Imaging and PET Analysis as a Prognostic Tool for Lu-PSMA-617 Treatment in Patients with mCRPC - Phillip H. Kuo

October 16, 2022

Phillip Kuo joins Alicia Morgans to discuss recent findings in a VISION sub-study. Dr. Kuo explains how this post-hoc exploratory analysis of the PSMA PET scans that were done for the VISION trial resulted in a better understanding of which patients are likely to respond better to treatment with lutetium-PSMA-617.

Biographies:

Phillip H. Kuo, MD, Ph.D., Professor of Medical Imaging, Medicine and Biomedical Engineering, the University of Arizona, Tuscon, Arizona

Alicia Morgans, MD, MPH, Genitourinary Medical Oncologist, Medical Director of Survivorship Program at Dana-Farber Cancer Institute, Boston, Massachusetts


Read the Full Video Transcript

Alicia Morgans: Hi, I'm so excited to be at ASCO 2022, where I'm speaking with Dr. Phil Kuo from the University of Arizona. Thank you so much for being here.

Phillip Kuo: Thank you for inviting me. In addition to being a professor at the University of Arizona, I want to mention that I'm the senior medical director for inviCRO, which is the imaging CRO that did the analysis on behalf of AAA Novartis for the VISION study.

Alicia Morgans: Great. Thank you so much. I'd love to hear you really give us some guidance on what you reported at ASCO 2022, and really how we can use imaging and the analysis of the PET scans that were used at baseline to help us understand who's going to respond better or less well to treatment with lutetium.

Phillip Kuo: It was really an extensive post-hoc exploratory analysis that we were asked to do by the FDA. So, this was separate from the VISION results, which as you know, already resulted in the FDA approval of Lutetium PSMA 617 for the therapy of metastatic castrate-resistant prostate cancer.

So, in this study, what we did was we did a quantitative analysis of the PSMA PET scans that were done as part of the VISION trial. In the vision trial, a PSMA PET was done at baseline for screening patients, for eligibility to randomization for the metastatic castrate-resistant prostate cancer PSMA 617 therapy, plus standard of care versus the control arm, which was standard of care alone.

So, we used the combination of artificial intelligence and manual segmentation to segment all the PSMA-positive disease within a patient. And it's important to note at this point that the patients were already by VISION underwent a eligibility visual analysis to remove patients who were PSMA negative. So these were only the patients who made it past that initial screening.

And in this case, to clarify, the definition of PSMA positive and negative was very specific to this trial. It wasn't whether or not it was prostate cancer or not. It's whether positivity was greater than liver by visual analysis and a PSMA negative was less than, or same as liver, by visual analysis. So we correlated our quantitative parameters with the prognostic information that was available from the vision trial.

Alicia Morgans: Well, that seems really, I think, potentially useful as we're trying to think about how patients might do on a given therapy. What did you find?

Phillip Kuo: With the VISION trial, when we used the PSMA PET for initial eligibility, we knew there was a lot more information there in those PSMA PET scans. We took a really blunt approach, initially, which was for enrollment into whether you'd be randomized or not, but there was so much more information there that we knew about. A lot of wonderful work had been done in earlier phase one and phase two trials around the world in PSMA-targeted therapy to show that this was likely going to be the case. But of course this was the VISION trial, which is the only phase three large trial that had the power to actually look at overall survival to a P value of less than 0.001 at a registration trial.

So, what we actually found was that the single most important predictor from all of our quantitative parameters was SUV mean. So a lot of oncologists I think, are used to the term SUV max, which is almost always exclusively what we report in the United States for a parameter. An SUV max is the single hottest voxel, which is a three-dimensional pixel, of all the disease. But what SUV mean is, which is a little bit harder to obtain, is after we segmented all the disease. So, in some patients, you can imagine it was bone, lymph node, liver disease. We segmented it all. We took all the voxels of all the segmented volume of disease and found the average of the SUVs of all those voxels. Okay? So it's not actually typically done in standard of care. But we hope to maybe change that soon.

If you had an SUV mean of 10 or greater, which was the top quartile. So we actually do a quartile analysis. And the top quartile was a SUV mean of greater than 10. That overall survival was much higher than the bottom three quartiles, the second, third, and fourth. The number was actually 21 months overall survival in the top quartile. And the other three quartiles range between 12.6 and 14.6. So a big, big difference. And when you actually compare that to the median of the whole study group, which was 15.3 months, for the treatment arm, and 11.3 for the standard of care arm, that's a 10-month greater OS compared to the standard of care arm alone and six months greater than the aggregate group. So a really big difference there in that top quartile.

It shows how that baseline PSMA PET has that prognostic information potentially, because that was not a primary endpoint to look at this, and that we can further stratify patients, and provide that personalized oncologic care to patients. And obviously, as oncologists, you can appreciate what the armamentarium of therapeutics that we hope to keep building in the future, you can personalize who would be best suited for which therapy by this imaging analysis.

