EURURO Vol. 72 No. 5

851 860

For example, Decipher was able to identify a particularly
poor prognosis group, as 23% of the patients had Decipher
high risk, and of these, 21% developed metastases by 5 yr,
and strikingly 9.4% of them died of PCa at 5 yr, which places
them at much higher risk of mortality than the typical
patient with intermediate- or high-risk PCa. Such patients
would likely want to pursue the most aggressive therapies
possible for their PCa. If they choose radiation, they would
need to combine with long-term ADT, which has been
shown to reduce PCSM compared with short-term ADT
, and may consider enrolling on a clinical trial
incorporating a novel agent or considering docetaxel, which
has been shown to improve 4-yr overall survival in men
with aggressive disease (1-sided p = 0.04, 2-sided p = 0.08)
. Those considering surgery would need to be prepared
for having a higher likelihood of requiring postoperative
adjuvant or salvage radiation and ADT, or might also
consider enrolling on a trial targeting those with highest-
risk localized disease.
This biopsy test may also allowmen with more moderate-
risk disease to tailor their therapy. For example, there is
currently controversy about whether all men with interme-
diate-risk PCa need ADT with their radiation. Patients with
intermediate-risk disease who have a low Decipher score on
biopsy might be adequately treated with dose-escalated
radiation alone and thereby forego the additional side effects
of ADT. However, as the majority of the patients in this study
who received radiation also received ADT, it is not yet
possible to make this conclusion. In a more general sense, it
must be acknowledged that this study has shown the biopsy
Decipher test to be a powerful prognostic marker, but it is not
yet known whether it is predictive for which patients will
benefit most from treatment intensification such as the
addition of ADT to radiation or the use of immediate adjuvant
therapies after surgery.
In previous studies, Klein et al and Nguyen et al
showed separately that biopsy Decipher outperformed
clinical risk models for predicting metastasis post-RP and
post-RT + ADT, respectively. However, the study performed
by Klein et al was a smaller (n = 57), single institution
cohort of mostly low- and intermediate-NCCN risk men, and
the study performed by Nguyen et al was a single
institution study of men who received only RT and ADT. The
current multi-institutional study of 235 men is the largest
study of the performance of the biopsy Decipher score and is
novel in that it incorporates both radiation and surgically-
managed patients, and is the first to show the ability of the
biopsy Decipher to predict PCSM in this combined population.
Previously, Freedland et al has studied the value of a
31-gene cell-cycle progression signature (Myriad) in biopsy
specimens of men who received RT or RT + ADT and found
that the signature was associated with biochemical recur-
rence and with PCSM, although there were only six PCSM
events. The current study is differentiated in that it looked
at a combined set of patients treated by either RP or
RT ADT and found that the genomic classifier could predict
for metastases and PCSM within this mixed-treatment cohort
It should be noted that Bishoff et al have also
evaluated the Myriad cell-cycle progression signature in
either simulated or actual diagnostic biopsies and found that
this signature could predict for biochemical recurrence and
metastases after RP.
The cohort size of this study was limited by access to
biopsy tissue from community and referral health centers.
Ninety-three percent of the unavailable cohort were either
unavailable or had inadequate tissue and 7.4% failed RNA
extraction. Of the 909 patients eligible for this study, only
235 had biopsy tissue available from the institution in
which the RPs or RT ADT were performed. A larger cohort
size with longer follow-up would have strengthened this
study and might have given more PCSM events to allow for a
MVA of predictors of PCSM rather than only univariable
analysis. Running multivariable models on a relatively small
number of events can also lead to issues with validity, and this
is why we fit the Cox models using an adaptation of Firth’s
penalized approach which was designed to minimize bias in
this scenario. Another limitation of our study was that the
majority of patients were NCCN intermediate or high risk and
we were not able to draw conclusions about Decipher among
low-risk patients, who only represented 10% of the patients in
the study. Such information would be useful to guide
decisions about treatment versus active surveillance. Of note,
a recent study of Gleason 6 prostatectomy samples found that
higher Decipher scores are associated with adverse pathology
at the time of prostatectomy Finally, ongoing work is
being performed to determine the concordance between
Decipher scores derived from biopsy versus prostatectomy
samples, which has been reported in prior small studies to be
64%, 75%, and 86% .
5. Conclusions
Despite these limitations, this is the largest analysis of
biopsy Decipher to date that shows Decipher predicts
metastasis post-biopsy, regardless of first-line treatment,
and also demonstrates its predictive ability for PCa
mortality. Our findings suggest that patients classified as
Decipher high risk have a very high 5-yr risk of distant
metastasis (21%) and PCSM (9.4%), and may therefore be
rationally subjected to multimodal therapy or enrolled into
clinical trials targeting men with highest-risk disease.
Author contributions: Paul L. Nguyen had full access to all the data in the
study and takes responsibility for the integrity of the data and the
accuracy of the data analysis.
Study concept and design: Nguyen, Ross, Feng, Klein.
Acquisition of data: Nguyen, Ross, Martin, Lotan, Spratt, Stoyanova,
Punnen, Kane, Pollack, Davis, Klein.
Analysis and interpretation of data: Nguyen, Haddad, Ross, Martin,
Deheshi, Chelliserry, Aranes, Margrave, Yousefi, Choeurng, Davicioni,
Kane, Feng, Klein.
Drafting of the manuscript: Nguyen, Haddad, Deheshi, Yousefi, Choeurng,
Davicioni, Trock, Feng, Klein.
Critical revision of the manuscript for important intellectual content:
Nguyen, Haddad, Ross, Martin, Deheshi, Lam, Chelliserry, Tosoian, Lotan,
Spratt, Stoyanova, Punnen, Ong, Buerki, Aranes, Kolisnik, Margrave,
Yousefi, Choeurng, Davicioni, Trock, Kane, Pollack, Davis, Feng, Klein.
Statistical analysis: Haddad, Yousefi, Choeurng, Trock.
E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 8 4 5 – 8 5 2 851