Transcriptome-Based Prognostic and Predictive Biomarker Analysis - The DECIPHER Genomic Classifier for Prostate Cancer - Ashley Ross
January 24, 2023
Ashley Ross, MD, Ph.D., Associate Professor, Department of Urology, Northwestern University, Feinberg School of Medicine, Chicago, Illinois
Alicia Morgans, MD, MPH, Genitourinary Medical Oncologist, Medical Director of Survivorship Program at Dana-Farber Cancer Institute, Boston, Massachusetts
ESMO 2022: Transcriptome-Based Prognostic and Predictive Biomarker Analysis of ENACT: A Randomized Controlled Trial of Enzalutamide (ENZA) in Men Undergoing Active Surveillance (AS)
An Ancillary Study of the STAMPEDE Trial Assessing the Clinical Qualification of Transcriptome Signatures for Advanced Prostate Cancer Starting ADT +/- Abiraterone Acetate and Prednisolone - Gerhardt Attard
ESMO 2022: Clinical Qualification of Transcriptome Signatures for Advanced Prostate Cancer Starting ADT with or Without Abiraterone Acetate and Prednisolone: An Ancillary Study of the STAMPEDE AAP Trial
Insights from the STAMPEDE Trial Platform - Marina Parry & Gerhardt Attard
Alicia Morgans: Hi, I'm so excited to be here at ESMO 2022 where I have the opportunity to speak with Dr. Ashley Ross of Northwestern University. Thank you so much for speaking with me.
Ashley Ross: Thank you.
Alicia Morgans: I've really been so delighted to hear all of the progress that you and your team have made in integrating Decipher analyses into ongoing phase III, phase II clinical trials to really help us understand biomarkers and these signatures that we might try to understand some prognosis, or perhaps, eventually predict outcomes for patients with different treatments. Can you tell us a little bit about the work you and your team presented here at ESMO? And then we could dig into some of the other work presented at ESMO and published recently that also uses this sort of analysis.
Ashley Ross: Thank you very much and I'd be happy to. As you mentioned, the Decipher platform has a couple things on it. It's a transcriptome-wide array, so there's 40,000 genes being analyzed, or transcriptomes being analyzed, and they have obviously their strong prognostic signature, which they often call their genomic classifier. And then they've, over time, developed other signatures that are maybe uncovering things about the tumor biology, whether that's antigen receptor activity scores, or scores that look at cell origin, there's the PAM50 based on breast cancer. They've started to analyze this retrospectively on prospectively conducted trials.
And so, we did two things here. One is, now that Decipher is so commonly used from the real-world data, they have over 50,000 men with biopsy specimens where they have the whole genome transcriptomics. In one poster presentation we just wanted to define the landscape., As you go across disease risk stratification in localized and locally advanced disease, what does it look like? And maybe as you would've expected when you have a lower stage tumor, say NCCN risks are low through favorable and intermediate risk, their antigen receptor activity scores tend to be higher. As you go towards NCCN high risk, NCCN very high risk, antigen receptor activity actually tends to go down a little bit, PSMA, as itself, goes up as you go across and immune activity measured computationally goes up as you go from low-risk disease towards a higher risk disease.
How does this now lead into different trials? We looked at the ENACT trial, and that, as you know, was a phase II trial that randomized people who were surveillance candidates to either getting a year of enzalutamide mono agent or just being on surveillance. That trial, which was recently published, was positive for its endpoint. It looked at a progression-free survival on surveillance, and it found that if you gave people 1 year of enzalutamide, then you had about a 50% reduction in how often they would progress on surveillance. The trial had some criticism on, well, is this too much toxicity for these men? These men often don't need treatment at all. And so this was one area where it would be rich for biomarker analysis.
We presented, here at ESMO, an analysis of roughly half of that phase II trial we had genomics on or could derive genomics, and what happened to those men? First, Decipher itself, the genomic classifier, was prognostic. So, as you'd expect, and as we knew from previous studies, would the molecular biology tell us about who would just do worse in general? That was true, but it was nice to see it validated in a prospective clinical trial setting.
And then we also looked at, were there men that derived the most benefit from enzalutamide? In the active surveillance population? And when you look at progression to treatment, it can be a mixed endpoint, and so the one thing that I thought for myself had maybe a lot of rigor would be negative biopsy, because you're getting biopsy at 1 year, biopsy at 2 year. Who would have a negative biopsy even off enzalutamide therapy at the 2-year endpoint? And, as we might have predicted, if you had a high antigen receptor activity or if you were luminal subtype, and these people, there's a high correlation there, that they're antigen driven, the chance you'd have a high or profound response to enzalutamide was very high. The odds ratio was about tenfold if you were a luminal subtype that you would have a negative biopsy at two years if you got enzalutamide.
It was small. This is a phase II trial of the patients that we could then analyze for biomarkers. You're now in the tens of patients, so it's about 100 patients being analyzed in total, but still provocative and hypothesis generating. And it's giving you the idea that, beyond the prognostic signatures that could help you risk stratify who might need intensification or deintensification of therapy, you might be able to design trials that are pre-qualifying on potentially predictive signatures. That might allow, maybe, more thoughtful design and more bang for your buck for some of these treatments, particularly in the localized setting. In ENACT, it was an extreme example because you're looking at people with favorable-risk disease.
Alicia Morgans: I think this is the perfect data, the perfect population to do this work, because I think one of the criticisms of ENACT is, is the bang worth the buck? Is it really giving us something that patients need? From a population wide perspective, perhaps there were more side effects. This was a bigger payout rather than a benefit. But if we are able to identify a subset, it seems like this is a possible strategy for patients potentially moving forward. I mean we do this in DCIS, we do this for women with that particular histology to prevent the development of invasive breast cancer. And so, the strategy, from my perspective, is one that is sound if you can find the right patients.
