A Systematic Review of the Decipher® Genomic Classifier in Prostate Cancer - Ashley Ross & Daniel Spratt
February 16, 2021
In this paper, the authors review the evidence of the Decipher genomic classification tool for men with prostate cancer. They found consistent evidence that the test helps identify which cancers are more or less aggressive, which in turn aids in personalized treatment decision-making.
Decipher Biosciences Learn More
Daniel Spratt, MD, Tenured Full Professor of Radiation Oncology and a leader in Prostate and Spine malignancies. Chief of the Genitourinary Radiotherapy Program, Associate Chair of Clinical Research, and the Laurie Snow Endowed Research Professor in the Department of Radiation Oncology. Co-Chair of the Genitourinary Clinical Research Team and co-Director of the Spine Oncology Program, in the Rogel Cancer Center, The University of Michigan, Ann Arbor, Michigan.
Ashley Ross, MD, Ph.D., Associate Professor, Department of Urology, Northwestern University, Feinberg School of Medicine, Chicago, Illinois
A Systematic Review of the Evidence for the Decipher Genomic Classifier in Prostate Cancer.
ASCO GU 2021: Use of Genomics to Guide Treatment Decisions in Post-Prostatectomy Patients
Validation of a Prognostic Genomic Classifier in Salvage Radiotherapy Prostate Cancer from a Prospective Randomized Trial - Felix Feng
Ashley Ross: Hello, and welcome to another episode of UroToday. Today we're going to be talking about a very influential paper that was just recently published by Dr. Dan Spratt. It's "A Systematic Review of the Evidence for the Decipher Genomic Classifier in Prostate Cancer". I'm Dr. Ashley Ross, I'm an Associate Professor of Urology at the Northwestern Feinberg School of Medicine.
Dr. Spratt is a Professor of Radiation Oncology at the University of Michigan. He is the Laurie Snow Endowed Research Professor and the Associate Chair of Clinical Research at the University of Michigan. In addition, Dan is an expert in many fields, prostate cancer in particular, with over 250 publications, and he's led over 10 clinical trials as a PI. A lot of his work deals in precision medicine and the study of omics in cancer. And today he's going to talk a lot about that for us.
Thank you for being here, Dr. Spratt. Please, go ahead.
Daniel Spratt: Thank you so much, Dr. Ross, for the introduction. I'm excited to talk today about the systematic review many people participated in. And I have no real relevant conflicts of interest for today's topic, but the others are listed here.
So first, it's important, I think, for the audience to know what is Decipher®. Well, it's a tissue-based prognostic gene expression classifier, where either pretreatment biopsy tissue or prostatectomy tissue, RNA is extracted from this tissue and undergoes a whole transcriptome microarray. And then a Decipher® score based upon 22 of these genes is calculated from a score of zero to one.
This assay is now over 10 years old, in the sense of it was initially trained for patients treated at the Mayo Clinic to predict the development of distant metastatic disease. And they ultimately identified 22 genes involved in seven cancer pathways to develop this Decipher® score. And since developing that Decipher® score, the main reason that that company existed, and companies like it existed, is that they were trying to address a current clinical challenge we have in prostate cancer.
And what I'm showing you here is sort of an extreme example of this. This is from a randomized trial called the SPCG-4 trial, and it [inaudible 00:02:23] men to get surgery versus undergoing watchful waiting. And this is men who have high-risk prostate cancer, a subset of prostate cancer we almost always would recommend treatment for. But these men underwent no initial treatment.
But what is very interesting is that in high-risk prostate cancer, 35% of men that underwent watchful waiting did die of prostate cancer. This is real, clinically significant disease. But about 30% of men, even by 18 years, who received no treatment did not die of prostate cancer, did not develop metastatic disease. So there are subsets, despite our clinical parameters calling it high-risk or aggressive, that still have indolent disease.
And so one of the challenges we have currently is if you were to have a hundred men with prostate cancer we see in clinic, we use a lot of tools: the PSA, the Gleason score, other risk group schemas that combine these. And we can risk-stratify patients to identify who will or will not develop metastatic disease. Unfortunately, though, most of these systems were off in 20, 30, sometimes even 40% of the time wrong in our estimation of who will develop metastatic disease.
