Roche Holding AG (ROG) Earnings Call Transcript & Summary

Ladies and gentlemen, welcome to Roche virtual event on analyst during ESMO IO. My name is Henrik, and I'm the technical operator for today's call. Kindly note that the webinar is being recorded. [Operator Instructions] At this time, it's my pleasure to introduce you to Karl Mahler, Head of Investor Relations and Group Planning. Karl, the stage is yours.

Yes. Thanks a lot, Henrik, for your kind introduction. Welcome to our ninth, actually, science call this year. So 9 events means that we really have lots of news flow -- positive news flow in the case of Roche, and this one is a particularly exciting one despite it's only in a Phase II [indiscernible] only in a Phase II. And we have next week our last one in the course of ASH, where we basically testimony our broad and maybe best-in-class portfolio in hematology. This is the program for today. We will have an updated -- and show to you an updated analysis from the Phase II CITYSCAPE study. Raymond Meng will do that. He is the Global Development Leader for tiragolumab based in San Francisco. Heather Stevens, with us. She is the Life Cycle Leader for tiragolumab. And we have also, I'm pleased about that Ira Mellman with us. Actually, Ira is the developer of the TIGIT data. We have close collaboration with Raymond Meng. What I learned during the call before is that Ira told Raymond don't mess it up, and he did mess it up. I mean, clearly, there is, maybe for the benefit of the listener today, a presentation, which Ira gave in the course of ESMO IO on the TIGIT -- on how TIGIT works. So maybe for those who want to be a bit more going in depth with CITYSCAPE and how it works, and in particular, on the mode of action, then please refer to the ESMO IO website. Before we go into the call, and of course, big thanks to our presenters and for those who participate in the Q&A. Also wanted to thank Sabine Borngräber, who did actually the heavy lifting today on the slides, developing the story line, [ Ifan Melani ] for the back office here and for the organization. Maybe you can kind of go to the Slide 5. So there's only basically one slide I wanted to show from my side, and this is basically a bit of a history on non-small cell lung cancer treatment. I mean, we started with adding a non-chemo agent in 2006. We had a 12-month -- generated a 12-month overall survival benefit at that time. Hazard ratio at that time was 0.79. There was really a breakthrough because we could first time really go beyond the 10 months medium survival. Actually, this baseline of the 12 months somehow translated into the baseline comparison of the other trials which came. You could see us doing the Tecentriq IMpower110 trial in 2018, median overall survival benefit, 20 months; KEYTRUDA, KEYNOTE-042 was basically the same, also median 20 months despite the baseline. Characteristics were a bit different here. Our comparator was 13 months. There was 12 months. Having said that, we had a hazard ratio of 0.59. We had a hazard ratio of 0.69. But in the end, it didn't matter because we were second to market. We were not first to market despite -- I would say, if anything, we look at least as good -- and our combination look at least as good as that what the competitor had. But as I said, first in class counts heavily. And we had seen in the chemo high expresses at the more than 50%. There is a final readout already on the KEYNOTE-189, hazard ratio of 0.59. It brought the overall -- median overall survival benefit to 27 months. So final readout of this one, good data, I have to say, with the addition of chemo. And now we have a new in the same setting a new attempt which basically is chemo-free. And we have passed the 30 months median overall survival benefit. We have a hazard ratio of 0.23. I mean, it only can get better here from this point onwards. And so these are really, let's say, super exciting data from my point of view. And with this one, over to you, Raymond. The floor is yours. Thank you.

