Read the Full Video Transcript

Nicholas James: Hi, I'm Nick James. It's a great pleasure to be chatting here this afternoon with my long-term friend and colleague, Gert Attard from University College London.

Gerhardt Attard: Hi, Nick.

Nicholas James: So we're going to show you a little bit of data first about patterns of care globally because this kind of plays into why these data are important. Then we'll look at the actual data and then come back to discuss how we should use it.

Gerhardt Attard: Yes. Thank you, Nick. This is, I guess, the breakdown of first-line treatment to metastatic patients over a period of time, I think it's about eight years, which calculated that about half of patients received in the US and then Germany and probably France as well, received ADT monotherapy. I think this is changing. I think a higher proportion of patients are being offered treatment intensification with androgen-receptor pathway inhibitor. And for a subset of those, also triple therapy with docetaxel and that ARPI. I think it takes a while for practices to change. Clearly the first trials to present this benefit were STAMPEDE and LATITUDE. I think you presented that, Nick, in 2017.

Nicholas James: It was.

Gerhardt Attard: The data for docetaxel presented in the two or three years before that. I mean, clearly in the UK, the big change happened around COVID when use of docetaxel dropped through the floor and was primarily replaced by ARPIs, androgen-receptor pathway inhibitors, mostly defined by funding access.

Nicholas James: So the thing that is very striking about this as well, and it's very striking if you go and give talks in Japan or China, is the very high use of first-generation androgen-receptor pathway inhibitors, so drugs like bicalutamide. If you talk to Japanese and Chinese urologists, they seem very convinced that those drugs behave differently in their populations to European or other ethnic group populations. I don't know if there's any hard evidence to back that up, but they're clearly backing it up by prescribing it as shown here. Again, as you said, it would be interesting to see if that's changed with the data on Abiraterone and enzalutamide and so on, as that becomes more solidified as the kind of standard of care, I guess.

Gerhardt Attard: Yeah. I mean, clearly a decade ago, Japan were the ones primarily using maximum androgen blockade. I know for abiraterone, there's higher toxicity in China, for example. So there are some genetic differences in side effect profiles which may make some of the ARPIs less well tolerated in certain areas.

Nicholas James: Yeah. Well, also actually I remember the first time I gave a talk in Japan after enzalutamide had been licensed, the question that I got asked over and over is how do you manage the terrible toxicity you see with enzalutamide? I was saying, "Well, we really just don't see it." I think it probably relates to the average BMI in Japan in the '60s and '70s to the average BMI in the UK, which is much higher. And I think that they were giving effectively much higher doses per kg and therefore seeing different toxicity issues to that which we saw. So I think there are reasons that are not necessarily ethnicity per se, but more sociological, if you like, that might end up in some of those differences at least.

Gerhardt Attard: So I guess to move on and summarize, we have a number of options. I think most of us will agree that for metastatic disease, we should be intensifying treatment with an ARPI or a new generation hormonal agent, accepting that there are differences globally in how that is used. So agreeing that that really should be the backbone of care now, ADT plus an ARPI. And accepting, yes, there are geographical differences in use and also reduced take-up, which I think with time, at least for patients who are fit and eligible for these treatments, will decrease.

Equipoise probably remains on who should receive docetaxel in addition to that backbone, so what we're now calling triplet therapy, and that's across the board, of course, for the low-volume patients—or maybe I shouldn't say of course, because some debate remains. But for low-volume patients, there's strong evidence from STAMPEDE primarily, that primary ADT therapy to the prostate will improve survival and that cannot be extrapolated to high volume. So really restricted to low volume as defined by conventional imaging.

Nicholas James: I think that's clear. So I think we can move on to the stuff that we actually presented. Well, not actually, but Emily presented.

Gerhardt Attard: Yeah, so a bit of background data. Clearly there's two pressing questions for physicians. The first is prognostic, and in our final analysis or trial-reported analysis from STAMPEDE last year, we reported nearly a quarter of patients on ADT and abiraterone were in remission when the trial closed and continued treatment, so that's eight to nine years of treatment. For such a patient, addition of docetaxel is going to cause side effects with no clear benefit, and given the improved outcomes, I believe we clearly need an improved prognostic test.

Then the second question is for patients who we have decided require treatment intensification. We need a predictive test to identify A, is that tumor sensitive to docetaxel, and then in the future, B, is it more sensitive to other treatments. I think metastatic volume captures some of that prognostic information, but not entirely, and we've therefore worked for at least six to eight years to molecularly profile tumors from patients directly randomized in the STAMPEDE trial, towards prognostic and predictive questions.

We initiated the work in partnership with the CHAARTED investigators who used a whole transcriptome microarray clinical test to derive three signatures from patients randomized to ADT or ADT-docetaxel. In total there were 160 patients in this analysis.

The primary result was of a differential benefit for luminal B patients in the basal-luminal PAM50 phenotyping, and this was repeated using a prostate-specific test called PSC. No differential benefit for AR activity, although that was shown. There's a hypothesis from the TITAN Trial and the SPARTAN trial, which tested apalutamide suggesting that AR activity identify differential benefit from ARPIs. And finally Decipher, which sits in the middle of the forest plot here.

