Genitourinar y Imaging Original Research

Lim et al.
PI-RADSv2 Categories and Pathologic Outcomes in Patients
With GS 3+ 4= 7 Tumors

Genitourinary Imaging
Original Research Prostate Imaging Reporting and
Data System, Version 2, Assessment
Categories and Pathologic
Downloaded from www.ajronline.org by 114.121.238.76 on 05/02/17 from IP address 114.121.238.76. Copyright ARRS. For personal use only; all rights reserved

Outcomes in Patients With Gleason

Score 3+ 4= 7 Prostate Cancer
Diagnosed at Biopsy
Christopher S. Lim1 OBJECTIVE. The purpose of this study is to assess associations between Prostate Im-
Matthew D. F. McInnes1 aging Reporting and Data System, version 2 (PI-RADSv2), categories and the presence of a
Trevor A. Flood2 tumor with a Gleason score (GS) of 4+ 3= 7 or greater or the presence of extraprostatic ex-
Rodney H. Breau 3 tension (EPE) at radical prostatectomy (RP) in patients with a GS 3+ 4= 7 tumor at biopsy.
Christopher Morash 3 MATERIALS AND METHODS. A total of 81 men with GS 3+ 4= 7 prostate can-
cer diagnosed by transrectal ultrasoundguided biopsy underwent multiparametric MRI and
Rebecca E. Thornhill1
RP between 2012 and 2015. Two blinded radiologists assessed multiparametric MR images
Nicola Schieda1 and assigned PI-RADSv2 assessment categories (categories 15) with the use of sector maps,
Lim CS, McInnes MDF, Flood TA, et al. which were compared with regard to the location of the tumor, the GS, and the presence of
EPE at RP. Comparisons were performed between groups with the use of chi-square and mul-
tivariate analysis. Diagnostic accuracy was assessed using ROC curve analysis, and localiza-
tion was compared using the Fisher exact test.
RESULTS. A total of 53.1% of men (43/81) had EPE, and 21.0% (17/81) had GS 4+ 3=
7 prostate cancer after RP, whereas 2.5% of men (2/81) had their tumors downgraded to GS
3+ 3= 6. No statistically significant difference in patient age, prostate specific antigen level,
or clinical stage existed between groups (p> 0.05). PI-RADSv2 assessment categories were
significantly higher for GS 4+ 3= 7 tumors (p= 0.03). PI-RADSv2 showed moderate accura-
cy for the diagnosis of GS 4+ 3= 7 tumors (AUC, 0.65; 95% CI, 0.540.77), with a category
of 4 or higher having a sensitivity and specificity for diagnosis of 94.1% and 23.4%, respec-
tively. No patient with a PI-RADSv2 category lower than 3 had a GS 4+ 3= 7 tumor. Accura-
cy of tumor localization ranged from 86.4% to 92.6%, with 88.2% of errors (15/17) occurring
in GS 3+ 3= 6 or GS 3+ 4= 7 tumors (p= 0.30). PI-RADSv2 categories were noted to be
higher when EPE was present (p< 0.001). Interobserver agreement was moderate (= 0.43).
CONCLUSION. For GS 3+ 4= 7 cancers detected at transrectal ultrasoundguided bi-
Keywords: active surveillance, Gleason score, opsy, higher PI-RADSv2 assessment categories are associated with upgrading to GS 4+ 3= 7
multiparametric MRI, prostate cancer, stage cancer and with the presence of EPE after RP. A PI-RADSv2 score of 3 or higher was 100%
sensitive for diagnosing GS 4+ 3= 7 tumors.
DOI:10.2214/AJR.16.16843

rostate cancer is a biologically curate assessment of tumor aggressiveness.

Received May 31, 2016; accepted after revision
November 10, 2016.
1
Department of Medical Imaging, The Ottawa Hospital,
The University of Ottawa, 1053 Carling Ave, Ottawa, ON
K1Y 4E9, Canada. Address correspondence to N. Schieda
(nschieda@toh.on.ca).
2
Department of Anatomical Pathology, The Ottawa
Hospital, The University of Ottawa, Ottawa, ON, Canada.
3
Department of Urology, The Ottawa Hospital,
TheUniversity of Ottawa, Ottawa, ON, Canada.
AJR 2017; 208:10371044
0361803X/17/20851037
American Roentgen Ray Society
P
diverse disease with varying ag-
gressiveness. Management deci-
sions are based on numerous in-
dividual patient factors; however, among
them, the Gleason score (GS) of the tumor
detected at transrectal ultrasound (TRUS)
guided biopsy is critical [1, 2]. For tumors
with a GS of 7 or higher, management typi-
cally involves radical prostatectomy (RP) or
radiotherapy [1, 2]. Low-volume GS 3+ 3= 6
tumors may be managed with active surveil-
lance [3, 4], and it has been proposed that se-
lected low-volume GS 3+ 4= 7 tumors may
also be managed with active surveillance [4
7]; however, this strategy is dependent on ac-
Active surveillance protocols that consider
patients with GS 3+ 4= 7 tumors may be
limited because, for a proportion of these tu-
mors, the GS may be underestimated by
TRUS-guided biopsy results [811], and clin-
ical staging of disease with the use of digital
rectal examination and clinical nomograms
can be imprecise [12, 13]. This may result in
the inappropriate surveillance of an appar-
ently low-volume GS 3+ 4= 7 cancer when
the patient harbors more aggressive disease.
In addition, although less common, GS 3+
4= 7 prostate cancers may also be down-
graded to GS 3+ 3= 6 tumors after RP [9, 14,
15], potentially resulting in overtreatment of