Sure. Thank you very much, Karl. Heather and I are very excited to be presenting to everyone today an update on CITYSCAPE. Next slide, please, Sabine. So as Karl reminded everyone, we initially did present this proof-of-concept Phase II study CITYSCAPE at ASCO in 2020. And at that time, we had data suggesting that there was a PFS benefit in all patients. And as mentioned by Karl, a lot of this was driven by the PD-L1-high patient population. Those patients assessed by 22C3 with a tumor proportion score greater than 50%, and that's what we've shown on the right side of the figure as well. Next slide, please, Sabine. The co-primary endpoints of the study were response rate in PFS, and we're showing the response rates. And like PFS, the benefit again in the ITT population in that first column, it's mainly driven by the PD-L1-high subpopulation. What was unknown to us at this time because we looked at this study in ASCO 2020 median follow-up at 10.9 months: Would these responses be durable over time? And I think that's something that we're very excited to present today where this response is durable. And what did the safety and tolerability look like as well? We're not showing that slide from ASCO, but the combination was relatively well tolerated at this time. The safety profile for tiragolumab plus [indiscernible] was very similar to atezolizumab at that time. Next slide, Sabine. So this is the presentation this morning that was given at ESMO IO by Dr. Cho from Severance Hospital in Seoul. There was an updated analysis on response rate, PFS as well as overall survival at the time of the ASCO presentation. We had not reported overall survival because the data was immature. And we also have a summary of the safety and also patient-reported outcomes as well. Special thanks to Dr. Cho. He's been a very great collaborator for us. He was on our Phase I study. Actually, he was one of the first investigators to pick up a potential signal in lung cancer which expanded into the Phase II study, so it's a wonderful opportunity for him to present on the study this morning. He did a fantastic job. This is the background slide we usually include in our tiragolumab presentation. So I won't go over all the details today as Dr. Cho presented it today. I'll just kind of pinpoint some of the highlights here. One of the reasons we had this slide is to remind us of the science underlying TIGIT. It's really the science that's driven our Phase I or Phase II development as well as our Phase III development. Fortunately, we have Ira here as well today to help address any of the scientific questions. But this target -- this TIGIT target was actually a target characterized identified at Genentech, a lot of the work done in Ira's lab as well. So at the same time, we're testing the clinical hypotheses in these scientific studies. We're also trying to explore the scientific hypothesis, and we're still doing a great deal of reverse translation from our Phase I or Phase II studies, and hopefully, our Phase III studies back to the laboratory. Next slide, Sabine. This is a reminder of the study design for CITYSCAPE, and this is a similar study design that we showed back at ASCO as well. This was a study of randomized Phase II proof of concept, really testing the hypothesis of what happens when you add tiragolumab, an anti-TIGIT antibody, on top of our anti-PD-L1 antibody, atezolizumab. This study was done in PD-L1-positive patients with non-small cell lung cancer, newly diagnosed patients, EGFR/ALK wild type. The diagnostic used in this was the 22C3 immunohistochemistry assay, and patients had to have a PD-L1 score greater than or equal to 1%. 135 patients were enrolled and randomized 1:1 to tiragolumab plus atezolizumab versus placebo plus atezolizumab. Patients were treated to progression or loss of clinical benefit. This was a randomized, double-blinded, placebo-controlled study. The stratification factors are shown as well, stratified by PD-L1 score, histology and tobacco use. Co-primary endpoints were response rate and progression-free survival. In the box, we give a reminder of what's different about this data set. And it's really the follow-up that was done. We've advanced the clinical cutoff date from June of 2019, at that time, the median follow-up of about 6 months, to an August 2021 data cutoff. So now the median follow-up has increased about 30 months, as mentioned by Karl in the introduction, so really giving us a sense whether these responses would be durable from CITYSCAPE. Next slide. A reminder of the baseline characteristics in the ITT population. This slide was shown back at ASCO. No significant differences as well. What's important to note from this slide, there weren't significant differences in the baseline characteristics, demographics of the 2 arms in this patient population. There were some minor differences, one aspect, in the second line, males, slightly higher number, 70% on the placebo plus atezolizumab arm versus 60% on the tiragolumab plus atezolizumab arm. The other subcategory, Asian patients, slightly more in the placebo plus atezolizumab, 34% versus 27%. On the bottom, we're showing the 3 stratification factors, which are relatively well balanced between the 2 arms. Next slide. And now, we'll go into the data. This is the updated PFS in the ITT population in all patients who are PD-L1 positive with TPS greater than 1%, and you can see the benefit similar to that shown at ASCO, 5.6 months versus 3.9 months. But really, we wanted to focus on the next slide that the benefit, again, was driven by the PD-L1 highs. So on this slide, the first column showing you the PD-L1 TPS greater than 50%. And at the time of ASCO, we have not yet reached the median PFS in the tiragolumab plus atezolizumab arm. This has now been achieved at 16.6 months versus 4.1 months with a hazard ratio of 0.29. And on the bottom, you can see the actual curves itself, tiragolumab plus atezolizumab in red and the placebo plus atezo arm in blue with a 12-month PFS rate of 51% versus about 22%. We're also highlighting, again, the overall response rate. It didn't change too much. We actually picked up one patient who responded on tiragolumab plus atezolizumab arm. Actually, a very interesting patient, a patient who had prolonged stable disease after about 1.5 years that then converted into a partial response. We're also showing the median duration of response as well of 15.7 versus 8.2 months. And for transparency, on the right side of the figure, we're again showing the PD-L1-low patient population, TPS 1% to 49%, essentially showing no difference in a median PFS, 4 months for tira plus atezo versus 3.6 months. And again, essentially similar, overall response rate and median duration of response and the curve -- PFS curves on the bottom essentially overlapping each other as well. So the benefit, primarily driven by the PD-L1-high patient population. On the next slide, we're actually updating now on overall survival, which was something that was not yet reported at ASCO given the immaturity. And now you can see that we do have overall survival in the ITT patient population, all patients who are PD-L1 positive, 23.2 months on the tiragolumab plus atezolizumab arm versus about 14.5 months on the placebo plus atezolizumab arm, hazard ratio of 0.69, and we're also showing the 12-month and 24-month PFS rates on this. But again, this benefit was really driven by the PD-L1 high sub-patient population. On the next slide, overall survival in the PD-L1-high TPS greater than 50% has actually not yet reached the median overall survival in the tira plus atezo arm. The lower confidence, 95% confidence interval is around 30.3 months. The placebo plus atezo months -- the placebo plus atezo median overall survival was at 12.8 months. So currently, a hazard ratio of 0.3. We're also showing the 12-month and 24-month OS rates, and you can see the curve separate relatively early for this. I will say the events have slowed down on the study. The study is still ongoing at this time. It's somewhat unclear to us when median OS will be achieved, but we're still continuing to track this in the PD-L1-high patient population. And again, for transparency, showing the PD-L1-low patient population of TPS 1% to 49% on the right side of the figure, where, essentially, like response rate, PFS, duration of response, we're not seeing that additional benefit of tiragolumab added on top of atezolizumab and median overall survival of 13 months versus about 14 months. The curve is essentially overlapping as well. Next slide, Sabine. Yes. In addition to efficacy, we also wanted to get a look at safety as well. And it's important to note that, essentially, we are doubling the treatment duration for patients on tiragolumab plus atezolizumab versus placebo plus atezolizumab as shown in the first line. But even though we're increasing the number of treatment duration, increasing the number of cycles, you can see that the adverse events are not significantly unbalanced between the 2 arms and something we saw at ASCO before. But now with longer durability, we have more data as well. All-cause adverse events are very similar between the 2 arms. And you do see more adverse events on the tiragolumab plus atezolizumab on Grade 3 to 4 adverse events, 52% versus about 40% on the atezolizumab arm. Some adverse events such as Grade 5 adverse events are actually more on the placebo plus atezolizumab arm as these are adverse events related to mainly progression of disease. Again, serious adverse events, more on the tiragolumab arm but not significantly more. We're seeing 52% versus 41%. What's interesting are the treatment-related adverse events in the next column, seeing again more treatment-related adverse events on the tiragolumab plus atezolizumab arm, 82% versus 70%. But what we'll show in the next figures, these are primarily Grade 1 to Grade 2 adverse events. If you look at the Grade 3 to 4 treatment-related adverse events, they're relatively well balanced on the 2 arms, 22% for tiragolumab plus atezolizumab versus 25%. We actually are also reporting the Grade 5 treatment-related adverse events. And here, we have 2 on the tiragolumab plus atezolizumab arm versus none on the placebo plus atezolizumab arm. Dr. Cho did not have time this morning to discuss this, but both of these Grade 5 adverse events appear to be infections. One may have been an upper respiratory infection, treated with antibiotics. This then progressed to potentially pneumonia. The second case was the urinary tract infection, treated with antibiotics, again, a question of recurrent fever. Whether this was pneumonia versus urinary tract infection, it was unclear to the investigator. These relatedness of adverse events are assessed by the investigator. And for both of these Grade 5 adverse events, the investigators could not fully discount that the addition of tiragolumab or atezolizumab could have caused some of that. We didn't have time to show all the Grade 5 adverse events, but they are relatively balanced between the 2 arms. We don't have a rationale for why inhibiting TIGIT might increase susceptibility to infections. If we look at serious adverse events related to infection, they're actually balanced on the 2 arms. So those were the Grade 5 adverse events. We're also reporting the serious adverse events, which are balanced. What's interesting are the immune-mediated adverse events, and this could be a potential pharmacodynamic effect as well as we're getting 2 immunotherapy drugs. We can see more immune-mediated adverse events on the tiragolumab arm, 76% versus 47%. But again, they are primarily Grade 1 to 2. The Grade 3 to 4 immune-mediated adverse events are relatively well balanced, 19% versus 16%. And again, those adverse events leading to dose modification or treatment withdrawal, are balanced between the 2 arms, slightly more on the tiragolumab plus atezolizumab arm. So overall, these results are consistent with what we saw at ASCO with longer durability. It seems that this combination is well tolerated over time. Next slide, Sabine. Here, we're actually reporting the adverse events and immune-mediated adverse events. With the 30 months of follow-up, we haven't picked up any new safety signals. For the most part, these adverse events, they seem to occur early in treatment, and patients usually seem to continue -- to be able to continue treatment through the adverse events. So the all-cause adverse events and the immune-mediated adverse events were relatively similar. Graphically, you can also see that most of the adverse events on the study were low-grade Grade 1, reflected by green bars; and Grade 2 in blue bars. And there are actually some adverse events more on the placebo plus atezolizumab arm, and these are primarily adverse events related to disease progression. Next slide, Sabine. Similar to the findings of safety and that even though we're seeing more safety events that was relatively well tolerated, we are reporting the patient-reported outcomes for the first time from the study. And again, these findings are similar to the safety findings that this combination seems to be well tolerated. Compliance was actually fairly high for patient-reported outcomes on the study. So the first set of patient outcomes are reported to baseline global health status. So on the bar graph, we're showing patients on the tiragolumab plus atezolizumab arm, red bars, versus those on the placebo plus atezolizumab arm. Basically, baseline function is relatively balanced between these 2 patients at the onset of the study. On the right side of the figure, we're actually showing what happens to those baseline functions over time on study. And you can see the curves are similar here and that we're not really compromising functional status with the addition of 2 active immunotherapy drugs over atezolizumab. And there may be a slight improvement by cycle 16, the caveat being that there are fewer patients on study at this time to give a significant assessment. Next slide, Sabine. This part of the patient reported outcomes, now we're looking at lung cancer symptoms itself rather than global health status. Again, a similar format. At baseline, the lung cancer symptoms are relatively similar between the 2 patient populations. And you can see that the patient populations are actually not that symptomatic from their lung cancer, fortunately. Over time, we're now seeing the progression of these lung cancer symptoms. And overall, it's similar on the 2 arms. There might be a slight improvement in the tiragolumab plus atezolizumab arm at cycle 16. But again, the numbers are low. But overall, we're not seeing a significant difference with the addition of 2 immunotherapies over atezolizumab at this time. Next slide, Sabine. So overall, to summarize, I think we're very excited to show that these results do seem to be durable with tiragolumab plus atezolizumab. We're seeing an improvement in progression-free survival response rate and overall survival as well in all patients on the study treated with tiragolumab plus atezolizumab arm. And primarily, the benefit, again, it's driven by the PD-L1-high patient population with TPS greater than or equal to 50%. Safety,of tiragolumab plus atezolizumab, it appeared to be well tolerated with longer durability of study. We did not pick up any new safety signals at this time. And in terms of patient-reported outcomes, it's relatively consistent with the safety data. It supports that -- it appears to be well tolerated. Patients do not have a compromising or functional scores. And lung cancer-related symptoms, there may have been a slight improvement over time with tiragolumab plus atezolizumab, although the numbers are small at that point. Overall, we're very encouraged by these results, especially by the overall survival data, which we believe continues to support evaluating this combination in frontline lung cancer, and we're actually running the ongoing Phase III study, SKYSCRAPER-01, right now. It's similar to CITYSCAPE. It's a randomized study, but we're focusing on the PD-L1-high sub-patient population with TPS greater than or equal to 50%. So next slide, Sabine. So now we've reviewed the CITYSCAPE data that I think is very exciting how does this translate to the overall tiragolumab development, and I'd like to turn it over to our Life Cycle Leader, Heather, now for a review of that. Thank you.

