乌鸦传媒

SAN FRANCISCO -- Genomic tests to help guide treatment decisions for prostate cancer frequently pointed in different directions when used to evaluate the same patient, a small retrospective study showed.

The Decipher and Prolaris tests produced discordant results a third of the time when used to evaluate patients who met National Comprehensive Cancer Network (NCCN) criteria for active surveillance. The Prolaris and Oncotype DX tests disagreed a fourth of the time. Used to evaluate only two patients, the Decipher and Oncotype DX agreed on one and disagreed about the other.

Individually the tests also differed with respect to how often they supported the NCCN criteria for active surveillance. Though limited to a total of 22 patients, the results underscored the complexities of using genomic tests to select patients for active surveillance, Joseph R. Wagner, MD, of Hartford Hospital in Connecticut, said here at the (AUA) meeting.

"What we can say from these results is that there are notable differences in favorable prognostic outcomes obtained from Oncotype DX, Prolaris, and Decipher," Wagner said during an AUA press briefing. "For patients who are candidates for active surveillance, Prolaris is most likely to support that option. Results with Prolaris and Oncotype DX were in agreement most often."

Emphasizing that the results did not suggest that one test is superior to the others, Wagner said, "This is less of a scientific presentation and more of a thought-provoking presentation."

The analysis had its origin in two changes to recommendations in 2018 NCCN clinical guidelines. The NCCN now recommends that physicians offer active surveillance to patients with favorable intermediate-risk prostate cancer (Gleason 3+4, PSA <10 ng/mL, low-volume disease) and that genomic testing may be considered for patients with low-risk and favorable intermediate-risk prostate cancer.

Each test incorporates different types of information to yield results that inform different aspects of decision-making process. Prolaris estimates a man's 10-year risk of prostate cancer death with active surveillance, 10-year risk of distant metastasis with definitive treatment, and the patient's standing in the AUA risk groups. Decipher and Oncotype DX provide information about a patient's 10-year risk of prostate cancer death or distant metastasis after surgery and the likelihood the patient will have high-grade disease (grade ≥3) or extracapsular extension by pathology.

Relatively little comparative information is available about the three tests. Given that clinicians may order any one of the approved tests (except as required by hospital or clinic policy), Wagner and colleagues sought to collect data that might help inform choices about the tests.

Investigators retrospectively reviewed medical records for patients with newly diagnosed prostate cancer from 2014 to 2017 and who underwent evaluation with at least two of the genomic tests. Multi-testing was by patient request, not the treating physicians.

The review identified 22 patients, all but two of whom met NCCN criteria for low- or intermediate-favorable risk and were considered candidates for active surveillance. Records showed that 12 patients had assessments by the Decipher and Prolaris tests, which produced concordant results supporting active surveillance in eight of the 12 cases. Eight patients had both the Prolaris and Oncotype DX tests, which agreed on the appropriateness of active surveillance for six of the patients.

Investigators also compared each test's results with the NCCN risk assessment for each patient. The Prolaris and NCCN results were in agreement for 15 of 20 patients (75%). The Decipher and NCCN criteria agreed on nine of 15 patients (60%), and the Oncotype DX and NCCN were in agreement for five of 10 patients (50%).

Wagner and colleagues also performed a kappa analysis to hone the estimated agreement between each test and the NCCN criteria (kappa value ≥0.6 reflects moderate agreement). The comparison of Prolaris and NCCN (n=20) yielded a kappa value of 0.21, and the Decipher-NCCN comparison (n=15) produced a value of 0.15. Too few patients (N=10) were included in the comparison of Oncotype DX and NCCN to calculate the kappa value.

Looking ahead, Wagner said suggested potential refinements to test criteria to improve each test's performance in the clinical setting. Examples: incorporation of mutation patterns to determine whether a patient might be better serviced by a specific type of local therapy; incorporate outcomes for radiation therapy and active surveillance into Oncotype DX and Decipher; and the addition of separate outcomes for radiation therapy and surgery with Prolaris.

With regard to deciding which test to use, Wagner said he explains to a patient the type of information that each test provides and then asks, "What do you want to know?"

"For most of us, we can choose which test to order or the patient can choose, but there is very little reason to choose one over another," said press briefing moderator Stacy Loeb, MD, of NYU Langone Medical Center. "When they all started out, they had completely different endpoints ... but over time, the reports have converged quite a bit."

Comment