AJR:208, May 2017 1037

 Lim et al.

disease. Furthermore, there is increasing evi-
dence that intermediate-risk GS 7 tumors
have a different biologic aggressiveness and
long-term prognosis, and it has been pro-
posed that GS 3+ 4= 7 and 4+ 3= 7 tumors
should be considered separately, rather than
Downloaded from www.ajronline.org by 114.121.238.76 on 05/02/17 from IP address 114.121.238.76. Copyright ARRS. For personal use only; all rights reserved
ing of prostate multiparametric MRI, to sim-
plify the reporting of prostate multipara-
metric MRI [24, 25]. PI-RADS, version 2
(PI-RADSv2), provides interpretation guide-
lines for peripheral zone (PZ) and transition
zone (TZ) tumors and five assessment cate-
the full multiparametric MRI technique was not
performed (n= 2), severe artifact from hip prosthe-
sis or motion was noted (n= 3), or the patient had
a history of androgen deprivation therapy (n= 1).
The final study group consisted of 81 patients. Data
on patient age, prostate serum antigen (PSA) lev-

grouped together [16].
Multiparametric MRI (the combination of
high-resolution T2-weighted MRI with func-
tional imaging [DWI and dynamic contrast-
enhanced MRI]) has emerged as an accurate
imaging test for the detection of clinical-
ly significant (defined by a GS 7) prostate
cancers [17, 18], although it is limited for the
detection of low-volume GS 3+ 3= 6 tumors
[19, 20]. Multiparametric MRI may also be
limited in the detection of certain low-vol-
ume GS 3+ 4= 7 tumors composed mainly
of Gleason pattern 3, which are difficult to
detect on MRI [2123]. The inability of mul-
tiparametric MRI to depict certain low vol-
ume GS 3 + 4 = 7 cancers implies that, when
nonradical treatment is considered for pa-
tients with GS 3+ 4= 7 tumors diagnosed at
TRUS-guided biopsy, multiparametric MRI
may have a role in potentially discriminating
tumors on the basis of their relative propor-
tion of Gleason pattern 4.
In 2015, the European Society of Urogeni-
tal Radiology and the American College of
Radiology issued a revised Prostate Imag-
ing Reporting and Data System (PI-RADS)
document for the interpretation and report-
gories indicating the probability of clinical-
ly significant cancer [24, 25]. The probabil-
ity scale in PI-RADSv2 recently has been
shown to be inversely associated with his-
topathologic downgrading of GS 3+ 4= 7
cancers after RP and to be correlated with
the risk of biochemical recurrence [26, 27].
The purpose of this study is to assess asso-
ciations between PI-RADSv2 categories and
the presence of tumors with a GS of 4+ 3=
7 or greater or the presence of extraprostatic
extension (EPE) at RP in patients with GS
3+ 4= 7 tumor at biopsy.
Materials and Methods
This retrospective study was approved by the lo-
cal research ethics board, which waived the need
for informed consent from all patients. We searched
the pathology and imaging databases at The Ottawa
Hospital for all patients with GS 3+ 4= 7 cancer di-
agnosed at systematic TRUS-guided needle biopsy
who underwent RP and preoperative multiparamet-
ric 3-T MRI of the prostate between January 2012
and May 2015. A total of 87 patients were identi-
fied. All patients had undergone biopsy that con-
firmed a diagnosis of GS 3 + 4 = 7 cancer at the
time of MRI. Six patients were excluded because
el, clinical stage determined from digital rectal ex-
amination, GS, and pathologic stage after RP were
retrieved from patient electronic medical records.
MRI Technique
The multiparametric MRI (T2-weighted MRI
plus DWI plus dynamic contrast-enhanced MRI)
technique used for prostate cancer imaging at
our institution is presented in Table 1. Endorec-
tal coil was not used for any patient. A total of
54.3% of patients (44/81) underwent imaging per-
formed using one clinical 3-T MRI scanner (Mag-
netom Trio Tim, Siemens Healthcare), and 45.7%
(37/81) underwent imaging performed with a sec-
ond clinical 3-T MRI scanner (Discovery 750
W, GE Healthcare). Because of the retrospective
nature of the study, high-b-value DWI was per-
formed at b= 10001500 mm2/s because acquisi-
tion at a b value of 1400 mm2/s or higher was not
instituted until January 2015, after it was formally
suggested in the PI-RADSv2 guidelines published
at that time [24, 25]. A total of 85.2% of patients
(69/81) underwent DWI at up to b= 1000 mm2/s
(because imaging was performed before January
2015) and 14.8% patients (12/81) underwent DWI
at up to b= 1500 mm2/s (because imaging was
performed on or after January 2015).

TABLE 1: Sequence Parameters for Protocol for Multiparametric MRI of the Prostate Performed With a Pelvic
Surface Coil on 3-T Systems
Slice Receiver
Imaging FOV Thickness/ Echo-Train Flip Acceleration Bandwidth Acquisition No.of
Sequence Plane (mm) Matrix Size Gap (mm) TR/TE Length Angle() Factor (Hz/voxel) Time (min) Excitations
T1-weighted TSE Axial 350 350 320 320 5.0/1.0 720/814 3 111 0 244 4 2
T1-weighted 3D Axial 240 240 292 224 4.0/1.0 4.8/1.11.3a NA 12 2 558 Breathhold 1
dual-echo GRE
T2-weighted TSE Coronal 220 220 320 256 4.0/0 38905250/ 2735 111 0 122 4 12
105125
Sagittal 3.0/0 4
Axial 3.0/0 4
DWIb Axial 280 280 128 80 3.05.0/0 4200/90 1 90 2 1950 5 410
T1-weighted GRE Axial 220 220 128 128 4.0/0 4.3/1.3 NA 12 2 488 6 1
dynamic
contrast-en-
hancedc
NoteClinical MRI examinations were performed using two 3-T MRI systems (Magnetom Trio Tim, Siemens Healthcare; and Discovery 750 W, GE Healthcare) with
four- to 16-channel integrated pelvic surface coils with eight- to 12-channel activated spine coils. TSE= turbo spin echo, GRE= gradient-recalled echo.
aTE1/TE2, TE1 = 1.11.3, and TE2 = 2.22.5.
bDWI was performed with spectral fat-suppression echo-planar imaging with tridirectional motion-probing gradients and b values of 0, 500, and 1000 mm2 /s with

automatic apparent diffusion coefficient map generation.

