NEW YORK (GenomeWeb) – Results from a retrospective study led by researchers at Dana-Farber Cancer Institute have shown that it's possible to predict outcomes of patients on hormonal prostate cancer therapy based on PCR measurement of PSA and AR-V7 transcripts in blood samples.
The results, published earlier this month in Clinical Cancer Research, suggest that this RT-PCR approach for assaying AR-V7 and PSA status — which used Bio-Rad's droplet digital PCR platform — could potentially be a simpler alternative to other methods for AR-V7 detection that require isolation of circulating tumor cells.
Phillip Kantoff, the study's lead author, now chairman of medicine at Memorial Sloan Kettering Cancer Center, told GenomeWeb that his team's work will have to be followed up by more studies before its clear whether a ddPCR or other RT-PCR approaches for AR-V7 detection in whole blood might have a place in the clinic either alongside of or in place of tests that examine individual CTCs.
This includes work to refine the best cutoff point for determining a predictive level of AR-V7 transcripts in whole blood, and testing this approach directly against current CTC-based methods.
CTC-based AR-V7 testing has advanced significantly in the last few years. Last month, Genomic Health and Epic Sciences signed an agreement giving Genomic Health exclusive right to commercialize Epic's non-EPCAM-based rare cell analysis AR-V7 test in the United States.
In June, Epic researchers and collaborators from MSKCC led by Howard Scher published a study in the Journal of the American Medical Association validating the assay's ability to distinguish men who might have better survival sticking with taxane chemotherapy versus anti-androgen drugs.
Researchers from Johns Hopkins have also shown that a method using Alere's AdnaTest technology with multiplexed qRT-PCR to detect AR-V7 could also predict patient outcomes on anti-androgen versus chemotherapeutic drugs.
In their new study, Kantoff and his Dana-Farber colleagues were working off of the hypothesis that because AR-V7 is thought to be produced exclusively or near-exclusively by prostate cancer cells, performing PCR on the whole mononuclear cell fraction could also yield a clinically meaningful measure of AR-V7 but without the need for CTC isolation or analysis of individual cells.
In much earlier work, Kantoff and collaborators were able to successfully detect PSA transcripts in blood samples with prognostic implications. Applying this approach to AR-V7 and to PSA in parallel in their new study, the investigators updated their approach somewhat, adopting Bio-Rad's ddPCR.
Whether it's actually necessary to have the high level of sensitivity ddPCR offers to measure AR-V7 in a way that best predicts patient outcomes isn't clear yet, but is something he and his colleagues are investigating.
In their study, Kantoff and his team perfomed ddPCR on blood samples from about 130 patients treated with either abiraterone or enzalutamide.
They detected PSA transcripts in 57 percent of abiraterone-treated patients and in 63 percent of enzalutamide-treated patients, and found that PSA-positive patients had a shorter time to treatment failure than PSA-negative patients in both cohorts.
Patients with a higher-AR-V7 transcript level also had a shorter time to treatment failure with both drugs in univariate analysis. In multivariable models, that association remained significant in the enzalutamide-cohort but statistically insignificant in the AA-cohort.
Both PSA and AR-V7 were also associated with overall survival. Patients with detectable PSA transcripts and high AR-V7 had the worst OS.
According to Kantoff, the main takeaway of the study is that the team's whole-blood approach is feasible and should be investigated further.
Moving forward, he said the team is now working on refining the cutoff point for when circulating AR-V7 transcripts cross the threshold to being predictive of patient outcome on hormonal therapies.
"What our study showed, and others have also found, is that AR-V7 is very prevalent," he said. Amongst the patients in the study, all but one had detectable AR-V7, as opposed to PSA which was only detectable in 50 percent.
"So the sensitivity is very high, but the question is what is the threshold above which it is predictive of outcomes," Kantoff said.
In the study, after testing different cutoffs, the researchers found that if the cohort were divided into thirds according to AR-V7 levels, the two lowest groups had significantly different outcomes from the top group.
"Whether that is the right threshold or not, that requires many more patients and validation," Kantoff added.
Also at question for further development of this testing method is the possibility that normal cells might also be expressing some AR-V7, potentially confounding an approach that doesn't isolate tumor cells from normal blood cells.
But according to Kantoff, the study results indicate that if this confounding is going on, it doesn't prevent the ability to predict outcome using whole-blood RT-PCR.
"That's a disadvantage of this, that we don't necessarily know where [the AR-V7] is coming from," he said. "But lo and behold it does have a prognostic and predictive signal."
In addition to working internally on the approach they used in the study, Kantoff said he and his colleagues also see the necessity for research that can directly compare their method with CTC-based testing.
"We've opened the door to doing a much easier test," he said. "But going forward what has to happen is a head to head comparison."
Such a comparison was taking place between Epic's platform and an AdnaTest-based CTC method being investigated by Qiagen and Tokai Pharmaceuticals as a companion diagnostic to Tokai's galeterone as part of the phase III trial of galeterone versus enzalutamide in treatment-naïve metastatic castration-resistant prostate cancer. But that trial was discontinued this year when it looked like it would not meet its primary endpoint.
Without a large prospective head-to-head comparison of all emerging AR-V7 testing methods, Kantoff said that he and Scher, his now-colleague at MSKCC, are planning a study to compare at least the whole-blood RT-PCR method and the method Scher has been researching using Epic Sciences' platform.
Regardless of what turns out to be the most accurate way to analyze AR-V7 status to predict patient outcomes, there still remain hurdles to convincing physicians to adopt such tests, Kantoff added.
"Because so few patients have undergone studies, there remain questions [about what the actual clinical sensitivity and specificity is]," he said. "So the threshold of comfort for using this to make a decision varies from physician to physician."