What we now know is that the biology of prostate cancer is extremely heterogeneous. And so this is an example of nearly 18,000 men with high-risk prostate cancer that underwent surgery, and they had Decipher® testing done. And you can see that despite this all being sort of a homogenous high-risk clinical state prostate cancer, it's extremely heterogeneous.
And so one of the challenges that motivated the paper I'm going to be discussing is guidelines, such as the ASCO guidelines and other guidelines, they often take multiple years to actually be published. And so this is an example that right now is undergoing various states of endorsement by professional organizations. But the data was polled now over two years ago, and this is in a field with rapidly acquiring data.
But even in this 2019 publication, you'll see that at the state of evidence, now about two years ago, guidelines do endorse the use of these genomic classifiers in situations where the assay results, when used in addition to your routine clinical factors, can affect a clinical decision. Which, obviously, you should only order a test if it's going to potentially change your clinical decision.
But this motivated us to actually then rapidly conduct a comprehensive contemporary systematic review, because there was just so much, dozens and dozens of papers coming out every few months, that we wanted to get the most up-to-date data on what is the evidence for this test. So we did a systematic review and ultimately identified 42 unique studies.
And what you'll see here is that actually, to much of our surprise because we didn't realize how much evidence was out there, Decipher® now has substantial evidence to support its clinical use. There are actually over 42 studies with 30,000 patients in these studies. And although, of course, many of these general classifiers have retrospective data, there now is data from prospective registries, prospective clinical trials, and now reporting out, and I'll show you an example of this, posthoc analyses of multiple Phase III randomized trials for the Decipher® assay.
It initially was approved and used in prostatectomy samples, but now there have been nearly 20,000 patients where the data has been published upon in biopsy samples. And the data now spans localized prostate cancer from low to high-risk, as well as post-prostatectomy. They're just starting to get into the more advanced prostate cancer setting as well.
And very relevant is the endpoints reported, is that there's a lot of clinical endpoints in prostate cancer. But I would say although things like adverse pathology may have clinical utility, a lot of the harder endpoints they've now reported on, and again as I said, prediction of distant metastasis is what the company mainly is focused on.
And very importantly, what you'd want a prognostic biomarker to be able to deal with consistently is independently prognostic on multi-variable analysis when you factor in things like a basic Gleason score or PSA. It improves the discrimination over our clinical models. It often reclassifies patients to a different risk group. And very importantly, it changes management. And so this is something that most of our molecular PET imaging or MRI tests, the way they get approved is actually simply through showing a change in management.
So this is just a summary of many of these studies. What it's showing you is the independent ability of Decipher® to prognosticate patients for different endpoints. Most of them, as I said, were for distant metastasis. And you can see it's highly consistent that Decipher® is around a hazard ratio of about 1.3, independently of PSA, Gleason score, T stage, and other clinical parameters. And also it improves discrimination for AUC as a test of discrimination.
And what you see here in gray bars is the standard of care clinical models we use. And the light blue is Decipher® itself alone. And then combining the two here, and you'll see in pretty much every study there's a substantial improvement when you add Decipher® to clinical variables.
Now, just to give a couple of examples of what some of these studies have been... Here's an example we reported out a few years ago, combining clinical parameters with Decipher®. We call it the clinical genomic risk grouping schema. And what you're able to see here is that patients with low clinical genomic risk, which you barely can see the curve here, it's a very large risk group. This is much larger than NCCN low risk. No men developed a metastatic disease by 15 years. And again, it identifies a very high-risk group here. Again, much larger than NCCN high risk is clinical genomic high risk, where at 10 years 50% of have developed metastatic disease.
And going back to those tests of discrimination, you can see here that with the incorporation of our clinical and genomic features, you're getting up into the 85, 90% accuracy range of identifying which men develop metastatic disease.
An increasing amount of data now is coming out, and this is some prospective data I'll show you that's now in submission. As you can see that in low-risk disease here, time on active surveillance, so men with low-risk disease that are on active surveillance, those with high Decipher® scores are much more rapidly likely to come off of active surveillance. But those even with lower risk or favorable intermediate-risk disease that undergo treatment, which some still do, the high Decipher® patients are the ones that are developing an earlier biochemical recurrence. And it just begs the question, should we be doing active surveillance differently on these low-risk patients with higher genomic scores?
