Alicia Morgans: Hi, my name is Alicia Morgans and I am a GU Medical Oncologist and Associate Professor of Medicine at Northwestern University. I'm so pleased to have here with me today, a friend and colleague, Dr. Michael Morris, who's going to be presenting a step towards personalized medicine, PET PSMA imaging in prostate cancer, and a specific talk on PyL PET imaging. Thank you so much for being here with us today, Dr. Morris. Dr. Morris is a Professor of Medicine at Memorial Sloan Kettering, as well as being the Prostate Cancer Section Head. Thank you, Michael.

Michael Morris: Thank you very much for that kind introduction, Alicia, and thank you for inviting me to talk about PSMA imaging and specifically in relationship to ¹⁸F PSMA imaging with the compound DCFPyL. So I think we're at a period right now in which we are seeing PSMA imaging really move from a field of data that was based on initially some really surprisingly clear images of prostate cancer disease and its distribution to then moving into retrospective studies and then subsequent to that, prospective studies with the intent of regulatory approval in the United States of these tracers. And these data that I'm about to discuss are the prospective data for an ¹⁸F radio labeled tracer known as DCFPyL, and hopefully these two trials, the OSPREY and the CONDOR studies, which are the registration studies for this will ultimately lead to FDA approval, the new drug application is in. So the hope certainly is that these two trials named after somewhat uncommonly birds, but I suppose majestic in their own way, will ultimately lead to regulatory approval and availability for patients across the country.

Here are my disclosures, and I think it's an important point to begin with is why do we image at all in prostate cancer? Really, the answer depends on where the patient is in the spectrum of disease. I've just put up the clinical state's model for prostate cancer from Prostate Cancer Working Group 3 to show how these questions pose themselves throughout the natural history of the patient, as he progresses from diagnosis, ultimately to the end stages of the disease. Certainly from a clinically localized disease in particular for high-risk patients, imaging, or accurate imaging is really important in terms of initial staging and treatment decision making for the patient. For those high-risk patients, whom in the past we really didn't know did they have strictly localized disease or did they have what was previously micro-metastatic disease or outright systemic disseminated disease?

PSMA imaging has the potential to more clearly identify the distribution of disease in those patients so we can more clearly stage them and then make decisions about their treatment, which are based on more accurate understandings of where their disease distribution is. For the rising PSA population, certainly accurate staging is crucial here because these patients could either be curable if they have a localized disease in the prostate bed or in the irradiated prostate, depending on what kind of upfront therapy they received, or it's an early appreciation of whether they have metastatic disease or a systemic disease or not and whether treatment should be focused on those areas, as opposed to receiving perhaps unnecessary local treatment to the area that the primary used to be in.

Then finally, for patients who have distant disease, imaging serves many purposes in terms of treatment prediction, response assessment, prognostication based on distribution of disease. For example, identifying visceral versus bony disease or low volume versus high volume disease. Then for drug development, identifying, targeting, and pharmacodynamic effects based on molecular imaging can be quite a useful tool. Now, all of these purposes, of course, are based right now on standard cross-sectional imaging and bone scintigraphy. As PSMA imaging is introduced into each of these contexts, the role of that imaging and the implications for treatment decisions needs to really be redefined. So we're on the cusp now of redefining really the staging of prostate cancer, the prognosis, and the risk stratification within each stage, and as well how to best treat these patients based on a newly comprehensive and accurate representation of where their disease is.

For those who may not be familiar, molecular imaging is dependent really on a few key concepts. One, the presence of a molecule on the prostate cancer cell, in this case, PSMA, that's expressed in order to target to bring in the radioligand, so that we can image. Of course, that target could be used for therapy as well, but that's a topic of a different day, I would think. But the way that the molecules work is you have a targeting agent to attach to the target, and that agent is carrying with it a payload, which in the case of PET imaging as either gallium 68 or ¹⁸F.

Now, how you target the molecule on the cell, which in this case for today is PSMA, can vary in size from molecule to molecule. So you can have antibodies, you can have minibodies, you can have just an FC component, or you could have a small molecule, like a DCFPyL and the size of these molecules and what you select to be the targeting agent can have really important implications in the value of that and the convenience of that radioligand.

So, for example, if you use an intact antibody to image PSMA, well, antibodies have a relatively long half-life. So the tumor to blood ratio is much longer and you might not get blood clearance for a week or more, and you wouldn't be able to image well, whereas a small molecule clears from the blood pool much quicker, so you can generate a good tumor to blood ratio in which the tumor really lights up on a negative background shortly after injection. Hence, the preferability for molecules, such as PyL or in the case of gallium, PSMA-11 of the small molecules. Generally, these have been developed by Marty Pomper at Johns Hopkins. They have a common core based on urea, and these are now the most commonly used molecules or the basic structure of them for PSMA-based PET imaging.