cTwo-dimensional dynamic fast spoiled GRE MRI performed with temporal resolution of 10 s after administration of gadobutrol (0.1 mmol/kg, Gadovist, Bayer HealthCare)

injected at a rate of 3 mL/s.

1038 AJR:208, May 2017

 PI-RADSv2 Categories and Pathologic Outcomes in Patients With GS 3+ 4= 7 Tumors

Histopathologic Findings
The standard nontargeted extended-template
TRUS-guided biopsy system used at our institution
uses an 18-gauge side-cutting needle and yields 12
core biopsy specimens. Prostate biopsy specimens
were obtained from 10 anatomic locations in the
Downloaded from www.ajronline.org by 114.121.238.76 on 05/02/17 from IP address 114.121.238.76. Copyright ARRS. For personal use only; all rights reserved
tumor focus within the same gland, and none of
these less than 0.5-mL tumor foci had EPE when
the dominant tumor focus was confined to the or-
gan. A total of 13.6% of the dominant tumor foci
(11/81) were located in the TZ, and 86.4% (70/81)
were located in the PZ. To create RPmultipara-
is highly unlikely to be present); 2, a low prob-
ability (clinically significant cancer is unlikely to
be present); 3, an indeterminate probability (the
presence of clinically significant cancer is equiv-
ocal); 4, a high probability (clinically significant
cancer is likely to be present); and 5, a very high

prostate (i.e., the right and left apex, right and left
midlateral and midmedial, and right and left base
lateral and base medial). All biopsies were per-
formed by experienced fellowship-trained radiol-
ogists at a tertiary care referral center for prostate
cancer. Tissue was fixed overnight in 10% neutral
buffered formalin. Three histologic slides were pre-
pared from each block, each with three serial sec-
tions cut to a thickness of 3 m and stained with H
and E. RP specimens were fixed in 40% buffered
formalin and then were serially sectioned into hori-
zontal sections with a thickness of 0.30.4 cm, with
the use of a step-by-step approach. All tissues were
paraffin embedded, and 4--thick sections were
cut and stained with H and E. All specimens from
TRUS-guided needle biopsy and RP were reviewed
by fellowship-trained genitourinary pathologists,
and the GS (from biopsy and RP), pathologic stage,
and overall percentage of Gleason pattern 3 for each
tumor (from RP) were recorded. A dedicated gen-
itourinary pathologist with 12 years of experience
reviewed the RP results and identified and mapped
dominant tumor foci with a volume measuring more
than 0.5 mL [19, 2831]. Tumors in the PZ and TZ
were both considered.
With the use of these criteria, only one domi-
nant tumor focus per patient was identified. None
of the tumors with a volume measuring less than
0.5 mL had a GS higher than that of the dominant
metric MRI histopathologic maps, one radiolo-
gist coregistered tumor foci identified on RP maps
to MRI approximately 6 months after the initial
blinded interpretation session. For tumors that
were not considered to be well visualized with
MRI, anatomic landmarks including the urethra,
the central zone of the prostate near its base, and
benign prostatic hyperplasia nodules were used to
localize tumors.
Visual Analysis
Two fellowship-trained abdominal radiologists
who had experience using the PI-RADSv2 scor-
ing system (both radiologists had used the updat-
ed PI-RADS classification since 2015 and had in-
terpreted between 150 and 200 examinations with
use of the PI-RADSv2 system) and who had 11
and 16 years of experience in genitourinary imag-
ing evaluated each MRI examination. They were
blinded to the final histopathologic results and the
location of the dominant tumor foci at RP; howev-
er, the radiologists were aware that all patients un-
derwent RP. All pulse sequences were available at
the time the imaging review was performed (in-
cluding T2-weighted MRI, DWI, and dynamic
contrast-enhanced MRI), and assessment of the
studies was performed according to PI-RADSv2
[24, 25]. For each patient, the radiologist assessed
each study for dominant tumor foci in the PZ and
TZ, with use of the PI-RADS sector map, and as-
signed an overall PI-RADSv2 category for the
probability of prostate cancer, with 1 denoting a
very low probability (clinically significant cancer
probability (clinically significant cancer is highly
likely to be present) (Figs. 1 and 2). Each exami-
nation was reviewed on a standard PACS worksta-
tion (Horizon Medical Imaging, McKesson) with
the use of a commercially available third-party
postprocessing software program (AquariusNet,
Tera-Recon).
Statistical Analysis
For statistical comparisons, patients were strati-
fied into two groups. The first group included pa-
tients with tumors for which the GS remained 3+
4= 7 or was downgraded to GS 3+ 3= 6 and com-
pared to patients with tumors with a GS of 4+ 3=
7 or higher after RP. The second group included
patients with organ-confined disease and com-
pared to patients with EPE. Patient age, the PSA
level, and PI-RADSv2 categories were compared
between groups with the use of chi-square and
multivariate analyses. The diagnostic accuracy
of PI-RADS for predicting disease with a GS of
4+ 3 or higher and the presence of EPE was as-
sessed using ROC analysis. Correlation between
PI-RADSv2 categories and the percentage of Glea-
son pattern 3 for each tumor was assessed using
Pearson correlation. The accuracy of dominant tu-
mor localization was compared between tumors
with a GS of 3+ 4= 7 or greater and those with a
GS of 3+ 4= 7 or less, with use of the Fisher ex-
act test. Ordinal data are reported as mean (SD)
values. Interobserver agreement was assessed us-
ing the Cohen kappa statistic. A p 0.05 was con-
sidered to denote a statistically significant result.

