Matthew Cooperberg: Good morning. I'm Matt Cooperberg from the University of California in San Francisco. It is a pleasure to bring you another installment of the UroToday Localized Prostate Cancer Center of Excellence video series. It is my pleasure today to be joined by Nikita Sushentsev, who recently completed his PhD at Cambridge and now is a resident in radiology at Cambridge University and has recently had a fascinating publication from the ProtecT trial looking at this issue of cribriform versus non-cribriform histology, which has been the subject of a lot of really important research in recent years and I think will really help us understand the results of this trial and their implications for treatment decision-making. So Nikita, welcome. Thanks for joining us.

Nikita Sushentsev: Thank you for having me, Matt. Pleasure to be here.

Matthew Cooperberg: So tell us a little bit about the study. How did this get conceived and maybe give us a quick summary of the overall results?

Nikita Sushentsev: Right. So if I could just share a couple of slides that I have prepared, this will give you a good idea of what we have been up to. So just a few thoughts on the background and a brief study overview, main outcomes and implications for future research, which hopefully will help us segue into the discussion of this work. So just I'd like to start with some definitions. So as you've mentioned, this is a subject of pretty important research and the cribriform-negative, cribriform-positive disease. So what's that about? So cribriform-negative disease generally includes Gleason pattern 3, so all grade group one tumors by default will be cribriform-negative. Gleason pattern 4, poorly formed glands, fused glands and glomeruloid structures to name a few. And then Gleason pattern 5 as well could also be cribriform-negative, cribriform-positive based on its morphology.

And the cribriform-negative pattern 5 would be comprised of single cells, cords, and small nests with intervening stroma. And then cribriform-positive disease can be grouped in Gleason pattern 4. We all know invasive cribriform carcinoma, intraductal carcinoma, these are the most common variants of this disease. And Gleason pattern 5, so medium to large solid fields, which are cribriform in their appearance without intervening stroma with or without comedonecrosis have also been classed as cribriform-positive. So what's the diagnostic relevance of these two broad subtypes? So it's important to stress that both ISUP and GUPS recommend reporting binary cribriform status in all needle and surgical biopsy specimens. And it's important to state that it is already in the guidelines and what needs to be reported is the binary status. So is there any amount of cribriform disease present in the specimen or not? It can be intraductal, invasive, all these types mentioned on this slide. And what is important is the binary status.

That's really key. And how is this used in practice? So EAU specifically advises against offering active surveillance to patients with any amount of cribriform disease detected at diagnostic biopsy. So this is something that's already in the guidelines and something that can be actioned today. In terms of research questions behind this study is so, first of all, the evidence from EAU and indeed from other guidelines which are emerging has actually not been based on any prospective studies, let alone trials that included patients with active surveillance or active monitoring. So we don't really know whether having cribriform disease confers adverse outcomes in this specific patient. So leveraging the benefits of ProtecT, which I'm going to talk about in a few seconds, first of all, we wanted to compare long-term oncological outcomes of patients with cribriform-positive and cribriform-negative disease that underwent active monitoring or early radical treatment in a randomized setting. Yes, it's a secondary analysis.

Cribriform morphology was not a randomization characteristic, but still ProtecT offers a really important insight in the long-term trial. And then the second question was an attempt to resolve the central paradox of ProtecT. So ProtecT, just so that I remind the audience, was a long-term trial with a 15-year follow-up. And this year we've just hit the census date for the 20-year follow-up, which hopefully will be out in a couple of years time, which randomized patients who were initially screened with PSA and then after undergoing biopsy were found to have clinically localized prostate cancer and randomized into three groups, active monitoring, surgery, or radiotherapy with neoadjuvant ADT that lasted between three to six months. And basically at 15-year follow-up, what happened is that prostate cancer-specific mortality was extremely low over 3% across all those three groups, so it didn't differ. So whether you treated patients radically or followed them up to begin with, it didn't really seem to prevent deaths in this cohort. However, early radical treatment did reduce the risk of metastases in these patients. So for us, it was really clear that there is a high-risk group hidden in the ProtecT cohort.

