Philip Koo: Hi, this is Philip Koo from the Prostate Cancer Foundation, and welcome to UroToday. Over the past several years, we've seen PSMA-PET really revolutionize how metastatic disease is diagnosed in patients with prostate cancer. It's been wonderful to see how it's been utilized, but it's also really, really exciting to see how it's going to be utilized in the future, especially as a prognostic and predictive tool. And today, we have with us Dr Rahbar from the University Hospital Münster, who's really been at the cutting edge of helping us understand the use of PSMA-PET as a biomarker to really help us learn more about the disease that we're dealing with. So Dr Rahbar, thank you very much for joining us.

Kambiz Rahbar: Thank you. Thank you for the invitation.

Philip Koo: Wonderful. So let's start off with your presentation and then we'll have some Q&A afterwards. I'll turn it over to you.

Kambiz Rahbar: All right, thank you. So first of all, I would like to thank everybody for the opportunity to present our work. So today, I will discuss the updated risk stratification in prostate cancer using the PSMA-PET imaging standardized by the PROMISE framework. So the data come from the PROMISE Registry, which was started by Wolfgang Fendler from the University of Essen, which is an international multicenter registry. So let me briefly summarize the background. As you know, PSMA-PET was introduced around 2012, and since then, as you already mentioned, it has evolved into a powerful imaging tool in prostate cancer. In 2018, the PROMISE criteria were proposed to standardize PSMA-PET interpretation. And in 2024, the first prognostic nomograms, the PPP1, were presented.

But these were just based on single center data. So the aim of this study and the results I'm presenting today, is to reassess and improve these nomograms using a large multicenter registry, which results in what we now call the PPP2. So we analyzed data from 6,128 male patients across all stages of prostate cancer. These patients underwent PSMA-PET at 20 hospitals in Europe, the US, and Australia, with a minimum of 36 months of survival follow-up. We used Cox regression models with the LASSO penalty to identify relevant predictors for overall survival and constructed the PPP2 nomograms. We then evaluated their prognostic performance using ROC analysis, and compared them with NCCN clinical trial classifications. For the PSMA-PET metrics, we used a visual and a quantitative model.

So we developed two versions of the nomograms. One based on quantitative metrics, such as total tumor volume. And another using visual interpretations, such as lesion count, PSMA expression scores. And both versions rely on PROMISE features for standardized reporting. Here, you can see the visual and quantitative PPP2 nomograms, and these allow us to predict three and five-year overall survival by combining metastatic spread, tumor burden, and PSMA expression levels. So both visual and quantitative PPP2 nomograms demonstrate strong discriminatory power. In the validation cohort, the Harrell's C-index was 0.8 for both, which indicates excellent prognostic accuracy. So using the PPP2 nomograms, we stratified patients into three risk groups, low, intermediate and high risk.

So these risk groups were shown to significantly correlate with overall survival in the total cohort and all disease stages subgroups, including biochemical recurrence, mHPSPC, and mCRPC. We also assessed the impact of PROMISE version I versus version II. The prognostic performance remained stable regardless of which version we used. And we also looked at the radiopharmaceutical used. And there was also no difference between these two. So which underlines the versatility of the models in clinical practice. So when compared to established clinical risk scores, such as NCCN and EAU classifications, the PPP2 nomograms showed also superiority in predictive accuracy. The area under the curve for PPP2 was significantly higher confirming its clinical utility.

So the PPP2 nomograms offer improved accuracy compared to conventional clinical scores. So they allow for reproducible and individualized survival prediction regardless of radiotracer or PROMISE version. So at the end I would say this is, on the left side, you see the publication according to the results I presented here. And I would like also to encourage the ongoing participation in the PROMISE Registry. You can visit PROMISE-PET.ORG to access more information. And if you're interested also to participate in this registry. And I would like to thank everybody to give us the opportunity to present the results. And thank you, Philip, for the invitation.

Philip Koo: Well, thank you very much. This is very interesting to see how you started off with a single center and then expanded it, and are seeing wonderful data to help support the use of this tool. So any signals with regards to which factors might be the main drivers of overall survival?

Kambiz Rahbar: So the main factors at the end would be the tumor volume. I think this is one of the major points which we have seen also in other publications previously. I mean, the larger tumor volume, the worse the survival. So I think this is the one major point that we should look at. And it depends also at different stages of disease, it might also have an impact on survival. So we included the first PET scan in each patient regardless of the stage where they were presented.

Philip Koo: Yeah. I like the fact that you had multiple different disease states that you were looking at. With regards to, you know, there's always a lot of talk about SUVmax, and that being a predictor. What's your general sense of that metric quantitative parameter as a tool for predicting prognosis?

Kambiz Rahbar: Yeah. First of all, as you better know, SUVmax might sometimes be only one pixel that indicates the SUVmax. So I think the most robust would be the SUVmean, or more the SUVpeak to use. So I think these are all the different metrics, but at the end I think the SUVmean of the whole tumor volume is the most important indicator. I don't think it is just one single lesion which predicts survival. I think it's about all tumors and the SUVmean out of that.

Philip Koo: So in general, would you say that the greater the avidity of the tumor, the greater concentration of the radiopharmaceutical, is that a worse signal or is it a better signal? Because I imagine in each disease state it might be a little different.

Kambiz Rahbar: It is different in each disease state. I think this is something we should look at in the future. Especially if you look at different results presented in the last couple of years, you see different SUV values presented also from the VISION study. So I think if you look at that study, there were SUVmean over the tumor, which was a predictor of response and overall survival and outcomes. And I think this depends on the stage. So if you are at earlier stages of disease, it depends what kind of PSMA expression you have. It also predicts the tumor volume. So I think it varies between the stages of disease. So we have to look at different variables. So it's not the single thing you can, or the single variable you can take, it's multifactorial.

Philip Koo: There's always struggle with getting the nuclear medicine physicians or radiologists to use these standardized templates, and then to make these calculations that we know are telling us certain things about the patient's tumor. From your perspective, how much extra time does this involve? And are there ways to automate and make this more efficient for practical use?

Kambiz Rahbar: Currently, this is a little time-consuming. I mean, I think we need more automated analysis of these results. But if you look at the templates from PROMISE, you have the molecular imaging TNM staging of PROMISE, which is very easy to use. There is also an application for that which you can download, and then you click just through this application, and you have the molecular imaging TNM staging. But according to tumor volume and SUVmean of the entire tumor, it can be time-consuming in patients with large metastatic spread. So we need in future automatic analysis, which is already available from some different companies, but we need more robust and more efficient software to do that. And we need standard software. So there are different software available.

There are freeware, there are software also from PET scanner providers. But we need a standard system. We need a standard way of use. So in this study we used the same software. But in the future, if you want to apply this to daily clinical practice, you need a single software, which is available for everybody. And does not have to be purchased with large amounts of money, so it can be used all over the world. So this is the future, what we need.

Philip Koo: Great. Well thank you very much, and congratulations. We look forward to seeing more and more work that really helps us get into this PSMA next level utilization, which I think is so promising. So thank you again, Dr Rahbar.

Kambiz Rahbar: Yeah. Thank you for the invitation. I would like to give thanks also to Wolfgang Fendler, who has started this registry. And he's the one, he's pushing this forward, and we all help him to do that. So I think there are many people helping in this registry. So the thanks go to everybody.

Philip Koo: Wonderful. Thanks.