This is singularly difficult because you can't have a histologic comparator easily because, by definition, there's nothing to be seen on standard imaging and qualifying imaging tracers in this space does require usually composite endpoints as truth standard to compare with the imaging, as opposed to the patient with high-risk localized disease, who's going to undergo a prostatectomy because there, you do have a histologic comparator or the patient with metastatic disease for whom you do have pathology comparative just by biopsying the patient. So CONDOR, as all trials do in this patient population of the rising PSA patient, faced some distinctive design challenges, which were discussed extensively with the agency in order to create a trial that was informative.

So let's just move state by state here in terms of the data that support PSMA imaging with DCFPyL in the prostate cancer spectrum. So let's start with the high-risk clinically localized disease patient. These patients on cohort A of OSPREY were intended all to go to surgery and so after that surgery, there was a comparison of the pathology that was then compared to the imaging. There was a specific focus in terms of the primary endpoint on the lymph nodes, because really it's the extra prostatic disease that was considered to be the greatest asset of PSMA imaging in terms of identifying extraglandular disease in patients preoperatively who are at risk of having extraglandular disease.

So the intent of the PSMA image in this case is not specifically to identify prostate cancer in the prostate. That's really a different task establishing without a biopsy. For example, whether a patient has prostate cancer or not, or high-grade prostate cancer. The focus here is on that practical question that faces the urologist as the patient who has high-risk disease is about to go to surgery and that is in treatment planning, does this patient really have strictly localized disease or does the patient have nodal or outright M1 disease and would perhaps require a different consideration for either altering the surgical template or planning the patient for multimodality therapy, which may or may not include a prostatectomy?

So the sites, there were 10 sites, eight in the US, two in Canada. There were three central and independent PET readers, and then one separate reader who was reading the conventional scans, all blinded to the clinical information and to each other. What's important here in interpreting these data is the fact that there was a histologic gold standard of truth here. So a very high bar, and I think it's worth pointing out that in some circumstances, of course, you would never expect a PET image to perform as well as microscopy in terms of interpreting the performance data that I'm going to show you. We don't think of PET as essentially imaging microscopy. We think of PET as having some threshold of detectability that will be somewhat worse than looking at lymph node packets sent to the pathologist, sectioned, stained, and then examined under a microscope.

So there were 385 patients in OSPREY, and we're talking right now about cohort A, which involved 252 men and cohort B, which we'll get to a little bit later, had 93 men in it. Again, all of these men had traditional NCCN high or highest risk disease. If you look at the patient characteristics, then unsurprisingly, you had the majority of patients here representing Gleason 8, 9, or 10 Gleason scores with the minority of these patients having Gleason 6 or 7. Again, not surprisingly, these PSAs were at a median of 9.3. So high-grade with PSA levels that if they didn't have high-grade, would bring them into the high-risk. You can see that the range here was from 1.2 to 86.2 in terms of the range of PSAs. Many of these patients did have extracapsular extension. About a quarter of them were seminal vesicle involvement as well. Similar to NCCN, you can make your way to high-risk by a variety of criteria.

So here are the summary data of the primary endpoint, which was sensitivity and specificity, but also the secondary endpoints of positive predictive value and negative predictive value. So what we did find was that the sensitivity of the tracer was 40.3% when we looked at all lymph nodes. Of course, if you remember what sensitivity is dependent on, it's really dependent on the number of false negatives that you have, because it's the true positives divided by the sum of the true positives plus the false negatives. Why would you have false negatives when comparing PET to histology? Because you would have lesions that are so small that they wouldn't be detected by PET, but they would be detected by microscopy at the pathologist's microscope.

So what we wanted to see was, well, what would happen if we set a threshold then of removing from that the nodes that were so small that we wouldn't expect PET to pick them up at all. There, the sensitivity came up. The number of false negatives declined as well. So I think that that's an important factor to keep in mind in terms of trial design, that is, we set the bar extremely high in terms of holding the PET accountable, especially for things that you can't visualize on a PET scan.

Now, moving to the lesions that you can visualize. So specificity, which is really the patients that actually don't have disease that are reflected by a negative scan. Specificity is dependent on the formula of true negatives over true negatives plus false positives. So really, the false positives are determining this? There are very few false positives with DCFPyL. You can see that specificity is virtually 98%. Meaning generally speaking, if you see a lesion, it is going to be a lesion, and that's reflected in the positive predictive value as well, which is the number of positive scans or positive findings that turn out to actually represent disease. It's quite high. This is true positives over true positives plus false positives. So 85% or so in terms of positive predictive value and very good negative predictive value as well.

