Urinary PCA3 to predict prostate cancer in a cohort of

Progrès en urologie (2015) 25, e1—e8
Disponible
en
ligne sur
ScienceDirect
www.sciencedirect.com
ORIGINAL ARTICLE
Urinary PCA3 to predict prostate cancer
in
a cohort of 1015 patients
Test
urinaire PCA3 et diagnostic du cancer prostatique : étude à partir de
1015
patients
V.
Vlaeminck-Guillema,∗,b, M. Devonecc,d,
D.
Champetierc, M. Decaussin-Petruccib,d,e,
P.
Paparelc,d, P. Perrinc,d, A. Ruffionb,c,d
aUnité
d’oncologie moléculaire et transfert, service de biochimie et biologie moléculaire
sud,
centre hospitalier Lyon-Sud, hospices civils de Lyon, chemin du Grand-Revoyet, 69495
Pierre-Bénite,
France
bInserm
1052 CNRS 5286, centre de recherche en cancérologie de Lyon — centre Léon-Bérard,
université
Lyon 1, 69373 Lyon cedex 08, France
cService
d’urologie, centre hospitalier Lyon-Sud, hospices civils de Lyon, chemin du
Grand-Revoyet,
69495 Pierre-Bénite, France
dFaculté
de médecine Lyon-Sud, université Lyon 1, chemin du Petit-Revoyet, 69921 Oullins
cedex,
France
eLaboratoire
d’anatomie et cytologie pathologiques, centre hospitalier Lyon-Sud, hospices
civils
de Lyon, chemin du Grand-Revoyet, 69495 Pierre-Bénite, France
Received
2 July 2015; accepted 10 October 2015
Available
online 14 November 2015
KEYWORDS
Prostate cancer;
Diagnosis;
Prostate
biopsies;
Prediction;
PCA3;
Urinary
test
Summary
Aim.
— To evaluate the performance of urinary PCA3 test to predict prostate biopsy outcome
in
a large French cohort.
Patients
and methods. — A urine sample was prospectively obtained from 1015 patients under-
going
prostate
biopsies to determine the PCA3 score. The predictive value of PCA3 was explored
using
receiver
operating characteristic curve analysis (ROC), multivariable logistic regression
analysis
and decision
curve
analysis.
DOI of original article:http://dx.doi.org/10.1016/j.purol.2015.08.005.
∗Corresponding
author.
E-mail
address:
[email protected] (V. Vlaeminck-Guillem).
http://dx.doi.org/10.1016/j.purol.2015.10.006
1166-7087/©
2015 Elsevier Masson SAS. All rights reserved.

e2 V. Vlaeminck-Guillem et al.
Results. — The median PCA3 score was significantly higher in patients with positive biopsies. The
PCA3
score AUC was 0.76 (0.73—0.79), significantly higher than that of PSA (0.55; 0.51—0.58). At
the
cutoff of 35, sensitivity was 68%, specificity 71%, positive and negative predictive values 67%
and
71%, and accuracy 69%. Using multivariate analysis, PCA3 score appeared as an independent
predictor
of biopsy outcome and its addition to a base model including usual clinicobiological
parameters
resulted in a significant increase in predictive accuracy. At the cutoff of 20, about
1/2
of the ultimately unnecessary biopsies would have been avoided while ignoring 7% of cancers
with
Gleason score ≥
7. PCA3 score did not correlate with Gleason score but did correlate with
tumor
volume (proportion of positive cores).
Conclusion.
— Urinary PCA3 is
a useful
test with high diagnostic performances for early prostate
cancer
diagnosis. Its correlation with cancer aggressiveness seems rather represented by a link
to
prostate volume than Gleason score.
Level
of evidence.— 5.
©
2015 Elsevier Masson SAS. All rights reserved.
MOTS CLÉS
Cancer de la
prostate
;
Diagnostic
;
Biopsies
prostatiques
;
Prédiction
;
PCA3
;
Test
urinaire
Résumé
But.
—Évaluer
la capacité
du
test urinaire PCA3 à prédire le résultat des biopsies prostatiques
dans
une
large cohorte
issue du
centre hospitalier Lyon Sud.
Patients
et méthodes.
