Risk Assessment in Localized Prostate Cancer - Insights from Real-World Data and SEER Registry - Jim Hu

May 9, 2023

Ashley Ross and Jim Hu discuss the use of genomic classifiers, specifically the Decipher Genomic Classifier, in the management of localized prostate cancer. They review its performance in larger datasets and its association with prostate cancer outcomes.  The conversation also delves into the use of the Decipher Prostate Biopsy Test in men with favorable risk disease undergoing conservative management in the SEER Registry. The role of genomics in guiding treatment decisions, including surveillance, surgery, and radiation therapy, is explored, with a focus on how the Decipher test can help guide decision-making toward improving patient outcomes. Additionally, they explore recent studies on tumor volume and real-world data and how these findings can be applied in clinical practice, particularly in the context of genomics and tumor volume. The discussion centers on real-world and SEER data, including the development of a longitudinal prostate cancer transcriptomic and real-world clinical data linkage.


Jim C. Hu, MD, MPH, Urologic Oncologist, Weill Cornell Medicine, New York, New York

Ashley Ross, MD, Ph.D., Associate Professor of Urology, Northwestern Medicine, Chicago, IL

Read the Full Video Transcript

Ashley Ross: Hi, I'm Dr. Ashley Ross. I'm an Associate Professor of Urology at the Feinberg Northwestern School of Medicine. I'm here at the AUA 2023 with one of my colleagues, Dr. Jim Hu, Professor of Urology and Urologic Oncology at the Weill Cornell. Dr. Hu, thanks for joining us. Today, we're going to be talking about genomic classifiers, particularly the Decipher Genomic Classifier and how it's performed in larger data sets, and some of the evidence presented here at the AUA.

Jim, really nice to have you here today. As you know, for localized prostate cancer risk assessment guides everything we do for the patient, whether it's initial conservative management with surveillance or treatment, and if it's treatment, what are their expectations and how intense will that treatment be? It's amazing that it's been about a decade that we've had the ability to use genomics on prostate cancer tissue that we get from biopsies or radical prostatectomies, and the Decipher genomic classifier has accumulated most of its data from, first it was retrospective registries, then some prospective registries, and now analysis of some phase three clinical trials.

In all those studies, it's been shown to have an independent prognostic ability for things like metastasis-free survival. Here at the AUA, they've now linked it into some large data sets, some work from both of our groups, yours, in particular, being presented. Before we go into some of that data, I wonder how are you using genomics in your practice currently?

Jim Hu: Sure. First of all, Ashley, thanks for having me on the show. I tend to follow the maxim of don't order a test unless you're going to do something with the results and so typically if someone's going to have surgery, radiation, of course, I don't think there's much of a role for a Decipher test. However, as you know for the men who fall into the guidelines, for example, they may have grade group two, they're older or even grade group one, for example, in an anxious patient. I think at least for a urologist, that's where the sweet spot is. Certainly, at our institution, I'm sure probably at yours as well, our radiation oncologist will use Decipher as well to risk stratify between say favorable intermediate risk or higher to determine whether or not to give androgen deprivation therapy as adjuvant with the radiation. That's how I think, certainly I'm using it in my practice.

Ashley Ross: Yeah, and it's interesting. Everyone has a different adoption strategy. For me, I would say it's a little bit more uniform for my practice. If I have the population you were talking about, certainly people who are questionable surveillance candidates and I want to figure out, is this really appropriate or are they going to have fast progression, I'll use it in that area. I'll use it somewhat on my lower risk patients. Not really very low risk, but lower risk to figure out what's the cadence of my surveillance and then as I get into the intermediate risk stage, as you mentioned, it's either me or my radiation oncologist doing the Decipher on the biopsy to figure out what's the patient's choice. Is it surgery or radiation or is it really surgery versus radiation plus some ADT? Postoperatively, I think a lot of us will use it to guide timing for radiation management, but as you said, it's been on for about 10 years, I think in really high clinical use maybe over the last few years as it's gone onto all the guidelines and accepted on payer networks. I think a lot of us are adopting it different depending on how we think about our treatment pathways.

That brings us to some of the real world data. How is it doing in the real world, apart from some of the smaller but really robust data series from phase three trials, et cetera? There was some work presented by Dr. Leapman from Yale in a multi-institutional collaboration with Veracyte who does Decipher and they looked at the Clarivate Registry and the Clarivate has 300 million records that's based on the electronic health record, based on insurance claims, pharmacy claims and they were able to link almost 100,000 patients who had Decipher, about 55,000 Decipher biopsy and the rest on prostatectomy specimens to look at what's going on and first it was amazing that they could link this dataset. I haven't really delved into the Clarivate database before this.

