Alicia Morgans: Hi. I'm so excited to be at ESMO 2022, where I have the opportunity to speak with Professor Gert Attard of the University College of London, as well as senior postdoc Marina Parry, also of the University College of London. Thank you both so much for being here.

Marina Parry: Thank you.

Alicia Morgans: Let's get started talking about STAMPEDE and multiple presentations at ESMO this year. Let's start with you, Gert first, you gave a first presentation on the metastatic population treated with abiraterone with or without enzalutamide. Can you tell us a little bit about that presentation?

Gerhardt Attard: The two separate phase three trials. In the first trial, 1000 metastatic patients were randomized between ADT and ADT abiraterone. And then the other trial, 1000 patients were randomized between ADT, ADT abi and enza and the two recruited sequentially. There's about a six-month gap between completing accrual to the abiraterone trial and start of the abiraterone-enzalutamide trials, so no shared controlled patients, but all recruited to the same protocol with the same eligibility criteria. We designed the abi-enza trial to test for a target hazard ratio of 0.75 and then using indirect approaches, meta-analysis methods compared to the abiraterone trail and if we're seeing a signal to trials that randomized to enzalutamide alone. And our target there was a difference in hazard ratio of 0.15, which is large, but we felt that's what was required to justify the increased toxicity and cost of two drugs versus just one. We've done that analysis. Adding abiraterone and enzalutamide significantly improve survival, but the hazard ratios are the same as adding abiraterone alone. There is no evidence of a difference in treatment, in fact. There's no evidence of between trial heterogeneity.

Our conclusion is we should not combine abiraterone and enzalutamide for metastatic HSPC. And this is in keeping with the data we presented last year for locally advanced disease. And it's also in line with data that's being presented in metastatic CRPC.

Alicia Morgans: Yes. And I think although, the hypothesis was there, that there might be some, whether its synergistic effect or at least, some lack of overlapping effects, you'd have something additive. It's important to know that that does not really seem to bear out in clinical practice. And so very, very important to have that data, so thank you. Gert, I know you also have been involved in other work, really digging into the biomarkers and trying to understand the transcriptome of patients. And I'd love for you to present Marina so that she can share a little bit about the work that you and your team have done.

Gerhardt Attard: Yeah. Great benefits for patients, but outcomes remain highly variable. I think this is one of the big challenges in treating advanced prostate cancer, such highly variable outcomes. And one of the advantages of STAMPEDE is we had this whole spectrum from locally advanced to metastatic. Today, presented the metastatic patients. But within the abiraterone trial, we had both groups. And going forward to make progress, we're going to have to better select patients. That's why it's really exciting. And the same session just straight after me, Marina today presented our work on transcriptome analysis of the abiraterone trial. And I'll hand over to you, Marina, to talk through that.

Marina Parry: Thanks. Yes, as Gert said, we looked at transcriptomes within the abiraterone phase three trial and we got hold of the diagnostic blocks from as many patients as we could. And we were able to obtain transcriptomes from almost 800 patients, both non-metastatic and metastatic patients, randomized to either abiraterone or ADT alone.

Alicia Morgans: That's wonderful. And really when you got all that data back, you were able, obviously, to look at comparisons between the treatment arms. And you could also look at differences within the transcriptome and understand within a treatment arm if there were differences in outcome and what did you find?

Marina Parry: With all this data, we wanted our analysis to be as robust as possible. Prior to any outcome analysis being performed, we predefined a statistical analysis plan. And this plan set out the objectives that we had. Our primary objective was to establish the Decipher genomic classifier as a prognostic marker in the non-metastatic and metastatic populations, separately. We were also interested in trying to see whether there were any predictive signatures because that's hugely important. And we selected four signatures for that analysis. And to increase our power in that analysis, we combined the M0 and the M1 patients. The signatures that we selected for that analysis, the predictive analysis, we selected ARA AR biology; the Decipher genomic classifier, PAM50 from which is the breast cancer classifier; and a recently developed classifier from Decipher called PSC. Our main findings for the predictive analysis was that there is a consistent benefit of abiraterone across all these molecular subtypes.

Another objective we had was to look at the Decipher genomic classifier as a prognostic marker. And there the results are really exciting, because we do find a difference in outcomes in patients, who are split by median Decipher score. And in our cohort, the whole cohort, the median Decipher score was 0.77. Looking at patients split not only by Decipher score but also treatment, we find differences in a hazard ratio of 1.18 in the metastatic and 1.2 in the non-metastatic patients where an increase in Decipher score is related to a worse overall survival in the metastatic patients and worse metastasis-free survival in the M0 patients.

Alicia Morgans: Okay. This was really prognostic as something so if you've higher Decipher score, you're probably going to do worse. Was there opportunity to have a predictive biomarker where you could say abiraterone will be helpful in this group, but will not be helpful in that group. Were you able to predict that effect of therapy?

Marina Parry: For the signatures that we selected a priori before we did the analysis, we found a consistent effect of abiraterone. No differences found in the signatures that we selected to test for that effect.

