkage followed on the next scan by growth. We think it's a rare phenomenon, but it's something that we're going to follow very closely. And of course, as we learn more about the biology of the disease here, we'll learn how best to deploy this drug.

Our next question comes from the line of Geoff Meacham from Bank of America.

It's Aspen on for Geoff. Just a couple of quick ones. I guess, first off, are there any notable differences in the studies here, the Phase I data, and maybe the baseline characteristics versus the Phase II study that's coming for the rest of the year that then might drive some different results? And then has FDA communicated any sort of thresholds to you for the Phase II study? And I guess how does -- how the data today kind of inform how you're thinking about that?

Yes. Let me address the second question, and then I'll ask Greg to talk about the outpatient characteristics in the Phase I versus Phase II, which are quite similar, in general. We don't comment on regulatory actions. I think one way to think about this is if we repeated in the Phase II data that were similar to what we've observed here in the Phase I trial, we'd be quite happy with those results, both in terms of response rate, also particularly in terms of duration of response and progression-free survival and the tolerability. So all of those, of course, are factors that regulatory authorities look at when affecting the benefit/risk of a new therapy. And like I said, if we can -- if we're in the ballpark of what we've seen in Phase I, we would be pleased with those results. Greg, do you want to speak to some of the nuances of the patient populations?

Sure. The short answer is that the populations are essentially the same. Of course, when we enroll to a Phase II, we have additional sites, and the geographic footprint might be a slight smaller mix, though all patients are being centrally confirmed with the QIAGEN G12C assay for this particular study. I would just add that there are some nuances in the entry criteria about prior therapies. But in practical purposes, we think that, that will result in an incredibly similar patient population where a great majority of the patients will have had prior PD-1 therapies, and everybody will have prior chemotherapy.

Great. Thanks. And before we move on, and we've got David Hong on the line, it might be good to have him just weigh in briefly in terms of his assessment given his vast experience on the efficacy data and tolerability profile. David?

Can you guys hear me?

Yes.

Yes.

Okay. So yes. So I think, number one, I'm more of an early-drug-at-the-moment Phase I guy, and my bread and butter is trying to identify toxicities in these new molecules. And this drug is incredibly well-tolerated. I don't know what else to say. The vast majority of these patients have really minimal adverse events of Grade 1 or 2. Even our patients who've had Grade 3-elevated LFTs, the vast major of these patients, if you stop the drug or dose-reduce them, the liver function tests return back to Grade 1 or lower. So I'm really pleased by, one, as an early phase drug-developing investigator, that this drug has been not only efficacious but has been well-tolerated. And I think that bodes well in the context that we are moving very quickly towards combination studies, which I think will test the hypothesis that this may actually, in combination, even more efficacious in the existing non-small cell lung cancer patient population and/or -- and possibly beneficial in the colorectal cancer population. And I don't necessarily focus on GI, but I do have an interest in GI. And I do think, just my best guess here, just -- is that this will have much more benefit in the context of the combination of chemo plus EGFR inhibitor with 510 in colorectal than alone. So I'm excited about that. Again, from an efficacy standpoint, I do think that the current population is -- does not -- is -- does not really show the potential of what this drug may be in either the second-line setting, as in CodeBreaK 200, and maybe even in the first-line setting. As you all know, Phase I populations tend to be very heavily pretreated. These patients often have much more higher tumor burden. And as Greg alluded to, the heterogeneity, oftentimes, these patients will have not only KRAS G12C clones, when they may have other clones within their tumor microenvironment that may lead them to faster progression, et cetera. So I'm hopeful. Obviously, this is still preliminary, but I'm hopeful that as we move up the line of prior lines of therapy, that we will see more efficacy.

We have another question from the line of Matthew Harrison from Morgan Stanley.

I guess -- I sort of have a broad question, which I know may be difficult to answer. But I think one of the questions people are all asking is, how do you think about this as a monotherapy profile if you're able to achieve something similar in the Phase II study from a registrable profile? Do you think this is enough? Or do you think you need to see more efficacy to be able to have something that's registrable?

Yes. Thanks, Matt. And as always, we don't speak to specific regulatory conversations. But again, I think if we replicated these results in the Phase II, we would certainly be very interested in moving forward with the drug from an approvability standpoint after the appropriate conversations with regulatory authorities. And again, it will be the entire mix of data not only the response rate but duration, progression-free survival and confirming the tolerability data that we've seen before. But again, I think as Dr. Hong just pointed out, for this patient population, it's still an overview of fairly impressive set of results.