One of the presentations by Dr. Leapman and Sprenkle was just kicking the tires on if we test it and we spot check, are we actually able to make this real world data linkage and they were, and then they looked further to say, "For the first analysis, first pass is Decipher an independent prognostic indicator when you look at NCCN criteria or CAPRA-S at post-op of metastatic potential," and they found that it was. It had an independent hazard ratio for prediction of those events. It's nice to see that works in the real world data.

Your team specifically led an effort to go even more granular in a specific subset, and that was in people who might be candidates for surveillance and you looked in SEER. I wondered, beyond that real world data in Clarivate that looked at and confirmed the prognostic ability, what you guys saw when you drilled down on that favorable risk population?

Jim Hu: I think it probably mirrors what we talked about earlier in terms of what do I use it for in my practice and looking at that population first. I think the other thing that led to that thought process to investigate that is just that, as you know, there's been recent controversy about grade group one or low-risk disease versus very low-risk disease. Is tumor volume an independent prognostic indicator, for example? Should high volume grade group one or Gleason six patients not do active surveillance? That led to us, again, through that linkage, which first happened in breasts and now in prostate to genomics tests, that led us to ask the question in the real world, like you said, for grade group one, grade group two, does the use of Decipher score, how is that as a prognostic indicator to look downstream at treatment versus conservative therapy treatment, specifically surgery? Then of course, we have radical prostatectomy outcomes, pathology, and SEER as well.

We have your typical variables that are measured, for example, age, you had tumor volume, PSA, and then as well, I think in future analysis we'll look at race as well and then see how Decipher compares in multi-variable analysis in terms of predicting adverse pathology, radical prostatectomy.

Ashley Ross: It's interesting, one of the highlights of your thoughts or how you framed it was you wanted to look at, there's lots of reasons and even a lot of our previous series in surveillance have been polluted by this idea of why do people take people off surveillance? You have this outcome of progression, but it's this wishy-washy outcome that can be triggered by lots of things and I like how you've looked at both what triggered progression and then what was your outcome for those who went to radical, what you were looking at as like, what do we consider? We don't want an event, a never event of someone we would be surveying. We didn't want to survey. You said 437 or someone with some vascular involvement.

Maybe you can tell us a little bit about, there's about, I think 2000 men you looked at on SEER that had genomics, and I think there was 50,000 men or something that did not have genomic testing. Maybe you could first talk towards what seemed to pull people off conservative management or steer them away from conservative management, not the outcomes, but what were the factors that maybe would trigger a provider to take someone off surveillance? You mentioned higher volume disease was one of them.

Jim Hu: Sure, absolutely. A higher GC score of course led to a greater likelihood of surgery. I think counterintuitively older age was associated with receipt of surgery. Our study population was just grade group two versus grade group one, and a higher grade also was associated with a greater likelihood of undergoing surgery versus conservative management.

Ashley Ross: It looked like in the SEER that people with higher volume grade group one, any grade group two, age, like you said, although the odds ratio was a little bit less and then for the genomic classifier score being higher, those took you towards treatment. The next question your folks asked was, "Which one is really appropriately correlated with outcome?" As you mentioned, you had done a previous study that suggested that, and I actually think disease volume has become even more wishy-washy in the MRI biopsy, perineal, transrectal. I don't really know what percent positive core really means anymore, but maybe go into now what predicted this adverse pathologist 437 or T3B?

Jim Hu: Sure. Adverse pathology was predicted really by the GC score independently separate from say, tumor volume PSA and so I think, again, it just shows it corroborates in real world evidence. Some of the things that have been looked at, for example, in the single institution by our repositories or as you said earlier, the randomized control trials. I think it gives as you know, and I think the other aspect that's interesting, and I think a lot of the viewership probably understands this, but real world means that you don't have, for example, the luxury of pathologists that were trained at experienced centers like under Jonathan Epstein and so forth. I think it just shows you there could be greater variation. I'm sure in your practice, for example, if you see a consultation, you're getting a second opinion pathology, but in the real world, that doesn't necessarily happen and so I think that that perhaps highlights even a greater value of the GC score in that setting.

