EURURO Vol. 72 No. 5

categorized as low, intermediate, and high risk by biopsy

Decipher, respectively (Supplementary Fig. 1).

3.3.

Evaluation of Decipher for prediction of metastasis

We performed stratified MVA Cox models to determine the

performance of biopsy Decipher for predicting metastasis.

OnMVA, biopsy Decipher remained a significant predictor of

metastasis when adjusting for clinical variables or risk

models (NCCN or CAPRA). Biopsy Decipher had a HR of 1.37

(95% CI: 1.06–1.78,

= 0.018) per 10% increase when

adjusting for age, biopsy Grade Group, clinical stage, and

PSA. Similar results were obtainedwhen adjusting for CAPRA

or NCCN

( Table 2 )

. We also evaluated the performance of

biopsy Decipher adjusting for clinical variables and first-line

treatment (RP or RT ADT). This analysis demonstrated that

biopsy Decipher was prognostic for metastasis (HR: 1.39, 95%

CI: 1.09–1.80) adjusting for first-line treatment and clinical

variables

( Table 3 )

. When Decipher was analyzed as a

categorical variable on MVA, patients with high-risk Decipher

score (

0.60) had a HR of 4.30 (95% CI: 1.63–11.49,

= 0.003)

compared with low-risk patients (Supplementary Table 2).

To evaluate the discriminatory performance of Decipher

and its added value to either NCCN or CAPRA, we employed

survival receiver operating characteristic analysis at 5-yr

post-biopsy. In this analysis, CAPRA, NCCN, and biopsy

Decipher had a c-index of 0.60 (95% CI: 0.50–0.69), 0.66

(95% CI: 0.53–0.77), and 0.74 (95% CI: 0.63–0.83), respec-

tively. Addition of biopsy Decipher to CAPRA improved the

c-index from 0.60 to 0.71 (95% CI: 0.60–0.82) and addition

of biopsy Decipher to NCCN improved the c-index from

0.66 to 0.74 (95% CI: 0.66–0.82;

Fig. 1 )

. On decision curve

analysis, biopsy Decipher remained superior to the CAPRA

model (Supplementary Fig. 2).

Next, we used cumulative incidence curves to stratify

metastasis risk by clinical and genomic risk models. NCCN,

CAPRA, and biopsy Decipher all significantly stratified

metastasis risk (all

0.01;

Fig. 2

). Patients with NCCN low,

intermediate, and high risk had 0%, 6.4%, and 14%metastasis

incidence by 5-yr postbiopsy, respectively

( Fig. 2

). Similar

results were observed for CAPRA

( Fig. 2

). Patients with

biopsy Decipher low, intermediate, and high risk had a

metastasis rate of 4.1%, 7.8%, and 21% by 5-yr post-biopsy,

respectively

( Fig. 2 )

In a sensitivity analysis, because we had a small number

of low-risk patients, we evaluated only the subset of

Table 2 – Multivariable analysis (MVA) using institution as a

stratification variable for prediction of metastasis

Model

Variables

MVA

Hazard ratio

(95% CI)

value

Model 1

Bx Deciphe

r a

1.37 (1.06–1.78)

0.018

Patient’s age at first

line treatment (yr)

1.04 (0.97–1.11)

0.25

Grade Group 1

Reference

Grade Groups 2–3

2.68 (0.50–17.2)

0.25

Grade Groups 4–5

4.29 (0.66–33.5)

0.13

Clinical stage T1c

Reference

Clinical stage T2a

1.07 (0.4–3.06)

0.9

log2 Pretreatment

PSA (ng/ml)

1.01 (0.68–1.48)

Model 2

CAPR

A b

1.08 (0.83–1.41)

0.6

Bx Deciphe

r a

1.44 (1.15–1.84)

0.002

Model 3

NCCN low

Reference

NCCN intermediate

8.85 (1.07–1151.1)

0.041

NCCN high

11.8 (1.16–1604.1)

0.034

Bx Deciphe

r a

1.39 (1.15–1.69)

0.001

CAPRA = Cancer of the Prostate Risk Assessment; NCCN = National

Comprehensive Cancer Network; PSA = prostate-specific antigen;

RP = radical prostatectomy; RT = radiation therapy.

Decipher reported per 0.1 unit increase

CAPRA is reported per unit increase.

[(Fig._1)TD$FIG]

Fig. 1 – C-index for metastasis at 5-yr post-Bx.

C-index of the combined

models was corrected for optimism.

CAPRA = Cancer of the Prostate Risk Assessment; NCCN = National

Comprehensive Cancer Network.

Table 3 – Multivariable analysis (MVA) adjusting for treatment and

clinical data for prediction of metastasis

MVA

Variable

Hazard ratio (95% CI)

value

Patient’s age at first line treatment 1.03 (0.97–1.1)

0.29

log2 Pretreatment PSA (ng/ml)

1.00 (0.68–1.48)

Grade Group 1

Reference

Grade Groups 2–3

2.80 (0.65–16.2)

0.17

Grade Groups 4–5

4.78 (0.92–32.2)

0.063

Clinical stage T1c

Reference

Clinical stage T2a

1.09 (0.42–3.06)

0.9

Bx Deciphe

r a

1.39 (1.09–1.8)

0.009

First-line treatment RP

Reference

First-line treatment RT ADT

0.73 (0.26–2.28)

0.6

ADT = androgen deprivation therapy; CI = confidence interval;

PSA = prostate-specific antigen.

Decipher reported per 0.1 unit increase

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

848