And our next question comes from the line of Yaron Werber from Cowen and Company.

If you don't mind, I have a couple. And maybe just to clarify, maybe David or Dr. Hong. I mean when we look at the 960 dose, 5 out of the 12 responders are still ongoing and very durable. So when you said that the overall MDOR is not -- you don't think it's substantially different from the overall cohort, I guess my real question is, why -- one would imagine it should be higher than 10.9 months as the data matures. And then second question, just again maybe to Dr. Hong's question or comments about -- in CodeBreaK 200, the baseline characteristic would be earlier. Any thoughts on the Phase II characteristics and enrollment? In this case, about 60% of patients had 2 or more prior lines of therapy. Would that be the same in the ongoing Phase II?

Yes. Let me actually ask Greg to discuss the issue of the 960-milligram dose and estimates of duration. Again, for your -- there's, of course, no reason for us to believe that, that should be any different or shorter than the overall population. And as you point out, some -- a number of those patients are -- remain on therapy. It's early days and, therefore, statistically hard or not possible to make a specific estimate of duration of response in that population. Greg, anything you would like to add?

Yes, Dave. I would just add that we have left data for a subset of those groups, the group that enrolled later on with the study. And so probably it was more fair to say it should be no worse than the overall. But I wouldn't overread that we said not significantly different. It's less mature, and we're being conservative in not wanting to overstate what we know and what we don't know based on space and time.

So it sounds like that arm is still enrolling. Can you confirm that?

From the Phase I perspective, the 129 patients represent a great majority of those that are going to be in that group. We do have a few additional clinical pharmacology cohorts. But I think what you can look at here is a great majority of the patients who will be enrolled there. Our eyes are now pointing to the additional studies that we have opened to ask additional questions. So this data represents a pretty full set. And again, we'll continue to follow those patients. And as you noted, there are several of them that are still on study. So we look forward to seeing how that data evolves. Of course, I just want to remind you all that we have fully enrolled the Phase II cohort, and we'll be turning that cohort over before the end of the year. And so from that standpoint, we're really looking forward to that data set as well.

Great. And Dr. Hong, there was a question regarding the Phase I population and then the population enrolling in Phase II. Any further comments you'd like to make about that?

No. Like I said, the Phase I population, as Greg and Dave already mentioned, I mean, these are really heavily pretreated population, median of 3-plus prior therapies. And these patients tend to have more resistant tumor burden disease, in general, for all therapies, whether it's targeted or immunotherapies. And so my hope and my guess, and that's just my guess, is that you're going to -- if you have second-line patients who are going to be randomized to docetaxel versus 510, that you're probably going to see a better response rate in those patients on the 510. But you just -- as we all know, you just don't know until you conduct experiment, right? But that's my guess.

Great. Thanks, David. And we would point out that we had roughly 90% of patients in the Phase I who have received prior PD-1 therapy, which is a requirement for the Phase II unless PD-1 is contraindicated. So we expect, again, roughly similar populations in terms of prior treatments and other demographic characteristics.

We have our next question from the line of Terence Flynn from Goldman Sachs.

Maybe -- I think, Dave, you had mentioned that PFS for Taxotere of 2 to 3 months and the fact that, that doesn't account for KRAS status. So when you -- do you guys have data that shows if you do account for KRAS status, what that would be? And then was wondering, in terms of the combination strategy, obviously, you did a detailed biomarker analysis here and it seems somewhat inconclusive. So as you think about the combination strategy here, you mentioned broadening that program, adding 3 or 4 additional combos. So how are you approaching that now given you've seen some of the early biomarker data? And when could we see some combo data from you guys in the clinic?

Yes. Thanks, Terence. In terms of the combination strategy, we are facing that on fundamental biology and, of course, standard regimens in different treatment types. So in colorectal cancer, combination with an EGFR inhibitor, for example. In non-small cell lung cancer and other solid tumors, looking at chemotherapy combinations as appropriate. So I think all of that is founded in both good, basic and clinical science. And we would expect data from the combination regimens over the first half or so of next year. And can you remind me what the first part of your question was again? All right. While we're waiting for Terence to come back, with that, Annie, can we have the next question?

