Andrew J. Armstrong, MD, MSc: If externally validated, an AR-V7 test could be useful in several ways. The first would be to predict response to standard-of-care therapies, abiraterone and enzalutamide, which is really the intent of the test. In my mind, the most useful clinical application would be following progression on frontline abiraterone or enzalutamide where the prevalence of the test becomes more common. We already know that the crossover therapy from abiraterone to enzalutamide and enzalutamide to abiraterone is fairly ineffective in most patients. However, some patients do respond, and having a rapid turnaround test that could spare the patient an ineffective drug could spare society a costly drug that would be ineffective for that patient and get that patient rapidly on to a more effective therapy would be the goal of that test.

If you think about it, these agents are quite expensive. They tend to average $8000 or $9000 a month. And if you can reduce the number of months that a patient is taking an ineffective therapy and get them to a more effective therapy, then that test really has some health economic benefits and some clinical benefits where you can reduce the exposure to ineffective therapies. So, the test utility is really a negative predictive biomarker for standard-of-care therapy.

The second place where these tests may be useful is in the development of new drugs. If you have a patient where you think a new androgen receptor blocker may be particularly active in AR-V7—driven tumors, you could design a trial to enrich specifically for AR-V7–positive patients using these tests. And several clinical trials have been designed selectively with eligibility criteria for AR-V7–positive castrate-resistant prostate cancer patients. One study was recently reported as not meeting its bar of success. This is the galeterone ARMOR-3 trial. In this study, the drug failed rather than the biomarker, although the biomarker indicated that in the frontline setting, AR-V7 was exceedingly rare—making drug development in that frontline setting pretty difficult when the test is so rare.

A second trial is comparing abiraterone or enzalutamide with taxane-based chemotherapy. Unfortunately, this trial closed early and won’t answer this important question, largely because of lack of interest in being randomized to chemotherapy or an oral agent. Future trials will need to be developed in these biomarker-selected patients to really explore novel AR-directed therapies, and there’s a number of them in the pipeline right now.

Emmanuel S. Antonarakis, MBBCh: It’s important to realize that all the data that I have discussed and presented so far represent single-center studies. So, neither of these studies truly validates the other because they used different methodologies and they were each done at a single center. In order to validate the prognostic ability or utility of AR-V7 detection, this needs to be now tested in prospective studies, preferably using a multicenter platform where 4, 5, or 6 centers are all involved in this together. And even better would be if we could compare 2 or 3 different AR-V7 methods head-to-head-to-head, so to speak, and see which one is the most prognostic. In fact, this type of effort is currently underway. My colleague, Andrew Armstrong from Duke University, is leading a study funded by the PCF, or the Prostate Cancer Foundation. In that trial, 120 patients who are beginning therapy for the first time with either enzalutamide or abiraterone are going to be prospectively enrolled, and all patients are going to have blood samples taken for 3 different AR-V7 tests.

The first is the Johns Hopkins mRNA-based AR-V7 detection, the second is the Epic Sciences protein-based detection, and a third assay is one developed at Weill Cornell Medicine at Cornell University, which looks at digital droplet PCR to measure the AR-V7 mRNA transcript. That one is the most exploratory of the 3, the other 2 being a little bit more validated analytically.

Going back to the study design: 120 patients who are about to start treatment with enzalutamide or abiraterone will each get all 3 biomarkers at baseline and at the time of clinical or radiographic progression. And this trial is being conducted at 5 centers. In fact, a few weeks ago, we completed the enrollment of the 120th patient. The primary endpoint of this trial is progression-free survival and how progression-free survival correlates with the presence of AR-V7 detection at baseline in the pretreatment samples. I would say over the next 6 to 9 months, we should hear data from that study, and this will truly be the first multicenter validation of AR-V7. Again, it’s prospective, and it also tests 2 to 3 different AR-V7 platforms comparing each against each other.

Howard I. Scher, MD: The biomarker is embedded in a number of phase III trials, but I think the important results will come out shortly from the Duke study, which is looking at about 100 patients who prospectively had samples collected at the decision point when a change in treatment was being given, and several assays are being compared to it, which are specifically for AR-V7.

There are also data sets that have been collected, and there are stored samples that are being what we call retrospectively, prospectively analyzed. And what that means is we have banked the specimens that can be analyzed, for example, by the Epic platform, and we know the clinical outcomes of those patients. So, those samples can be run where the laboratory is blinded to the results and then essentially conduct the trial without having to start over again.

Andrew J. Armstrong, MD, MSc: Epic has put a lot of work into the controls for the AR-V7 test. The AR-V7 protein is somewhat finicky. It can stain the cytoplasm, for example, and the nucleus. And the positive test is really a nuclear stain rather than a cytoplasmic stain. The nuclear stain seems to be much more associated with treatment outcomes. It’s more predictive for resistance and quick progression in the clinical settings. And these have been largely validated at Memorial Sloan Kettering through single institutional experiences.

As I mentioned, we’re doing an external validation study where we’ve defined a positive test using the Epic AR-V7 nuclear protein expression and prospectively looking at, in 120 men, whether it’s validated as a predictive biomarker in men treated with abiraterone or enzalutamide. We follow them over time to see if their circulating tumor cells evolve or adapt and if these resistance mechanisms become more dominant over time. So, the test is very sensitive. It has a very good positive predictive value, meaning, if the positive test is detected, the chance of a benefit with these drugs is very low.

One of the big limitations of AR-V7 is its lack of negative predictive value, meaning that patients can still not respond to the drug even if the test comes back as normal. So, many patients with AR-V7—negative disease still have other mechanisms of resistance that we can’t forget about.

Transcript Edited for Clarity