PSMA PET CT Derived Risk-Stratification Tool for High-risk Prostate Cancer - a UroToday Journal Club - Christopher Wallis & Zachary Klaassen

April 30, 2022

Christopher Wallis and Zachary Klaassen discuss an article published in JAMA Network Open titled "Performance of a Prostate-Specific Membrane Antigen Positron Emission Tomography/Computed Tomography-Derived Risk-Stratification Tool for High-risk and Very High-risk Prostate Cancer." This study aimed to evaluate the prognostic significance of a nomogram that models an individual's risk of nonlocalized upstaging on PSMA PET/CT and to compare its performance with existing risk-stratification tools.


Christopher J.D. Wallis, MD, Ph.D., Assistant Professor in the Division of Urology at the University of Toronto.

Zachary Klaassen, MD, MSc, Urologic Oncologist, Assistant Professor Surgery/Urology at the Medical College of Georgia at Augusta University, Georgia Cancer Center

Read the Full Video Transcript

Christopher Wallis: Hello and thank you for joining us for this UroToday Journal Club discussion. We're discussing here an interesting paper on the role of PSMA, entitled Performance of PSMA PET/ CT Derived Risk-Stratification Tool for High-risk and Very High-risk Prostate Cancer. I'm Chris Wallis, an assistant professor in the Division of Urology at the University of Toronto. Joining me today is Zach Klaassen, an assistant professor in the Division of Urology at the Medical College of Georgia. This is a citation for this recent publication, published in a JAMA Network Open and led by Dr. Kishan.

Prostate cancer in general, and even within the subset of high-risk and very high-risk disease, is a heterogeneous condition with a variety of clinic trajectories. A metastatic failure in patients with high-risk and very high-risk disease is a more common outcome than local failure, suggesting that initial occult metastatic disease at presentation may underline or underpin the eventual disease trajectory and natural history. And while we have NCCN guidelines regarding the investigation and imaging of these patients, an inability to accurately stage and detect metastatic disease may affect our long term treatment outcomes. As you can see here, bone imaging and pelvic imaging are recommended in high and very high-risk disease.

However, we do see that novel treatment approaches may be useful in improving our imaging approaches. And so novel PET radiotracers have increased sensitivity, both in the detection and diagnosis of metastatic disease and in other settings, including biochemical recurrence. While there is a wealth of literature demonstrating the ability of these radiotracers to identify disease, it is less clear how this affects clinically meaningful outcomes for patients.

And so the UCLA group developed a PSMA risk calculator as a nomogram to calculate the probability of non-localized disease on a PSMA scan for patients who have clinically non-metastatic disease based on their conventional imaging. In the present study, the authors sought to assess the significance of this nomogram on clinically meaningful endpoints.

And so to do this, they used three different cohorts of patients. First, they assembled a multi-institutional cohort from 15 different tertiary referral centers of patients diagnosed with prostate cancer between 1995 and 2018. To be included, patients had to have high or very high-risk disease, non-metastatic disease based on conventional imaging and have under undergone curative intent radical prostatectomy or radiotherapy.

Additionally, they assembled two large database cohorts, including a SEER cohort of patients with high and very high-risk prostate cancer diagnosed and treated between 2010 and 2016. And here, because surgical nodal staging supersedes clinical nodal staging for surgically-treated patients, patients undergoing radical prostatectomy were excluded from this cohort. In contrast, in the NCDB cohort, again the authors accrued patients with high and very high-risk disease treated over the same interval, but included all treatment modalities as clinical staging could be utilized here.

So the authors performed prognostication based on the baseline characteristics, including age, PSA level, clinical T stage, Gleason score, number of positive and total number of involved biopsy scores and treatment type. And they identified a number of important outcomes, including prostate cancer, specific mortality, and overall survival, as well as biochemical recurrence and distant metastases only in their institutional cohorts. These are not available at the database level.

The nomogram was originally derived based on 262 men with non-metastatic disease based on conventional imaging who were enrolled in two prospective clinical trials. The nomogram was derived to estimate the probability of any non-localized disease using a logistic regression model, which included initial PSA, biopsy Gleason grade group, percentage of positive cores involved and clinical T stage. In the initial derivation and validation, the accuracy of this nomogram showed an AUC of 0.75. The authors also used clinical risk stratification tools, including CAPRA groups, STAR-CAP stage groups, and the MSKCC pre-prostatectomy nomogram as clinical comparisons to the imaging-based nomogram.

In terms of statistical analyses, using the multi-institutional cohort as alluded to previously, the authors assessed biochemical recurrence defined in a standard fashion, just in metastasis and prostate cancer-specific mortality, whereas in SEER/NCDB, the authors extracted vital status information. The model performance was evaluated using time dependent ROC curves, as well as the concordance index or the C-index.

The association between nomogram risk and clinical endpoints was modeled continuously using Cox or Fine-Gray regressions models as appropriate with age adjustment. The authors identified PSMA nomogram cut-offs using a stepwise method with the endpoint of eight-year distant metastasis. These cut points were used to divide the group into four nomogram-derived risk groups and the authors then assessed time to event outcomes between these four groups using Gray's test and the log-rank test.

Now I'm going to hand it over to Zach to walk us through the results of this study.

Zachary Klaassen: Thanks so much, Chris. This is the baseline patient characteristics, table one, for the multi-institutional cohort, NCDB and the SEER cohort. As you can see up top, these are listed across the X axis, and you can see the NCDB cohort included 88,909 patients; the SEER cohort, almost 24,000 patients; and the multi-institutional cohort, just over 5,200 patients.