A B C
Fig. 157-year-old man with Gleason score 3+ 4= 7 prostate cancer involving three of 12 core biopsy specimens. Prostate serum antigen level was 6.2 ng/mL, and
clinical stage was T2a.
A, Axial stained whole-mount histopathologic image (H and E, 2) obtained at low microscopic power through mid and apical prostate shows two areas of tumor (black-
outlined area) in left peripheral zone (PZ) and transition zone (TZ). For purposes of this study, only tumors with volume greater than 0.5 mL were considered to have
potential clinical statistical significance; therefore, PZ tumor was dominant tumor focus in patient because TZ tumor volume measured less than 0.5 mL. Dominant tumor
was 80% Gleason pattern 3.
B and C, Axial fast spin-echo T2-weighted MR image (B) and apparent diffusion coefficient map (C) show low T2 signal (arrow, B) in left PZ but no corresponding
restricted diffusion (arrow, C) on apparent diffusion coefficient map. TZ tumor also is not visible. Neither radiologist correctly identified dominant tumor in this patient;
accordingly, both radiologists assigned Prostate Imaging Reporting and Data System, version 2, assessment category of 2.

AJR:208, May 2017 1039

 Lim et al.
Downloaded from www.ajronline.org by 114.121.238.76 on 05/02/17 from IP address 114.121.238.76. Copyright ARRS. For personal use only; all rights reserved

A B C
Fig. 271-year-old man with Gleason score 3+ 4= 7 prostate cancer involving five of 12 core biopsy specimens. Prostate serum antigen level was 10.7 ng/mL, and clinical
stage was T1c.
A, Axial T2-weighted image shows well-defined ovoid nodule in left anterior apical horn of prostate (arrow).
B and C, Axial apparent diffusion coefficient map (B) and trace echo-planar image (C) (b value, 1000 mm2 /s) show ill-defined lenticular-shaped lesion (arrow, B and C) with
very low signal intensity on apparent diffusion coefficient map (arrow, B) and marked signal hyperintensity on echo-planar image (arrow, C). Lesion measured 18 mm and
was graded by both radiologists as Prostate Imaging Reporting and Data System, version 2, category 5. Histopathologic findings after radical prostatectomy confirmed
dominant tumor focus in left anterior peripheral zone with Gleason score of 4+ 3= 7 and pathologic stage of T2.

Statistical analysis was performed using statistical in age (mean, 64.5 6.0 vs 61.9 6.3 years, No statistically significant difference was
software (Stata, version 13.0, StataCorp). respectively; p= 0.15) or PSA level (mean, noted in the distribution of disease according
7.4 3.1 vs 6.2 3.0 ng/mL, respectively; to GS (p= 0.68) or the presence of EPE (p=
Results p= 0.16) was noted. When patients with EPE 0.87), when patients who underwent imaging
A total of 21.0% of patients (17/81) had a were compared with those with organ-con- on either MRI scanner were compared.
GS 4+ 3= 7 tumor after RP. For 2.5% of pa- fined disease, no difference in age (mean age, Table 2 summarizes the distribution of
tients (2/81), the tumor was downgraded to 63.2 5.7 vs 61.6 6.9 years, respectively; PI-RADSv2 categories compared by GS and
GS 3+ 3= 6, and for the remaining 76.5% of p= 0.44) or PSA level was noted (7.2 3.1 vs the presence of EPE. Both radiologists noted
patients (62/81), the tumor GS remained 3+ 5.9 2.7 ng/mL; respectively; p= 0.09). With that PI-RADSv2 categories were higher for
4= 7. A total of 46.9% of patients (38/81) had regard to tumor category, a total of 58.0% patients with GS 4+ 3= 7 cancer (p= 0.03).
organ-confined disease, and 53.1% (43/81) of patients (47/81) had clinical category T1, The AUC value for the diagnosis of upgrad-
had EPE. 38.3% (31/81) had clinical category T2, and ing the GS with the use of PI-RADSv2 was
When patients with a GS 4+ 3= 7 tu- only 3.7% (3/81) had clinical category T3, 0.65 (95% CI, 0.540.77), and the sensitivi-
mor were compared with those with a GS with no difference noted compared by GS ty and specificity associated with the use of
3+ 4= 7 or 3+ 3= 6 tumor, no difference (p= 0.21) or the presence of EPE (p= 0.41). various PI-RADSv2 thresholds are present-
TABLE 2: Distribution of Prostate Imaging Reporting and Data System, Version 2 (PI-RADSv2), Assessment
Categories Compared With Pathologic Outcomes for 81 Tumors After Radical Prostatectomy
Gleason Score of Tumora,b Pathologic Stage of Tumorc
PI-RADSv2 3+ 4= 7 or 3+ 3= 6a 4+ 3= 7 Organ-Confined Disease Extraprostatic Extension
Assessment
Category Radiologist 1 Radiologist 2 Radiologist 1 Radiologist 2 Radiologist 1 Radiologist 2 Radiologist 1 Radiologist 2
1 4 (6.3) 3 (4.7) 0 (0) 0 (0) 3 (7.9) 3 (7.9) 1 (2.3) 0 (0)
2 2 (3.1) 4 (6.3) 0 (0) 0 (0) 2 (5.3) 2 (5.3) 0 (0) 2 (4.7)
3 9 (14.1) 16 (25.0) 1 (5.9) 3 (17.6) 6 (15.8) 9 (23.7) 4 (9.3) 10 (23.3)
4 21 (32.8) 23 (35.9) 4 (23.6) 4 (23.5) 17 (44.7) 17 (44.7) 8 (18.6) 10 (23.3)
5 28 (43.8) 18 (28.1) 12 (70.6) 10 (58.8) 10 (26.3) 7 (18.4) 30 (70.0) 21 (48.8)
All 64 (79.0) 17 (21.0) 38 (46.9) 43 (53.1)
NoteData are number (%) of tumors. PI-RADSv2 assessment categories were defined as follows: a score of 1 denoted a very low probability of prostate cancer
(clinically significant cancer is highly unlikely to be present); 2, low probability (clinically significant cancer is unlikely to be present); 3, indeterminate probability (the
presence of clinically significant cancer is equivocal); 4, high probability (clinically significant cancer is likely to be present); and 5, very high probability (clinically
significant cancer is highly likely to be present) [24, 25].
aGleason score 3+ 4= 7 (n= 62) and Gleason score 3+ 3= 6 (n= 2) were considered together because of the low number of patients whose tumor scores were downgrad-

