Alicia Morgans: Hi, my name is Alicia Morgans, and I'm a GU medical oncologist at Dana-Farber Cancer Institute. I am so excited to have here with me today Dr. Eric Klein, who is an emeritus professor at the Glickman Urological and Kidney Institute at Cleveland Clinic. Thank you so much for being here with me today.

Eric Klein: Thanks, Dr. Morgans. Pleased to be here and to have the opportunity to actually discuss two papers that are related to a unique cohort of patients with early stage prostate cancer. So the project goes back about 20 years, and it's a description of 20-year outcomes based on pathology and Oncotype GPS score in a little over 400 patients that were used to develop the Oncotype GPS score. And the context is we were trying to find a tool that could improve our ability to predict patients who were safe for active surveillance. The endpoint of the first study was adverse pathology, which as you see here, is high-grade prostate cancer or disease outside the prostate. And the reason we picked those is that generally speaking, when we know a patient has these features of adverse pathology, whether that's an original diagnosis or while they're being followed for active surveillance, that they're not good candidates for active surveillance. And the data that I'm going to show you now from the paper actually bears that out.

So this shows 20-year risk of distant metastasis on the left and 20-year risk of prostate cancer-specific mortality in this cohort as predicted by the definition of adverse pathology there. So you see in the red line, these are individuals who had adverse pathology and at 20 years of almost 25% who had those features had metastatic disease, and about 7% had died of prostate cancer. So the question then comes up is what does this have to do with active surveillance? So we did a subset analysis, looking at patients who had biopsy and clinical features of intermediate, favorable intermediate, and low-risk disease are generally considered safe for active surveillance now. And what we found was that even if you start with those characteristics, and it seems you're eligible for active surveillance, if you have adverse path, even when you start out with good clinical features, you still have a significant risk of adverse outcome.

So in this particular cohort of low and intermediate risk patients, about 17% had metastatic disease at 20 years, and about 4.5% died of prostate cancer. So that led us to thinking about that, now that we have the 20-year outcomes, led us back to the original [inaudible] Oncotype and ask the question can Oncotype predict and be validated for the presence of adverse pathology? And the answer is yes, and that data's all been published. So if you have a high Oncotype score, you're at much higher risk of having these features that we now show are associated with bad outcome at 20 year. And so from the same cohort, we had published in JCO Precision Oncology last year 20-year outcomes, just based on Oncotype score. And what we found from this 17-gene expression profile based on biopsy was great predictability, that's judged by the area under the curve, both for metastasis and for prostate cancer mortality.

And as you see, what's mapped out here on the X axis is Oncotype GPS score, and on the Y axis is risk of metastasis or mortality. And you can see Oncotype score predicts for these things, in addition to predicting for adverse pathology, very well. And this is also true in the low and favorable intermediate risk group, who we consider for active surveillance. So there's a nice alignment here of observation. We know that adverse pathology is a bad thing in terms of long-term outcome. We know that Oncotype can predict for the presence of adverse pathology, but here we show a direct correlation between Oncotype score and long-term outcomes. And it's very striking. And I want just too show in the larger scale what these curves look like. And what's interesting is that it's not linear. It seems like there is an inflection point right around an Oncotype score, if you look closely, at around 29 or so for metastasis and for prostate cancer death.

And so what that suggests to us, that an Oncotype score below 29 in a patient who otherwise meets characteristics or criteria for active surveillance is at a pretty low likelihood of having adverse pathology and the low likelihood of having metastatic disease and is probably a good candidate for surveillance. The same could be said for a prostate candidate. On the other hand, once the Oncotype score gets into the 30s and higher, almost regardless of their clinical presentation, even if they presented with low or favorable intermediate risk disease, they're still at risk for significantly bad outcomes down here. To me, that helps solidify the data that suggests that a gene expression profile test like this can actually be useful in excluding patients from active surveillance. If you have a low score, it doesn't absolutely mean that you don't have something bad going on, but if you have a high score, I don't think that there should be any question that those patients are not good candidates for surveillance. And that would be true at the time of diagnosis or if the Oncotype score is done later on in the course of disease.

And so you know that the other tool that we use to assess the extent of tumor in the prostate is MRI, which is great for guiding biopsies and certainly has made biopsy a lot better than it used to be. But there are some head-to-head studies that have compared the ability of MRI or things like Oncotype or Decipher to predict for these adverse pathologic features that we know at 20 years are associated with higher risk of metastasis or death. And if you read the papers carefully and were to vote which did a better job of [inaudible] gene expression profiling in all three. And I think this is surprising to most urologists and certainly to most medical oncologists. Urologists are very focused on MRI. And the way I think about this is that these two tests are important, and they give us complimentary information. And this is how I look at it. MRI helps you find the shark fin. If there are bad things going on that are visible above the water or the surface, MRI is certainly going to find it. But genomics helps you see all the garbage that's underneath the surface there. And again, to me, they're complimentary. They're not mutually exclusive, and probably both are necessary when qualifying a patient for active surveillance.

Alicia Morgans: Thank you so much for going through that. I do have to commend you on your choice of images, because those really, I think, drive your point home so effectively. And I have to say, as someone who practices and helps patients through the decision-making process in active surveillance, it's a very trying time. I recommend actually genomic profiling pretty regularly, but not all the time, but it's not uncommon necessarily for urologists not to use that type of an approach and really to rely on the clinical features alone and then of course do use MRIs in terms of identifying and following up with the biopsies, as you've said, in most of our active surveillance protocols. But how do you advise folks to start using or integrating these kinds of risk profile systems into their practice if they're not familiar with them?