For medium follow-up, it was slightly longer for the multi-institutional cohort than the registry cohorts, at 5.1 years. We see that in terms of median age, the SEER cohort was slightly older at 71 years versus the mid sixties or the other cohorts. In terms of Gleason grade group for all three cohorts, Gleason grade group four was the most common, at almost 50%. In terms of clinical T category, we see that the most common across all three was T1, at roughly 50%. In terms of median, PSA was just over 10 for the multi-institutional cohort and NCDB cohort, whereas upwards of 13.8 for the SEER cohort. In terms of median, positive cores was 50% for all three cohorts. In terms of primary therapy, in the multi-institutional cohort, surgery was 55%, external beam radiation therapy was 32% and external beam radiation therapy plus brachytherapy was 14%. As Chris mentioned, the SEER cohort just included radiation patients and the NCDB cohort had 42% that underwent surgery and 39% that underwent external beam radiotherapy.

In terms of PSMA upstage risk, we see that for any non-localized disease was 16.7 and then for regional nodal was 15.0 and for distant metastatic was 5.1 in the multi-institutional cohort, which was comparable to the registry cohorts. The PSMA nomogram risk group, this is important for understanding some of the future figures we'll discuss. So group one was described as less than 14% upstage risk, was 42% in the multi institutional cohort, and roughly 40% in the registries. Group two, which was 14% to 27% upstage risk, was 26% in the multi-institutional cohort and the same in the SEER and NCDB cohorts. Group three was 27% to 41% upstage risk, which was 16% across all cohorts. And in group four, which was greater than 41% upstage risk, was 16 in the institutional cohort and NCDB cohort and 20 in the SEER cohort.

This is the prognostic significance of the PSMA nomogram in the multi-institutional cohort. And this is looking at concordance indices across several outcomes. We can see that the eight year C-indices for biochemical recurrence with 0.63, for distant metastases was 0.69, and for prostate cancer specific mortality, 0.71, and overall survival is 0.60.

These are the survival curves: looking at the multi-institutional cohort, looking at biochemical recurrence-free survival, distant metastasis survival, prostate cancer-specific mortality, and overall survival. This is stratified by these four groups that we just discussed on the table one. You can see that there's nice separation of these curves across essentially all of these outcomes with patients in group one having better outcomes than patients in group two, three and four.

This is the concordance indices looking at the registry-based cohort and this is looking at five-year C-indices: for prostate cancer-specific mortality in the SEER database was 0.71; overall survival in the SEER database, 0.61; and overall survival in the NCDB was 0.62. Again, looking at the survival curves by the registry-based cohorts, we have the four groups which we've discussed as well, as well as two additional metrics looking at clinical N1M0, as well as M1. And again, across prostate cancer-specific mortality, as well as overall survival, we see a nice separation based on these PSMA upstaging risks, as well as clinical N1M0 and M1 on its own.

This is the performance of the PSMA nomogram up against the STAR-CAP stage group, MSKCC nomogram and the CAPRA risk group. And this is looking at biochemical recurrence and distant metastases. For our biochemical recurrence on the left, we see that the PSMA nomogram did very well compared to these other metrics, with being the most accurate upwards of 10 years. And looking on the right, the distant metastasis concordance, again, PSMA nomogram doing very well, with the STAR-CAP stage group being the next most accurate, below PSMA nomogram.

This very similar curve is looking at prostate cancer-specific mortality and overall survival. We see in the prostate cancer-specific mortality concordance, essentially very similar, if not exactly the same, concordance indices for PSMA nomogram in the STAR-CAP stage group. And again, we see the PSMA nomogram, especially when we get out to 10 years for overall survival, performing very well compared to the STAR-CAP, MSKCC and CAPRA risk groups.

This looks at the performance of the PSMA nomogram, the STAR-CAP risk groups, CAPRA and MSKCC assessed by concordance indices. And this is specifically looking at prostate cancer-specific mortality. On the left here, we see that the five-year indices for PSMA was 0.71, was 0.72 for STAR-CAP, down to 0.69 for MSKCC, and 0.57 for the CAPRA risk group. So specifically for this prostate cancer-specific mortality, STAR-CAP was designed to predict this and so it's not surprising that we see STAR-CAP doing well for PCSM and also reflective of PSMA nomogram doing very well.

There are several important discussion points from this study. First, direct prospective evidence between PSMA PET/CT and clinically important endpoints is still at least several years away. And until such data is available, this PSMA nomogram serves as a proxy for long-term prognostic and clinical significance of PSMA PET/CT findings. The external validity of these results is supported by the inclusion of 15 institutions in the primary cohort and by the external validation in the registry-based cohorts of SEER and NCDB. The PSMA nomogram showed improved risk stratification, outperforming all the other models for all endpoints except prostate cancer-specific mortality, for which it had similar performance as STAR-CAP. More precise risk stratification may enable more personalized treatment. As we've seen in the proPSMA trial, PSMA PET/CT led to more changes in management compared with conventional imaging, at 28% versus 15%.

So in conclusion, the estimated probability of non-localized upstaging on PSMA PET/CT was significantly prognostic of long-term clinically meaning endpoints, and the nomogram performed favorably versus existing risk-stratification tools. This was true despite PSMA nomogram being trained purely on radiographic findings at initial diagnosis, while the comparison models were specifically designed to predict clinical endpoints. Previously occult, PSMA PET/CT-detected disease may be the main driver of outcomes in high-risk patients. And finally, prospective investigation centered on making use of PSMA nomogram as part of risk-adapted treatment strategy is warranted. Thank you very much. And we hope you enjoyed this UroToday Journal Club discussion.