Read the Full Video Transcript

Gerhardt Attard: I draw your attention to the bottom of my, I guess, non-financial disclosures that are relevant to this talk. So I split my talk into two: treatments and biomarkers. What does the future bring? And I’ve got 10 minutes to cover the future. Well, the future certainly doesn’t bring immune checkpoint inhibition for unselected patients.

I hope we’ll develop a way to use the immune system to treat our patients, but clearly, there’s no scope for unselected patients. So looking through clinicaltrials.gov, I identified a number of randomized pharma-sponsored trials. Pleasingly, we’re seeing patient selection for the majority of them—and biologically based patient selection.

So AKT inhibition for patients with a PTEN deficiency as ascertained by centralized IHC score, PARP inhibitors for patients with deleterious alterations in BRCA2 and a few other genes involved in homologous recombination repair, Lu617 for PSMA-positive disease. And then I guess what—I emphasize these because the estimated primary completion date is coming up.

They all started randomizing in the past four to five years. So expect in the next APC we’re going to have data for these trials. One thing I’ll emphasize, however—the future is not here yet. I’m not going to disclose the voting already, but this was one of the questions we were asked, and I believe strongly we should not pre-empt the results of trials.

We were asked whether we would introduce a PARP inhibitor for a patient who’s found to have a BRCA alteration. I don’t think we have that data yet. I guess more to discuss later. So now, I’ll talk to three challenges—considerations that we’ll come across or be faced with as these trials report. And the first is the primary outcome measure.

All the trials I listed have RPFS as the primary outcome measure. Overall survival is a secondary endpoint with a time frame for follow-up of about 48 to 80 months. These trials have not been pre-powered to detect a significant benefit for OS. I think that could come with some challenges that we need to think about—some considerations.

So first, Susan Halabi’s recent work showing that RPFS is a reasonably good surrogate of overall survival for patients starting ADT or ADT–docetaxel. So this is upwards of 6,000 patients. But I’ll draw your attention to the recruitment period for these trials, which ended over 10 years ago. About 30% had ADT plus docetaxel.

And interestingly, in all these trials, RPFS was defined by the investigator, did not require confirmatory lesions or additional lesions as required by PCWG3. So as you’ll see in the graph, the y-axis is the five-year overall survival rate and the x-axis is the three-year RPFS rate, and the circle size refers to sample size.

So some considerations. mHSPC therapeutic trials are effectively a test of early versus late, at least for the effective therapies in that list—PARP inhibitors, Lu617. So showing an RPFS benefit doesn’t necessarily mean using that treatment earlier is going to improve survival. What we would expect it to do is increase toxicity, increase exposure.

So we need to be careful about that. The second is the surrogacy data has been established in ADT, ADT–docetaxel, and we need to be cautious about extrapolating that to drugs with a different biological mode of action. Now this, of course, is the chicken and the egg. We want to use surrogates to show efficacy for new drugs, and those surrogates have been developed for old drugs.

But what we can certainly do—and I know this is ongoing, and Chris already intimated this—is surrogacy analysis using ADT and ARSIs. And we have the data, or stopgap has the data, from academic studies. But we really need our pharma partners—especially a call out to those in the room—to increase access to the pharma trial data.

And I think we really need that to understand—I guess I didn’t point out, but it’s been mentioned so many times over the past day and a half—the backbone for all those trials was ADT plus ARSI. And that analysis was done in patients, none of whom received ADT–ARSI. So the second challenge is patient selection.

This is data we published last year: this is failure-free survival for patients in STAMPEDE randomized ADT versus ADT–abiraterone. Clearly abiraterone is effective; all ARSIs are. But there’s a notable proportion of patients who do not progress. When we closed the trial about a year and a half ago, a quarter of patients continued abiraterone with no evidence of progression.

So those patients do not want to go on our trials, and we do not want to treat them in our trials. How do we identify them, and how do we interpret that data? I don’t have an immediate answer. I’ll come to that later. There are a number of prickly questions that the PEACE program is looking to identify, including should we use darolutamide in vulnerable patients? Should we treat oligometastatic disease with SABR? Do we de-escalate veterans? Already talked about this. And that’s especially pertinent given the previous slide.

