Predictive Modeling for 177Lu-PSMA-617 Therapy in Patients with mCRPC - Ken Herrmann

June 21, 2023

Phillip Koo interviews Ken Herrmann about his presentation focusing on a predictive model for 177Lu-PSMA-617. Herrmann explains the difference between a prognostic tool, which indicates a patient's overall prognosis, and a predictive tool that specifies whether a patient is likely to benefit from a particular therapy. He presents a model based on data from the Vision trial and explains how this model uses parameters to predict patient responses and survival rates. This model notably considers the whole body SUVmax, a predictive marker, as a major contributor to the score. They further discuss the potential role of artificial intelligence in refining these models and the impact of different PSMA tracers on results.


Ken Herrmann, MD, MBA, Professor and Chair of the Department of Nuclear Medicine, Universitatsklinikum Essen, Essen, Germany

Phillip J. Koo, MD, Division Chief of Diagnostic Imaging at the Banner MD Anderson Cancer Center in Arizona

Read the Full Video Transcript

Phillip Koo: Welcome to UroToday, and our coverage of ASCO 2023. We're very privileged to have with us Dr. Ken Herrmann from University of Essen, who's presenting data regarding a predictive model for Lutetium-177-PSMA. Thank you for joining us. Before we get into to the data that you're presenting, can we talk a little bit about what is a prognostic tool versus a predictive tool?

Ken Herrmann: That's an excellent question. So the prognostic tool mainly says... It's a tool telling us if a patient is going to have a good prognosis or a bad prognosis. A predictive tool rather, tells us if a patient is going to benefit from the therapy, or not. And the really important thing is... Because certain factors, for example, age or certain wellbeing parameters, by definition prognostic. But we really want to find out which are the patients, who really benefit from the therapy. And that's why we are looking mainly for predictive markers.

Phillip Koo: So to be able to develop something that tells us specifically whether or not they'll respond, or not respond, to a specific therapy.

Ken Herrmann: Exactly.

Phillip Koo: So you're presenting a model that's using some of the data from the VISION trial, to help us figure out what patients will benefit and who won't. And I know that's been a really hot topic, with regards to theranostics. Can you tell us more about this study?

Ken Herrmann: The groundwork was actually laid by the UCLA group, including Johannes Czernin, of course all the other people providing retrospective data to establish the first predictive nomogram, published in 2021, in Lancet Oncology. But like I said, this was based on retrospective data, mainly. And now what we did is, we really used the advantage of this fantastic pivotal, approval leading, VISION trial to really establish a new predictive nomogram, more patients, much better clinical data. And that is what we did. So we tried to actually predict, not only PSA 50 likelihood to respond, but also the likelihood to be alive after one or two years, both according for overall survival, but also for rPFS.

Phillip Koo: Great. Could you talk about some of the parameters where you're seeing some signal that tells us whether or not they will respond?

Ken Herrmann: We started with 29 predictive parameters, all prior to baseline. And then we performed the univariate analysis. Then we also corrected a little bit for collinearity. So for example, we usually have a collinearity and then we perform the multivariate analysis. And the interesting thing is that for example, for PSA 50, there were only three parameters predictive. For overall survival, up to 10 parameters per predictive.

Phillip Koo: Can you talk about the three that were predictive for PSA 50?

Ken Herrmann: For PSA 50? So of course, and the most important point is actually for all three different categories, the most dominant marker was the whole body SUV. And also when you talk about the three markers, it was also lymphocytes. The third one, I forgot, I think it was ALP, but not hundred percent sure. But their contribution to the overall effect was quite small. So especially if a PSA 50 with only three markers, the dominant marker was a whole body SUV.

Phillip Koo: And then what about for overall survival?

Ken Herrmann: Also, there were seven respectively, 10 markers including also factors for example, like ALP, like ADH. Also, patients had a long time or short time since the primary diagnosis of the prostate cancer. But again there, the major contributor to the score, the overall score was actually the whole body SUVmax.

Phillip Koo: So there's some early signal that whole body SUV and quantifying that can help us choose patients better. Where do we head next with really getting, developing a predictive tool that is reliable that we can start deploying in the clinic?