ed, which prevented any meaningful subgroup analysis.

bThere was a statistically significant association between PI-RADSv2 categories and the presence of GS 4 + 3 =7 cancer after radical prostatectomy as determined using

the chi-square test (p = 0.03).

cThere was a statistically significant association between PI-RADSv2 categories and the presence of extraprostatic extension of tumor as determined using the

chi-square test (p < 0.01).

1040 AJR:208, May 2017

 PI-RADSv2 Categories and Pathologic Outcomes in Patients With GS 3+ 4= 7 Tumors

ed in Table 3. PI-RADSv2 was highly sensi-
tive for GS upgrading, and no patient with a
GS 4+ 3= 7 tumor had a PI-RADSv2 score
of less than 3 assigned by either radiologist.
The mean percentage of Gleason pattern 3 for
each tumor was 65.2% 21.2% (range, 10
Downloaded from www.ajronline.org by 114.121.238.76 on 05/02/17 from IP address 114.121.238.76. Copyright ARRS. For personal use only; all rights reserved
TABLE 3: Diagnostic Accuracy of Prostate Imaging Reporting and Data
System, Version 2 (PI-RADSv2), Categories for Diagnosis of Cancer
With Gleason Score (GS) of 4+ 3= 7 or Higher and the Presence of
Extraprostatic Extension Among GS 3+ 4= 7 Cancers Diagnosed at
Transrectal UltrasoundGuided Biopsy
Finding GS 4+ 3= 7 or Highera Extraprostatic Extensionb

100%). A weak negative correlation existed

between the percentage of Gleason pattern PI-RADSv2 category Sensitivity Specificity Sensitivity Specificity
3 within each tumor and PI-RADSv2 cat- 1 100.0 0.0 100 0.0
egories ( = 0.18; p= 0.02). Interobserver 2 100.0 6.3 97.7 7.9
agreement between the two radiologists for
3 100.0 9.4 97.7 13.2
PI-RADSv2 categorization was moderate
(= 0.43). 4 94.1 23.4 88.4 28.9
PI-RADSv2 categories were statistical- 5 70.6 56.3 69.8 73.7
ly significantly higher when EPE was pres- AUC (95% CI) 0.65 (0.540.77) 0.72 (0.620.87)
ent (p< 0.001). The AUC value for the diag- aTumor was upgraded to GS 4+ 3= 7 in 21.0% of patients (17/81).
nosis of EPE was 0.72 (95% CI, 0.620.87), bPatients were stratified into two groups: those with organ-confined disease and those with extraprostatic

and the sensitivity and specificity of various extension.