So understanding what kind of histological features, particularly those that are already in the guidelines, could identify that high-risk group. So that's why we bet on cribriform in this particular analysis. So in terms of the study overview, the centralization of materials in ProtecT is still ongoing. So out of 1,643 participants who were randomized in the trial, we found biopsy slides from 712 patients. So this represented all the specimens from these patients, so we were sure that we were not missing anything. And out of those, you can see the breakdown of how many patients were assigned to different study groups. And at the bottom of it, you can see that only roughly 13% of the cohort we have reviewed harbored cribriform-positive disease on biopsy, so 87% didn't have it. The primary outcome that we looked at in this particular analysis was metastases at 15-year median follow-up. If we're talking about randomization, so whether it was preserved in this particular analysis because we have not reviewed the whole cohort, there was no statistically significant difference in randomization characteristics such as age, PSA, revised Gleason score across the three groups. And why I say revised Gleason score is because the original ProtecT grading was based on 2005 ISUP guidelines. So in this review, we applied the 2019 guidelines.

And even having done that uniformity to all patients, it didn't really make a difference. And the really key thing is that here we did two types of analysis, the intention-to-treat analysis based on the original randomization and the per-protocol analysis, which accounted for the crossover between the two different groups, so the actual treatment that these patients underwent. And if we look at this, well, there is some trend to the fact that patients who underwent radiotherapy with neoadjuvant ADT had higher prevalence of cribriform disease. Well, this is obviously something that happened by chance because cribriform disease, again, was not a randomization characteristic then. So if we look at outcomes of patients with cribriform-negative disease, and this comprises 87% of this ProtecT cohort, we can see that both in the intention-to-treat and the per-protocol analysis early radical treatment did not seem to result in statistically significant reduction of 15-year metastases compared to active monitoring. And what's absolutely key here is that if we look at this multivariable Cox analysis in the intention-to-treat setting, which would be the same in the per-protocol group, patients with grade group two disease who were cribriform-negative had the same risk of long-term metastases as patients with Gleason score 6 or grade group one disease. And I think that's one of the really key, if not the key, finding of this particular review.

And then if we look at patients with cribriform-positive disease, again, the same intention-to-treat, per-protocol analyses. You can see in the intention-to-treat analyses, particularly at a kind of middle-of-the-road outcome points, you can see that ProtecT was more impactful, but even then, the really key takeaway here is that by 15-year follow-up, there was actually no significant difference in metastasis-free survival between the two different groups. And when you make an adjustment per protocol, again, the risk reduction is preserved. However, you look at the absolute risk reduction, so you take a 10-year time point, for example, the cumulative incidence rate in the active monitoring group was 34%, in the radical treatment group it was 22%, so absolute risk reduction was only 12%. And if this cohort represented, for example, 13% of patients overall, then the likely benefit is actually quite incremental. So that's what we had kind of taken from this particular study. And just overall, even for patients with grade group two disease, so cribriform-positive disease, so that's grade two again, the outcomes seemed absolutely atrocious. 36% metastases rate at 15 years in the active monitoring group, compared to 5% in the radical treatment group. So this is a huge difference. But again, I would say even then, by 15-year mark, there is a catch-up in terms of outcomes regardless of which treatment you choose.

I think that's the most recent publication by Paul Abel, which I'm going to mention briefly, highlights that actually you treat or you don't treat patients early, but then by 20 years, you probably still need to treat them. And I think that's the idea here. So if we looked at the outcomes overall across the three groups, so active monitoring, surgery, radiotherapy in this particular cohort. What we can see is, unsurprisingly, patients who were cribriform-negative had the same overall survival across the three groups, kind of in line with the idea that their oncological outcomes in terms of metastases were the same, and that's just another measure of the efficacy of different interventions. However, for cribriform-positive disease, even though radiotherapy, surgery and active monitoring were more or less the same within the ProtecT cohort that we have analyzed, interestingly, if we compared cribriform-positive with cribriform-negative disease, the cribriform-positive disease, even if you had a radical intervention early, still had worse overall survival compared to cribriform-negative disease. So even if you treat these patients, they're more likely to die from something else.