Eric Klein: So you have to understand what they show, and you have to understand what MRI shows. So to understand how to use genomic testing, it's useful to understand the history of how we have defined adverse pathology or high-grade prostate cancer that needs to be treated. So that's based for historical reasons on histology and the development of Gleason scoring. And until recently, that's where we started to be able to detect the presence of aggressive cancers. There was a lot of pattern 4 in the biopsy. And it's only after that that we got biochemical PSA detection and radiographic detection, meaning that the molecular changes that cause the cancer occur, first they become evident histologically. First, if you're lucky enough to have a biopsy, then you get an elevated PSA, then you might see it on MRI, and then you get clinical detection. The development of these gene expression [inaudible] profiles have shifted this to the left. And now we have genomic and gene expression changes that precede what the pathologists can see under histology. And that's what these tests are measuring. They're measuring the biologic potential of the cancer that's not yet evident in histology.

And there are several publications, for example, that show that cribriform histology and [inaudible] histology promote aggressive cancer, and that's true in [inaudible] active surveillance [inaudible], but most pathologists don't see those and don't report those because they don't have the expertise on it. And Oncotype and Decipher can substitute for that pathologic expertise. That's really why they're useful. So here's how I use both histology and gene expression profiling in my practice. Histology is useful in certain circumstances where you don't need genomic testing or gene expression profile testing. If you have a microfocus of low-grade prostate cancer and a negative MRI, genomics or gene expression testing isn't going to add any value. You don't need to do it.

If you already have adverse pathology, it's in grade group three are higher, you don't need genomic testing. And if you have cribriform or intraductal, we know those are bad. You don't need genomic testing. The sweet spot for genomic testing is here. It's when you have multiple cores of Gleason [inaudible] cancer or low-volume Gleasons 3 + 4 cancer, because the genomics testing has revealed many of those are safe for active surveillance. And the common scenario now is someone with a PI-RADS 4 lesion on MRI has multiple cores of tissue taken from that lesion, and it's all grade six. And the MRI doesn't tell you anything about whether or not that cancer needs to be treated, but gene expression profiling will. So that's the sweet spot for genomic testing. And so if you're going to adopt them, think about the specific clinical scenarios where genomic testing has added clinical value and use them in that circumstance. And if patients have a high score, they're probably not a good candidate for surveillance. And if they have a low score, it's just one additional piece of evidence that it's safe for them to start on surveillance.

Alicia Morgans: I think that's a great way to think about it. So as you have it labeled here, the less useful time for histology is really that high-volume Gleason grade 1 or the low-volume Gleason grade 2, to really kind of sort things out. And it's interesting, as we think about risk stratification, as we're specifically biopsying lesions that we see on MRI and we have our risk profiles around the numbers of cores positive, that doesn't really help me if most of the cores are specifically taken from the MRI lesion-

Eric Klein: [crosstalk]. Exactly right.

Alicia Morgans: So we have a lot of evolution to go through, I think, in order really make sure that we're advising our patients. But I personally find in my practice that there is a role for these in exactly the space that you're mentioning. And so I really appreciate that. I appreciate you sharing that. And also would just encourage urologists to think about it, because you can send these samples on the biopsy tissue. That's how they're designed. You do not need so much tissue that you need a prostatectomy to get these tests done. They're designed for biopsy samples, and it can be really, really useful. So that makes a lot of sense as you think through histology and genomic profiling, but I'm curious how you think through the complimentary roles that MRI may play with genomic testing.

Eric Klein: Yeah, that's a great question. When these technologies first came out, people thought that they were at odds with each other, and you had to pick one or the other. And the reality is they're complimentary, and they give us complimentary information. So here's how I think about it. The added clinical value of MRI is it makes biopsy better. I don't think that there's any question about that. On the other hand, the added clinical value in genomics is that it reveals the biologic potential of the tumor, which is not something that MRI can do. And, again, in particularly the example I used, which is very common in practice now, a PI-RADS 4 lesion with four or five cores in it that's all [inaudible], the MRI is not informative there. MRI is highly operator-dependent. I know lots of people outside of academic medical centers who don't have experienced radiologists to read MRI.

On the other hand, one of the underappreciated advantages of genomics is that it's objective, and it can augment regular pathology and substitute for some expert pathology. If the patient's already had a biopsy and not had an MRI, in order to make MRI useful requires a second biopsy. On the other hand, even without an MRI, genomics can be performed on the initial biopsy. And then finally, based on the data that I showed you, one other advantage of genomics is that MRI is not validated for any outcome. And in fact, there's data out of Michigan where they tried to look at a gene expression profile that was associated with MRI visibility and to see if that would predict [inaudible] outcome, and the answer was it wouldn't. So there are lots of advantages to each. They answer different questions.

Alicia Morgans: So as you think about the general concepts of these genomic profile tests and of course the use of histology and the papers that we've talked about, what would your take-home message be to the audience as they're trying to sort through ensuring that the right patients get on active surveillance and making sure those patients who may not be the right fit get the treatment that they need?

Eric Klein: So in summary, when I think about someone for active surveillance, which is certainly the management strategy of choice for the right patient, based on the data that I showed you, you really want to be sure that they don't have adverse pathology. There are a number of ways of ensuring that. One is looking at histology. One is looking at the MRI and doing guided biopsies. And the last and important piece that improves our ability to predict the presence of adverse pathology by about 20% in every study that's been published, genomics or gene expression profile.

Alicia Morgans: Well, thank you so much for talking us through that. I do hope that some folks who are listening think about integrating this data that you've shared with us from your two publications into their clinical practice, and perhaps, as you said, using MRI histology and adverse pathology and our genomic profiling systems and together to really help us make those best recommendations for our patients. So thank you so much for your time and for your expertise. Keep up the good work.

Eric Klein: Thanks.