Alicia Morgans: Hi, my name is Alicia Morgans, and I'm a GU medical oncologist and Associate Professor of Medicine at Northwestern University in Chicago, in the US. I'm so excited to have here with me today, Dr. Tom Powles, who is a Professor of GU oncology, and the Director of the Barts Cancer Center in London, England. Thank you so much for being here with me today Dr. Powles.

Thomas Powles: Alicia, thank you.

Alicia Morgans: Wonderful. Well, Tom, I really wanted to catch up with you on, actually, quite a nice presentation at ESMO 2020, our virtual meeting, you spoke in one of your presentations about the DANUBE trial, the outcomes of that Phase III trial. Could you tell us a little bit about the study design and what you ultimately found?

Thomas Powles: So the DANUBE trial is a randomized Phase III study. It's a study looking at frontline, metastatic, urothelial cancer. Patients who have not been previously treated with chemotherapy. Chemotherapy is the standard control arm, that's what we give to patients. Six cycles of gem/cis or gem/carbo, and the median survival is about 12 to 14 months for these patients. Immune therapy is the standard second-line therapy. Although more recently, maintenance durvalumab directly after frontline chemotherapy looks attractive, too. In the DANUBE trial, it's one-to-one-to-one randomization of chemotherapy versus durvalumab, which is a PD-L1 inhibitor, versus durvalumab plus tremelimumab which is a CTLA-4 inhibitor.

We know the PD/PD-L1 inhibitors have established activity in, even in frontline disease, atezolizumab and pembrolizumab. We don't know much about the effect of the addition of the CTLA-4 inhibitor. The trial was robust with a thousand patients, 41 months of follow-up. The results showed, in the primary endpoint for the durvalumab alone, was durvalumab in the biomarker positive population and the hazard ratio was 0.89, which was not statistically significant. The response rate was 23%, significantly lower than that seen with chemotherapy, and the progression-free survival was shorter as well. The shape of the capsular marker showed the chemotherapy did really well at the beginning, reflecting that high response rates. Immune therapy kind of catches up, but it's not enough for a positive trial.

So essentially, durvalumab alone, the biomarker didn't really help that much select the patients because the durvalumab in the ITT population and exploratory analysis was 0.99. So there's about a 10% enrichment associated with the biomarker. That's not enough. And so essentially what happened was the chemotherapy did better initially, the biomarker didn't select the patients as much as we had liked, the immune therapy catches up in the end, but it wasn't enough. Durvalumab and tremelimumab, the combination was tested in the ITT population. That's the all-comers. And again, the shape of the curve was very similar. Chemotherapy did best at the beginning, the durvalumab and tremelimumab caught up. The hazard ratio was 0.85, which was encouraging because it was in an ITT population and it didn't have the advantage of that PD-L1 enrichment, which durvalumab alone had.

Actually in an exploratory analysis, when you looked at durvalumab plus tremelimumab in the biomarker positive population, the hazard ratio went down to 0.74, and that is not statistically significant. But had we picked the biomarker positive population, as we did in the monotherapy, the durvalumab and tremelimumab, had that being the endpoint in the arm, which we did talk about many times, had that been the case, well, the trial would have been positive. The reason we didn't do that. Is back in the day, right? At the beginning of the immune therapy journey, we thought CTLA-4 would pick up the biomarker negative population. So biomarker positive gets single-agent therapy, biomarker negative, combination therapy. That turned out to be an oversimplification. It looked like the CTLA-4 patients actually had further enrichment in the biomarker positive cohort, and actually, it looked like the immune therapy, single-agent PD/PD-L1's were not as active as we originally hoped. We were hoping for hazard ratios of 0.75, not 0.89. And that essentially summarizes what happened from the efficacy perspective of the study.

The tolerability showed durvalumab was better tolerated than chemotherapy, but actually, the addition of tremelimumab to durvalumab, increased the toxicity quite significantly, still less than chemotherapy, which I think is relevant. And so to summarize very briefly, the trial missed its two co-primary endpoints. Durvalumab was not better than chemotherapy in the bio-marker positives, largely because the chemotherapy was better at the beginning. Durvalumab and tremelimumab, in unselected patients, didn't beat chemotherapy but had we use the biomarker positive population, we would have gotten a different result. We didn't do that so it was a negative trial.

