Read the Full Video Transcript

Daniel Spratt: So I'm hyper-simplifying the NCCN guidelines for what I'm going to talk about today—decisions of when we add short-term hormone therapy, long-term hormone therapy, and now a consideration of Abiraterone.

And the basis—what is the primary biomarker in NCCN risk group? So it's a Gleason score, and this is the Gleason system taken from the work from about 60 years ago based upon seven PCSM events. Obviously, it should not surprise any of you that a system 60 years ago does not perform very well in a contemporary randomized Phase 3 trial. Patients that have a C-index or discriminatory value of barely better than a coin flip to use Gleason, which is how we're deciding this.

So if you look at MARCAP, which is the largest independent patient data of almost all the randomized trials of radiation with systemic therapy, you can see that the use of hormone therapy here on the left versus not with radiation—the top, just to simplify it—improves MFS as well as OS. And these are the number needed to treat based on risk group of intermediate or high risk. And this is the same for prolongation from short to long-term ADT. Similarly, pretty low number needed to treat to prevent a distant met at 10 years with, again, improvements in MFS and OS.

New kid is Abiraterone based on the STAMPEDE data, as you see here. I like to highlight just to compare and contrast. Well, this is phenomenal work. These are not like any of the TROG, EORTC, or NRG trials. Twenty-five percent of patients did have PSAs, and this is based on the selection criteria that Dr. Gleave already went over, and there were patients with PSAs into the thousands, but these are M-zero patients.

So it's actually interesting because I have Nick James and Vedang Murthy sitting next to each other. So I'm overlapping here. This is not meant to be; this is really for illustration. The POP-RT trial, which is a pretty high-risk contemporary trial—almost all were PSMA-PET staged to rule out N or M disease—versus the STAMPEDE's control arm. So these are both trials that got radiation with long-term hormone therapy, no Abiraterone. And the difference here is pretty pronounced. There's only about a 5% biochemical recurrence at five years in the POP-RT trial versus almost a 15% death from prostate cancer at the same time point in STAMPEDE. These are not the same patients.

So let's get into how do we personalize these decisions with contemporary tools. So the use of hormone therapy—and I won't be going into full detail—I would recommend you actually read a pretty expanded section we've now added to the NCCN guidelines that talks about genomic classifiers. There are three. The primary one that is used is the Decipher 22-gene. It has Simon Level evidence 1 versus the others which don't have this type of evidence. And it has a section now that actually goes into detail if you have Level 1 evidence to the trial data to support it.

But really, this Level 1 data comes from about 11 Phase 3 randomized trials. Some were pre-specified, some are post-hoc, shown here. And what's relevant in this use of hormone therapy setting is these are two studies. One on the left is a registry from Princess Margaret. The right is a randomized Phase 3 trial. We just reported both of these were intermediate-risk patients, mostly unfavorable intermediate risk, got radiation alone, did not get hormone therapy. And the Decipher low patients, as you see on the top, almost none developed distant mets long-term; Decipher high almost 20%. Obviously, ADT cannot provide probably a large benefit in the Decipher low patients.

Shown at the bottom left here, you can see here, similar to what I showed you earlier, the AUC for just the NCCN favorable versus unfavorable has 0.54—again, barely better than a coin flip. The clinical genomic risk groups with Decipher 0.89. So very good discrimination. And the right is just showing you that Gleason score. Again, fairly poor at really identifying; there's a massive on the Y-axis. These are the Decipher GC scores. A lot of heterogeneity here.

So the newest, which I've been very fortunate to be a part of using the MMAI—multimodal artificial intelligence—which is based on digital pathology features, now also has Level 1-B by Simon criteria for not just prognostic but predictive. It's the first predictive test in our guidelines and non, I guess, mCRPC setting. And again, this is here for your reference. I will not read all of this or I'll get the cowbell, but this is the study we just published last year that trained and validated the predictive biomarker that I'll show you the results here.

So it was trained off of four Phase 3 NRG RTOG randomized trials, about 4,000 patients, to train for a differential benefit of the use of ADT. It was locked. And then my statistician independently validated this work in RTOG 9408. It's the largest Phase 3 trial that has tested the use versus no use of hormone therapy in over 1,600 patients. And of course, to show it's predictive, you need to demonstrate a biomarker treatment interaction, and that's exactly what we showed. It was trained for distant metastasis as the endpoint, and you can see the difference here where you have a sub-distribution hazard ratio if you're biomarker positive of 0.34. So a very large relative and absolute benefit. If you're biomarker negative, which is two-thirds of these patients, the hazard ratio is 0.92.

So getting into now, what about prolongation of ADT, which is the standard of care in our guidelines for high-risk prostate cancer. Again, using the 22-gene genomic classifier, there's some pretty promising data here to help personalize treatment decisions. This is data from the NRG group led by Dr. Paul Nguyen that you can see in these high-risk trials. And these are what I would consider very high-risk trials from the nineties. A lot of DRE T3 disease, some of them up to 75% high PSAs. You can see here that the Decipher low patients, very few of them are developing distant mets. If you look at that Y-axis closely, that's out to 20 years of follow-up. Decipher high is approaching 50%.

And so how can you use this? Well, we talk about shared decision-making. Well, in Decipher high patients, the number needed to treat to reduce one distant met at 10 years of short-term versus long-term ADT is just nine. So that's pretty favorable. If this was Decipher low patients, that's 33. So based on patient's age, comorbidity status, preferences, this is pretty meaningful.

And now Dr. Armstrong's in the audience, he presented this at ASCO. The paper is written, and this will be submitted. This is not yet commercially available, but very similarly, a different novel biomarker has been trained and validated in RTOG 9202. Again, a randomized trial of short versus long-term ADT showing we can identify about a third of high-risk men that derive absolutely no signal of benefit. On the left and on the right, a pretty pronounced benefit of using long-term ADT.

So of course, there is data that is out. This is pre-print data thanks to Dr. Attard and the STAMPEDE group. But there have been many trials actually in the metastatic setting, which I'm not going to be talking about using these tools, but he reported into. The pre-print is online of that STAMPEDE sub-study of use of Abiraterone. I'll bring your attention to the right-hand side. So they ran the 22-gene classifier test in this cohort, and you can see here that below the median score, which is around 0.8 for those familiar with it, there is absolutely no benefit of using Abiraterone in this population. So not only is this, again, an ultra, ultra-high risk, half the patients if you just do a simple cut at the median, are not benefiting from years of this therapy.

There's ongoing biomarker-driven trials. These were accruing very rapidly. O-10 is for unfavorable intermediate, O-9 on the right is for high risk. They're both primarily focusing on de-intensification—so shortening the duration or omission of ADT because many of these patients, as I mentioned like from POP-RT, are doing very well—and then a subset regarding intensification, usually with ARPI addition.

The point of all of this, and I think that patient's talk at the very beginning, the opening, was fantastic, is why does all this matter? Well, in modern high-risk with PSMA-PET, which is a biomarker, if you screen out the node-positive and M-positive disease, patients do extremely well. And that's why many trials are focused on de-escalation.

ADT prolongation has been shown in randomized trials to cause two to three-fold increase in cardiocerebrovascular disease. We're putting out and will be submitting soon as well that prolongation increases other cause mortality for every month you prolong ADT. And then this was presented at GU-ASCO, and this is in the non-metastatic—this is the metastatic setting of all the Abiraterone trials that were reported by older patients. There's really not a signal of benefit. So we need to just be careful in this M-zero setting. And I will await very much so the data from ATLAS, ENZARAD, and DASL-HiCaP before I routinely use these agents in this setting.

Thank you so much. Appreciate it.