PSMA-PET for Overall Survival Outcomes Risk Stratification of Prostate Cancer "Presentation" - Wolfgang Fendler
April 9, 2025
At the 2025 UCSF-UCLA PSMA Conference, Wolfgang Fendler demonstrates how PSMA-PET provides spatial biomarkers predicting overall survival in prostate cancer. Drawing from the 10,000+ patient PROMISE registry, he presents visual and quantitative prognostic nomograms that stratify patients into risk groups across all disease stages, outperforming established clinical systems. He introduces ePROMISE, a free application for standardized reporting with forthcoming risk prediction tools.
Biography:
Wolfgang Fendler, MD, Professor, Department of Nuclear Medicine, University of Duisburg-Essen and German Cancer Consortium (DKTK), University Hospital Essen, Essen, Germany
Biography:
Wolfgang Fendler, MD, Professor, Department of Nuclear Medicine, University of Duisburg-Essen and German Cancer Consortium (DKTK), University Hospital Essen, Essen, Germany
Read the Full Video Transcript
Wolfgang Fendler: PSMA-PET for overall survival prediction in prostate cancer. And the basic topic is that we image features of cancer not only at one site, but the entire body, and whatever we can capture there, in our view, is a spatial biomarker. It's not only localization of disease, but it gives us properties of the disease.
We try to categorize them in standardized reporting systems published under the name of PROMISE, but also other systems, beginning in 2018, and we believe that disease distribution can be independently associated with relevant patient outcomes, such as overall survival throughout the entire disease spectrum.
And I think this is hugely important. And we wanted to explore this deeper in a larger registry trial by looking at patients that have overall survival outcomes of more than three years in the entire disease spectrum, and then applying standardized reporting to those patients based on PROMISE criteria, which basically captures visual features of prostate cancer, disease distribution, number of lesions, the involvement of organs, but also captures some novel metrics such as total tumor volume and SUV mean metrics.
And the question is, do these PSMA-PET features outperform our current clinical data and clinical nomograms such as STAR-CAP for initial staging or EAU for BCR or the NCCN risk groups that we just heard of as part of the guidelines, or some novel nomograms? We really tried to map this using a registry trial and published it in Lancet Oncology last summer as a first system to use PSMA-PET for risk stratification in patients, and since then have tried to give updates at the congresses.
And the last update was presented at ASCO GU for a larger group of patients. This is an update from February 2025 that includes more than 10,000 patients worldwide. Most of the continents are involved, and there's a large number of patients coming from Germany because PSMA-PET was adopted very early there and the overall survival outcomes are more available.
In total, there are 49 institutions that contribute to this registry trial effort, and we were able to collect, until February 2025, these patient numbers and also event numbers. More than 6,000 patients are in there, more than 1,000 patients for each of the subgroups, including initial staging, BCR, nmCRPC, mCRPC, and mHSPC—so all the relevant prostate cancer stages.
And depending on prostate cancer stage, this registry includes an overall survival event count of between 20% in the early stages and up to more than 80% in the late stages of disease—so really relevant and partly mature overall survival outcomes for analysis.
Looking into the database—this is what we presented at ASCO GU—we wanted to find out if the PSMA-PET features are associated with overall survival in the entire group, but also in the different subgroups of the prostate cancer disease spectrum. And these are the submetrics, starting with local tumor stage, T stage, going to locoregional stage, N stage, but also all the metrics for organ involvement. There's tumor volume in there, there’s number of lesions in there, and there’s also a PSMA expression score in there.
Without going into all the details—there’s way too much on this slide—this is just to highlight that most of the features, nearly all of those features, are associated with overall survival in these patients. Only very few local disease features, such as unifocal local disease, are not associated with overall survival. All the other features—so more disease on PET—means worse outcome in terms of overall survival.
And this can be used to create nomograms and to combine those with regression analysis, also selecting those criteria which are most relevantly associated with overall survival. These are two approaches: separating into a visual nomogram that everyone can apply with any imaging system, any software, and a quantitative nomogram that additionally includes tumor volume whole-body assessments.
These assign total points, and the total points are translated into risk groups—into those three: low, intermediate, and high-risk groups. And those nomograms show a quite high association with overall survival, a C-index of more than 80%, and will be further detailed in a poster talk that we also have today from Flora Fankhauser.