Thank you very much, Ray, for taking us through that data. If we go to the next slide, Sabine, just as a quick reminder of how we see the evolution of cancer immunotherapy, we do know that combinations will be necessary to continue increasing the percent of patients that respond and can benefit from cancer immunotherapy. At the same time, in the past, there have been challenges with combining multiple cancer immunotherapy agents due to toxicity which can lead to treatment discontinuations. And finally, overall survival really is king. This is what patients are looking for, and OS does remain somewhat modest across many tumor types despite the advances of cancer immunotherapy. If we move to the next slide, I'd like to just discuss briefly our entire sort of portfolio strategy for tiragolumab across different diseases. It stems from our scientific observation and preclinical work and increasingly in clinical data sets like the one that Ray shared that there's a convergence between the PD-L1 and TIGIT pathways that leads us to our core hypothesis that where PD-1/PD-L1 inhibitors show benefit, we may be able to boost or amplify that benefit with the addition of tiragolumab. Our strategy therefore falls in 3 different areas. First, we're seeking to build on disease areas and settings where we've observed the benefit from Tecentriq. Secondly, due to the somewhat favorable toxicity profile, we're looking to advance rapidly into early disease and really accelerate our plans to move to areas where patients could benefit from potentially curative regimens. And lastly, we're looking to explore additional indications where we have not necessarily played before but where we think the addition of TIGIT could deliver benefits for patients. And our strategy is really in 2 different categories. We have our randomized studies, our SKYSCRAPER studies, which you're familiar with. And, for example, SKYSCRAPER-01 in the PD-L1-high non-small cell cancer, which built on CITYSCAPE; as well as SKYSCRAPER-02, which builds on our successful IMpower133 regimen. In early disease, we're also moving into Stage 3 non-small cell lung cancer with SKYSCRAPER-03 and also locally advanced esophageal cancer with SKYSCRAPER-07. And then finally, our China strategy is really focused on frontline esophageal cancer, where we have an Asia-Pacific-focused trial in SKYSCRAPER-08, an additional study, SKYSCRAPER-04 in PD-L1-high cervical cancer; as well as SKYSCRAPER-06, our Phase II study in all-comer frontline non-small cell lung cancer in combination chemotherapy. I just want to highlight that we also have additional signal-seeking efforts underway across all of these different areas as well to prepare us for a potential next wave of signals we can explore further. And finally, we are also now embarking on novel combinations. The most recently announced just in the last few weeks is a combination with our PD1-LAG3 bispecific, which we're studying in melanoma. Next slide, please. So digging in a little bit further, specifically to lung cancer and our strategy here. Next year, we're very excited that we will have our first 2 Phase III studies reading out SKYSCRAPER-01, which replicates CITYSCAPE; and SKYSCRAPER-02 in small cell lung cancer. Additionally, we have SKYSCRAPER-03, which will read out a bit later due to the disease state; and have 2 signal-seeking studies in neoadjuvant/adjuvant non-small cell lung cancer in SKYSCRAPER-05, where we're evaluating the benefit of adding chemo or potentially chemo-free regimens in this neoadjuvant-to-adjuvant setting; and as I mentioned previously, SKYSCRAPER-06, where we're looking at frontline chemo combo. If we go to the next slide, I'd like to finish just by talking about where tiragolumab sits in our overall lung portfolio. And there are 3 key sort of macro trends that influence our thinking about the opportunities in lung cancer. The first, of course, which is the focus today is really cancer immunotherapy, where we began with monotherapy Tecentriq. We then built on Tecentriq by adding standard-of-care combinations, chemotherapy, bevacizumab and now are really moving to partner Tecentriq with novel NMEs, such as tiragolumab. We think a further evolution in the fourth wave may be based on individualized neoantigen therapies. On the precision medicine side, we also have a suite of highly effective, highly targeted personalized therapies, and we continue to make advances there as well. And then finally, across both cancer immunotherapy and precision medicine, we see a trend towards early and earlier treatment, where we hopefully can impact even more heavily patient benefit and potentially look towards curative regimens in lung cancer. With that, I will end there and I believe hand it over to a Q&A session. Thank you very much.