One thing I’ll add to that information is that whereas on ADT, over 95% of low-volume metastatic patients die from prostate cancer, on ADT–abiraterone, it’s about 70%. So 30% of patients, at least in the low-volume group, are dying of a competing cause; in the high-volume group, it’s about 15%. And the last point that Kareem referred to earlier was how do we treat—or can we intensify treatment—using PSA?

So the third challenge: treatment sequencing or overlay. So this cartoon—we use triplet or quadruplet therapy. Nick talked about triplet yesterday. I won’t go into that. But clearly, if all the trials on that first slide are positive, we’re going to start to face the challenge of a year, a year and a half of aggressive treatment with further combinations going forward. We’re trying to—or we will be—addressing some of these challenges in STAMPEDE2, which has started.

This borrows a lot from STAMPEDE1. It’s a multi-arm design but restricted to metastatic patients, primarily because the high-risk non-metastatic now have such good outcomes. The prostate cancer–specific death rate for those M0 patients we discussed yesterday is 7%.

So for metastatic patients, they are divided into oligometastatic and—if they’re oligometastatic, can they have SABR? In which case they are randomized to receive SABR or not. If they’re SABR-ineligible or “poly-metastatic,” they’re then offered randomization to Lu617 or not.

All patients undergo next-generation clinical testing for BRCA2 and a few other HRR genes, and those who are positive are randomized to niraparib maintenance therapy. And of course, this is on the backbone of ADT, ARSI. And we also allow docetaxel, and of course we’ll stratify by intent to use docetaxel. So onto biomarkers.

I think we need prognostic tests to inform on that intensification, and then we need predictive tests to work out who the drug will work for. And I’ll focus on docetaxel, which is a question—a problem—we’ve been looking to interrogate for five, six years now.

So to discover, validate, and then clinically implement a biomarker, we need large data sets with prospective follow-up. We need to use a test that works on the challenging tissue we have access to, so it has to be fit for purpose. To ask a predictive biomarker question, we need randomized patients.

And of course, we need—of course—but we certainly need large numbers of events to prove a statistically significant interaction between the biomarker and the drug. And we need cross-validation across trials. And really, the only way to achieve this is cross-consortia collaboration, which necessitates using a test. There are a number of approaches that are being developed.

The one which is really leading the pack currently is transcriptome-wide testing in collaboration with Veracyte. This cartoon shows the flow: there’s the ITT population, we collect blocks from about 40% to 60%, then tumor areas are macrodissected, the cDNA is hybridized to microarrays.

We get data on 40,000 genes, extract signatures, and then we can start testing for associations with outcome and subsequently prediction. So the CHAARTED trial provided really intriguing hypothesis-generating data. Matthew Smith presented this yesterday. And then, based on that, we proceeded to study the STAMPEDE abiraterone trial. Again, Matthew presented this data, so I won’t present it again.

And it’s in a preprint referenced at the bottom. And now, we’ve proceeded to study the STAMPEDE docetaxel trials and work on “enza metastasis.” So we’re going to have thousands of patients—or tumors from thousands of patients—all analyzed with the same test that I hope will allow us to achieve those three targets in my title.

Finally, artificial-intelligence–derived models on digitized histopathology images. The advantage here is that the resource is not finite; we can work with multiple partners. This is data from STAMPEDE working with ArteraAI. What this multi-modal model does, and it includes digitized images, T stage, age, PSA, and Gleason score, is prognosticate.

And how can this be helpful? So I’ll really draw your attention to the gray box. And this is the M1 low patient. The numbers there are the five-year prostate cancer–specific mortality estimates. In brackets are 95% confidence intervals. And you can see for a low-volume patient, we can distinguish one who has a 90% chance of prostate cancer death at five years versus 37% for the “bad” ArteraAI prognostic group.

So to conclude, I think—or I hope—we will have positive phase III trials. As for RPFS, I think that will change practice, but we will need to follow up patients to show an OS benefit. And I am confident we will be incorporating biomarkers to guide docetaxel use. Thank you.