I mean, you've mitigated the risk of metastases if you've seen it early, but they're so much more likely to die from other causes. And I think probably the mechanisms are more likely to do with the metabolic syndrome, perhaps shared risk factors, perhaps just comorbidities being diagnosed much earlier, presenting much earlier in patients who have cribriform-positive morphology. And there has been one study which actually correlated it and maybe this metabolic syndrome phenotype and cribriform disease. So it's probably worth following up on. So just to conclude the summary of the analysis, patients with cribriform-negative grade group two disease have similar long-term oncological outcomes compared to grade group one disease patients, even if they undergo active monitoring. Presence of cribriform morphology in biopsy tissue identifies a subset of patients who benefit from early radical therapy, but notably, if you look at the 15-year results, the benefit is not actually sustained, which is again in line with that recent ProtecT publication.

What is also seen is that even with early radical treatment, patients with cribriform-positive morphology appear to have worse overall survival with more deaths from other causes, which is interesting. And I think the final point is active surveillance, which obviously has much more stringent enrollment criteria, is not the same as active monitoring. So this finding shouldn't be extrapolated directly to the active surveillance population. So one thing that I probably haven't said before is that approximately 50% of patients in the ProtecT cohort had grade group two disease, whereas approximately in the active surveillance cohorts, the equivalent figure is approximately 15%. And one common way patients got into the active surveillance cohorts in addition to being grade group two is that they must have low volume pattern 4. That's not necessarily the case in the ProtecT cohort. I think grade group two had like 40% median Gleason pattern 4 in those patients.

So that's why we really need to be careful. And there are a couple of implications for future research, obviously, ongoing quest to try and understand which patients with biopsy Gleason 3+4 without cribriform disease are likely to benefit from active surveillance. It is 87% of the ProtecT cohort I'm hoping would help with that. And obviously the radical prostatectomy specimen pathology review for patients who underwent surgery would allow us to understand the extent of undersampling or indeed which pathology features do we find in the final lesion compared to the biopsy. And then 20-year outcome is another step. So I'll probably stop here and happy to answer questions.

Matthew Cooperberg: Just amazing, detailed work. I'm really impressed with how much data was pulled from this. A couple of questions on the pathology details. So there was a big study out last year on favorable versus unfavorable histology and histopathology. This is Jen Wynn and Jess McKinney. And there've been a few papers out, but I think this one really crystallized this difference between large or expansive cribriform and simple or small cribriform. I know you did not make that distinction here, but do you have any sense of, along those lines, were most of these cases in the cribriform category actually large?

Nikita Sushentsev: Well, that's a great question. I think several points are worth making. So on the one hand, yes, the definition of unfavorable histology that includes specifically large cribriform disease, I think it's pretty compelling. It's not yet in the guidelines because the definitions have not been standardized. And we have discussed that previously. Okay, you can take 0.25 millimeter as a cutoff, somebody will take 12 lumens as their cutoff. We're probably talking about the same thing and I guess the rationale here is clear, but we deliberately almost didn't look into this in this paper because I mentioned in the overview binary cribriform status is something that's already in the guidelines, is something that pathologists should be reporting. So we stuck with that in the first instance.

We, however, when we were doing the analysis, we applied these definitions to some of them, so we looked at whether they were large or small. And that didn't really seem to make much of a difference. I would say, again, how many large cribriform glands do we need to have to classify them as large? Is one enough? Is two enough? Well, what if 99% are small and only one is large? It's a bit hard. But yes, I mean, there was certainly a trend that patients with larger cribriform disease tend to do worse. And actually, patients with larger cribriform glands also had other non-cribriform patterns present in their specimens. And that's something maybe we can touch upon later, talking about particularly patients with non-cribriform grade group three disease and why some of them experienced those poor outcomes in this study.

Matthew Cooperberg: That's exactly my next question, actually. So the four plus threes, the grade group threes with non-cribriform did not do well. And when I think about those, the Canary paper, most four plus threes you typically think of as having one of the adverse histologies, cribriform or intraductal or something similar, and that analysis would say the glomerular fused gland subtype four plus three, if you happen to have that, should not be that bad. So what did you see in those four plus threes? Because they were, if anything, worse than the cribriform grade group twos.