Yes. Thanks a lot, Heather and Raymond. So here is Ira as well, who joined the team, I would like to say. And we have about 300 people online, which is a lot for such a call, so thanks for your interest. And I would say let's get started with the first question over the phone from Wimal from Sanford.

Great. Thank you, everybody, for the presentation. So I guess my first question, please, is just post this update, can I get a sense of your level of conviction heading into the Phase III readout next year? Is it fair to assume that you now have quite a high probability of success, just given the trial design in high PD-L1 [indiscernible]. And then just tied to that, what would success mean for Tecentriq in non-small cell lung cancer? Do you expect tiragolumab usage to be mirrored by Tecentriq? Or rather, do you expect physicians to stick with the currently used PDX therapy and then add tiragolumab to that? And then my second question is success in other indications. Is there any read across to small cell, in particular, or any other solid cancers? Or is it very much an unknown wait and see?

Okay. Raymond, maybe one for you. And maybe on the overall franchise, Ira, if you could answer with this one.

Thank you, Karl. Thank you, Wimal. Yes, thank you for that question. So I think our overall confidence in tiragolumab based on these updated study results is it's kind of similar that we presented at ASCO. We were encouraged by the results. But I feel like it's important to realize this was a small study, a proof-of-concept study, and we hope to replicate it in the Phase III study. So I don't know necessarily if our confidence has increased. It's encouraging to see these results, certainly. And then the other question on other indications as well, I think that's -- one of the goals of running a proof-of-concept study in lung cancer is that most of immunotherapies have been approved in lung cancer. It's an indication similar to melanoma. And our goal was to try and get a result in this lung cancer. As a proof-of-concept study, on the flip side, we also wanted to know whether this combination would work or not. If it hadn't worked in lung cancer, I think we were confident to say it wouldn't work as well. So I think this is in an informative go/no-go decision for us. But I'll turn it over to Ira as well about the relevance of the study to other indications as well.

I think not that much to add, except to say that in the case of immunotherapy, you're dealing with a common set of mechanisms, regardless of what the indication is. And because of the complexity of different types of cancer, of course, you can't expect the same types of responses as you move from indication to indication. But from, I think, a basic perspective in terms of how these drugs work, there's more that's similar than dissimilar. So if you see a very positive result in one indication that -- limited to a group of patients that have a particular immunological phenotype, in this case, the PD-L1 highs, then even without having randomized Phase II data to back you up, there's a very strong justification to at least try it in these other indications because the fundamental mechanism of the immune response in those patients is similar to where you do see it working in a randomized study.

Yes, yes. I mean, 30 months plus -- it's 30 months plus. I mean, it's unheard of so far. So we have very strong arguments, I have to say, to see what we see. So seeing is believing. I have a question here from Richard Parkes, who asked a question over the chat. He asked: Are you able to speculate why there is no benefit in the 1% to 49% expresses? And I wanted to add here with one question from Andy Acker, who was asking about the subpopulation, the more than 50% baseline characteristics. Have you seen any kind of differences here? Or is it very similar to the expression? So maybe you can answer both in one. Yes.

