Coleman Explains New Biomarker for PARP Inhibitors in Ovarian Cancer

Robert Coleman, MD

In the ARIEL2 study, rucaparib demonstrated robust clinical activity in patients with BRCA-negative relapsed ovarian cancer with a distinct genomic biomarker based on next-generation sequencing. The biomarker consisted of genomic "scarring” caused by homologous recombination deficiency (HRD) that can be quantified using genome-wide loss of heterozygosity (LOH) in single-nucleotide polymorphisms.

An overall response rate (ORR) of 45% (RECIST or CA-125) was shown in the biomarker group, which included patients without a BRCA mutation who had high LOH. ORR was 13% by RECIST and 21% with CA-125 in the biomarker-negative arm. As expected, patients harboring BRCA mutations had a high ORR of 82% (RECIST or CA-125).

Progression-free survival (PFS) was 9.4 months (90% CI, 7.3-not reached) in the BRCA-mutated group and 7.1 months in the biomarker arm (90% CI, 3.7-10.8). In the biomarker-negative group, the median PFS was 3.7 months (90% CI, 3.5-5.5).

The study’s findings are the first to suggest a clinical benefit with PARP inhibitors is possible in BRCA-negative patients, said lead investigator Robert Coleman, MD, professor, Department of Gynecologic Oncology and Reproductive Medicine, University of Texas MD Anderson Cancer Center. In an interview with OncLive Coleman explains the impact of these findings and the next steps in understanding this biomarker in ovarian cancer.

OncLive: What were the goals of the study?

Dr Coleman: This is a phase II single-arm trial conducted in women with platinum-sensitive recurrent ovarian cancer who had measurable disease. The idea was to not only take a brief look at the activity of this particular PARP inhibitor in women who carry the germline or somatic BRCA, but also to look at a group of patients that were identified as not having a germline or somatic BRCA and see how the drug performed in that group of patients.

We’ve known for some time that the group with the highest expectations for activity was those who carried a germline or somatic mutation of BRCA. We have also known that there are some patients with high-grade serous ovarian cancer that have responded to PARP inhibitors that do not carry the BRCA gene. What this study was doing was trying to identify a cohort of women through a genomic analysis using a prospective biomarker that do not have the BRCA mutation but still response to PARP inhibitors.

We capped the number of germline BRCA—positive patients to 15 early in the trial. We then opened up the trial to the BRCA-negative patient population and stratified them on the basis of whether or not they had a large amount of what we call genomic scaring, caused by homologous recombination deficiency (HRD). This was essentially represented by a diffused amount of genomic aberrations across the genome through a SNP array.

What were the results?

What we found is that there was probably a critical cutoff that was generated through two independent data sets as a way to discriminate women who had a lot of heterozygosity and those who did not. Through that dichotomization, we developed a decision point for women going into the trial who would be anticipated to be more likely to have a response. We termed those patients loss-of-heterozygosity (LOH)— high and LOH-low. Our three arms became LOH-high, LOH-low, and BRCA.

What we were excited to see is that the model performed very well. For patients with the BRCA mutation, we saw the high response rate we expected: 70% by RECIST criteria and 82% by CA-125. In the BRCA-negative group there was a 13% ORR and a 21% ORR with CA-125. However in the middle group, which consisted of patients that were BRCA-negative but that had LOH, we saw an ORR of 30% and even higher when the CA-125 responders were added.

So overall, what we found is that this was a tolerable regimen that has activity as expected in the most vulnerable group of patients, the BRCA mutation carriers, but we also found this large group of patients in the middle who carry the LOH aberration in their DNA who are also responding who would have not otherwise been identified.

What will be the most significant impact of these findings?

PARP inhibitors are a new class of agents that have been restricted to a very small cohort. Right now the olaparib approval is only for patients who have had more than three lines of prior therapy who are also germline positive. This study potentially opens up this class of agents to other populations.

What are the most common toxicities for PARP inhibitors?

Across this whole class, and there are multiple agents in the field, we’ve come to have a fairly good understanding of toxicities. The most difficult one is the gastrointestinal toxicity. We are very aggressive about informing patients to anticipate this particular toxicity and we prepare them with multiple medicines to help to manage it. There are some other minor adverse events that don’t really result in a need for dose modification.

The good news is, progression-free survival for patients that carried the biomarker was over 9 months in the BRCA-positive group and 7 months in the BRCA negative group, so these patients were exposed to the treatment for a long amount of time. Getting over the hump can really lead to lasting tolerance of the drug.

What questions remain to be answered regarding utilizing this biomarker with PARP inhibitors?

We did this as something that was integrated into the protocol. So we want to use this as a prospective decision tool. In the ongoing ARIEL3 trial, that is exactly what is happening. We are now going to be stratifying these patients on the basis of their genomic score, their LOH score, and BRCA mutation status. With that predictive biomarker now integrated into the trial, it will help to move it forward.

We are also going to expand the ARIEL2 trial to allow for more heavily pretreated patients, that didn’t quite fit this platinum-sensitive cohort, to try and understand how this biomarker helps to discriminate patients. I would love to say it was perfect, that we had the perfect biomarker with 100% response. But we clearly are not there. Even patients that do carry BRCA don’t respond 100% of the time, and the question is, “why not?” It can always be improved. So we are going to go back and look in that cohort and see if we can tweak the model and come up with a better algorithm.

We also need to find the right partners. Chemotherapy is a decent partner, but it is also more toxic. There are probably other more relevant partners like checkpoint involvement, immunotherapy, and angiogenesis that could make a difference. As we also better understand the mechanism of resistance and emergence of resistance, I think that is going to open up a new field for us going forward.