As a radiation oncologist, something that's very relevant to us is that for men with unfavorable intermediate-risk prostate cancer, we almost always add hormone therapy. That's what national guidelines recommend. And this is a nice prospective registry from Princess Margaret with eight years of follow-up in men with unfavorable intermediate-risk disease primarily. And what's very important is about 60% of those patients, despite having unfavorable intermediate-risk disease, had a low Decipher® score. And none of them, despite them just having radiation alone, developed metastatic disease, meaning why would you add ADT to these men? And just, again, reinforcing this point, almost a 90% accuracy in identifying with men would develop a metastatic disease here.
So this is hot off the press. This is from a paper coming out in JAMA Oncology using a Decipher® assay in a randomized phase three trial, the RTOG 9601 trial. This is a trial of men getting salvage radiation plus or minus two years of high-dose a bicalutamide. And what you'll see here is that despite this tissue being almost 30 years old, men with a low Decipher® score on this trial... And again, this is in recurrent and fairly aggressive prostate cancer, that almost none of these men with long-term follow-up ever died of their prostate cancer. And those of course with higher Decipher® scores, over 20% died of the disease.
And very importantly, what you can see here is that those with a low Decipher® score had basically no benefit to two years of hormone therapy. You'll see that there's almost no improvement in distant metastasis, prostate cancer, specific mortality. And actually, they had worse overall survival when you gave them hormone therapy. And that's simply because if there's no oncologic benefit, you're just experiencing the side effects and morbidity of long-term high-dose bicalutamide. But in contrast, the higher Decipher® score patients did derive that benefit.
And so when using the criteria, and these are two common schemas, the Simon criteria or the AUA criteria, what you'll see here is that for the Decipher® biomarker, there is for various disease states of prostate cancer, not all of them, there's actually very good evidence in that the biomarker is prognostic for long-term metastasis survival. With prospective data, it changes management. And actually, the absolute benefit from treatment is different by the biomarker results.
We actually don't have this same level of data despite endorsement by various guidelines for PSMA PET, for example, which recently got FDA approved. As I said, it got FDA approved simply because it changed management. It's not been shown to be prognostic for long-term outcomes or a differential benefit of the treatments we provide.
And so in summary, there is Grade A or B by AUA criteria evidence to support Decipher® testing in most clinical settings where it will change management decisions. The strongest evidence presently is in guiding ADT use with salvage radiation therapy, as well as guiding ADT use in intermediate-risk prostate cancer. There's moderate evidence in its guiding use of active surveillance in low and favorable intermediate-risk disease. And now, this is more of a controversial point given some recent data we may discuss about guiding adjuvant versus salvage radiation therapy.
But, very fortunately, there are over 30 completed or ongoing clinical trials, many of which are randomized Phase III trials utilizing Decipher® across the world that'll hopefully fill up these evidence gaps and help us understand the utility of this test in prostate cancer. So thank you so much.
Ashley Ross: Thank you, Dr. Spratt, for your presentation. It's a great synthesis of the work by you and many others that's been bringing this type of multigene testing into our clinical practice for prostate cancer. As with the adoption of all new technologies, there can be lots of barriers for providers. How available is this test and is it covered by insurance?
Daniel Spratt: Yeah, so the Decipher® test is one of a few different commercial gene expression tests. Right now it has the broadest approval for use. So MolDX, which basically governs CMS or Medicare approval, patients with low to high-risk prostate cancer, now it has to be paid for, as well as men post-prostatectomy with recurrent disease, it'll be covered.
There are a number of growing private insurance companies that do cover it, but you should check with your local insurance carrier if they do. There also are reduced rates, that patients can contact the company if the test is not covered by the insurance, the company will offset the cost of the test for patients.
Ashley Ross: Oh, go ahead. You mentioned that there are a few different multigene tests out there, and people hear about Prolaris® or Oncotype DX® prostate. Are all these tests equivalent?
Daniel Spratt: Yeah. As you said, so there are three tests in the market: Prolaris®, Oncotype®, and Decipher®. They each are very different in the technology they use and also the evidence to support them, in how they were trained, and sort of the endpoints that they're trying to predict for men with prostate cancer. There is no good study comparing them head-to-head. So instead, it's similar to in breast cancer where there are numerous assays of breast cancer.