Sure. I'll answer the second part of that actually, Karl. Did we find any imbalance because, as Andy noted, we're only showing the baseline characteristics for the ITT population. We actually do have it for the 1% to 49s and the greater than 50% as well. We didn't have time to show it. They're relatively similar to the ITT as well. We didn't pick up any significant imbalances. Something else that's interesting to us, we stratified by PD-L1 1% to 49% and then greater than 50%. We also wondered could we have somehow unbalanced the TPS scores, so maybe within the TPS greater than 50%, we had more TPS, 80% or 90% on one arm. We've looked at that as well, and it's relatively well balanced, fortunately, on the study, the -- within the TPS greater than 50%, also within the TPS 1% to 49swithin the different stratifications as well. But I think going back to Richard's question, why we did not see a benefit in the 1% to 49% population, I'll actually turn it over to Ira for a scientific rationale for that.

Thanks, Ray. I'm sure you could have answered that yourself. But all I'd say, again, it really comes back down to mechanisms. I think current evolving concepts of how checkpoint inhibitors, in general, work is that they really have benefit in those patients who are already exhibiting an ongoing immune response to their tumors. And I think the mystery of PD-L1 expression is emerging as less of a mystery and more just a pretty accurate readout as to the extent to which a patient does have an ongoing immune response. And so the more PD-L1 you see, the better the response; the better the response, the more there is to amplify. So I think in the first instance, seeing these things in that group is emerging at the expected response. And in terms of whether or not you would see it at lower levels, remember, this is, in some ways, just an arbitrary cutoff that was defined by just aggregate experience. But I think probably in future studies, we may be exploring other cutoffs as well the types of characteristics.

Okay. Actually, the similar question came from Marietta Miemietz from Charles Pitman. I just wanted to picture here. The -- people ask questions and they have freely, don't depress it. So next one would be from Elizabeth Walton.

Can you hear me okay, Karl?

Yes. Perfect. Thank you.

I've got two questions, if that's okay. The first one is just on the timing of the SKYSCRAPER study in the frontline non-small cell lung cancer PD-L1 high. I think previously, you talked about that being a second half of 2022 event. It looks like in the slides, that's moved forward. I see at 3Q, you talked about a filing in that population in 2023. So I just wanted to check that, that has, in fact, moved forward. And then my second question was just -- in fact, we could dig a little bit more into your confidence on the small cell lung cancer readout because I believe that's the one we're going to get first. One of the points that discussant made this morning was around the role of TIGIT in hot versus cold tumors. We tend to think about small cell as being slightly colder than the non-small cell lung cancer. Should we be concerned? And are you also stratifying by PD-L1 status in the small cell lung cancer study as well?

Thank you. You want to take it, Ira? You want to go for confidence in small cell lung cancer?

Yes. That's an excellent question. I mean, small cell lung cancer has its predominant immune phenotype, the so-called immune desert, which is in absence for response, which is exactly the opposite of what I just told you just a moment ago. I think there, the key variable, though, is the introduction of a prior chemotherapy, which, depending upon which one you use, is capable of causing an inflammation that at least temporarily can cause an immune response. And the hypothesis, what's happening in this case is that the chemotherapy does that. And then by adding the combination, which gives us a very effective one-two punch to the early checkpoint that prevents T-cell proliferation and would turn off T cell prematurely. This particular combination just seems to work at least with obviously much smaller data set that we have for non-small cell lung cancer. So honestly, I was surprised by that result, but you have to trust what the clinical data say and then backfill and understand whether or not this makes any sense mechanistically. And I think we're coming to an explanation of why this has make sense.

And the timing, maybe, Heather, one for you. Thank you.

Thanks for the question. In reality, the trial hasn't moved significantly. We were always looking at midyear next year. We were fortunate that we had very robust enrollment on the study, and it actually came in about 4 months earlier than originally projected. All that said, just a reminder that this is an event-based trial, so we'll be closely tracking events to finalize kind of the ultimate readout timing for SKYSCRAPER-01. Thank you.

Yes. Thank you. The next one would be from Sachin Jain. Sachin, hope we could address your questions today. And the next one would be Sachin.

Just a couple, please. So first on SKYSCRAPER-06, which you've referenced is chemo combo first-line, broader PD-L1, and it's listed on Slide 24. I think on a prior call, there have been mention of a potential interim or really look at the study even its Phase II signal finding through '22. So I wonder if you could comment to that and what you need to see to start a registrational study? Or are we thinking about registrational study not starting until '24? So I just wanted to clarify time lines there. The second question was just to continue the vein of questions on solid tumors and extent of evidence we have there and so sort of a two-part question really. So firstly, your -- as I said, the only TIGIT in Phase III in solid tumors outside of lung despite there being a lot of companies in the TIGIT space. So just interested in your thoughts as to what you've seen others haven't despite this being a super competitive space. And the second one related, there was a U.S. biotech company, I think [ Mario ] recently reported some initial responses in some gynecological tumors, which I think is the first clinical data we've seen outside of lung. So just interested in your thoughts there and whether it validates your thesis, particularly as we think about cervical data also coming early next year.

Yes. Thank you, Sachin. Just to close the picture here, also Charles Pitman, actually, had another question here on the line in exactly on SKYSCRAPER-06, so maybe we can just go for this one. Raymond, would you like to take it?

Sure. Yes, I'd be happy. I think those are great questions. Actually, I may turn it over to Heather because I think you kind of covered a little bit of this in the development overview.

Sure. Thanks, Ray. So with SKYSCRAPER-06, the core question is really across an all-comer population, looking at different PD-L1 subgroups. We're looking to see if the addition of tiragolumab to atezolizumab in chemotherapy is able to drive responses. So that's really why we've set up that study and the results of the study will drive further steps in the frontline chemo combo setting. And Ray, maybe back over to you for the rest of the question.