When you think about which of these numbers is really most important to us as clinicians, it's really, even though the primary endpoint here is sensitivity and specificity, but when you think about how we work and how we think, it's really positive predictive value. That's the most important number to us because we come into the patient with negative standard imaging or equivalent standard imaging in terms of distance disease, and then we take the PyL scan or the PSMA PET and then you wonder whether the positive findings that you see should change your management in terms of, can I trust that that positive finding on the scan actually represents disease? I think that the answer to that is in most cases, yes, you can. So I think that from a clinician standpoint, it's really the PPV that's most important.

Moving to the rising PSA population, we're just going to talk about the CONDOR trial for a little bit. So we're going to switch from OSPREY to CONDOR. I described what the dilemma is for most clinicians and the rising PSA population. I think all people that have treated prostate cancer have seen this dilemma. You have a rising PSA. Usually, that PSA is relatively low. You have negative or equivocal standard imaging, and you don't know what to do. You don't know whether to give the patient salvage local therapy after prostatectomy. Generally speaking, that's going to be salvage radiation therapy, or whether that patient will receive that therapy to no effect because they already have micrometastatic disease.

So the CONDOR endpoints, because these patients don't have any standard imaging, generally as a standard comparator, and it's hard to biopsy these patients. The primary endpoint is a composite endpoint that was proposed by the agency that essentially equates to positive predictive value, but with the added requirement of anatomic location matching, right? So a true positive there, let's say you had a positive finding in a left-sided rib, but the actual correlate is in the right pelvis. That shouldn't count as a true positive, right? So there has to be some synchrony between the PyL finding and whatever's going to constitute truth anatomically for the correlation. So what does constitute truth was really a composite truth standard and that composite truth standard, I've outlined on the next slide, which really is hierarchical. So the first and highest level of truth is, of course, histopathology. So if you saw something on the PyL scan that you could put a needle into, or you could verify what it was, perhaps by a pelvic lymph node dissection, then certainly that's a gold standard of truth because you have tissue correlation, but for most patients, of course, you won't necessarily have that.

So the second tier of the composite truth standard was some other combination of conventional imaging. So that could have been a standard CT or focused MRI on a specific area that perhaps the PyL was detecting, or it could be bone scintigraphy or because the FDA had approved fluciclovine and choline, it could be another type of molecular imaging as well. Then finally, if none of these would suffice to identify a lesion as a true positive, then you could irradiate that lesion if you saw 50% PSA decline, then that too would define the truth composite. So any one of these or any permutation of these could define a truth standard for a positive on the PyL scan. This trial specifically addressed the issue of positive predictive value. So only the positive lesions were counted here, so to speak and the only those positive lesions were then applied to this truth standard.

So the way that the study worked was the patient who had biochemically relapsed standard AUA or Phoenix definition would get the PyL scan. Right before that PyL scan would be performed, the clinician filled out a management questionnaire as to how he or she was planning to manage the patient. Then if the patient had a positive lesion on the PyL, then they underwent this subsequent imaging or biopsy or radiation to define the composite standard of truth. Those who had no findings on their PyL scan, they did not undergo any additional testing, but everybody got a post-imaging questionnaire as well to see whether or not the image made a difference in terms of the treatment decision.

There was a lot of blinded groups here. There was a blinded group of radiologists reading the PyL PET scan. There was a separate blinded group of independent arbiters of the composite truth standard. These groups never interacted with each other. Only the statistician would have the blinded result of the central reads and the result of the composite standard of truth, and then would match the data together. So this group of patients, as you might expect for most biochemically relapsed patients, had a median of a PSA of 0.8, relatively low. Although, as you can see from this range, there were certainly some outliers that would have, if we express this in a mean skewed the data, but some of these patients had very high PSAs, but very few up here. You can see that most of the patients in the trial had a PSA of less than two.

If you add all of these strata up 70% of these patients had a PSA less than two. So you can ignore most of the higher outliers, but the median was 0.8, where most people are struggling with this decision at those low decimal point values of making a decision as to whether to give these patients a salvage therapy or not. So here are the top line data from CONDOR. So the first thing I want to point out is this top red box. These are the three independent readers who were looking at the scans, and I've just presented the data from each reader on to himself and into a separate column. So remember that all of these patients had negative or equivalent standard scans as performed by the patient's treating physician. So of these patients, about 60% of them did have a PSMA PET finding. So you can see how much more informative a PSMA PET is than a standard CT or bone scan, generally speaking, in the sense that for seven out of 10 of these patients, you did have something positive to guide your decision-making in comparison to standard scans.