We have our next question from the line of Michael Yee from Jefferies.

Question for Dr. Hong and question for Dr. Reese. Dr. Hong, maybe you could just comment on what you think the level of clinical meaningfulness is versus the alternatives here? Given the durability you've seen here in the overall responses, maybe just talk a little bit about the magnitude of clinical meaningfulness, I guess, versus options. And then for David, can you just remind us what the Phase IIb? Is there something that triggers the completion of that? Is it just kind of arbitrary, more than 6 months of follow-up on all the responders? Maybe just talk about what actually drives you to decide, stop it, lock it and release the data?

Thanks. And maybe I'll start, and then I'll ask Dr. Hong to weigh in. First, let me address the first part of Terence's prior question on KRAS status. We are conducting large real-world evidence studies to date. We don't see any dramatic change in behavior in terms of response based on KRAS G12C status in the lung cancer population. We're continuing to accrue more data there. But certainly, to date, we've not really seen anything suggesting an enhanced response to, for example, standard chemotherapy. In terms of the Phase IIb study, we -- all patients will be followed for a minimum of 6 months after a first response. And so remember, scans are done every 6 weeks. So if you have a response at 6 weeks or 12 weeks, all patients will be followed for at least 6 months after that response to get solid estimates of duration of response. And let me -- with that, let me ask David Hong to comment on your views of the clinical meaningfulness of the data.

Yes. So I'm not a lung guy, so I'm not an investigator who focuses on lung. I focused primarily on early drug development and so forth. But I've spoken to many of my lung colleagues and our colleague, Bob Li, who is also the co-author on New England Journal of Medicine paper. All of this -- in this population, this population of non-small cell lung cancer patients, 3 lines or more who've been mostly -- all have been chemorefractory and immunorefractory, the median progression-free survival of 6.3 months is very significant. And the fact that -- and if this pans out in a Phase II, would -- in a larger Phase II set, would definitely be informatory of this. So I think all these things -- and that's why I think it got published in the New England Journal, right? I mean if this was not significant, I'm sure Dan Longo and the editorial staff would probably have dismissed this paper. But the fact that we're seeing these response rates and also particularly the meaning of progression-free survival, which is a fairly, I think, accurate surrogate for overall survival in the non-small cell lung cancer setting, is promising.

We have another question from the line of Ronny Gal from Bernstein.

Congratulations on the data. And I have two. First, I'm a bit confused about how to think about the relationship between KRAS and PD-1. You have preclinical data showing an enhancement of PDL-1 standing by KRAS. Merck presented data suggesting the KEYTRUDA works the longest in KRAS-mutated patients. And today, you have told us that PD-L1 standing is not a good biomarker for the activity of anti-KRAS agents. So I was wondering if you can just let us know with kind of like what is your thinking right now. What is the hypothesis about the interaction between those 2 agents? And then for Dr. Hong, as somebody who looks at the early-stage data, I'm sure you kind of looked at the comparison along the various KRAS G12C targeting agents. I wonder if you can comment, in your mind, where does the science stand today, especially in the relationship between agents targeting the RAS on- and off-state with the KRAS-mutated targeting agents.

Thanks, Ronny. Let me ask Greg to talk about the relationship between response and PD-1 expression. And I think the take-home message here was that we saw responses across a range of PD-1 expression. Small data sets and a lot more to learn here. We'll have extensive data in the Phase II trial as well. But let me ask Greg if he wants to provide a little additional color.

Thanks, Dave. Yes, when we looked at a variety of molecular epidemiologic studies, there's a clear pattern that these KRAS G12C patients are more commonly smokers. They have higher tumor mutational burden. And unsurprisingly, they have, de novo, a slightly higher PDL-1 expression levels compared to an unselected population. Now that's, again, an epidemiologic phenomena. What we have to keep in mind is that the patients who are flowing into the study that we've just presented data on are those that in whom PDL-1 and PD-1 therapies have failed them and that they've progressed through those. So that may explain when we show our data. The numbers are small, but you see that there's -- it looks like there are very few patients in the study, numerically, who have high PDL-1 expression. Whether or not this is going to be a hard trend that plays out or not, this is the data. And what's fair to say is the patients for whom those therapies are no longer active, no longer working, our data shows that, that doesn't preclude the possibility of response in later lines of therapy.