Validation of Performance of the Decipher Biopsy Genomic Classifier in Intermediate-Risk Prostate Cancer: The Phase III Randomized Trial NRG Oncology/RTOG 0126 – Jeff Michalski
February 24, 2022
Jeff Michalski, MD, MBA, FASTRO, Carlos A. Perez Distinguished Professor, Vice-Chair, Department of Radiation Oncology, Director, Clinical Sciences Division, Chief of the Genitourinary Service, Washington University of St. Louis, St. Louis, MO
Daniel Spratt, MD, Vincent K. Smith Chair in Radiation Oncology, University Hospitals Seidman Cancer Center, Case Western Reserve University School of Medicine, UH Cleveland Medical Center
ASCO GU 2022: Validation of the Performance of the Decipher Biopsy Genomic Classifier in Intermediate-Risk Prostate Cancer in the Phase III Randomized Trial NRG Oncology/RTOG 0126
ASCO GU 2022: Development and Validation of a Prognostic AI Biomarker Using Multi-Modal Deep Learning with Digital Histopathology in Localized Prostate Cancer on NRG Oncology Phase III Clinical Trials
ESOU 2022: State-of-the-art Lecture: Molecular Advances in Risk Stratification of Localized Prostate Cancer
ASTRO 2021: Prognostic and Predictive Performance of Routine Clinicopathologic Variables in 10,535 Men Enrolled on Randomized Phase III Trials in Localized Prostate Cancer
Dan Spratt: Thanks so much for having me here today. My name is Dr. Dan Spratt. I'm a professor and chair of radiation oncology at University Hospitals Seidman Cancer Center in Cleveland. And it's a true pleasure to have Dr. Jeff Michalski here who's a professor and vice chair of radiation oncology at Wash U St. Louis University. And what we'll be discussing today is a translational kind of secondary study of the pivotal NRG Oncology RTOG 0126 trial, which Dr. Michalski is the principal investigator for. And so just as a quick background, this was a very large phase three randomized trial in men with intermediate risk prostate cancer, randomized to what at the time was a standard dose. Some may call it a low dose today of 70.2 gray versus dose escalated radiation, which is more standard today of 79.2 gray and its primary endpoint was overall survival, which was not significantly improved.
However, numerous other secondary endpoints were improved such as biochemical failure. Kind of regardless of the definition, you can appreciate here that use of dose escalation significantly reduced the development of biochemical failure. And one of the things that's relevant to what we're going to present is that actually a lot of men here you could argue and this has eight years of follow up, but that there were a significant number of men who got the lower dose radiation who did not recur. And similarly, it actually showed an improvement on distant metastasis. And you can see here though that similarly, some men appear to benefit and some men don't. And one of the challenges we have clinically in prostate cancer is that our clinical tools we used to risk stratify patients, even though all of these patients were in a fairly tight risk group, things like PSA and Gleason score, or even some of our risk groups like NCCN, they're sort of moderately prognostic in terms of their ability to risk stratify to accurately identify which men will or won't recur or develop metastatic disease.
And what's been developed for sort of the past decade and with a lot more recent evidence that's a lot stronger from randomized trials, especially in the postoperative setting or in the high risk setting is there's a lot of heterogeneity in the genomic risk, even when you're within a defined clinical entity, like a high Gleason score or patients with seminal vesicle invasion. And this tool, this decipher tool has been studied in numerous studies. Most of them are retrospective studies and it seems to consistently be independently, prognostic above and beyond our standard clinical tools to improve risk stratification. And so this has been combined, the clinical tools with this genomic information to sort of have, it's not perfect, but improved ability to discriminate or risk stratify our patients to identify those who are more or less likely to recur. A small study, but that sort of was some of the motivation for this study that we're going to present was a study from Princess Margaret, where they had a prospective registry of patients.
So it's not a trial. That they treated them all with dose escalated radiation. It's about 120 patients, all with intermediate risk disease. None had hormone therapy. So it's very similar in that regard to the RTOG 0126 trial. And you could see here that in this small cohort, those with lower genomic scores, actually none of them develop metastatic disease. And if you look here at sort of the AUC that NCCN alone, which is intermediate risk, so it was basically favorable versus unfavorable in this smaller cohort, didn't really discriminate risk that well. But when you combine it with the genomic features, it was very accurate at discriminating risk.