The results suggest two things. Number one is that the biomarker is not selecting patients well enough, this biomarker is not selecting patients well enough to beat chemotherapy as it currently stands. So we either have to develop better drugs or combinations, or better biomarkers. And one of the exploratory analysis suggested actually tremelimumab, the CTLA-4, added to the activity. And there are ongoing trials looking at ipilimumab and nivolumab in this setting, and perhaps if they use a biomarker-driven approach, those trials may be positive for the future. And that in a meta-analysis would be really powerful. So you'd have a two randomized trial showing CTLA-4 plus PD-L1 in biomarker positives outperforming chemotherapy.

So, I hope DANUBE is the springboard to changing practice for the future. I hope that we've shown that tremelimumab adds activity. I hope that we've shown that we need a more specific biomarker approach if we want to use monotherapy. And I hope trials in the future will take CTLA-4 inhibition forwards. I hope we can replace chemotherapy in that biomarker positive population, and I also hope that we can move durvalumab, tremelimumab, CTLA-4, PD-L1 earlier in the disease process and cure more patients. So a negative trial, with really positive, I hope, are messages for the future.

Alicia Morgans: I agree. And you know, we talked minutes before we started recording about just the compliments and the congratulations that should be given to everyone, especially the patients for enrolling in another Phase III trial in the metastatic urothelial cancer population setting where we had had really a lot of difficulties getting patients into trials for many years. So congratulations on completing this trial, and ultimately, as you said, just because you didn't meet those two primary endpoints does not mean we will not learn anything from this trial. So along those lines, similar to the way that you and the team did in the JAVELIN trial, for example, are you exploring other biomarkers to understand whether there may be other ways, maybe more sensitive and better ways essentially, to identify patients who may respond either to a PD-L1/PD-1 approach, or a CTLA approach, or the combination. Again, similar to the way that you guys did in JAVELIN?

Thomas Powles: Yeah. So it's a really interesting question. If I may just take a second to step back from where we are. So the 361 trial, the pembrolizumab study had an almost identical design from the monotherapy perspective, pembrolizumab versus chemotherapy, in the atezolizumab study, that 130, that had atezolizumab versus chemotherapy. Those three trials have shown that but all three biomarkers doing actually completely different things. And I was speaking to some colleagues earlier today, and what we talked about was, saying that we need to stop using the saying PD-L1 positive. We need to start using, which antibody was used, and how was it used? So 142, 22C3, 288, 142, whichever it might, 263, whichever it might be because 263 and 142 look like they're doing completely different things, and by bunching them all together, we're probably doing, we're simplifying things, but it's an oversimplification because the 142 results, the atezolizumab results, their hazard ratio is currently 0.68. Whereas the pembrolizumab results, come in at 1.02. And there's a huge difference between those two.

There's no question that pembrolizumab is less active than atezolizumab, no one's suggesting that. So it must be the biomarkers that are driving those changes, and those differences look too big to be statistical noise. And actually, when you look at the ITT populations, for all three trials, not the biomarker select, pembro comes in at 0.93, durvalumab is 0.99, and atezolizumab at 1.02. So they are all within a basket of each other. They're all doing very similar things. So what this tells us actually, quite robustly, is the drugs have got more similarities than differences, but the biomarkers are all over the place. And actually what we need to do is get better biomarkers.

And I put it to you that even the 142 biomarker is not yet statistically significant for survival. It may or may not be in the future. I don't know. But is it common? Am I confident that I should be treating my patients with this approach rather than chemotherapy followed by maintenance durvalumab? No, I'm not. I'm much more confident. And the reason why is, for every 10 patients I treat with a frontline disease, two will respond with single-agent immune therapy, but five will progress. And of those five that progress, three won't be on sequence chemotherapy and I'll lose those three. That's my opinion. I've treated quite a few patients now with this approach, and therefore it's much easier to give the chemotherapy first, getting controlled disease, and then give maintenance durvalumab. And we have to clarify the frontline biomarker issue.