Ashley Ross: I would say that there is a maelstrom, at least on social media and other forums, on, should we be studying this population at all? I think that what we've realized in the favorable-risk disease is, maybe shifting a little bit towards the more worrisome favorable men, if you will. That could even be a high Decipher person who has favorable-intermediate risk disease or even high-volume low-risk disease. And then, just as you mentioned, if they're very biomarker selected or at least now we have some hypotheses of what we would select them on, maybe then it is worth a trial in that setting. I think that the future might be many exploratory trials that are showing signal, that are then bringing it out to a bigger audience. And so that was right.
It was also nice hearing you mention that, this morning, both Dr. Parry and Dr. Attard talked about some of the STAMPEDE data. I think that the first thing, again, was shown there, and this is obviously a later stage disease, but again, that the prognostic power of some of these molecular markers like Decipher, really large differentiations. If they stratified around, they're mostly were high risk Decipher, but around the median, so it was about 0.7 higher and 0.7 lower for Decipher score, if that genomic classifier was showing you a low score or high score, particularly among the local locally advanced men or N1 men, you could see some folks that had low Decipher and really did quite well in ADT alone and some men that definitely needed intensification. And then among the Decipher-high score men who had metastatic disease, you could see people that, even with the two therapies, abiraterone and ADT, that there was still a need to do something else in those individuals to get a cure.
I think the predictive signatures are still being more intensely studied there and I think that they still have a lot to do with development. As I was mentioning, one of the other landscape abstracts we were showing was an idea of, well, where should our cut points be? The Decipher genomic classifier has been heavily studied and prognostic value is well developed, but we have to link it to more therapeutics to figure out what should our cut points be. And I think understanding the landscape of some of these signatures as we do trial design will help us be thoughtful about who we're recruiting in, do we want to make at least sub-stratify around some of those biomarkers?
Alicia Morgans: I couldn't agree more. I think that not only in which patients and with which therapies, I guess, is really your point, but in which patient population, because it was interesting in the STAMPEDE data that the relative benefit seemed to be pretty similar across the population. But when you look in certain populations that have a very high risk, the absolute benefit may be higher than in patients where the event rate is so low anyway, that, to your point, a relative improvement may be such a small increment for so few events that it may not actually be worth it for the patient. So, understanding this across the population in these different disease settings and by treatment is something that I really look forward to hearing in the future.
Ashley Ross: Yeah, absolutely. The one other thing I would say that was provocative from the STAMPEDE data is, as they went from M0 to M1, they saw differences in some of these gene expression signatures. We were talking earlier about one of them was the immune signatures that seemed to be less critical as you go towards M1, but in M0, were somewhat enriched. We saw, as we looked at early stage disease, like American Joint Cancer Commission stage 1 through stage 3, that they don't seem to be that critical in early-stage disease, you're low-risk patient or your favorable-intermediate risk patient, but the immune activity then kind of spikes. So there might be a window of opportunity there.
As a final point, as you mentioned now going to the higher spectrum of high-volume hormone-sensitive disease, even, some of the data showed this morning that had not been biomarker analyzed as much, was showing that with triplet therapy, if there's a subgroup that's fairly sizable, even getting chemo, that if they're not reaching a good PSA nadir, they just don't do well.
Docetaxel, even that trimodal therapy, may not be enough. And I think some of the other work that's come out that is trying to classify who has RB loss, what's the PPV3 status, will help us be more thoughtful about what we have to intensify on in that hormone-sensitive setting. So, as you point out across the spectrum, very exciting, and I'm sorry to spiral a little bit, but there's a lot of data here that's bringing up that we have to be very thoughtful about, not just the patient's prognostic risk, but some of these other molecular signals. Every trial design should have it, and I think that sooner than later we're going to start to see some actionable results in our clinic from that.
Alicia Morgans: I completely agree. And I think that's just the population, that patient population, where, as we see they're not getting sufficient outcomes, even from a triplet combination, perhaps driven by p10, RB loss, p53, maybe we're using the wrong chemo. So studies are being designed around using carbo-cabazi, for example. Or maybe we need something completely different and there are other studies looking at abemaciclib in this population. And so, how can we make up that difference, because maybe docetaxel is just not the right agent, or maybe we need something else with the docetaxel? So whatever it is, we have work to do and it's really exciting that the biomarkers are helping us understand the biology so we can be thoughtful about those next steps. Yeah. If you had to sum it all up, what would your summary be?
Ashley Ross: I think that we now see the power of the molecular prognostic markers that really go, I think, far beyond, and add to, what we already knew about the prostate cancer. This is really going to become mainstream, as it's already been, but not only in the localized setting, but also in the metastatic setting with the STAMPEDE data. And then the second summary is, I think that we have to really be driven into precision medicine, because we have the tools now to make that happen. It's just about understanding it and matching the patients with their biology. And we're seeing some early signals from that. Trials like ENACT, as controversial as it might have been, in some ways, also was very rich for showing us some of these signals with the Decipher platform and androgen receptor activity and I think we're just going to see more of that as we go forward.
Alicia Morgans: As usual, I agree, and I think no trial is done in vain and we need to understand from those trials why they succeed, why they fail, which patients may benefit even when there is an overall questionable signal. And in that case, this was a positive trial, so there's even more to dig into. But every study, I hope, will be able to really understand at a deeper level, because that will inform the study of tomorrow. So thank you so much for continuing to support this work and for sharing your thoughts today.
Ashley Ross: Thank you so much for having me.