—Le
score
PCA3
a été déterminé prospectivement chez 1015 patients
adressés
pour
biopsies prostatiques.
La capacité
prédictive du score PCA3 a été évaluée par la
comparaison
des aires
sous
les courbes ROC, de modèles de régression logistique et une analyse
par
« decision
curve analysis
» (DCA).
Résultats.
—Le
score PCA3
médian
était significativement
plus élevé chez les patients avec
biopsies
positives. L’AUC était de 0,76, significativement plus élevée que celle du PSA à 0,55.
Au
seuil de 35, la sensibilité était de 68 %, la spécificité de 71 %, les valeurs prédictives positive
et
négative de 67 % et 71 %, et l’efficience de 69 %. En analyse multivariée, le score PCA3 était
prédicteur
indépendant du résultat des biopsies et son addition à un modèle de base comportant
les
données clinicobiologiques classiques apportait un gain diagnostique significatif. Au seuil de
20,
près de la moitié des biopsies a posteriori inutiles auraient été évitées, tout en ayant ignoré
7
% des cancers avec score de Gleason ≥ 7. Le score PCA3 n’apparaissait pas corrélé au score
de
Gleason, mais était bien corrélé au volume tumoral (évalué par la proportion de carottes
envahies).
Conclusion.
— Le test urinaire PCA3 présente des performances diagnostiques élevées pour le
diagnostic
précoce du CaP. Sa corrélation avec l’agressivité du cancer s’exprime à travers le
volume
tumoral plus que par le score de Gleason.
Niveau
de preuve.— 5.
©
2015 Elsevier Masson SAS. Tous droits réservés.
Introduction
The first urine test for the quantitative assessment of
prostate
cancer gene 3 (PCA3) RNA transcripts in patients
with
suspected prostate cancer (PCa) was published more
than
10 years ago [1]. The PCA3 RNA product is almost
exclusively
expressed in cancerous prostate tissue. It is not
expressed
in other tissues and only very weakly expressed
in
healthy prostate tissue or in non-malignant prostatic dis-
eases
[1—3].
The discovery and diagnostic use of PCA3 came
about
in
response to
the lack of specificity of serum prostate-
specific
antigen
(PSA)
testing for the diagnosis of PCa.
To
date, approximately 100 studies, involving cumu-
latively
close to 30,000 patients, have been conducted
to
evaluate urinary PCA3 RNA measurement (PCA3 test)
in
clinical practice [4—6]. The PCA3 test was approved
for
diagnostic purposes by the United States Food and
Drug
Administration in 2012, as a decisional aid for repeat
prostate
biopsy
in patients aged at least 50 years who have
had
one
or more previous negative biopsies. In France,
as
of
this
writing, the PCA3 test is only available for
clinical
research
studies or
in the setting of a voluntary,
patient-payed
diagnostic process. Published clinical stud-
ies,
variable in scientific quality, tend to show that the
urinary
PCA3 test can be of significant value when deciding
whether
or not to perform prostate biopsy. In contrast to the
serum
PSA, which is affected by prostate volume, the PCA3
score
does not increase in the presence of a non-malignant
prostatic
disease [7]. It does however correlate with the
risk
of positive biopsy and its use may reduce the number

Urinary PCA3 to predict prostate cancer in a cohort of 1015 patients e3
of ultimately-negative (and thus, a posteriori unnecessary)
biopsies
by half to two-third [4—6]. Correlations are assumed
to
exist
between the PCA3 score and histoprognostic criteria
identified
on biopsy
or prostatectomy
specimens.
The
purpose of the present work is to report our experi-
ence
with the diagnostic and prognostic performance of the
urinary
PCA3 test in a large cohort of patients referred for
prostate
biopsy following a suspicion of prostate cancer.
Patients
and methods
Patients
All patients referred to our urology department between 20
December
2007 and 5 May 2014 for prostate biopsy were
consecutively
included in this prospective, single-center
study.
Reasons for referral were serum PSA ≥ 4 ng/mL
and/or
an abnormal digital rectal examination (DRE) and/or
a
family history of PCa. Patients with a personal history of
PCa
were excluded, but those with a history of one or more
negative
prostate biopsies were included. One hundred and
sixty
of the included patients had been included in an earlier
study
[8]. All patients provided written informed consent.