Alicia Morgans: And that's also highly important, right? Because as we're thinking about choosing these treatments for our patients in clinic and abiraterone being one of the less toxic treatments, it's good to know that at least with this analysis, which is I think the largest analysis of its kind, in this type of a population. So far, this is very, very encouraging that we don't see a difference. At the same time, of course you have work to do in case there is a signature. And I think the ARA population or the ARA signature has been shown in other populations to potentially suggest benefit or lack of benefit in hormonal treatment. And I wonder, Gert, if you have any comments?

Gerhardt Attard: Yeah. ARA was the prime one we tested and there's no difference. The hazard ratios in ARA low, ARA high are identical for benefits of abiraterone. As Marina said, the relative benefit of adding abiraterone is consistent across all the molecular subtypes, but the absolute benefit is very different. And this is most pertinent in the locally advanced. At 96 months, the lower risk Decipher patients, very few have events and the benefit of adding abiraterone as a few percentage points was than the higher risk Decipher patients. The 40% or so have an event, have metastasis or death and that has improved as a result of adding abiraterone. I think in metastatic patients, who are going to stick with using ADT plus abi-enza- [inaudible 00:08:15], et cetera. But in localized disease, as we start to weigh the cost of treatment over treatment, as we look to treat a larger population, outwit the very high risk STAMPEDE group.

That's where I think we're going to have clinical utility for the Decipher signature score of identifying those who are high risk and they're going to have real benefit. And if you sit in front of a patient, who has mild cardiac heart failure or other considerations or you've started treatment and there's toxicity, should I stop or should I push on? That's where this is really going to be powerful, that ability to predict more accurately what's going to happen, where we have robust data up to 96 months.

Alicia Morgans: That's a great point. Really, if the incremental benefit is so small and the time on treatment would be two years for these patients, it's going to be so important to use these tools. At this point in time, I'd love to hear, and I'll direct this to Gert, and then I'll give you the second opportunity. Do you see that this data can be imminently, clinically useful in the situation such as you described or in any other? Is it something you can use in clinic today, Gert?

Gerhardt Attard: I can't, because I don't think I'm going to have funding to use Decipher. But I know many colleagues who've seen the data over the past two days, who are working in the US and have been using the Decipher assay extensively for their patients at diagnosis are talking about extending that use to metastatic patients. Think about how we're going to implement this in Europe. We're probably going to need to analyze additional cohorts. And I think that's, well, we're already having those discussions. Over the next few years, we will generate additional data. And I'm pretty confident that in the locally advanced patients transcriptomes are going to improve our ability to identify who should receive treatment, for whom is that balance of toxicity versus benefit going to weigh in favor of treatment versus those who we probably can avoid that additional treatment.

Alicia Morgans: That's very, very helpful and something of course for us to look forward to work for you to do in the lab. Marina, and as you think about that future work, which direction do you think your team is going to go in next? Where will you take this onto next steps?

Marina Parry: We're soon going to be having data from the docetaxel arm or STAMPEDE, so we'll be looking at that independently but also in combination with the abiraterone arm data. But as a group, we're also interested in other types of molecular classifiers, so not just transcriptomes, copy number changes and pathology based assays. And I think ultimately as it has been shown in many types of cancers, maybe a multimodal assay will be interesting. And I think the way this is going is a large integrative study of all these biomarker features to really drill down into the biology of prostate cancer, which having worked in this field a few years is still relatively misunderstood.

Gerhardt Attard: The other discovery, which is really exciting us, is we had 58 signatures, we've tested them all. As Marina said, there were four primary signatures. And then, the other 54 we had the more stringent p-value for calling significance. That was a cutoff of 0.001. There are signatures in the metastatic patients that come out as prognostic, cell cycling, loss of P10, loss of P53. There are ones that are not prognostic like ERG for example, probably starting to draw lines under AR-V7 biological pathways that are not associated with worse outcome was the others that's now the strong biology to support that observation and now clinical qualification with our dataset. In the non-metastatic patients, we see different signatures associating with the outcome primarily related to immune activation. And we've performed a formal test for interaction between outcome and those signatures and metastatic stage.

And we find that we have statistical evidence that signatures, three in particular, that are prognostic in the localized are not prognostic in the metastatic. That starts to introduce this paradigm shift that the biological pathways driving progression in localized disease are potentially different from metastatic disease and particularly immune escape resulting in metastatic seeding. Now, that could have therapeutic implications. And our current way of developing drugs is testing advanced disease and slowly move backwards and then onto localized. But there are different biological parties we should target that we need to start thinking differently about the two disease states. This is speculative, but I think the more we look at this result, spent a couple of months chewing it over, we're excited by that. And I think as we discuss it with our colleagues in the community, we may really develop some new insights.

Alicia Morgans: Well, I think that makes sense too. As we think about localized disease, having different treatment approaches, different things being successful, and even when we look at things like radiating the prostate that it might be better in one situation but not when we reach a certain separate situation. It can't be based only on the distribution of where the cells are, right? It has to be perhaps because of the drivers of these different disease states. They really are different disease states. And I am very excited to hear about this and I appreciate you sharing that with us. And I really look forward to seeing where you and the team take us because STAMPEDE has been giving us data in large magnitudes over the last number of years. But this is really, I think, one of our biggest and most exciting areas to go next. Where do we better have the opportunity to understand the biology? And you too, I think may be just scratching the surface. Thank you so much for sharing all of this today and congratulations on a fantastic ESMO 2022.

Marina Parry: Thank you.