thresholds of PI-RADSv2 categories are pre-
sented in Table 3.
Radiologist 1 correctly identified the
dominant tumor location in 92.6% of cas-
es (75/81), and in 7.4% of patients (6/81),
no dominant tumor was detected. In no pa-
tient was a dominant tumor identified but as-
signed to the wrong location. Radiologist 2
correctly identified the dominant tumor loca-
tion in 86.4% of cases (70/81). In the remain-
ing 13.6% of cases (11/81), either no domi-
nant tumor was identified (8.6% [7/81]) or the
dominant tumor was assigned to the wrong
location in 4.9% (4/81). Errors occurred
mainly in patients with GS 3+ 3= 6 or 3+ finding compares favorably with the findings
4= 7 cancers, with 88.2% of errors (15/17)
occurring in association with these tumors
(p= 0.30) (Fig. 1).
Discussion
The present study evaluated the utility of
prostate MRI, with the use of PI-RADSv2,
for patients with GS 3+ 4= 7 prostate cancer
diagnosed at TRUS-guided core needle bi-
opsy. Our results show that PI-RADSv2 cat-
egories are higher for patients with GS up-
grading after RP. In this study, a PI-RADSv2
assessment category of 3 or higher was high-
ly sensitive, and category 1 or 2 excluded
GS 4+ 3= 7 cancers. A strong association
was also noted between the PI-RADSv2 cat-
egory and the presence of EPE of the tumor.
These results are important because they
suggest that there is a role for MRI in further
risk stratification of the diverse group of pa-
tients with GS 3+ 4= 7 tumor diagnosed at
TRUS-guided needle biopsy. The ability to
more accurately predict disease aggressivity
in this patient population is needed and may
be clinically useful because, in a proportion
of patients with GS 3+ 4= 7 cancer diag-
nosed at biopsy, disease aggressiveness may
be incorrectly assigned on the basis of biopsy
results and clinical staging. Moreover, accu-
rate disease classification is required if active
surveillance is contemplated in low-volume
GS 3+ 4= 7 tumors.
In the present study, PI-RADSv2 catego-
ries were significantly higher for patients who
had histopathologic upgrading of tumor after
RP. For approximately one-fifth of patients
with GS 3+ 4= 7 cancer, upgrading to GS
4+ 3= 7 cancer occurred after RP, and this
of Epstein et al. [32], who found that 21.1% of
patients with a GS 3+ 4= 7 tumor at biopsy
(333/1577) eventually had that GS upgraded.
PI-RADSv2 was highly sensitive for the diag-
nosis of GS upgrading, and no patients with
categories of 1 or 2 had their GS upgraded
after RP. These results suggest that, in a pa-
tient with GS 3+ 4= 7 tumor diagnosed at
TRUS-guided biopsy, a negative MRI find-
ing effectively excludes the possibility of GS
upgrading after RP. Although the specificity
of the diagnosis of upgrading of disease with
the use of PI-RADSv2 was low, it could be
argued that the desired outcome of MRI of
patients with GS 3+ 4= 7 prostate cancer di-
agnosed at TRUS-guided biopsy is a highly
sensitive test result that can exclude the pos-
sibility of more aggressive disease when non-
radical therapy is considered.
Previous studies have shown promising
preliminary results of the use of quantita-
tive analysis of multiparametric MRI for dis-
crimination between GS 3+ 4= 7 and GS
4+ 3= 7 intermediate-risk tumors [33, 34];
however, these studies applied advanced
quantitative MRI texture features, which
may be difficult to apply in routine clinical
practice and which require validation on a
larger scale. With the use of subjective anal-
ysis, it has previously been shown that Likert
probability scales or the PI-RADS sum score
correlates with the aggressiveness of the tu-
mor [35, 36]; however, these studies did not
attempt to evaluate intermediate-risk GS 3+
4= 7 and GS 4+ 3= 7 tumors separately but
generally considered them as a single group
when comparing them to low-risk (GS 3+
3= 6) and high-risk (GS 8) tumors.
Regarding histopathologic downgrading
of GS 3+ 4= 7 cancers, a previous study
by Gondo et al. [14] showed that multipara-
metric MRI was able to predict pathologic
downgrading with good degrees of accuracy.
More recently, Woo et al. [26] showed that
PI-RADSv2 categories were associated with
downgrading of GS 3+ 4= 7 cancers de-
tected on biopsy after RP. In our study, only
2.5% of tumors were downgraded to GS 3+
3= 6 after RP, and, therefore, we were not
able to reevaluate the results by Gondo and
colleagues or Woo and colleagues; however,
our findings support the conclusions by Woo
and colleagues by showing a similar use of
the PI-RADSv2 assessment categories in
that higher categories were associated with
GS upgrading of the tumor and the presence
of EPE after RP.
In the present study, interobserver agree-
ment for PI-RADSv2 categorization was
moderate. This compares favorably to
prior studies that evaluated the original
PI-RADS system [37] and, more recently,
the PI-RADSv2 system [38], but it is lower
than the agreement reported in a larger mul-