After
the collection of urine samples for the PCA3
test,
the patients underwent a prostate biopsy series in
accordance
with the recommendations of the European
Association
of Urology, i.e., at least six cores taken from
each
lobe under transrectal ultrasound guidance, this latter
permitting
also an evaluation of prostate volume. The fol-
lowing
anatomopathological data were recorded: Gleason
score,
percentage of positive cores, percentage of tumorous
prostatic
tissue, unilateral or bilateral involvement, pres-
ence
or
absence
of perineural invasion, presence or absence
of
extracapsular
extension.
Urine samples and urinary PCA3 test
First-catch urine samples (25—30 mL) were collected after
standardized
DRE [9]. The samples were immediately trans-
ferred
to specific specimen tubes (Progensa Urine Specimen
Transport
Kit, Gen-Probe), and stored at −20 ◦C until the test
could
be performed. The PCA3 test (Progensa PCA3 assay,
DTS400
system, Hologic Gen-Probe) was performed at the
hospital’s
biochemistry service as per the manufacturer’s
recommendations.
The PCA3 score was calculated as the
ratio
of PCA3 to PSA RNA copies, multiplied by 1000. When
less
than
10,000 copies of PSA mRNA were detected, the
urine
sample was considered as non-informative.
Statistical analyses
The normal distribution of quantitative variables was veri-
fied
to permit their comparison using the Student’s t-test.
Proportions
for qualitative variables were compared using
the
Chi2test. Correlations between quantitative variables
were
identified using linear regression.
ROC
curve was established for PCA3 score to calculate
the
area under the curve (AUC). For a given threshold, test
performance
was evaluated in terms of sensitivity, speci-
ficity,
positive and negative predictive values, and accuracy
(rate
of correctly-classified patients) by comparing PCA3
scores to biopsy histological results. The contribution of
the
PCA3 score in relation to existing clinical and biological
data
was evaluated in multivariate (nested logistic regres-
sion
models)
and decision curve analyses (DCA). Calculations
were
performed
using Stata Statistical Software Release 11
(StataCorp,
College Station, Texas, USA). A P-value < 0.05
was
considered statistically significant.
Results
Baseline characteristics
In the study period, 1029 patients had a urinary PCA3 test
before
undergoing prostate biopsy
for suspicion of PCa.
Urine
samples were informative for 1015 (98.6%) of them.
The
characteristics of the 14 patients with non-informative
samples
did not differ from those of the other patients
except
for a higher rate of 5␣-reductase inhibitor use: 21%
vs
2%
(P = 0.003). Further analyses were thus performed in
the
1015 patients
with
informative samples (Table 1). Sig-
nificant
differences between
the 480 (47%) patients with
positive
biopsies
and the
535 (53%) with negative biopsies
were,
for the
former:
older age, higher serum PSA levels,
lower
prostate volume,
less frequent history of negative
biopsies,
DRE more
frequently
suspicious (Table 1).
Diagnostic performance
of the urinary PCA3
test
The median
PCA3 score
was significantly higher in patients
with
positive biopsies,
i.e., 53 vs 20 (P < 0.0001) in patients
with
negative
biopsies (Table
1), and the risk of positive
biopsy
increased
with the
PCA3 score (Fig. 1). The AUC of
the
PCA3 score
was significantly higher than that of the
serum
PSA,
respectively 0.76
(95%
CI: 0.73—0.79) and 0.55
(0.51—0.58)
(Fig. 2). The cutoff of 35, usually used in the lit-
erature,
was indeed a good compromise between sensitivity
(68%)
and specificity (71%) for an accuracy of 69% (Fig. 2).
The
risk of positive biopsy (67%) in the 482 (47%) patients
with
PCA3 score ≥ 35 was more than twice that of the 533
patients
with PCA3 score < 35 (P < 0.001).
The
diagnostic performance of the PCA3 test was similar
in
patients with or without a history of negative biopsies,
Figure 1. Increasing risk of positive biopsies with the urinary PCA3
score.

e4 V. Vlaeminck-Guillem et al.
