Michael Zelefsky: Thank you very much. Thank you, Dr. Spratt, really, for the opportunity here to talk about what I think we all know is an extraordinarily interesting report on really a large number of patients who've had Decipher scores. And what I think it really provides us is extraordinarily new insights in what the Decipher score is about and really highlighting some of the discordances that we know exist between the Decipher score and the classic NCCN risk group. And you have done really important work in this area, highlighting this discrepancy, in particular for intermediate-risk patients, which you published on where patients have had intermediate-risk NCCN disease, which we would call Gleason 7, 3+4 or 4+3. And nevertheless, they had some discordance with the Decipher score where they had higher-risk genomic classifier scores. And this has opened up really a clinical conundrum as to how to manage these patients with this kind of discrepancy. And really, another question that I think is really the thrust of this report is, what's driving this discordance? Why are we seeing patients that have NCCN risk group of, let's say, intermediate but high-risk genomic classifier scores? And what's behind the discrepancy?
So, for this study, over 200,000 prostate biopsies were evaluated, taken from 2016 to 2024, where we had the information on the genomic classifier, the Decipher scores, stratified by their score of low risk, intermediate risk, high and very high risk. And we also had an opportunity to look at behind the scenes in the grid analysis of the adverse molecular feature signatures that were present. And in particular, the most prevalent ones were the p53 mutation versus wild-type, PTEN inactivity versus active, the RB gene loss versus intact, neuroendocrine-like gene versus adenocarcinoma, and the androgen receptor activity, low or not. And the prevalence of these adverse molecular features were examined within the Decipher score and looking at the NCCN risk groups as well.
Now, what was right away interesting, which is consistent with what we know as well, is this discrepancy between the Decipher score, the genomic classifier, versus the NCCN. And what's highlighted here is that if we look at the very high-risk or high-risk genomic classifier scores, we see that 14% of the NCCN low-risk classifications or designations actually had high or very high genomic classifiers. And so, when a NCCN low-risk classification is there, 14% of those patients had higher-risk Decipher scores. And when we look at, as well, for low- and intermediate-risk patients, where we are looking at the high-risk, let's say, NCCN, 33% of NCCN high-risk patients, as we see here, 33% of them, had lower- or intermediate-risk Decipher scores, despite the fact that they were called high risk by NCCN. That's a significant number. So, we're seeing 33% of the NCCN high risks are really reclassified as lower or intermediate risk. And when we look at the very high risk, the genomic classifiers, 14% of those low-risk patients are actually having high or very high risk. And that, obviously, presents as a clinical conundrum because you have patients that have a low-risk disease. It's Gleason 6. And suddenly you get a Decipher score. And in 14% of those patients, they are reclassified as high or very high risk.
Along these lines, when you dig a little deeper into the data and you're looking at the adverse molecular features, and specifically these mutations that are noted here, p53, PTEN activity, inactivity, and the RB gene and the like... And here we're looking at the discrepancies. This side of the slide is focusing on NCCN low or favorable intermediate, and the columns on the right side, NCCN very high or high-risk group. And when we focus on really the discrepancy, and let's turn our attention to NCCN low or favorable intermediate, the p53 mutation seems to be a major driver here. This is, again, low or favorable intermediate risk, and these have high or very high genomic classifiers or high Decipher scores. And the driver here, especially in very high-risk genomic classifier, 48% demonstrate a p53 mutation, although they were classified as low risk or favorable intermediate risk. And when we look at the other gene genomic abnormalities, the PTEN inactivity was also an important driver and, in particular, the low AR activity as well.
Now, if we turn our attention to the discrepancies noted for those who were designated as having high or very high NCCN classification, the classic Gleason 8 to 10, the discrepancies with lower- or intermediate-risk Decipher was basically driven by these really lower-than-expected genomic aberrations. And this is the p53 in a much lower percentage, despite the classification of high or very high risk. And you see this across the board. And so, what seems to be the driving, real reason for these discrepancies are the presence of these mutations or lack of them.