Sure. So I think there is also a question as well about [ Mario ] reporting some cervical data. And I think it's great, all the accumulating data. I think with our study, with other studies in the anti-TIGIT space, if we could bring benefit to patients, that would be great. We are running a study in cervical cancer, SKYSCRAPER-04 as well. It's a randomized Phase II study. A similar study design as CITYSCAPE as well, patients are randomized to tiragolumab plus atezolizumab versus atezolizumab. I think we're exploring cervical cancer as well. For a lot of these indications, a lot of it is based on the science as well. Is there a scientific rationale to pursue these solid tumor mutations, PD-L1 expression? Something that remains somewhat unknown to us as well is what is the contribution of the other components of the TIGIT [ CD ] pathway, TIGIT expression, the receptor TdR as well. And I think going back to one of my points, this is something we're trying to learn with Ira's team as well with our reverse translational group. So I think development in other solid tumors, this is very interesting, and we look forward to seeing our results as well as others' results, certainly.

I guess if I understood Sachin correctly, and I feel that a bit with the other questions as well is, I mean, I guess that we are 1.5 years ahead of the competition, and we know that we really have to be first to market. I mean, we saw it in the first line on small cell lung cancer that actually the data -- to say the least, we are not worse than that of the competitors. But if you come here second to the lunch, the lunch is eaten. I have to say. The others have taken advantage of what you left or what you didn't eat. So the question is how can you make sure that you are, let's say, in front here and you keep the 1.5 years and not, let's say, falling behind it? I guess it's a bit what I feel here.

Can I just clarify? The SKYSCRAPER-06 question was a timing question. Is there data in '22? And when do you anticipate registrational study starts?

I can't really speculate about registrational study starts because that will be data dependent, but I can confirm we will have an interim analysis next year in 2022.

Usually, we do not comment on the interims, but now it's out. Okay. But how can we make sure that we are in the lead? I guess this is a bit of we take all the opportunities. I guess this is more a question to you, Raymond. I mean, you have a view on the competitive landscape, and I guess you feel confident that you are in the -- maybe take the opportunities there where you really, let's say, see major opportunities and really go full steam ahead of the competition.

Yes. It's a good point, Karl. It's a balance, right, of data and an element of risk taking, as you mentioned as well. So as Heather reviewed our SKYSCRAPER space, I think the lung cancer development is somewhat grounded upon the results of CITYSCAPE as well. We do have studies in esophageal cancer as well, so SKYSCRAPER-07, SKYSCRAPER-08 as well. And that was driven by -- we had some preliminary data in Phase Ib, which we actually presented at ESMO GI over the summer. Zev Wainberg from UCLA presented on our esophageal cancer cohort. We -- interestingly, it looked like the lung cancer data set, we saw responses in PD-L1 positive. There's also some PD-L1-negative patients. We saw responses in squamous and adenocarcinoma histologies as well. Even though it was a Phase I data set, the durability look good in that preliminarily. So I think taking that Phase I data set, taking what we know from lung cancer, with an element of translation of lung cancer, esophageal cancer, I think we did somewhat take a risk there and advance into the Phase III studies without doing a randomized Phase II study. So I think we are trying to balance data and risk taking right now.

Okay. That's very clear. Thank you. Next question would be from Stephen Scala from Cowen. Thank you for the question, Sachin.

Karl, can you hear me?

Yes.

Is there any additional insight into the reasons for the underperformance of the Tecentriq monotherapy arm? Roche previously suggested that could be due to selection of poor prognosis patients into chemo-free trials. But if that is the case, then could tiragolumab plus Tecentriq efficacy go even higher in SKYSCRAPER? So that's the first question. Second question is when we see the data from SKYSCRAPER, what does Roche view as the most relevant OS benchmark in the PD-L1 greater than 50% tumors? Would it be median OS of 20 months from KEYNOTE-042 or median OS of 30 months from KEYNOTE-024?

I can go ahead, Karl. Maybe I can do the first part of that. So in terms of the performance of the Tecentriq atezolizumab arm, and that's something we saw as well when we first saw the results prior to ASCO in 2020. It's something we've looked at. We haven't been able to pinpoint anything specific on the study. So I think, fortunately, nothing stood out that was imbalanced. Stephen bought up a good hypothesis, could be the poor prognosis patients. It's difficult to do these trial comparisons into trial comparisons. But I think within the Phase II study itself, we do feel like the arms are balanced, what we're seeing on the atezolizumab arm, could we be seeing that on the tira plus atezolizumab arm? It's unknown, but I don't think we've found anything specific to account for this performance. And I think we are doing the Phase III study, SKYSCRAPER-01, so we will have an answer on that. In terms of the comparison to overall survival, whether it's on KEYNOTE-024 or KEYNOTE-048 (sic) [ KEYNOTE-042 ], let's defer to Heather on that?

Not much to add, Ray, other than I think this community is probably aware when we talk to external TAEs, they tend to really benchmark KEYNOTE-042 and feel that, that is really the kind of standard watermark for what we can expect for monotherapy, PD-1/PD-L1. So I think that supports what we've sort of described previously. The OS in the CITYSCAPE study is indeed very encouraging.

Yes. Thank you. Next one -- I hope we could address your questions. Next one would be from Richard Vosser from JPMorgan. Richard?

Yes. Sorry. It's just come through, Karl. It was a bit slow. My fault, my side, I think, so apologies. So just a quick question. The discussion brought up I think today at ESMO IO, the idea of removing the Fc region and whether that had an impact on the level of CD8+ T cells and whether that was better or worse for a TIGIT molecule. And I think Ira, you've made comments in the past about this, but I wonder if you could clarify those comments and what you think -- because there are some competitors out there without Fc regions. So just your thoughts there, please. And then just one quick one on -- or maybe not quick on less inflamed tumors. Just your thoughts on where you might be going there to try and get at those tumors beyond TIGIT because TIGIT is obviously for more inflamed, so just thoughts.

[indiscernible] actually had a bit of similar, going the same lines, which kind of development work do you do in order to address the broader patient population with the TIGIT. I guess this is all going to you, Ira.