Alicia Morgans: That's really fascinating. And I wonder, and you may not have investigated this, you mentioned the top quartile and that group really having of this extended survival. That's sort of an arbitrary cutoff. You picked the top 25%. Is there a cutoff of SUV mean, a numerical cutoff, where you might say, "Patients above this threshold have X expected prognosis versus below this cutoff," because that sort of numerical cutoff might be something that might be easier to use at least within a clinical practice?

Phillip Kuo: Right. So we did not present the analysis using the continuous data versus the quartile. You can understand why the quartile is obviously very easy.

Alicia Morgans: Yes, of course.

Phillip Kuo: It's easy to understand, but remarkably, actually it's pretty similar.

Alicia Morgans: Interesting.

Phillip Kuo: Yeah. And it's all working out very well. So Dr. Michael Hoffman presented some additional work at this meeting to build on what he'd presented before and in his work, not doing quartile analysis, [inaudible 00:06:25] more continuous analysis, SUV mean 10 actually also came to the fore. And so we've shown it here too. And in his studies, they used the same radioligand therapy, lutetium 177 PSMA 617, but they used it in a different way. So it's nice to see that even across trials, with some important differences in how the therapy was applied, but the same radioligand therapy. That SUV mean of 10 really is turning out to probably be a real number we can hang our hat on.

And in his work, it was around 30% of patients tend to have that. It seems like sort of in that same metastatic castrate resistant prostate cancer population, we're finding that due to the tumor biology, there's about that 30% that really are the high expressors of PSMA. And we found that as well likely in our study. So even if you don't use a quartile analysis, that number holds pretty well, which is really interesting because it does support that across potentially trials, there is this kind of maybe this consistent 30% population, even though Dr. Hoffman's group is Australian, New Zealand and the VISION trial was very multinational, and so had a much broader patient population. Although obviously, they, like many trials, still done a little bit better on diversity, it shows that that number really might turn out to be, and it's kind of a nice round number so it to turns out pretty well.

But the important thing to remember from this study and the VISION as well as other great ones coming out of Germany, Australia, and the United States is that those are the patients who have been screened already. Okay? So in the vision, if you had even one PSMA negative lesion of sufficient size by the size criteria, meaning that its uptake did not exceed liver, you were out. And so we did not include these patients in our analysis.

We are actually doing that analysis now. So we will have that data to really see how it compared to just standard of care alone, which is actually obviously a very interesting question separately. So we kind of removed that as they did in the Australian New Zealand studies, removed those PSMA negative if you will, patients. So you're only looking at those patients who were most likely to respond already. And then within that group, the ones with the SUV mean of 10 or greater.

It's important to point out that even that bottom quartile, bottom three quartiles, second, third, and fourth, still had median survivals that were better than the standard of care alone arm. So every patient who met that visual eligibility criteria still benefited, even that bottom quartile SUV mean, still did better than the standard of care arm alone.

Alicia Morgans: So that's definitely important to point out. And thank you for bringing up the therapy SUV mean of 10, because that was exactly where I was trying to go.

Phillip Kuo: Yeah.

Alicia Morgans: And interesting, of course, they had different inclusion criteria for the trials. So they had an FDG PET as well as the PSMA PET. And so perhaps they had 30% because they were enriching a little more specifically for the higher PSMA expressors. But who knows? Smaller study as well. But really, really nice to see some consistency. And as you said, a nice round number, which makes it easy. So, if you had to sum up your findings, this analysis, that is I think, really going to help us better understand and guide our patients and what to expect during treatment with lutetium, what would that be?

Phillip Kuo: Well, I think this is really important that we've shown that beyond that first step of excluding the patients who are really least likely to benefit, we can now further stratify patients with that baseline PSMA PET scan, that who's really going to do well with that SUV mean of 10 and that we should consider, although practically right now, it's difficult for the average imaging center to do and radiologists to do, to develop an SUV mean. But we are developing algorithms and software tools that will make that more readily achievable. That actually could really be a useful parameter for nuclear medicine physicians who are treating the patients, of course, on their oncologist and their whole treatment team.

When they're trying to decide what tools, which medications, will be best for this patient, and at that top quartile or SUV mean at 10, they really have a much better overall survival. And in VISION specifically, have 21 months versus the 15.3, median of the overall group and the 11.3 months for the standard of care arm alone.

So, we've really shown we can take that personalized approach with the imaging even by itself. And as we know, we can add other clinical factors in a nomogram-type of fashion to even then further develop what is the best treatment for the patient.

Alicia Morgans: Wonderful. Well, thank you so much for your time and your expertise today.

Phillip Kuo: Thank you.