AJR:208, May 2017 1041


ticenter study by Rosenkrantz et al. [39] in
which the reproducibility of PI-RADSv2
was evaluated. The lower agreement report-
ed in our study, compared with the study by
Rosenkrantz and colleagues, most likely is
associated with our patient population, which
Downloaded from www.ajronline.org by 114.121.238.76 on 05/02/17 from IP address 114.121.238.76. Copyright ARRS. For personal use only; all rights reserved
consisted almost exclusively of patients with
intermediate-risk cancers and primarily of
GS 3+ 4= 7 tumors.
The rate of dominant tumor localization
noted with the use of PI-RADSv2 in the pres-
ent study is concordant with the rate noted
in a previous study that evaluated the accu-
racy of PI-RADS for dominant tumor local-
ization (range, 8892%) [40]. Most errors in
localization (either because of nonvisualiza-
tion or incorrect localization) occurred in pa-
tients with GS 3+ 3= 6 or 3+ 4= 7 tumors, overall percentage of Gleason pattern 4, and
and a negative correlation existed between
PI-RADSv2 categories and the percentage of
Gleason pattern 3 for each tumor. These re-
sults are also concordant with those of previ-
ous studies that have shown that the use of
multiparametric MRI to identify tumors with
cancers mainly composed of Gleason pattern
3 may be limited [2123] and that findings on
multiparametric MRI (e.g., lower apparent
diffusion coefficient values) are more pro-
nounced in tumors with a higher GS [41]. We
did not assess other factors, such as the den-
sity and distribution of tumor cells and their
potential impact on the visibility of interme-
diate-risk tumors, which may be a topic for
future research [42].
This study showed significantly higher
PI-RADSv2 categories for patients with EPE
of a tumor. Staging of prostate cancer with
MRI with conventional T2-weighted imag-
ing findings of tumor spread is limited, with
a recent large meta-analysis showing only a
modest degree of accuracy [43], inconsistent
results, and repeatedly reported poor interob-
server agreement [4447]. The role of MRI
in assessing other indirect features that may
predict pathologic stage has been the top-
ic of recent studies showing benefit in mea-
suring tumor volume [44, 48] and the length
of capsular abutment [48, 49] and in deter-
mining apparent diffusion coefficient values
[34, 44, 50] (which are quantitative mark-
ers of GS). Our results are concordant with
previous studies by Kayat Bittencourt et al.
[51] and Schieda et al. [52], who showed that
PI-RADS scoring was predictive of EPE.
Further studies evaluating the use of the
PI-RADSv2 system for prediction of patho-
logic stage in larger patient populations, in-
cluding varying GS tumors, is required.
1042 AJR:208, May 2017
Lim et al.
This study has limitations. The single-
institution retrospective nature of the study
introduces bias into our results, which may
limit their application to other populations.
The sample size was relatively small, and our
results require validation on a larger scale.
Only two of the patients had histopatholog-
ic downgrading of the GS after RP, and we
were not able to validate the previously re-
ported findings of Gondo et al. [14] and Woo
et al. [26] indicating that multiparametric
MRI is valuable for predicting downgrading
of GS 3+ 4= 7 tumors. We did not eval-
uate additional histopathologic parameters
that can be detected at TRUS-guided nee-
dle biopsy, including the number of core bi-
opsy specimens with Gleason pattern 4, the
logic parameters, which may further improve
the subtypes of pattern 4, because these have
been previously shown to be associated with
histopathologic upgrading [53]. All patients
in the present study underwent RP (which
was necessary to extract the histopathologic
outcomes that we used for analysis), which
may have further biased our results because
patients who had GS 3+ 4= 7 cancer diag-
nosed at biopsy and who were treated with
active surveillance would have been exclud-
ed from the present study. Moreover, because
the study population consisted of patients
who underwent RP, most patients would be
expected to have PI-RADSv2 categories of 4
or 5 (given the high negative predictive val-
ue of multiparametric MRI for the detection
of clinically significant cancers [17, 18]). The
inclusion of consecutive patients who were
undergoing multiparametric MRI for indica-
tions such as an increased PSA level or ab-
normal digital rectal examination findings
for biopsy-naive men, patients with previ-
ously negative biopsy results, and men being
considered for, or managed with, active sur-
veillance presumably would have yielded in-
creased numbers of cases with PI-RADSv2
assessment categories 13, perhaps enrich-
ing the results of our study. Nevertheless, the
need for RP to extract our histopathologic
outcomes restricted the selection of the pa-
tient population in our study.
In conclusion, the present study shows that
using multiparametric MRI of the prostate
with PI-RADSv2 has potential value for fur-
ther risk stratification of GS 3+ 4= 7 pros-
tate cancers detected at TRUS-guided nee-
dle biopsy. This is important because, for a
proportion of these cases, disease aggres-
siveness may be misclassified on the basis of
TRUS-guided needle biopsy results and clini-
cal staging. PI-RADSv2 categories were asso-
ciated with histopathologic upgrading of the
tumor. A PI-RADSv2 category of 3 or high-
er was highly sensitive for the diagnosis of
GS upgrading, and a PI-RADSv2 assessment
category of 1 or 2 excluded the possibility of
tumor upgrading after RP. PI-RADSv2 cate-
gories were also higher in patients who had
tumor EPE. The specificity of the PI-RADSv2
system to diagnose GS upgrading was low;
however, a higher sensitivity may be consid-
ered more desirable in this patient population
when nonradical management decisions are
contemplated. Further analysis is required in
a larger patient cohort to validate our findings
and potentially incorporate advanced model-
ing, including additional clinical and histo-
disease classification.
References
1. Mohler J, Bahnson RR, Boston B, et al. Clinical
practice guidelines in oncology: prostate cancer.
JNatl Compr Canc Netw 2010; 8: 162200
2. European Association of Urology. Guidelines on pros-
tate cancer. European Association of Urology website.