I mean, first, concerning the role of the Fc domain. I mean, I think it's fair to say we don't understand it. It's something completely -- although, we are learning a lot about it. So the first result came from preclinical studies which have been repeated by us and by others now. We have another paper about to come out about this shortly, I hope. And it was unexpected, but now I think we understand it increasingly in terms of the other activities associated with TIGIT that are not primarily on the T cell. So I think one of -- the one hypothesis that we're getting fairly close to being able to confirm is that the presence of the Fc domain on TIGIT enables the myeloid cell with which T cells interact at the time of TIGIT and PD-L1 inhibition, enables the myeloid cell to become more activated. And as a consequence of the activation of the myeloid cell, so this would be a dendritic cell or a macrophage, the T cells with which the myeloid cell is engaged actually are stimulated to become more active and more cytotoxic. Now I think that's reasonably clear now from mouse studies and I think from our most recent human biomarker studies taken from the CITYSCAPE trial, in fact. One is starting to see some pretty interesting and even compelling evidence to say that this may also be happening in humans. In other words, you see evidence for macrophage activation as a consequence of TIGIT treatment, TIGIT PD-L1 -- TIGIT T-cell treatment in responders and not in nonresponders. So there may very well be an important access there. So I think, certainly, the Fc domain's presence does no harm. We originally thought that it might result in depletion of cell types. We have not seen that. And I think our feeling is a reasonably high degree of confidence that it does play a positive role. So there may be some activity without it, but maximal activity, I think, is derived when there is an Fc domain. And I've now forgotten the other half of the question. It has to do with loss of activity in less inflamed tumors. Everything gets back to -- I think the point I made again earlier is that I think the emerging consensus about how checkpoint inhibitors work is that they don't cause an immune response or rather amplify an ongoing immune response, which many patients have. But in most patients, that immune response is -- falls well short of what is needed to actually control tumor growth. So if PD-L1 expression in the tumor or in infiltrating myeloid cells into the tumor is an indicator of an ongoing immune response, which is now, I think, what we and others believe, then -- and my overall idea that I just gave you is correct, then almost invariably, you would see the best activity and possibly the only activity but certainly the best activity of this particular combination in inflamed tumors. Now that doesn't mean you couldn't combine with something else to amplify the ongoing immune response and amplify the degree of inflammation. That could be chemotherapy. It could be another type of agent entirely. But I think to understand the current results, I'll stick with this particular explanation which is most consistent with everything we know scientifically.

Yes. Thank you. Hope we could address all the questions. Thanks for the questions, Richard. Peter Welford would be next.

I just got some very quick ones. Firstly, on SKYSCRAPER-01, would you confirm is crossover allowed in that study? Because I think one of the things why crossover wasn't allowed because it wasn't allowed security to be allowed to cross over in that study. Secondly then, just regards to the cervical cancer. I think this was asked before, but I'm not -- maybe I missed the answer. But should we expect cervical cancer data in '22? Or I think in clinical trials, I think it's early '23 is to be done. Is it till '23 until we get cervical data? And then thirdly, just wondering with regards to combination with radiotherapy or that's -- whether that's benefited or at least is there any rationale to potentially believe that, that could be more efficacious to the TIGIT combo?

Raymond, over to you.

Sure. Great questions. I can start with the SKYSCRAPER-01. So crossover is not allowed on that study, similar to CITYSCAPE. It's really some of the clinical -- preliminary clinical data we generated from Phase I. One hypothesis is this combination work in patients who have been treated with immunotherapy. So in patients who are treated with various immunotherapies, it was a mixed patient population in Phase I. We didn't really pick up a signal in these immunotherapy-treated patients. We're trying to standardize the experiment now. So that's one reason we haven't allowed crossover on these studies.

Yes. The questions on cervical cancer? Heather, maybe for you?

Thank you. So while we haven't determined the venue just yet, we do intend to present the cervical cancer data in the second half next year.

Second half next year. Okay. Good. And the last one here, combination with radiotherapy. I have to admit, I really don't know whom to give the question. Maybe, Ira, you. I don't know if it was something -- or maybe Raymond?

Well, I think there's increasing interest in radiotherapy, particularly low-dose radiotherapy as a way of triggering the type of inflammation that you would like to see in order to trigger an endogenous patient response. So I think there's more and more data, not so much from lung cancer yet, but breast cancer and lymphoma, which actually looks quite compelling in that regard. So in a separate series of studies it has at the moment, nothing to do with tiragolumab. We are investigating that. And at least at the moment, I think it's a viable hypothesis, but it still is a bit of a hypothesis.

Okay. Hope we could address your questions, Peter. Next one would be Diana . Thank you for the questions, Peter.

It's Diana Na from Berenberg on behalf of Luisa Hector. Apologies if this has been discussed. I joined the call a bit later. But in terms of the safety profile, there were 2 deaths in t he TIGIT arm. Could you perhaps just explain what the cause was for this? And is this something to be concerned about? And then my second question is what is your read of this OS data into the role of TIGIT in the adjuvant setting, please? And then my third question is could you perhaps just talk about how the launch of Tecentriq in adjuvant lung is going in the U.S. and what ramp we should expect in 2022?

I can go first, Karl, I can take Grade 5 events.

Yes. I'd be happy to. Already discussed, but please, again. Yes.

Sure. No problem, Diana. We actually discussed it in the beginning. So Dr. Cho do not have time to discuss these Grade 5 events, but they are both infections, one, potentially a pneumonia; one, potentially urinary tract infection. Patients were on study, developed a fever, were treated with antibiotics. And then they subsequently developed these potential infections. We don't have a scientific rationale for why inhibiting TIGIT may increase infections. These were multifactorial Grade 5 events, unfortunately, so the investigators could not rule out tiragolumab or anti-TIGIT as a cause contributing to this. If we look at those infections resulting in serious adverse events resulting in deaths, they're actually balanced on the 2 arms. Infections on the whole were balanced as well. But for these 2 cases, the investigators could not rule out that tiragolumab might have been an attribution. So for transparency, we have included them on the list.