Nikita Sushentsev: Yeah, they were. The first thing is that there was a substantial ... Well, because the size of, for example, patients who had the size of the cohort with grade group one disease is so big that actually if you look at the number of patients who had metastasis with having grade group one disease at systematic biopsy, it's pretty big. And this tells you about sampling error, which probably occurs. So even in these patients who had grade group three disease and no cribriform-positive glands in their specimens, well, they could have had them or they could have had some other adverse pathology that we just haven't sampled. 

I looked at this specifically when I was preparing for this and having reviewed this four plus three with non-cribriform, the interesting thing we noticed is that these seem to have had multiple non-cribriform phenotypes present. So for example, they could have poorly formed glands with fused glands intermixed together or with some glomeruloid structures, which could be simple or complex. Doesn't matter. However, if you look at most grade group two cases or even grade group three non-cribriform cases which did do well, they would be almost monoclonal or monophenotypical. So they would have some poorly formed glands and that's it. They wouldn't also have fused glands or some glomeruloid stuff mixed in.

So it's this heterogeneity of morphologies perhaps can point you to the fact that maybe there's something else lurking in the background. Maybe that tumor has evolved to be comprised of different phenotypes and that maybe increases the risk of cribriform being present there as well or maybe pattern five or something else. Just we haven't sampled it. So in short, I would say probably this could be due to the sampling error and the radical prostatectomy review, which would be similar to that paper that you've mentioned, would help us understand the real breadth of what's in those lesions.

Matthew Cooperberg: Yeah, the sampling, the 2023 ProtecT paper included in the supplement, the patients that went to surgery, the undersampling is remarkable.

Nikita Sushentsev: It is remarkable.

Matthew Cooperberg: These grade group ones who had T3B disease, who had grade group four on final pathology. These were 10-core non-image-guided biopsies in some rural centers for some of the ProtecT sites. And it really has highlighted the point. Given how narrowly those curves split, it tells us something about just how slow moving even some of the more aggressive ones may be. But the flip side is you're seeing some of these patients with metastasis at two and three years after enrollment. Histology aside, this is clearly a patient who's been pretty substantially undersampled, whatever arm they're in.

Nikita Sushentsev: Absolutely.

Matthew Cooperberg: Somebody fated to have a met at two years, it doesn't matter whether we do radiation or surgery or anything, that cat is well out of the bag. So these are all scanned in now. Any plans? I know there are some plans underway to start looking at biomarkers and look at the RNA tools in ProtecT. Any plans to let these loose with some of the PATH AI initiatives that are out there, either Panda or some of the companies developing tools?

Nikita Sushentsev: Yeah. First things first, we have to get the ethics sorted to enable this kind of analysis to happen. So hopefully, while there've been some very positive developments on that front recently, this coincides with the ongoing attempt to centralize everything, perhaps recut slides, regrade them properly. And once all of these pieces of the puzzle come through, I don't see why all of this work couldn't be undertaken. In fact, with a cohort that's so rich in terms of follow-up, in terms of the randomized setting, it would be a shame not to deploy all of these different tools to try and identify, well, what are the markers that can help?

Matthew Cooperberg: Completely agree. Hopefully that's a topic we can stay tuned for and bring you back in a couple years with more results.

Nikita Sushentsev: Yeah, exactly. Well, 20-year results should be definitely a point for conversation. One thing that I maybe would like to highlight, and we mentioned briefly in the sampling error we know from several studies and even our own practice that even in the current era of MRI-targeted biopsies, there's miss of up to 50% or more of patients with cribriform disease, so we still haven't nailed that down. So of course we have seen randomized studies, particularly in radiology, where they've shown that MRI-targeted biopsies detect more patients with cribriform disease than the systematic biopsies. But again, the numbers are with sensitivity in the order of 50%. It's still not good enough, so we need to do better.

Matthew Cooperberg: No, this is critical work and I think it's really pushing us very much toward answering this question of which prostate cancers really matter, which ones are actually cancer and in which we can watch from a much healthier distance. Listen, thanks so much for joining us. This is really terrific work. Great space for us to continue to follow and hope to have you back in the future. Thank you.

Nikita Sushentsev: Pleasure. Thank you. Thank you, Matt.