Table 1 Baseline characteristics of the 1015 patients with informative urine samples, and correlations with presence
of
cancer on biopsies. For quantitative variables, median, IQR (interquartile range), mean and standard deviation are
indicated.
Whole cohort Patients with
negative
biopsies
Patients
with
positive
biopsies
Significativity
(P)
Number of patients n = 1015 n = 535 (53%) n = 480 (47%)
Median
age
(n
= 1015)
64
years (59—69)
(mean:
64
± 7)
63
years (58—67)
(mean:
63
± 6)
66
years (61—70)
(mean:
66
± 7)
<
0.0001
Prior
negative biopsies
0
n = 825 (81%) n = 406 (76%) n = 419 (87%) <
0.001
1
n = 144 (14%) n = 95 (18%) n = 49 (10%)
2
n = 31 (3%) n = 22 (4%) n = 9 (2%)
≥
3 n = 15 (1%) n = 12 (2%) n = 3 (1%)
Previous
administration of 5˛-reductase inhibitors
Yes
n = 20 (2%) n = 11 (2%) n = 9 (2%) =
0.836
No
n = 995 (98%) n = 524 (98%) n = 471 (98%)
Digital
rectal examination
Suspicious
n = 145 (14%) n = 41 (8%) n = 104 (22%) <
0.001
Not
suspicious n = 870 (86%) n = 494 (92%) n = 376 (78%)
Median
prostate volume
(n
= 1003)
40
mL (29—54)
(mean:
44 ± 21)
45
mL (32—60)
(mean:
49 ± 23)
35
mL (26—48)
(mean:
38 ± 17)
<
0.0001
Median
number of biopsy
cores
12 (12—12)
(mean:
12 ± 1)
12
(12—12)
(mean:
12 ± 1)
12
(12—12)
(mean:
12 ± 1)
=
1.000
Total
serum PSA
Median
6.6 ng/mL
(5—9.4)
(mean
6.2 ± 4.3)
6.4
ng/mL
(4.9—9)
(mean:
7.6 ± 4.8)
6.8
ng/mL
(5.1—10)
(mean:
16.4
± 57.2)
=
0.0004
<
2.5 ng/mL n = 25 (2%) n = 17 (8%) n = 8 (2%) =
0.018
2.5—3.99
ng/mL n = 67 (7%) n = 36 (7%) n = 376 (6%)
4—9.99
ng/mL n = 697 (69%) n = 382 (71%) n = 315 (66%)
≥
10
ng/mL
n
= 226 (22%) n = 100 (19%) n = 126 (26%)
Median
urinary
PCA3
score 32
(16—66)
(mean:
57 ± 67)
20
(11—41)
(mean:
35 ± 47)
53
(28—100)
(mean:
81 ± 78)
<
0.0001
i.e., AUCs of 0.74 (0.67—0.82) and 0.76 (0.73—0.79) respec-
tively,
P = 0.643. It differed however as a function of PSA
values.
Whether
the PSA values were between 2.5 and 4,
between
4 and
10, or greater than 10 ng/mL, the diagnostic
performance
of the PCA3 test was homogenous, with AUCs
between
0.73 and 0.76. However, in the group of patients
with
PSA < 2.5 ng/mL, performance was particularly high
with
an
AUC of 0.95 (0.85—1.00).
Integration of the PCA3 score in a nomogram
Age, DRE findings (suspicious vs non suspicious), history of
negative
prostate biopsies (no history vs at least one neg-
ative
biopsy series), prostate volume, history of physical
treatment
for benign prostate hyperplasia (BPH), serum PSA
and
PCA3 score were all predictive factors of biopsy results
in
univariate analysis (Table 2). They were all independently
predictive
in multivariate analysis also, except for physical
treatment
for BPH. All the clinical and biological variables
predictive
of
biopsy results in multivariate analysis (except
the
PCA3 score) were integrated into a base predictive
model,
the AUC of which was 0.75 (0.72—0.78) (Table 2).
When
the PCA3 score was added to the base model, the
AUC
increased significantly to 0.80 (0.77—0.83) (P < 0.001)
(Table
2). To compare the net predictive benefit of the PCA3
test, a
decision curve analysis (DCA) was performed. The
model
incorporating the PCA3 score produced a predictive
benefit
greater than that of the base model (Fig. 3).