The second thing is the CLR. This is the positive predictive value with anatomic localization. You can see that that is 85% in one reader, 87 in another, and 85 in the third reader. So pretty uniformly, quite good. From a statistical standpoint, the threshold for success was defined by 20%. So this is well over the threshold that had been agreed upon with the agency for defining success. The reason for defining success at 20% is that the standard most sensitive study that has been FDA approved for detection of disease in this group of patients is Axumin®. We were looking at a PSA less than two in Axumin® and what the success rate would be for those patients and set that as the bar.

You can see that the positive predictive value here, where the CLR is pretty good no matter what the PSA stratum is from low PSA values to intermediate PSA values, to the outright, very high PSA values. So I think that the takeaway is here, even with patients with low PSAs, the data are reliable. So in terms of the change of management, 64% of these patients had to change in intended match management. What's important is that some of these changes are really important to the patient's life and outcome. So for 20% of these patients, that treatment plan changed from a systemic therapy plan, where there was a non-curative treatment intent to a curative plan. That's not to say that the patient was necessarily cured, but at least the plan changed, which changed to a treatment goal, which was cure involving salvage local therapy.

So 20% of those patients brought definitive therapy back into the therapeutic equation. On the converse, about 28% of the patients had a change in their treatment goal from salvage local therapy alone to either salvage local therapy with some systemic therapy or to systemic therapy alone. Then around 24% of patients had a change from just observation to actually initiating therapy, presumably because the target had been identified and 4% had the opposite plan treatment plan, which was changed then to just surveillance and follow-up. So of course, we don't know whether the changes that were proposed by these physicians are the best changes. We don't know whether those changes resulted in cure or not. We don't know whether they resulted in prolonging survival or not. That would be the topic of a different clinical trial, but nonetheless, you can see how people do incorporate this information.

Remember that it is primarily the positives that drive the decision-making because the patients had negatives on their plain old conventional scans to start with, so a negative on a PyL doesn't bring you any lower level of information that they would have had, but the higher level of information is the positive finding. Of course, we do need to do the future clinical trials to find out whether this decision-making is in fact the best decision-making and how molecular imaging should ideally impact the treatment decisions that clinicians make.

Then I'll just wrap things up by looking at the metastatic patients, and the reason that cohort B was added to OSPREY, so we're moving back to OSPREY for just a moment. The reason that it was added, just to get more data on the performance characteristics at identifying each type of metastatic lesion from bone to lymph node to viscera. We know that biologically, that these different areas of disease do carry with them different biologic features. We also know that they carry different prognostic implications for patients. We wanted to see how PyL performed in each of those different locations. I'll limit this to just one slide, where you can see divided by overall bone nodal disease and viscera and note that there are very few visceral soft tissue specimens as is anticipated, given that this is a prostate cancer population. You can see that both sensitivity and positive predictive value pretty much hold from organ system to organ system, from metastatic site to metastatic site.

So with that, I think we have to say that first of all, as of about two weeks ago, PSMA imaging, at least with gallium 68, PSMA-11 should no longer be considered to be investigational therapy in the United States. It was FDA approved, specifically at UCLA and UCSF. What I think is very important in regards to the label here is that PSMA-based imaging was not just approved for the purposes of detection of disease at the point of relapse, which is what currently the molecular imaging studies such as Axumin® are approved for in terms of their label, but rather it includes initial staging of patients, as you can see here at my cursor, who are potentially curable by surgery or radiation who have suspected prostate cancer metastasis.

I think that that's important to recognize also the broadness of that wording. The label does not define for you who those patients are. You just need to suspect that they may have distant disease. So I think that's a very positive movement because really, I think that it's not just the relapsed patient that can be helped by PSMA imaging, but the patient with high-risk localized disease, and this label opens that up in terms of molecular imaging for the first time, at least in the United States. The second thing that I would say is that PyL was last week awarded for priority review of the data that I just showed you for its NDA. So we're hoping, of course, that this is a positive NDA and that it will result shortly within the next several months. Hopefully, it will be approved as well. We can certainly discuss what the advantages or disadvantages of one tracer versus another might be.

There is a third NDA also in review that will also add to the armamentarium of these imaging agents if two is approved. So in the US in the next year, I would think that PSMA will really transition in full force from just an imaging modality available to those in clinical trials or those who could afford to go to other countries to get them to something that's available more ubiquitously in the United States. If PyL is approved, it won't apply to specific sites. It will be a national approval. Hopefully, for all of these tracers, these will be ubiquitously available to the prostate cancer patients who need them.

Just to wrap things up, I think that molecular imaging really is an assistive device for making good decisions in prostate cancer. It certainly should reflect the true status of the patient's disease as much as possible. I think PSMA is the closest that we've come to that. PSMA-based imaging does give reliable information about the distribution of disease, because are just the performance characteristics of these tracers and they are superior to current standards, but how we use that information, well that still has to be determined in future studies. With that, I think that will wrap things up. I thank you for your attention. 

Discussion