Ken Herrmann: So when you ask me this, one of the big challenges obviously is the whole body SUVmax, right? So right now SUVmax, we usually get ourselves even. So some clinics complain that nuclear medicine physicians don't provide it. It's still easily available. The whole body SUVmax right now is quite complicated. It's a tedious job.

So the first step would be to establish a tool where you really get it automatically. You have the scan and you automatically get the whole body SUVmax. The second thing would be then also to establish this nomogram to make it online available. The third one would be to get this tool obviously into the MDT documentation because technically you discuss a patient at MDT, it's not going to be a binary discussion, right? Pluvicto, yes or no, it will be the overall story. And having this information automatically available will help us guide the patients who we want to really treat with Pluvicto and where we would say, "Okay, this is maybe a case where we could also discuss something else."

Phillip Koo: One of the things I'm learning more about is artificial intelligence and being able to combine some of the data points that we have from imaging with the clinical data points, and being able to put all that together and look at real world data, real world evidence, and really be able to provide a more robust toll that iterates and gets better as we move forward. What are your thoughts on that in the future?

Ken Herrmann: So I'm a huge fan of that. So first of all, artificial intelligence will help us to actually probably make the automatic the interpretation of the whole body SUVmax. So it's number one. Number two is you're absolutely right. We have a tool, but this tool should not stop from now. It needs to learn for the future. So technically, if we are able to use the tool and really understand what happens to the patients along we will probably including the real world data, we find the tool for the future.

Another challenge obviously is all this was done with VISION. So we know VISION was only with Gallium PSMA 11. Now the big question is what happens if I use Pylarify? What happens if I use copper and PSMA? It's the whole body SUVmax makes it be different. And these are exactly the questions you could potentially answer much better, if we use a tool, we evaluate it prospectively, get all the information and... Because it's a lot of information, we probably need AI to help us to find out how is a corresponding effect.

Phillip Koo: That's a great point that you bring up, and I'm kind of torn with this idea of VISION had only used Gallium 68 and PSMA 11 and there's been a lot of discussion about using PYL and that being fine. And I think it's probably okay because we're looking at it in a binary fashion. But clearly as we move forward and we start looking at dosimetry and predictive tools, we probably need to look at it a little more differently as opposed to just putting all PSMAs in one bucket. But your thoughts on that?

Ken Herrmann: It's absolutely right. And even if you just look for the therapy selection, of course, because you have a different bio distribution, for example, the uptake in the liver is different. When you look at the promise criteria, we make a differentiation between Pylarify for example, or PSMA 11. You're absolutely right, there's something we have to take into account. We might also maybe see that the whole body SUVmax is different for different traces. Even so, we have a class of imaging with PMSA, they have subtle differences and we probably have to establish all this and have to plexus into the future and nomograms if we really want to use it for any patient who is potentially suited for protocol.

Phillip Koo: So you've had a couple days here in Chicago to explore and learn at ASCO. What are some of the highlights that you've seen and what are the things that you're going to walk away most excited about with regards to theranostics?

Ken Herrmann: So I'm of course super, super excited to see Lu177-PSMA-617 tomorrow. We both have seen the abstract, so I'm really, really curious to see the full dataset. I think this is one of the very appealing things, really combining a DNA and damage substance together with reliant therapy. The second one, which I find is really interesting for us, that for the first time we have shown tumor dosimetry, the substudy of VISION. You might remember that last year we have showed the renal dosimetry, which I think is very important because of the protection and overall the normal organs. But now for the first time we really see how the tumor dosimetry is.

And just to give you my personal one second or two second take home messages that we see it, that actually there's a 10 to 15 fold higher uptake in the tumor dose, the mean tumor dose compared to the kidney. And this is a quite impressive number, especially if you think about 23 great times 10 or 15. We talk all a sudden about 230 to 300, 245 grains of tumor dose and then we can really expect a nice responses.

Phillip Koo: Thank you very much. I think I'm excited about those two aspects as well, and hopefully we see continued research and investment in those spaces, and see some progress in the field. So thank you very much for joining us and we look forward to talking to you again.

Ken Herrmann: Pleasure to be here. Thank you.