And along these lines, with looking at p53, PTEN, and AR driving these discordances, you could see that for the high or very high Decipher scores, especially p53 is a major driver, PTEN inactivity and the low AR activity. And when we, in particular, look at the NCCN very high-risk group, we could see here that the major drivers are the really greater prevalence of the p53 wild-type and PTEN activity instead of inactivity and they higher AR activity. So, these seem to be the important genomic drivers that are explaining the discrepancy.
And so, one of the, also, important features of this report is a really exciting linkage that was able to be done as part of this collaborative process, where the what we call real-world data was linked to the genomic scores. And so, of these 200,000 prostate biopsies, link data to these real-world data, which is based on the medical claims, pharmacy, and EHR, the electronic health records, across payers/sites, as we can see here, 15,000 were included. And albeit that there was a follow-up that was relatively short, of only 1.1 years, there was still some really interesting data that had come out of this, especially looking at the distant metastasis-free survival in this group that had a linkage to the adverse molecular genomic features that we just mentioned and the genomic classifier score.
And so, in particular, when we look at the NCCN high and very high risk, the specific genomic combinations of those adverse features really gave us greater insight as to stratifying further the genomic classifier score. And so, if we take a moment just to look at this, let's focus over here, the lower genomic classifier scores, which scores of less than 0.85. Now, this dotted line over here is what would be the expected distant metastasis-free survival with a high genomic score, greater than 0.85. But if you start combining some of these adverse molecular features, you start getting a further substratification of this high or very high-risk group. And so, shown here, and again, these are lower genomic classifier scores, this curve in red represents the combination of the GC high and PTEN deletion.
And as you move across here, you can see that with various combinations, we can stratify further for the lower genomic classifier score. And there are differences in the distant metastasis-free survival. In particular, if we look at, for instance, the GC very high-risk group, the high Decipher scores, this would be, at five years, a distant metastasis-free survival of 76.8%. And then, with other various combinations of the adverse molecular features, you start getting improved if they have less, really, worrisome or aggressive features, the differences in distant metastasis-free survival. The same thing over here. When it comes to the very high-risk genomic scores that are over here, we can see also that the combination of the various genomic adverse molecular features can really impact upon the stratification of what is within very high risk a more ominous long-term outcome.
And so, the take-home messages are shown here. There's been, what we can see and has been shown already, significant discordance that's noted between the NCCN risk group and the Decipher scores. And in particular, 14% of patients with low-risk NCCN scores, so these are Gleason 6 disease, could have a high-risk Decipher score. And 18% of patients with high-risk NCCN scores, such as Gleason 8 to 10, would have and could have low-risk Decipher scores.
Also, when we dig deeper into what's driving these discordances, the adverse molecular features really are p53 mutations, PTEN inactivity, and low AR activity. And when we look at these real-world data, albeit with short follow-up, we begin to realize that maybe these adverse molecular features and maybe combinations of these features could further refine the Decipher score stratification within the high-risk group. And so, patients that have high-risk Decipher scores are not all the same. And ultimately, we may further refine these by combining the various genomic abnormalities, and these may give us even further insight and further prognostication about what a high-risk genomic classifier is all about. Thank you very much.
Daniel Spratt: Well, thank you so much, Dr. Zelefsky. That was very impressive work. This makes me start to question many things when we think about Gleason grading and how that's been sort of the backbone of risk stratification for decades. And then we developed these tools with Decipher and other tools that now we've got greater confidence in its accuracy. And now, building on top of that, you even further now are refining prognosis.
And so, to ask, I guess, first, I don't know if this is a controversial question, but right now, we still anchor off of whether it's Gleason score or NCCN risk group. Do you ever think we'll get to a point in some cancers where you anchor on the genomic profile, let's just pretend it's Decipher, and that's the anchor point to where you then add things on top of to refine? I'm just curious to your thoughts.