At the time, we didn't make too much of this, but we noted that the upper quartile of the Decipher signature had a numerically lower hazard ratio of 0.41 for addition of docetaxel. To me, critically, this analysis confirmed that this approach could deliver useful information on the majority of our samples. And given we really had one shot at this and no one's going to randomize patients again to docetaxel or no docetaxel—or at least there's a number of challenges in doing so and certainly no one's going to perform more randomizations ARPI versus no ARPI—we prioritized, I guess after much discussion in the TMG and biological monitoring group, which tests to use, and this was clearly the front-runner because of its reliability and also the high success rate.

Nicholas James: Yeah. And actually just in terms of the stats on that, I think one of the key things that we get told over and over again is that we can't just go trawling through hundreds of signatures and say, "Ah, that one looks good." In order for us to be credible, we had to pre-specify our best shots or shots restricting ourselves to a small number. And as you say, there were two things we wanted to know: could we better the prognostic information that we have already from clinical work, but also very critically could we have a predictive test? And I think it's worth saying here that I don't think there are any predictive tests in any disease setting that I'm aware of that allows you to select patients who do or do not benefit from docetaxel. So I think that makes what we're going to show you now particularly striking, given both the numbers of patients that we have in STAMPEDE and the size of the effects that Gert's going to show you.

Gerhardt Attard: So we decided to start with the Abiraterone Trial first. So of course STAMPEDE's a platform protocol. Each one of what are called comparisons for regulatory reasons is an independently powered phase three trial. So 2,000 patients in the Abiraterone Trial, and this we presented two years ago in STAMPEDE, and here we analyzed tumors from about 800 patients. So actually we started with 1,000. Once we excluded cases that had insufficient tumor, that were not prostate core biopsies, and where the test failed—and they failed in about 10%, which is excellent considering these are historic samples—we had about 800 patients and we showed Decipher was strongly prognostic. So we're proving the implementability and importantly, the strong prognostic association in metastatic disease, which hasn't been shown in this way previously.

But yes, can certainly have value for identifying patients who require treatment intensification who are metastatic. And potentially, if our data with abiraterone in localized high-risk patients is extrapolated or starts to extend to lower-risk patients, I think there could be a role for prognostic tests like Decipher. The STAMPEDE localized high-risk patients have a high chance of dying from prostate cancer when treated with ADT alone.

But as one moves further to the left, reduces the risk, starts to treat patients at a lower risk, the chance of that person dying from something else rather than prostate cancer pleasingly tips, so there's a lower risk of dying from prostate cancer. So the benefit of intensified treatment decreases, and this is what we're capturing in these plots where the red is above median, the green is below median, and very early on, the high Decipher patients derive most of the survival benefit from abiraterone, so those are patients dying within 36 months.

Clearly those are the quartile of patients with the worst prognosis. But overall, we're seeing a benefit for the vast majority of patients of addition of abiraterone, and there's a fit selected STAMPEDE population. But in localized, you'll notice a different profile there. Relatively few patients die or have a metastatic event in the first 72 months and then when patients start to develop metastasis or die at a later time point. And we can see the small difference between the high and the low Decipher, and with increasingly beyond this point, start to observe competing risks of death. So I guess we haven't pursued this too much further, at least in the localized high-risk. But I think there's value for the test, as I said, as other trials emerge studying ARPIs in lower-risk patients.

Nicholas James: Yeah, no, I think that's super interesting. And so basically we can't identify a population that doesn't benefit, but as you say, as your risk of death from prostate cancer goes down simply because your overall prognosis is better, your absolute benefit is diluted by just dying from something else, which I guess mostly is a good thing. I guess if you're dead, you're dead and you're not very happy about it. But from our point of view, it means that we've kept you alive from the prostate cancer point of view. So moving on, we've got the docetaxel data, which as we'll see, looks different and therefore we think is very interesting.

Gerhardt Attard: So onto the new data we presented at ESMO, and here, having analyzed the 1,000 patients in the Abiraterone Trial, we moved on to the 1,000 patients in the docetaxel trials. These are actually two trials, docetaxel alone and docetaxel plus zoledronic acid shared the same control and two-to-one randomizations, so we had equal numbers who received docetaxel plus or minus zoledronic acid versus ADT. And we analyzed both non-metastatic and metastatic patients. We tested PAM50, PSC, and Decipher as our primary signatures.

So this is pre-specified in our statistical analysis plan prior to inspection of any data. As one looks through the hierarchy of robustness for evidence, pre-defining prior to randomization and using the test to influence that decision is the highest level of evidence showing—if that shows an improvement in survival. I guess we're one below that, referred to as level 1B, which different regulators and different funders and jurisdictions take, I guess, different views of.