Now, if durva/treme or I hope if ipi/nivo comes in with a hazard ratio of 0.70 or something terrific in the biomarker positive population with long-term durable remission with that approach, then that becomes competitive for the maintenance durvalumab approach. And I think that's a really important message. So your question was, are we, where are we with other additional different biomarkers? Well, the first thing is, we haven't sorted out the first biomarker because it's all over the place. So let's do that first because clearly, 22C3 is not enriching for responders at all, so that we probably need to put one side. We probably need to actually look at the 288 biomarkers, the 142. And then as you know, the 130 data, the atezolizumab data builds tumor mutational burden on top of 142, and there's only one in eight patients, but the hazard ratio is 0.3, and that looks very provocative.

We did the same in the JAVELIN 100 trial. We were unable to reproduce it in the same way, but again, the PD-L1 biomarker was slightly different. So we do need a huge approach to try and identify these winners. It would be really nice to give patients upfront immune therapy. I'm not against it, I'm all for it. I've been doing it for a long time, but currently, I'm not confident enough with the biomarkers that we have, to do it to unselected patients. We need better drugs and better biomarkers, and until we get that, we should be giving chemotherapy first, getting control, and then once you're in control, give the immune therapy.

Alicia Morgans: I certainly don't argue that point at all and I use chemotherapy first, actually, every time that I can. I think my, my question is really around the poorness, as a single biomarker of a PD-L1 or a PD-1 antibody. And I think that, yes, when we coupled that with TMB, we did seem to have some better predictive value there, but I think that we still have a long way to go. And I ultimately think that for patients who are robust enough for chemotherapy, are fit enough for chemotherapy, an approach that would give chemotherapy and then potentially use a biomarker to determine whether or not those patients can actually benefit from additional immunotherapy may be the best way to go because certainly these agents also have side effects. We've had some pretty catastrophic events that have occurred, rarely of course, but they can occur.

So if we ultimately could make a biomarker, whether it's a biomarker single, biomarker, or maybe a combination of biomarkers, perhaps a signature would be required, whatever it is, it must be more complex than the single antibodies that we are using now, that really are not telling us what we ultimately want to know. And I think it would be great if we could just even use chemotherapy, and then one of these agents. Perhaps one day replacing chemotherapy, but I think that the body of evidence really suggests, as you said, that chemotherapy is effective and when patients are fit, we should be using that first. What do you think about the evolution of biomarkers, I guess? Not just relying on these because we had them for years. They don't seem to be doing the trick.

Thomas Powles: Yeah. I love the phrase you used, which is, they are not telling us what we want to know. It's a really nice simplification of where we are with it. What do I think about the evolution of biomarkers? Well, I was involved in 211, which was the first biomarker problem pro-child. And that was the 142 biomarker, which by the way, is working well now in the frontline setting, but didn't work well in the adjuvant setting. It's unusual to have a biomarker and the same disease that's working well in one setting, but not in other settings. That makes me nervous in its own right. But I am, in terms of where we are moving on the evolution of these biomarkers, I am not convinced that TMB as an individual biomarker is any better than PD-L1. So that's the first thing to say.

I think there is as much inconsistency. We've done neoadjuvant work with TMB and we haven't shown it to work particularly well. We did some work in the JAVELIN 100 program. It actually looked pretty good there, to be honest. And there was some work that we were involved with others, of course, with the pembrolizumab program, which showed that TMB is there. Doesn't look better than PD-L1. There may be if you combine them together, it's a bit better. But what about the second generation of biomarkers? Well, I'm interested in different biomarkers, and one of the things which I'm impressed with is that I'm impressed with CDA. I'm impressed with TFX with gene signatures. I like both of those. I'm also, from a mutation perspective, I like the DDR signatures associated with, which seemed to be associated with tumor mutational burden, but maybe more specific. So I'm quite attracted to those.