I think what's apparent is that, right now, and this could change, but right now Decipher® seems to have by far and away the most robust evidence and has linked up with academia and industry to conduct all of these clinical trials to provide academicians like ourselves with evidence we can tangibly use in practice and to shape the way we practice. So I can't say that the other ones don't necessarily work. It's just there's less robust evidence of their ability to perform well.
Ashley Ross: And in speaking on the evidence, and this was highlighted in your talk and well highlighted in your review, if you had to pick one clinical scenario where you would say, "This is basically a must-use test," can you speak to that one thing that if people are going to take home from this, when they should really use Decipher®? The one clinical scenario that you think is the most important?
Daniel Spratt: Yeah. So right now, because we have that RTOG 9601 trial, because this is based upon a randomized trial, as well as a lot of other retrospective data combined with it... To me, in recurrent prostate cancer, men that you're trying to guide, if they should receive hormone therapy, to me, this is something that is a... I'll never say anything should be a reflex test or always ordered, but this is something that I think if this were my patient or my family member that I think this provides tremendous value to help risk-stratify them.
Ashley Ross: Yeah. The result was impressive to me as well. And to my knowledge, it's the first genomic type test in prostate cancer to be linked to an overall survival outcome and had very important messaging on de-intensification of therapy in the correct patients. So thank you for highlighting that.
In addition to the genomic classifier, which is what stratifies men's risk for metastasis and prostate cancer mortality, as you mentioned, when I order the Decipher® in my practice, I'll also get multiple other signatures that are in various stages of validation that is on what is called a research only grid report. A lot of our practitioners have seen that.
Can you talk a little bit about the grid signatures? How should I be using them in my practice? Should I be using them in my practice? Your review highlights some of the trials that are using those signatures and exploring them, but are there any I should look at? Or should I just wait and just leave it for academic research?
Daniel Spratt: Yeah. So that's one of the stand-out features of the Decipher® company. It's really the product, because the way the test is done, it's on a full transcript or micro-ray so you get thousands and thousands and thousands of genes, and they provide you with, I think, over 40 or 50 other signatures in various states of validation in this great report that you get for free. And none of the other companies are able to do this because they just assess a handful of genes on a PCR test, so they can really only report their outcome.
The value is, and you'll see if you go through the grid report, is it's really nice, Decipher® doesn't claim that the 22 genes it measures will necessarily solve all of prostate cancer. But they have, you'll find, that it reports out, I think, 15 to 20 other prognostic gene expression scores, as well as the genes involved in Prolaris® and Oncotype®. So you can kind of have derived versions of those for free, and you can compare it and contrast the prognostic score across these 20 signatures, and it will provide you an overall prognosis estimate.
And I think sometimes helps me to see, are these all concordant? Or are they giving me a different... Is there some discordance there? There's some really exciting ones that we've completed randomized trials, that we're waiting for follow-up, sort of luminal and basal subtyping, where we've got a variety of signatures about hormone therapy responsiveness.
Or sometimes I view it, I call it the tea leaves, in that if you have someone, let's say, that appears to be more androgen-driven and is maybe more androgen-responsive has an unfavorable prognosis, you know maybe that's someone you need to really start thinking if you're on the borderline of whether you're going to add hormone therapy, for example. Maybe that's someone to really consider adding hormone therapy to.
Ashley Ross: That's a wonderful summary. I will say two things just to, again, highlight some of that. One is, I use it exactly like you do, for my patients where I'm thinking about surveillance. Or if they have an intermediate Decipher® risk, I will sometimes look at the grid and look at all of the other published signatures that they report on risk to decide are they can concordant or not?
And as you highlighted, I think a lot of interest is going to be in this androgen response or luminal versus basal phenotype, which has been established in breast. And there's a lot of parallels between breast and prostate, and are in the studies of ongoing trials.
Thank you so much for your time. I would really recommend to our readers, take a look at this systematic review. Not only because it really gives solid support for the use of genomics in our practice, which is going to become more and more commonplace because the evidence is strong, but also has wonderful things like a table of all the ongoing paired clinical trials. And it's just really well put together. It's fantastic work by you and your colleagues, Dr. Spratt, and thank you for sharing it with us today.
Daniel Spratt: Thanks so much for having me.