Read the Full Video Transcript

Christopher Wallis: Hello, and thank you for joining us for today's UroToday Journal Club discussion. We're talking about the PRIMARY Score using intra-prosthetic PSMA PET/CT patterns to optimize prostate cancer diagnosis. I'm Christopher Wallis, an assistant professor in the Division of Urology at the University of Toronto. With me today is Zachary Klaassen, an assistant professor in the Division of Urology at the Medical College of Georgia. This is the citation for this recent publication. The Journal of Nuclear Medicine led by Dr. Emmett and Dr. Hofman.

Prostate cancer, as most will know, is common with one in every six to eight men being diagnosed in their lifetime. An additional one in every five men wanted to go biopsy as a result of elevated PSA. And we know from a variety of studies, including precision highlighted here, that MRI triage of men with elevated PSA prior to biopsy can improve diagnostic accuracy and avoid biopsy in a substantial proportion of men.

However, both the positive predictive value and negative predictive value of MRI have some limitations and as such there's work to optimize the pre-biopsy diagnosis pathway. Once such approach use advanced prostate cancer imaging. And we know that that imaging forms a critical component of prostate cancer care through all stages of disease. Novel PET radiotracers may provide increased sensitivity and for diagnosis and detection of prostate cancer. And we've seen these initially used in the biochemical recurrence setting and followed by local staging prior to initial primary treatment. And now, we can consider earlier options. In this data here, you can see results from the pro PSMA trial in a comparison with conventional imaging, PSMA-PET/CT has substantially greater accuracy with both improved sensitivity and specificity in a randomized comparison. This is the primary report of the PRIMARY study accessing the combination of PSMA and MRI.

This study includes approximately 300 patients who have undergone MRI and have clinical suspicion of prostate cancer result of an abnormal DRE or elevated PSA, and then subsequently undergo a PSMA-PET and multiparametric MRI followed by a confirmatory prostate biopsy. And you can see that the combination improves measures including a negative predictive value from 72% with MRI, 80% with PSMA-PET to 91% when combined. So the rational for the combination of both MRI with PSMA-PET, is the PSMA intensity alone may have difficulty distinguishing clinical significant prostate cancer from other lesions in the prostate, including BPH, prostatic intraepithelial neoplasia and low grade prostate cancer. And this is in part driven because the interpretation of PSMA depends on a comparison of the tumor region to the background. In contrast, the MRI-based PI-RADS scoring system uses intra prostatic-anatomy. And so if you incorporate the two with the anatomic localization of PSMA and pattern characterization, we may get improved diagnostic accuracy. And so this is a post hoc analysis of the PRIMARY trial focused on the value of intra-prostatic PSMA-PET/CT patterns.

And as alluded to, this is a prospective, multi-center, phase II imaging trial performed at three academic centers and including patients with clinical suspicion of prostate cancer as a result of an abnormal PSA or abnormal DRE who are planned for prostate biopsy. Patients were excluded if they had a prior diagnosis of prostate cancer, previously had undergone a biopsy or MRI, or had a low-risk MRI and were not planned to undergo biopsy. So, MRI was performed and reported locally by a urology investigator, whoever the urology investigator preferred as their subspecialist prostate MRI radiologist, it was predominantly performed on three T magnets and interpreted according to PI-RADS version two. PSMA-PET was performed with a pelvic-only PSMA-PET with a low-dose, non-contrasted CT. And these images were also used for the target biopsy. The authors segmented the prostate into four regions, by quadrants and centrally reviewed all images by two reviewers.

Now, they determined these PSMA patterns, which were categorized in four levels, including pattern A with a diffuse transition zone activity, pattern B with symmetrical central zone activity, pattern C, focal transition zone activity, and pattern D, vocal peripheral zone activity. You can see here that these were used to derive a PRIMARY score, which was assessed by two blinded independent reads performed on a random sample to determine inter-rater variability of the scoring system here highlighted. And as like the PI-RADS, score ranges from one to five with increasing scores associated with an increased probability of prostate cancer or clinically significant disease. All patients underwent just systematic transperineal prostate biopsy with a minimum of 18 cores, as well as targeted biopsy for imaging based abnormalities.

The authors assess the AUC of the five level PRIMARY score and PI-RADS, and they compared these using DeLong's test. They looked at inter-rater agreement using Cohen's kappa and assessed the interaction between SUVmax and PSMA pattern type in a logistic regression model. The marginal probabilities of clinically significant prostate cancer that were plotted against SUVmax for each pattern type. Now going to hand it over to Zach to walk us through the of the study.