And so this was sort of the motivation to say, well, how about in a rigorous very well done, large randomized trial? Can we determine if similar results can be achieved? And so the RTOG 0126, part of the NRG biobank for patients that also agree to bank tissue, there was about 850 patients that had tissue samples available. And given the age of the tissue, right, this is almost 20 years old, some of the tissue, a subset of this was high quality enough to sort of pass the quality control. And in the end, you see here that in the two randomized arms, they were very balanced, that it was about 110 patients in each of the arms of the trial.
And so in this 215 patient cohort, what you see here is that in either arm, first of all, the arms had similar genomic risk scores, which is good, there's not an imbalance. But you see, as I was saying before this significant heterogeneity and genomic risk, again, despite this being a fairly tight windowed intermediate risk cohort. And probably the big takeaway from the study we did was that for each of the endpoints assessed from biochemical failure, receipt of salvage therapy, disease progression, distant metastasis, prostate cancer specific mortality, metastasis free survival, and even overall survival, that on multi-variable analysis, the genomic score on a continuous scale from zero to one was independently prognostic for all of these outcomes when adjusting for our standard clinical tools. And in a different way that maybe is more clinically relevant is that the absolute benefit of dose escalation was different between those with lower decipher scores, versus those with intermediate or high, I'll just call it higher scores.
And you can see that there is benefit. So example, biochemical failure. There's 7% benefit here for lower decipher, but it's almost 20% with higher scores. And probably one of the biggest ones that sort of came out was the metastasis where you really see some benefit from dose escalation, but a really massive benefit here in the higher decipher patients. And although this is more some would say hypothesis generating for one of the endpoints, it actually not only was the absolute different significant the relative difference or benefit from dose escalation was also different with a significant interaction. And what that means here is for the decipher lower patients, there really wasn't any meaningful improvement. Again, sample sizes is subset of subset improvement here from dose escalation, but a pretty substantial improvement in metastasis free survival from dose escalation. So the sample size limits more robust analysis, but it's I would say at minimum thought provoking, if not, especially for these patients with higher decipher scores that dose escalation is definitely warranted.
So I think just a quick summary here is I think the study we did showed that this genomic classifiers independently prognostic in the context of a randomized trial for patients with intermediate risk disease. Those with lower genomic scores drive less absolute benefit and potentially less relative benefit from dose escalation. And there's multiple ongoing randomized trials listed here, GU009 and 010 that are in intermediate and high risk. That's further aimed to refine how to optimally treatment based on these clinicogenomic risk groups. So with that, thank you so much. And now we'll get into, I'd love to hear Dr. Michalski thoughts on his interpretation as he's the PI of the overall trial and how these results, do these have clinical relevance? What do you think, Jeff? How do you interpret all of this?
Jeff Michalski: Thanks, Dan. That was a great introduction to the study and also a great analysis done by you and the team, looking at the decipher scores. When we first designed the study, we tried to pick the sweet spot of patients who would benefit from dose escalation. It's worth emphasizing that none of these patients receive hormone therapy. This was strictly a local therapy question. And we realized that picking high risk patients intensifying local therapy alone, while it might improve local control, it would have very little impact on distant metastasis, because many of them may already have established spread of their disease. And likewise, the low risk patients, we felt that standard dose radiation would be appropriate for them. And there was really no value to dose escalate them, put them at risk of more side effects, if there was no benefit.
And actually today you would agree that those low risk patients probably zero dose is most appropriate for them. Active surveillance is probably the best way. So we picked a group of patients with intermediate risk disease based on the established criteria at the time, PSA, Gleason score, and clinical stage that we thought if we could just improve the local control, that would then prevent subsequent development of metastases. But as you pointed out, whether it's, you call it the D'Amico criteria or the NCCN criteria, relying strictly on those clinical factors, there is still quite a bit of heterogeneity. And oh, don't I wish I had a tool to predict who would've not required dose intensification and who we could've pulled out either from the higher risk side or the lower side. Because I think perspectively, if we could've stratified by these now genomic classifications, this would've been a very positive study.
Now going forward, I don't think anyone is going to invest in repeating the study because studies like this, it took 10 years to accrue and then another five years more of follow up to report. But the question is what can we do with the information on these genomic classifications going forward? And I think right now we're using them in current NRG trials as you pointed out, but most of those are using systemic therapy. And we've kind of accepted that for the higher intermediate risk or unfavorable intermediate risk patients that the addition of another nine gray of radiation, another five fractions and 7920 has become the new standard of care for those patients. And now, quite honestly, we've adopted a hypofractionation for them, but still, an intensified local therapy. And we're now asking questions about whether or not a hormone therapy with deescalation of the hormone therapy, perhaps instead of 24 months, 12 months would be sufficient in the more favorable groups of patients.
