Ovarian Cancer: Evolving Concepts Around Systemic Therapy - Episode 7
Bradley J. Monk, MD: In PRIMA the primary end point included HRD [homologous recombination deficiency]. In PAOLA it was not even an end point, but because it helps inform decisions, it was reported. Tell us, Katie Moore, about this HRD assay that they used and what it means. We’ll get into whether we should use it. If we’re using it in all comers, if we believe that PRIMA is an all-comer opportunity, then maybe we don’t need to test it. But if we believe that in PAOLA it didn’t really work in patients who didn’t have the molecular signature, then it’s mandatory. Which is it?
Kathleen N. Moore, MD: Let’s just start without the assay. What’s HRD? Homologous recombination is the high-fidelity way that all our cells fix their double-strand DNA breaks. That function, in a lot of tumors—epithelial ovarian cancer is 1 of them—has some Achilles’ heels. You can lose this homologous recombination function through point mutations, most commonly in BRCA1 and BRCA2, through epigenetic mechanisms, or through loss of other genes. There are increasingly identified ways that a tumor can lose its capability to complete homologous recombination and become homologous recombination deficient. The clinical biomarker for homologous recombination deficiency is platinum sensitivity. That gives you your answer of who is homologous recombination deficient, but you’d like to know that up front and use it as a predictive biomarker for selection of DNA-damaging assets, such as PARP inhibitors.
There are a number of companies that have tried to commercialize this. Two are on the market. One looks at loss of heterozygosity [LOH] alone as a percentage. It’s basically the percentage of the chromosome that has lost heterozygosity between the 2 chromosomes, and if it’s 16% or higher, it’s considered LOH positive. Then the assay that was used in these 2 studies, in PRIMA and PAOLA, uses 3 components—LOH not measured in the same way, telomeric allelic instability, and large state transitions—and combines them. They have a cut point of 42 in these 2 studies. That’s based on some science that is hand waving. You dichotomized a function of DNA by a point score to say these women qualify and these women don’t. That’s what was used in these 2 studies and is the test that’s been most widely studied among the most patients with ovarian cancer.
Bradley J. Monk, MD: Whether you send it to Foundation Medicine, Inc and get LOH or send it to Myriad Genetics, Inc and get the 3 scores, it also gives you the somatic BRCA status.
Kathleen N. Moore, MD: Right.
Bradley J. Monk, MD: This is a tumor test, and there is useful information: HRD and somatic BRCA. Do we need this for everybody, or is it not a very good test and you’re going to use PARPs anyway?
Kathleen N. Moore, MD: That is the big question.
Bradley J. Monk, MD: I know, that’s why I asked it.
Kathleen N. Moore, MD: The truth is, we don’t have a good answer. It is a test. I think it’s a reasonable test. I have a lot of problems with it, though, and I’ll just go on record as saying that. When you talk about PRIMA, which was what Tom Herzog discussed, the analytical end points were HRD and ITT [intent to treat], and they met both end points. It was designed as a registration-enabling study, so they had their premeeting with the FDA. It should get labels in both settings, in my opinion. PAOLA was not designed as a registration study. It was an investigator-initiated trial that was very well done, and I’m really proud of Isabelle Ray-Coquard for doing it. I’m jealous that they can do these things in Europe. They had a great response and have filed it without the usual premeetings and things that go into designing a registration trial. Is that how you would have designed the trial if you were going through registration approval? I don’t know; probably not. Now we’ve stuck this assay in there as a secondary analysis, and it’s not hypothesis tested. It wasn’t 1 of the prespecified analytical end points. To make a registration decision based on that makes no sense to me. But the concession is that they hadn’t talked to the FDA before, so that just may be what they did. I think it should be an ITT indication. I don’t know if I’m going to get my way.
Bradley J. Monk, MD: But are you going to do the test? These are great points, but are you going to do the test? I don’t mean to make it complicated.
Kathleen N. Moore, MD: Why would I? This is what I believe. I’m going to reference 1 of the studies that came out of SGO [Society for Gynecologic Oncology Annual Meeting] that I know we’re not going to talk about. VELIA was another big phase 3 PARP study. They have an analysis that Liz Swisher presented, where they looked at HRD as well as a score of 33 to capture more people, but then they also looked at it across the continuum. They looked at the predictive capability of the actual score—1, 2, 3, 100—and they found that it’s not predictive of a response to veliparib. It’s prognostic. It’s prognostic to identify a new group of patients who are probably in trouble and for whom we need to develop better trials, but it wasn’t predictive. The magnitude of benefit of veliparib was the same above or below 33, above or below 42. These cut points are arbitrary. I don’t like to accuse it of not being a biomarker, I just want a good biomarker. If you make me pick 42, well, someone who’s score is 40 doesn’t get a PARP, and I think that’s dangerous. You’re leaving people behind.
Transcript edited for clarity.