But also in the JAVELIN 100 program, and we presented some data, the biomarker data from JAVELIN at ESMO that just went, and essentially what that showed was that not just could we identify those responding biomarkers, PD-L1, TMB, TFX's signatures, DDR signatures, innates and immune signatures, a broad spectrum of immune signatures. But we also showed that there were mechanisms of resistance as well. Angiogenic signatures were attractive, not signaling TGF beta. So the picture, actually, it looks like there are multiple players of the innate and adaptive immune system involved in the response. There are also some breaks in the process which may be independent of the widely established immune repertoire, such as angiogenic type signatures, and pulling that together, actually, if we're going to pick these responders, it was always going to be an over-simplification just looking at PD-L1 or TMB. I kind of like circulating biomarkers and certainly ctDNA is something which I'm interested in as well, and that story I think, we'll run.

Alicia Morgans: Well, I agree. And like I said, I love the work that you did in JAVELIN with all of those different, even the FC portion of these antibodies. I mean, this is all really exciting. And so despite the primary endpoints not being positive in DANUBE, you have I think an opportunity to look at all of these within this data, if this is something you and the team want to pursue and really educate the field, maybe reanalyzing multiple studies with these really intriguing biomarkers in any events.

Thomas Powles: What DANUBE I think has done, is it has really crystallized that frontline approach because DANUBE and 361 have been pretty conclusive and it really has said, "Hold on a minute everyone, this isn't right, let's find a better alternative." And that's, I know it doesn't publish as well, but it is as important as an equivocal result, a badly performed trial, a hazard ratio of 0.65 with wide confidence intervals with imbalanced populations. I'm not suggesting another trial has that, but DANUBE has come in and said, "Listen, frontline immune therapy is not as good as chemotherapy, it is not as productive as getting in control of the disease, the biomarkers don't look great, go and find a better way of doing it, and by the way, the additional CTLA-4 might just be that." And that's a really important finding for the medium and long-term, not for today and tomorrow, but for the medium and long-term.

Alicia Morgans: I agree. And I wonder, as we start to wrap up, there was a really nice, although chemotherapy, of course, did do better in the beginning, the tail of the curve suggested that this combination of the durva plus the treme, may like you said, the response rate is certainly enhanced with that addition and that these patients may have a nice tail on the curve. You did mention a study that is, I think looking at ipi/nivo in this setting. Do you know if durva/treme will potentially move in that direction as well with an upfront chemotherapy approach or other studies upfront chemotherapy, and then a sort of a maintenance approach with a combination?

Thomas Powles: It's my personal opinion that we need to just be looking at this slightly differently, and the antibody-drug conjugates CTLA-4, PD-L1, they all seem valid targets to me at the moment.  And you put erdafitinib in there, and of course chemotherapy. And there are lots of ways in which you can beat frontline chemotherapy as it currently stands. The opportunities and I've been saying this a lot, we've come to the end of a first chapter of bladder cancer. It has not been as successful as I would have liked. We've established chemotherapy is difficult to beat upfront. We've established immune therapy is better than chemotherapy second line. We've established sequencing immune therapy directly after chemotherapy is better than waiting for cancer to come back. We've done that successfully. We've not established effective biomarkers. We've not established effective immune combinations yet, and we've not yet established a program where we can confidently beat frontline chemotherapy.

But we do now have antibody-drug conjugates, we've got personalized therapy without erdafitinib, and indeed immune therapy to some extent, and we've got to add other ADCs, SG, EB, for example. We now have the tools in our armory to beat frontline chemotherapy, but as you suggested, also go into the perioperative space and cure a whole lot of patients in that place too. Remember we're not doing terribly well. Most patients don't get neoadjuvant chemotherapy and the benefits of it, neoadjuvant chemotherapy and a meta-analysis are somewhere between 5% and 13%, depending on which one you read. So, the opportunities in bladder cancer, if you compare it to kidney cancer or prostate cancer, your mean survival is only 14 months, the opportunities to improve outcomes for patients are huge.

Alicia Morgans: I completely agree and I love hearing your vision for the future, and of course, hearing about what you and the team have been presenting at ESMO this year. Thank you so much for sharing your insights and for giving us a fresh take on what at first glance may not seem as exciting. It is absolutely as exciting as it appears when viewed from your perspective. So thank you for your time.

Thomas Powles: Thank you very much. It's lovely seeing you. Keep well.