And in the less favorable groups of patients, we're looking to see if adding more hormone therapy on top of the dose escalation will help. But will anyone use this data to omit hormone therapy altogether in the intermediate risk case patient? Perhaps. I would love to have stronger data to support that. And perhaps, there are other similar trials that we could seek done in Europe, and Spain, and Canada, et cetera, to see if they might pool our data, to get a little bit stronger conclusion than what we had. Unfortunately, when the samples get as old as they were, you're going to lose some cases just because of the decay of the DNA in those samples. But I still think it's an important contribution to literature that we'll be presenting at ASCO. And I think it will help inform people as they try to make decisions. Do I add hormones to the dose escalation or not? And or maybe I could back off on the dose in patients and select patients who might, you might not want to do that.
Dan Spratt: Yeah. No, that's great overview and sort of perspective. And sort of on that last point to be, so I guess, provocative and thinking about other cancers, like head and neck cancer, right, there're always, can we deescalate? Can we deescalate? I mean, now, as you said, and you've led a lot of the work. Right now, we use whether it's image guidance and rectal spacer, hydrogels. I mean, our toxicity is now even lower. Is there really an appetite or is it necessary? Like, do you think this should spawn an era of trying to identify patients to deescalate? What do you think?
Jeff Michalski: Yeah, yeah, that's what I was just thinking, is that what patient population would you want to choose? Because the radiation has evolved from the time that we did the trial. It started with just 3D, IMRT was added later. And actually this study, the 0126 helped inform radiation oncologists, how to prescribe and how to restrict dose to normal organs. So now that we've established safe limits to the organs at risk, and that we can do this safely with modern radiotherapy and using IMRT or other modalities, using image guidance and rectal protection. Yeah. I don't think there's much appetite to deescalate. And you'd have to do a non-inferiority trial, which would be enormous only to find out that you prove what you already knew.
Dan Spratt: Yeah.
Jeff Michalski: There really is not much toxicity and not much loss in deescalating. So I think it's probably not going to happen. Maybe if you have a patient who has comorbidities where you think dose escalation might be a challenge for the patient with inflammatory bowel disease, Crohn's, or ulcerative colitis. In those patients, you might feel more comfortable backing off a few fractions if they had a low risk genome classification. But I think those would be exceptional cases. I think for the most part, there's not much added cost or both in terms of financial aspects, but also in terms of toxicity to do that.
Dan Spratt: Yeah. No, completely agree. Yeah. I think then the flip side I think people, when they see this will ask, so right, we now can go higher dose. Right. Whether its various ways of boosting, whether it's adding brachytherapy. Yeah. Although obviously that wasn't tested at all in this trial, but people will want to start making conclusions. What do you think? If you had a low or a high decipher, do you think people should, or how will you [crosstalk 00:16:57].
Jeff Michalski: Yeah. And that's a great question. And my colleague, Brian Bowman has proposed, we take the flame approach where you do intratumoral boosting or intraprostatic boosting of high concentration of tumor cells, the dominant intraprostatic lesion. That is appealing. And again, it's a new era of radiotherapy approaches with stereotactic or maybe limited boost with brachytherapy. So that might be the next generation of questions to be addressed, is maybe we should still dose intensify, but not for the entirety of the gland, but for components of the gland where the highest volume of cancer is.
Dan Spratt: Excellent. All right. Well, oh yeah. I think this is, we always look back as it's a great, fantastic trial and then, right, all this new technology comes out and then now these new trials are starting. And then I'm sure when they're done, we'll say at least we had this.
Jeff Michalski: I've learned that the questions you should've asked are always smarter than the ones you did. And it's just natural. We've made progress in understanding the risk factors. The genomic analysis that we are doing today, weren't available when these studies were put together. So yeah, hindsight is 2020, but we have to keep pushing forward and introducing these new tools into our stratification for trials going forward.
Dan Spratt: Excellent. Well, thanks so much, Jeff. It was fantastic to hear your perspective and thanks to URO Today and thanks everybody.
Jeff Michalski: Yeah. I appreciate being invited. Thanks for doing this to help us inform the membership of ASCO and ASTRO and the attendees at the GU ASCO symposium.
Dan Spratt: Great. Thanks, Jeff.
Jeff Michalski: Thank you.