Ashley Ross: Yeah. There was very robust data that your group went through, but also, at least in Gleason grade group two, we don't know how much the percent is involved. As you said, it's subjective. It was interesting that Decipher came out as that powerful indicator. It made me also think about the MUSIC registry data that was presented from the Michigan groups and they had a surveillance series that people with higher Decipher scores, not only were they more likely to seek treatment, about half of them had to get treatment in one year, about 75 percent at three years. If you looked at what happened to them after treatment, it was about three times as many had biochemical recurrence such that the biochemical recurrence rates were over 30 percent at three years and for your data that shows that the hygienal classifier scores in SEER can predict this adverse pathology and for that data together that says, "If you put them on surveillance, potentially after a year you can still cure them."

They're still a localized disease, but a lot of them will need a one-two punch now for cure. 30 percent of them might need that. If you treated them up upfront, you might have been able to get away with surgery with better nerve sparing, et cetera. Now it all begs the question of how do you decide to manage your patient with a high Decipher score? It looks like that's a powerful prognostic indicator. In your practice, if I was your patient on surveillance, then obviously it's going to be on my expectations, but say I have 336 disease and four cores. My genome classifier score is about 0.7, so high risk. I'm in that, I think it was about 15 percent of your population was like that. How do you counsel me on what I should do as a patient?

Jim Hu: I think that it is interesting, and I'm sure you as well, having had some recent patients with that, if you will, the dichotomy between the light microscopy grade as well as the GC score. I think a lot of patients are pretty sophisticated about that, particularly, as you know, things are resulted at 10 or 15 year outcomes as well as the likelihood of adverse pathology as we've talked about. I'm sure the company, Veracyte has been very thoughtful about how to present that data and so when patients as you know, look at those proportions individually, I think they come away of course, not having, for example, a relative difference. They only get one Decipher score in contrast to the physician seeing several, but I think that those percentages, of course, they come away with that if you have at least a 30 percent chance of upgrading upstaging adverse pathology, that's pretty significant. In my practice, it's helpful because of course, with shared decision making, that just makes that easier. It's not necessarily the physician pulling numbers out of say the Partin tables or the nomogram, which you remember and it's individualized of course, which has great value for the patient.

Ashley Ross: Just for our listening audience, again, when Jim and I are talking about upgrading, upstaging, we're really thinking about patients we wouldn't want to survey. Those are people who predominantly have 437 or have T3B disease. I know that in my practice, if I take someone to radical and they just have like 347, and it's T2, it doesn't really bother me that much. Their curability is very high. Even T3A with a negative margin has that, but the T3B guys, 437s, that's what we mean when we're saying these are people, we're trying to weed them out. If there's a high probability of that, I often will counsel them early. Certainly I think about risk counts. If you're 347 and you have a high Decipher, that's a couple different yellow flags, and I often steer them away from surveillance. If it's a low grade disease and a high Decipher in my practice, I'll tell them, "We can do surveillance, but you definitely need your confirmatory biopsy at one year, maybe even another at two years and we can think about treatment depending on what your expectations are."

Some people say, "Let me have treatment now," where the nerve sparing is going to be great, where everything is going to recover better. Some will say, "I want to defer that." I think it's still an individualized choice even though surveillance is preferred for our low risk patients, there's still an individualized choice there. I think there's much more wiggle room in our favorable intermediate risk patients.

Jim Hu: Absolutely. Just to add on to one aspect of it, we talked about some of the subjectivity in looking at pathology scores, and so perhaps the future areas, for example, if you saw someone with a grade group one or a Gleason six and that person had Cribriform features or Intraductal, those are again features that your more inexperienced pathologists who are looking at everything may not necessarily capture, but yet maybe reflected by a higher GC score.

Ashley Ross: Certainly if it was a cult, we sometimes we'll get that read on our 347s and that's another place, like you said that in the pattern four component, catching Intraductal, catching Cribiform, particularly on biopsies where there might be a limited amount of tissue, it's a good homogenizer, and actually, it was interesting to see that sometimes the data actually, like you mentioned, looks more robust in real world than if you look at a focus series that comes from a tertiary institution where everything, their MRI, their pathologist, their radiologists are all just the top level which may be something we should all aspire to, but might be unrealistic in our practices.

Jim Hu: Absolutely.

Ashley Ross: Really a pleasure of being here at the AUA with you, Dr. Hu. I don't know if you have any other closing thoughts, but I think that it was great work that your team has done and continues to do, showing how we use these tools in our practice and what's the power of them.

Jim Hu: Thank you so much, Ashley. Always great to connect with you and UroToday.