Michael Zelefsky: Yeah, it's a great question, Dan. And I truly believe that it's not going to be too far from now where we are going to anchor off these genomic classifiers more than the NCCN risk group. And the Gleason score, while it's always been important, and I agree, it's been a backbone of our clinical practice, I believe very soon, we're going to be really relying more on these opportunities, like genomic classifiers and other approaches that are coming into reality right now.
And I think an interesting idea would be, do we really call these still high-risk patients? Or are these going to be in the future, if we rely on genomic classifiers, this is a p53 mutant of prostate cancer, and no longer defined as high risk, but maybe defined as the presence of various mutations that would drive our clinical practice more? I don't think it's going to be too far from now, and I think we're still learning. And I would also say that I am really challenged in the clinic, and I'm sure you are, as you see many patients, how we manage these patients that have these discrepancies and what we put more emphasis on. And should we discount it if we have such discrepant values, or should we not take any chances and treat the patient more aggressively?
Daniel Spratt: Yeah. No, I mean, absolutely. I think anyone that is practicing but also using these tools, it is... I mean, you showed it perfectly. It's not like this is 1% or 2%. I mean, this is coming up in 10, 20, 30%. I mean, this is a common phenomenon. And we have very robust data now that you've showed and others that Decipher and these adverse molecular features are more accurate than NCCN risk groups and Gleason grading.
And so, yeah, I think that it's... I, too, am struggling to... We know our outcomes, especially with the stage and grade migration, and how good our treatment options are, how good from your trials with radiotherapy and SBRT. It's really challenging. There's probably high risk... Well, I would say there is, no question, high-risk men that don't need hormone therapy. But as long as we define them in the historical way, it's going to be very challenging. And so, I agree with you that future trials and work, if we can think of it. And I think the way you've outlined this is the future. I mean, I really do, because the DNA landscape, mutational landscape is just fairly quiet in prostate cancer. But as you showed, it's transcriptionally very rich.
Any other closing thoughts on this work for the audience?
Michael Zelefsky: Well, I do think the trials that are in progress and finishing up are going to utilize these approaches to be not only prognostic but predictive markers and guide clinical decisions. And I know you have been instrumental in this regard in the oncologic community. And we're going to need better ways to figure out what is the optimal duration of hormonal therapy or intensification of hormonal therapy. And perhaps as well, we may be even further stratifying that by the number of genomic abnormalities that are going to be within a genomic classifier score. And maybe if they have multiple abnormalities, this may require a certain amount of aggressive treatment versus perhaps certain high-risk patients may not need any hormonal therapy at all. I know we're decreasing it down to 12 months and maybe 6 months, but maybe there's some features that will be identified in the future that will tell us, "Yes, this was called high risk, but you don't need any further therapy rather than local therapy."
Daniel Spratt: Yeah. No, absolutely. I joke sometimes in that when we think in the late 1990s, we had a nice, simple, the, quote, unquote, "D'Amico risk groups." Easy to memorize. And then you and we put out the favorable, unfavorable intermediate risk, and it definitely took a lot more memorization for people. And when you look where we are today, you're talking about, of course, all the different NCCN risk groups, but now how Decipher overlays and these adverse features. And it is. When you look at other cancers, like lung cancer, I mean, this is part of the way they treat patients. So, appreciate. You're always moving the field forward. Great speaking with you. And I'll let you have the last word.
Michael Zelefsky: Always great speaking with you. And I think this is one of the most interesting directions facing the field of physicians treating prostate cancer. As we know more, we may be more confused. But I think in the very near future, as you alluded to, I think this will only help our patients get better personalized therapy with an emphasis on reducing the impairment of quality of life that we need to treat with more therapy when it's not necessary, and on the other hand, when we may need to intensify the therapy, ultimately, when there really are these adverse features there that we never realized before.
Daniel Spratt: Yeah. Well, thank you so much.
Michael Zelefsky: Thank you.