Zachary Klaassen: Thanks very much, Chris, for that great introduction. This is the table one, looking at the baseline characteristics for these patients. There was 291 patients, as you can see here. The median age was 64. Median PSA was 5.6 and the most common T-stage was T1c at 68%. In terms of PI-RADS on the multi-parametric MRI, we see that there was 18% PI-RADS three. 31% PI-RADS four. And 18% PI-RADS five. With regards to the grade group biopsy, most commonly was grade group two at 35%, but we also see some high risk disease at grade group five at 4.8%. In terms of the PSMA pattern, no pattern 16%, diffuse transitional or central zone at 33%, focal transition zone at 18% and peripheral zone at 53%. As Chris alluded to this was the PRIMARY score. He delineated in the methods, and we can see here on the far right, this is the percent of clinically significant prostate cancer stratified by PRIMARY score.

So we see, as you would expect the PRIMARY score one 8.5% clinically significant prostate cancer all the way up to a hundred percent for PRIMARY score five. And we see a nice gradient as we move up in the PRIMARY score with the percent of clinically significant prostate cancer. This is a nice set of images looking at the PRIMARY score and examples of what this looks like on PET/CT scan. So score one, no pattern, as you see here, no signal in the prostate. Score two, which is the diffuse transitional zone, we see some signal in the transitional zone, nothing particularly focal, again, score two, but looking at the central zone, scattered diffuse pattern here. And then moving over to PRIMARY score three, this is focal transition zone, we see a brighter SUVmax in the transition zone. Similarly, patterns four in the peripheral zone, a very discreet high SUVmax lesion in the peripheral zone of the prostate. And then score five with intense uptake, more than SUVmax of 12 with a very bright signal in the prostate as well.

This table looks at the diagnostic performance for PRIMARY score, looking at one to two versus three to five on PSMA, and then compared to the MRI PI-RADS scoring of two versus three to five. And we can see here amongst all the patients, we have a generally higher sensitivity, specificity, positive predictive value and negative predictive value for PRIMARY score on the PET/CT scan compared to the PI-RADS score. This was again evident when multiple readers were assessed and then looking specifically for the PRIMARY score PET reader one versus two, we see similar results between these two readers with a cap of PRIMARY score 0.76 and 0.64. With so good inter-reader correlation between reading these PRIMARY scores for PET/CT scans. This figure looks at the distribution of overall ISUP grade group by PRIMARY score and PI-RADS score. On the right, we see the PI-RADS score stratified by cumulative percentage on the Y axis. And on the left, we see the PRIMARY score.

And the color coding scheme is so no cancer is in gray. Grade group one is in purple. Grade group two is in light pink, Grade group three in a little bit darker pink. Grade group four, darker pink. And grade five, the darkest pink. And a couple of things to highlight here on this figure. So, if we look at the PRIMARY score, no cancer, certainly as we have low PRIMARY scores, there's more than 50% no cancer. And this decreases with a gradient up to PRIMARY score four. We also see this in the PI-RADS scoring, but to a lesser extent in terms of the overall cumulative percentages. What's interesting is if you look at the PRIMARY score five and the PI-RADS score five, we see no instances where there was no cancer with the PRIMARY score five.

And so, more than two thirds of these patients with a PRIMARY score five, had ISUP grade three to five disease. Whereas we see a little less of a tight specification for high grade disease when we're looking at the PI-RADS score five on the far right over here. This final figure looks at the predicted probability of clinically significant prostate cancer versus SUVmax, looking specifically at pattern uptake in these scans. And so, what this is essentially telling us is what we've already talked about before in the presentation, is that with a low SUVmax and no specific pattern, the predicted probability of clinically significant prostate cancer is very low. When we look at no focal pattern, but increasing SUVmax, we do see a little bit higher predicted probability for clinically significant prostate cancer, but where the real take home messages is in these very focal lesions with increasing SUVmax, as we can see here, the SUVmax greater than 12 on this study had almost a hundred percent predicted probability.

And certainly as we get higher in the SUVmax, this becomes a hundred percent predicted probability of clinically significant prostate cancer. So, several discussion points from this paper. This study use key patterns within the PRIMARY score to improve specificity without compromising sensitivity for the diagnosis of clinically significant prostate cancer. Use of pattern and intensity within a five level PRIMARY score was reproducible between readers with a higher equivalent diagnostic accuracy for detection of clinically significant prostate cancer compared to multi-parametric MRI. Importantly, the PRIMARY score uses a combination of pattern, focal versus diffuse, zonal location and high SUVmax. And it relies more on the pattern than intensity, only using a high SUVmax of greater than or equal to 12 as the highest score, which is PRIMARY score five, due to its hundred percent specificity for presence of significant malignancy. And this makes the PRIMARY score more applicable across a range of PET cameras and PSMA ligands. Finally, the inter-rater reproducibility for differentiating PRIMARY score low from high risk was substantial between independent readers.

So in conclusion, a five level PRIMARY score incorporating intra-prostatic pattern and intensity on PSMA PET/CT shows potential with high diagnostic accuracy for clinically significant prostate cancer. Of course, further validation of this scoring system in the screening population is warranted prior to its clinical implementation. We thank you very much for your attention. We hope we enjoyed the UroToday Journal Club discussing the PRIMARY score for PSMA PET/CT scan.