Yes. This is standard practice, I have to say, but you must do it, I mean, for transparency. And the other 2 questions, the confidence in adjuvant setting -- in the adjuvant setting. I don't know. I have a bit of speculation here, but maybe, Ira, your call. Ray, yours. I don't know, crystal ball. I'll hand it over to Ray. We agreed that you take all the difficult questions.

I can pass it on to Heather.

Yes, yes. That's a good one, exactly, yes. I'll give you the try, Ira. Maybe, Ira, we agreed on this one and then maybe Ray as well.

Yes. Well, I think we kind of addressed this a little bit earlier, which is increasing -- I'm trying to get everyone to think in terms of indications as distinct immunological diseases. So within lung cancer or cervical cancer, anything else for that matter, you're going to have some patients who will be susceptible based on their immunological phenotype, immunological profile to these types of inhibitors more so than others. And the number of those patients will vary. But nevertheless, if you can find them, and at the moment, the only way we really have to do that is by looking at PD-L1, which, again, we're taking as an indicator of an ongoing immune response. That then becomes the subgroup of patients. And that may sound like it's obvious, but let me give you an example. It took many years, I think, for the oncology community to start thinking about tumors as genetic diseases, where the genotype of the tumor is the most important feature that specifies how you are and go about treatment. So for example, if you have a non-small cell patient with a G12C KRAS mutation and one that does not have that KRAS mutation, you would not treat the patient with the wild type or normal RAS with a RAS-mutant inhibitor, okay? We all take that now as obvious and something that we should do. It wasn't always that way. I think we have to kind of change our thinking about immunotherapies in a way to really put front and center as up there as important as the indication, well, what is the immunological phenotype associated with that, rather than just treating everybody in cervical cancer and seeing some large or small response and coming to the conclusion that, well, this is a susceptible indication. It's still an important consideration, but I -- when we decide what to do internally, our conversation is much about indication as it is about what do we believe the underlying immunology is in a subgroup of patients that can be defined within that indication. So I think, in principle, anything is on the table, but you will find different groups. I mean, the other example is MS-high colorectal cancer. We normally think of colorectal cancer as being refractory to immunotherapy, well, except for that one small group. Why? Because that one group manifests an immunological phenotype that's indicative it's an ongoing immune response. And I think it's quite clear that that's why those individuals which really represent only a few percent of the total of CRC patients respond. And -- but normally, again, we don't think of CRC as a response of tumor to immunotherapy.

Yes. Thank you. I think we should leave it there. And as that -- and thank you. All right. We are doing fine. We have one more question and one on the line. So we have one more minute. Keyur, you have the privilege of the last question on line here. Thank you for the questions, Diana.

I'll limit myself to one given the time. Just kind of given the strength of the OS signal on this study, how are you guys thinking about the potential for you to change the statistical plans kind of on the Phase III? Is it kind of something you are actively thinking about? Kind of, obviously, for a lot of the world beyond the U.S., the OS data will have a lot of kind of commercial relevance. So just kind of help us think through that.

Well, that's another one where I really have struggled with to whom to give the question. Maybe both Heather and Raymond, do you want to change the statistical plan based on what you have seen here for the trial? Yes or no? Well, you're the decision-makers here.

Yes. So I think that's a great question. It's really about what are we learning from the ongoing studies to help us with the Phase III studies, either from the CITYSCAPE study, even from our Phase I combination as well, I think it's certainly encouraging the overall survival results, but would we necessarily change anything at this time with the readouts coming up? And I think our team is discussing these things, but we have to have a good rationale to make these changes in these pivotal studies. So I think it's a careful discussion that we do have. And, Heather, other comments about that?

Just briefly, I'll comment in our SKYSCRAPER-02 small cell trial. OS and PFS are co-primary endpoints, so we already have a plan to look at OS there. And I would just comment that although the OS looks great in CITYSCAPE, the PFS also looks good. So we are learning from this study. But at this point, no major changes needed to the size of our trial.

Okay. Thanks a lot. A last, last one for you, Ira, from Marietta Miemietz. She was asking, maybe you can give you 1 or 2 minutes answer on a very broad question. What do we see about IO therapies in development which could really turn hotter or the -- change the development field to turn, let's say, low and cold tumors into hot tumors? What is, at the moment, on your plate, how you can basically fundamentally change what is out there?

Yes. I think that is the question that really stands between us and the next quantum leap in the field. How do you change these immune phenotypes, how do you generate a situation where a patient can mount an endogenous response, such that it can then be amplified. So when I think of cold tumors, I think of 2 distinct things. I think for the so-called immune-excluded tumor, which is representative of a patient who does have an ongoing immune response but the T cells are physically restrained and then inactivated by the tumor microenvironment, specifically the tumor stroma. And I think of the immune deserts, where there is no over -- indication that there is an immune response in general. Both of those represent different mechanistic challenges. First one, as I think we've spoken publicly about before, we think the TGF-beta pathway plays a major role. And so I think we just entered one agent into the clinic. There are some already that are out there. I don't know how the -- what theories behind their development, but this is an area that we're very interested in. But I think the bigger and perhaps even more interesting question from the point of view of initiating immune response of these immune deserts, where there is no evidence of an immune response at all, I think the most attractive idea for those is that these are tumors who are devoid of a really critical cell type that's important for initiating everything that happens in the cancer immunity cycle. And these are macro -- the dendritic cells. They're just simply not there. If you don't have a dendritic cell there, you can't initiate a response. So we have programs in place that are just now entered into development to try and solve that part of the problem. In other words, what can we do to actually make more and activate the dendritic cell compartment such that a patient can at least begin to mount their own response to their tumor that we then can use checkpoint inhibitors on top of that in order to amplify that response. That's quick, but I hope that would be reasonably there.

Absolutely. Thanks a lot. Heather, Raymond, Ira, many, many thanks for your help today on presenting the data and clarify any questions. Keyur, thanks for the last question from your side. And Marietta, thanks for your question online. Wishing you a nice day, a nice weekend. Stay safe. All the best. Thanks for your interest in Roche. Bye-bye.