I just want to highlight that if we apply these nomograms to the different disease groups—both the overall cohort of more than 6,000 patients, but also the initial staging BCR group and the advanced staging groups—the nomograms can separate patients with low, intermediate, and high risk significantly. They can separate different risk groups. And these are the survival plots for the more advanced disease stages: nmCRPC, mHSPC, and mCRPC.
Again, it's the same nomogram, one visual and one quantitative, applied across the entire disease spectrum. And it also works throughout the entire disease spectrum. So the critical question now for a prognostic nomogram is, does this give additional value? Is this superior to the current clinical systems out there, which are quite simple, taking information from the clinical routine—PSA level, local disease stage, histopathology?
We wanted to find this out based on the large registry cohort, and this is a comparison between the clinical nomograms (NCCN, both for the entire cohort, as the upper row, and for the initial staging patients only, where it’s meant to be used) and the EAU risk stratification proposed for the BCR patients. And you can see, by applying the PSMA-PET nomogram in a validation cohort (so a separate cohort not involved in the creation of the nomograms), the area under the curve—so the accuracy—of PSMA-PET-based risk prediction is higher through the entire spectrum than the clinical nomogram.
So it’s always significantly higher for both the visual and the quantitative risk assessment using PSMA-PET criteria, as compared to NCCN and EAU risk stratification criteria.
So in summary, looking at large registry trials (the large PROMISE registry), we can say that PSMA-PET-based nomograms—PPP, updated PPP, two nomograms—accurately stratify prostate cancer patients into different risk groups. And we get enough information to stratify patients into low-risk, intermediate-risk, and high-risk for overall survival throughout the entire disease spectrum. These nomograms, based on PSMA-PET, yield superior accuracy when compared to clinical nomograms. Here, as an example, we show NCCN and EAU, given at different stages. And we try to continue the follow-up and collect more patients and especially more overall survival endpoints in the PROMISE registry, to be able to get a higher level of precision of risk prediction in the future.
This also needs to go out, of course, as a clinical or usable tool. And for this, we picked up the idea with Technical University to create an online app, the so-called ePROMISE app, that can be reached over this webpage. It includes standardized reporting of prostate cancer based on PSMA that can be used on any platform. It’s free, it’s academic, and it will also include—once new risk systems are published (they are currently under review)—risk stratification into low, intermediate, high risk, and also prediction of overall survival in this one application to be used and explored online.
And with this, I thank you very much for your attention. I invite every department to also contribute to this registry if you’re interested. You have the webpage here. It’s a large registry trial, and I thank all the supporters of the registry trials—that’s the only way to gather this number of patients to be analyzed in the future. And thank you very much for your attention. I’m looking forward to the next talk from Irène Buvat.
Wolfgang Fendler: PSMA-PET for overall survival prediction in prostate cancer. And the basic topic is that we image features of cancer not only at one site, but the entire body, and whatever we can capture there, in our view, is a spatial biomarker. It's not only localization of disease, but it gives us properties of the disease.
We try to categorize them in standardized reporting systems published under the name of PROMISE, but also other systems, beginning in 2018, and we believe that disease distribution can be independently associated with relevant patient outcomes, such as overall survival throughout the entire disease spectrum.
And I think this is hugely important. And we wanted to explore this deeper in a larger registry trial by looking at patients that have overall survival outcomes of more than three years in the entire disease spectrum, and then applying standardized reporting to those patients based on PROMISE criteria, which basically captures visual features of prostate cancer, disease distribution, number of lesions, the involvement of organs, but also captures some novel metrics such as total tumor volume and SUV mean metrics.
And the question is, do these PSMA-PET features outperform our current clinical data and clinical nomograms such as STAR-CAP for initial staging or EAU for BCR or the NCCN risk groups that we just heard of as part of the guidelines, or some novel nomograms? We really tried to map this using a registry trial and published it in Lancet Oncology last summer as a first system to use PSMA-PET for risk stratification in patients, and since then have tried to give updates at the congresses.
And the last update was presented at ASCO GU for a larger group of patients. This is an update from February 2025 that includes more than 10,000 patients worldwide. Most of the continents are involved, and there's a large number of patients coming from Germany because PSMA-PET was adopted very early there and the overall survival outcomes are more available.