Prediction
of prostate cancer aggressiveness
In the 480 patients with positive biopsies, the median PCA3
score
did not differ significantly between the 225 (47%)
patients
with Gleason 6 score and the 255 (53%) patients
with
Gleason score ≥ 7 (Table 3). Equally, there were no dif-
ferences
according to the presence or absence of perineural
invasion
or extracapsular extension (Table 3). Conversely,
PCA3
score was significantly higher when lesions were bilat-
eral,
when ≥ 33% of cores were positive or when ≥ 10% of
prostate
tissue was tumorous (Table 3).
An
issue in PCa diagnosis is to only identify significant
cancers.
Using a cut-off of 20 for the PCA3 score to perform a
biopsy,
332 (33%) of the patients in this study would have not
undergone
the biopsy. A third of the ultimately unnecessary
biopsies
would
have been avoided, but 67 of the 480 cancers
(14%)
would have
missed. Among those 67 cancers with PCA3
score
< 20, 17 were Gleason 7 and 2 were Gleason 8. Among
the
48 unidentified Gleason 6 cases, only one had a rate of
positive
cores ≥ 33%. Among the 67 missed cancer cases,
14
had abnormal DRE and/or PSA level ≥ 10 ng/mL.

Urinary PCA3 to predict prostate cancer in a cohort of 1015 patients e5
Figure
2. Diagnostic performances of the urinary PCA3 score and
comparison
with serum PSA. AUC: area under curve ROC; 95% CI:
95%
confidence interval; Se: sensitivity; Spe: specificity; PPV: posi-
tive
predictive value; NPV: negative predictive value; Acc: accuracy
(proportion
of correctly-classified patients).
0.00 0.10 0.20 0.30 0.40 0.50
0 20 40 60 80
Seuil de probabilité de biopsies positives en %
suotnucua
Net benefit
Probabili ty thresho ld for posi ve biopsies (%)
none
Base model
all
Base
model + ¨P CA3 score
Figure 3. Evaluation of the ability of the PCA3 score to predict
positive
biopsies (net benefit) using decision curve analysis (DCA);
base
model: age, digital rectal examination, previous negative biop-
sies,
prostate volume, serum PSA. DCA examined the theoretical
relationship
between the threshold probability of the outcome of
positive
prostate biopsies and the relative value of false-positive
and
false-negative results. Here, we estimated the magnitude of
benefit
resulting from altering clinical management in patients with
different
threshold probabilities of positive biopsies.
Table
2 Multivariate analyses evaluating the ability of clinicobiological variables to predict prostate biopsy outcome
and
comparison with the addition of urinary PCA3 score (n = 1015 patients).
Univariate analysis Multivariate analysis
Base model Base
model
+
PCA3 score
OR
(IC
95%)
P
AUC OR
(IC
95%)
P
OR
(IC
95%)
P
Age 1.06
(1.04—1.08)
<
0.001 61.6%
(58—65)
1.07
(1.05—1.10)
< 0.001 1.05
(1.02—1.07)
< 0.001
Previous
biopsies 0.46
(0.33—0.64)
< 0.001 55.7%
(53—58)
0.45
(0.30—0.67)
< 0.001 0.48
(0.31—0.75)
0.001
Prostate
volume 0.97
(0.96—0.98)
< 0.001 65.5%
(62—69)
0.97
(0.96—0.97)
< 0.001 0.97
(0.96—0.97)
< 0.001
Digital
rectal
examination
3.33
(2.27—4.90)
< 0.001 57.0%
(55—59)
2.15
(1.39—3.34)
0.001 2.03
(1.31—3.15)
0.002
Serum
PSA 1.04
(1.02—1.06)
< 0.001 54.9%
(51—58)
1.07
(1.02—1.12)
0.005 1.07
(1.02—1.12)
0.006
PCA3
score 1.02
(1.01—1.02)
< 0.001 75.8%
(73—79)
— — 1.01
(1.01—1.02)
< 0.001
Prediction
accuracya
74.6%
(72—78)
80.1%
(77—82)
Increase
in
performances
+5.5%
P
< 0.001
OR: odds ratio.
aEstimation
using the AUC (%).