But yes, that's an important point that Nick alluded to earlier, that we didn't just go trawling through every possible signature. We pre-specified our three primary signatures of interest based on the CHAARTED data, so prior data to substantiate that hypothesis. And PAM50 and PSC showed no treatment interaction, but Decipher showed a statistically significant interaction. And the bar here is really high for statistical significance for biomarker treatment interaction. So we're first focusing on the differential outcomes. So this is for survival, 14-year survival follow-up in the metastatic group. Hazard ratio for high Decipher is 0.64, and the hazard ratio for the low Decipher is close to 1, with a biomarker treatment interaction p-value of 0.039.

I'll quickly go through the split by high volume and low volume, and then I guess we can discuss the implications, Nick. So this is exploratory because we are not pre-powered to study Decipher for biomarker treatment interaction in the subgroups. But of course, it's an important question where we first showed this to our investigators, everyone including us agreed, okay, is this just volume? And we knew from the distribution of Decipher scores that there was no difference between low and high volume. So that was unlikely to explain our primary result.

And yes, this was confirmed by the consistency of effect we observe in low volume shown here, hazard ratio of 0.53 in the high versus 0.78 in the low, and then in the high volume—and that's by CT or bone scan patients—Decipher high hazard ratio 0.72 and in low hazard ratio of 1. So as you see a consistency of effect in both high and low volume.

Supporting use of the test across all metastatic patients, not split by disease volume. Effectively, the high Decipher signature is identifying patients with poor prognosis on ADT and ARPI, which I think can be extrapolated to other ARPIs, although we'll see data from trials like ENZAMET that look at the prognostic association there, and then that same classifier identifies docetaxel-sensitive tumors. We've said there's no clinically significant heterogeneity of effect with ARPI. So I guess after much discussion as a group, we've agreed that this provides a rational biomarker for selection of patients for ADT, ARPI, and docetaxel.

Nicholas James: Yeah, and as you say, we really agonized over the wording of point number five here before Emily presented it. Emily Grist, the research fellow who presented it. And I think it's a really important point because at the moment, people are deciding whether to give docetaxel as part of triplet therapy partly on a sort of gut feeling—this looks like a young patient with bad disease—partly because they believe the data, and there's a body of people who don't believe the data and don't believe we should give docetaxel to any of these patients. But this actually gives us an independent, orthogonal way of identifying the ones who are likely to be docetaxel-sensitive, and also very importantly the ones who are really not likely to benefit from docetaxel. So there are high-volume patients who are not going to benefit from docetaxel, who should not be treated with it, I think.

And so I think this gives us really good data, and yes, you're right, the gold standard way of doing it would be to do a further trial where we use the test to select for treatment or not. But at the very least, that trial will take a decade. And so in the meantime, we just have to make decisions one way or the other. And this seems to me, I think, a better way of making a decision than what people use at the moment, which is mostly volume, I think. But I'd be interested in your take—you're more of a docetaxel skeptic than I am anyway, Gert, I think. So I mean, is this going to persuade you to use docetaxel and triplet therapy?

Gerhardt Attard: Yeah, so despite being a docetaxel skeptic because we don't have randomized data for docetaxel, I still find myself in clinic with a patient with very high-volume disease, as you said, who may be young or maybe old, but young is more likely to tolerate treatment well. And they're saying, "Give me everything you have." I do have the triplet therapy conversation. We talk about quality of life, but some patients say, "Give me everything." So we do use docetaxel, and I think that's the challenge we will have with randomization now. And I think the key patients who we want to understand and the randomized patients who will maybe be excluded because physicians or the patient will not feel comfortable with that randomization step. So I think we need to keep that in mind, and there are ways we can control for that in our trial design. But there's a high chance that a key segment of our population will be missed in randomized trials now.

Nicholas James: Absolutely. And I think the other thing is costs. I don't know what the test costs, and I guess what it costs depends a bit on healthcare systems as well, but the cost of docetaxel are pretty well established, and whatever the drug itself costs, you've got a lot of infusion time costs, one in six or so of these patients will get neutropenic sepsis—it's absolutely nailed on. They will all get toxicity. So to treat patients who are looking unlikely to benefit on the basis of a test, which will for sure be not prohibitive compared to the cost of treatment, it would seem to stack up quite much in favor of doing the test, I have to say. I should say as well, I have no direct financial interest in the test, saying what I've just said. So yeah, I think this is really interesting data. We are obviously giving our personal views on this, but the 0.5 here, we put a lot of thought into this. It will be very interesting to see how it plays in the wider community.

Gerhardt Attard: Agreed. Thank you, Nick. Thank you for giving me the chance to talk through this. We've discussed it so many times before presentation, I guess it's good to start to air it more widely. Finishing off with our financial disclaimer or disclosure of interest, in that UCL that I'm employed by and that sponsors STAMPEDE, has licensed the STAMPEDE data to Veracyte and therefore could benefit commercially from these data through Veracyte's commercial ownership of the test.

Nicholas James: Yeah, thanks for this. I think that's important to state that as part of the presentation. So it's been great talking to you about this work, and obviously we're both heavily invested in the whole thing from an emotional and intellectual point of view, if not a financial one. Yeah, that would be very interesting to hear reactions to this, please do give feedback to one or both of us directly if you want more information or comments.

Gerhardt Attard: Certainly.