In total, there are 49 institutions that contribute to this registry trial effort, and we were able to collect, until February 2025, these patient numbers and also event numbers. More than 6,000 patients are in there, more than 1,000 patients for each of the subgroups, including initial staging, BCR, nmCRPC, mCRPC, and mHSPC—so all the relevant prostate cancer stages.
And depending on prostate cancer stage, this registry includes an overall survival event count of between 20% in the early stages and up to more than 80% in the late stages of disease—so really relevant and partly mature overall survival outcomes for analysis.
Looking into the database—this is what we presented at ASCO GU—we wanted to find out if the PSMA-PET features are associated with overall survival in the entire group, but also in the different subgroups of the prostate cancer disease spectrum. And these are the submetrics, starting with local tumor stage, T stage, going to locoregional stage, N stage, but also all the metrics for organ involvement. There's tumor volume in there, there’s number of lesions in there, and there’s also a PSMA expression score in there.
Without going into all the details—there’s way too much on this slide—this is just to highlight that most of the features, nearly all of those features, are associated with overall survival in these patients. Only very few local disease features, such as unifocal local disease, are not associated with overall survival. All the other features—so more disease on PET—means worse outcome in terms of overall survival.
And this can be used to create nomograms and to combine those with regression analysis, also selecting those criteria which are most relevantly associated with overall survival. These are two approaches: separating into a visual nomogram that everyone can apply with any imaging system, any software, and a quantitative nomogram that additionally includes tumor volume whole-body assessments.
These assign total points, and the total points are translated into risk groups—into those three: low, intermediate, and high-risk groups. And those nomograms show a quite high association with overall survival, a C-index of more than 80%, and will be further detailed in a poster talk that we also have today from Flora Fankhauser.
I just want to highlight that if we apply these nomograms to the different disease groups—both the overall cohort of more than 6,000 patients, but also the initial staging BCR group and the advanced staging groups—the nomograms can separate patients with low, intermediate, and high risk significantly. They can separate different risk groups. And these are the survival plots for the more advanced disease stages: nmCRPC, mHSPC, and mCRPC.
Again, it's the same nomogram, one visual and one quantitative, applied across the entire disease spectrum. And it also works throughout the entire disease spectrum. So the critical question now for a prognostic nomogram is, does this give additional value? Is this superior to the current clinical systems out there, which are quite simple, taking information from the clinical routine—PSA level, local disease stage, histopathology?
We wanted to find this out based on the large registry cohort, and this is a comparison between the clinical nomograms (NCCN, both for the entire cohort, as the upper row, and for the initial staging patients only, where it’s meant to be used) and the EAU risk stratification proposed for the BCR patients. And you can see, by applying the PSMA-PET nomogram in a validation cohort (so a separate cohort not involved in the creation of the nomograms), the area under the curve—so the accuracy—of PSMA-PET-based risk prediction is higher through the entire spectrum than the clinical nomogram.
So it’s always significantly higher for both the visual and the quantitative risk assessment using PSMA-PET criteria, as compared to NCCN and EAU risk stratification criteria.
So in summary, looking at large registry trials (the large PROMISE registry), we can say that PSMA-PET-based nomograms—PPP, updated PPP, two nomograms—accurately stratify prostate cancer patients into different risk groups. And we get enough information to stratify patients into low-risk, intermediate-risk, and high-risk for overall survival throughout the entire disease spectrum. These nomograms, based on PSMA-PET, yield superior accuracy when compared to clinical nomograms. Here, as an example, we show NCCN and EAU, given at different stages. And we try to continue the follow-up and collect more patients and especially more overall survival endpoints in the PROMISE registry, to be able to get a higher level of precision of risk prediction in the future.
This also needs to go out, of course, as a clinical or usable tool. And for this, we picked up the idea with Technical University to create an online app, the so-called ePROMISE app, that can be reached over this webpage. It includes standardized reporting of prostate cancer based on PSMA that can be used on any platform. It’s free, it’s academic, and it will also include—once new risk systems are published (they are currently under review)—risk stratification into low, intermediate, high risk, and also prediction of overall survival in this one application to be used and explored online.
And with this, I thank you very much for your attention. I invite every department to also contribute to this registry if you’re interested. You have the webpage here. It’s a large registry trial, and I thank all the supporters of the registry trials—that’s the only way to gather this number of patients to be analyzed in the future. And thank you very much for your attention. I’m looking forward to the next talk from Irène Buvat.