uroweb.org/wp-content/uploads/07_Prostate_Cancer-
pocket-version-2012.pdf. Updated 2012. Accessed
January 19, 2017
3. Thomsen FB, Brasso K, Klotz LH, Rder MA,
Berg KD, Iversen P. Active surveillance for clini-
cally localized prostate cancer: a systematic re-
view. JSurg Oncol 2014; 109:830835
4. Morash CTR, Tey R, Agbassi C, et al. Active sur-
veillance for the management of localized pros-
tate cancer: guideline recommendations. Can
Urol Assoc J 2015; 9:171178
5. El Hajj A, Ploussard G, de la Taille A, et al. Patient
selection and pathological outcomes using cur-
rently available active surveillance criteria. BJU
Int 2013; 112:471477
6. Bul M, van den Bergh RC, Zhu X, et al. Outcomes
of initially expectantly managed patients with low
or intermediate risk screen-detected localized
prostate cancer. BJU Int 2012; 110:16721677
7. Cooperberg MR, Cowan JE, Hilton JF, et al. Out-
comes of active surveillance for men with inter-
mediate-risk prostate cancer. JClin Oncol 2011;
29:228234
8. Siddiqui MM, Rais-Bahrami S, Truong H, et al.
Magnetic resonance imaging/ultrasound-fusion
biopsy significantly upgrades prostate cancer ver-
sus systematic 12-core transrectal ultrasound bi-
opsy. Eur Urol 2013; 64:713719
9. Noguchi M, Stamey TA, McNeal JE, Yemoto CM.
Relationship between systematic biopsies and his-
tological features of 222 radical prostatectomy
specimens: lack of prediction of tumor signifi-
 PI-RADSv2 Categories and Pathologic Outcomes in Patients With GS 3+ 4= 7 Tumors
cance for men with nonpalpable prostate cancer.
JUrol 2001; 166:104109; discussion, 109110
10. Kksal IT, Ozcan F, Kadioglu TC, Esen T,
Kiliaslan I, Tun M. Discrepancy between Glea-
son scores of biopsy and radical prostatectomy
specimens. Eur Urol 2000; 37:670674
Downloaded from www.ajronline.org by 114.121.238.76 on 05/02/17 from IP address 114.121.238.76. Copyright ARRS. For personal use only; all rights reserved
11. Turkbey B, Mani H, Aras O, et al. Prostate cancer:
can multiparametric MR imaging help identify
patients who are candidates for active surveil-
lance? Radiology 2013; 268:144152
12. Boyce S, Fan Y, Watson RW, Murphy TB. Evalua-
tion of prediction models for the staging of prostate
cancer. BMC Med Inform Decis Mak 2013; 13:126
13. Fanning DM, Fan Y, Fitzpatrick JM, Watson RW.
External validation of the 2007 and 2001 Partin
tables in Irish prostate cancer patients. Urol Int
2010; 84:174179
14. Gondo T, Hricak H, Sala E, et al. Multiparametric
3T MRI for the prediction of pathological down-
grading after radical prostatectomy in patients
with biopsy-proven Gleason score 3+ 4 prostate
cancer. Eur Radiol 2014; 24:31613170
15. Freedland SJ, Kane CJ, Amling CL, Aronson WJ,
Terris MK, Presti JC Jr. Upgrading and down-
grading of prostate needle biopsy specimens: risk
factors and clinical implications. Urology 2007;
69:495499
16. Epstein JI, Zelefsky MJ, Sjoberg DD, et al. A con-
temporary prostate cancer grading system: a vali-
dated alternative to the Gleason sscore. Eur Urol
2016; 69:428435
17. de Rooij M, Hamoen EH, Futterer JJ, Barentsz JO,
Rovers MM. Accuracy of multiparametric MRI
for prostate cancer detection: a meta-analysis.
AJR 2014; 202:343351
18. Hegde JV, Mulkern RV, Panych LP, et al. Multi-
parametric MRI of prostate cancer: an update on
state-of-the-art techniques and their performance
in detecting and localizing prostate cancer.
JMagn Reson Imaging 2013; 37:10351054
19. Stamatakis L, Siddiqui MM, Nix JW, et al. Accu-
racy of multiparametric magnetic resonance im-
aging in confirming eligibility for active surveil-
lance for men with prostate cancer. Cancer 2013;
119:33593366
20. Schoots IG, Petrides N, Giganti F, et al. Magnetic
resonance imaging in active surveillance of pros-
tate cancer: a systematic review. Eur Urol 2015;
67:627636
21. Kim JY, Kim SH, Kim YH, Lee HJ, Kim MJ,
Choi MS. Low-risk prostate cancer: the accuracy
of multiparametric MR imaging for detection.
Radiology 2014; 271:435444
22. Turkbey B, Shah VP, Pang Y, et al. Is apparent dif-
fusion coefficient associated with clinical risk
scores for prostate cancers that are visible on 3-T
MR images? Radiology 2011; 258:488495
23. Yerram NK, Volkin D, Turkbey B, et al. Low sus-
AJR:208, May 2017 1043
picion lesions on multiparametric magnetic reso-
nance imaging predict for the absence of high-risk
prostate cancer. BJU Int 2012; 110:E783E788
24. Weinreb JC, Barentsz JO, Choyke PL, et al.
PI-RADS Prostate Imaging - Reporting and Data
System: 2015, version 2. Eur Urol 2016; 69:1640
25. Barentsz JO, Weinreb JC, Verma S, et al. Synopsis
of the PI-RADS v2 guidelines for multiparametric
prostate magnetic resonance imaging and recom-
mendations for use. Eur Urol 2016; 69:4149
26. Woo S, Kim SY, Lee J, Kim SH, Cho JY. PI-RADS
version 2 for prediction of pathological downgrad-
ing after radical prostatectomy: a preliminary
study in patients with biopsy-proven Gleason
Score 7 (3+4) prostate cancer. Eur Radiol 2016;
26:35803587
27. Park SY, Oh YT, Jung DC, et al. Prediction of bio-
chemical recurrence after radical prostatectomy
with PI-RADS version 2 in prostate cancers: ini-
tial results. Eur Radiol 2016; 26:25022509
28. Montironi R, Hammond EH, Lin DW, et al. Con-
sensus statement with recommendations on active
surveillance inclusion criteria and definition of
progression in men with localized prostate cancer:
the critical role of the pathologist. Virchows Arch
2014; 465:623628
29. Turkbey B, Mani H, Aras O, et al. Correlation of
magnetic resonance imaging tumor volume with
histopathology. JUrol 2012; 188:11571163
30. Amin MB, Lin DW, Gore JL, et al. The critical
role of the pathologist in determining eligibility
for active surveillance as a management option in
patients with prostate cancer: consensus statement
with recommendations supported by the College
of American Pathologists, International Society
of Urological Pathology, Association of Directors
of Anatomic and Surgical Pathology, the New
Zealand Society of Pathologists, and the Prostate
Cancer Foundation. Arch Pathol Lab Med 2014;
138:13871405
31. Johnson LM, Choyke PL, Figg WD, Turkbey
B. The role of MRI in prostate cancer active surveil-
lance. Biomed Res Int 2014; 2014:203906
32. Epstein JI, Feng Z, Trock BJ, Pierorazio PM. Up-
grading and downgrading of prostate cancer from
biopsy to radical prostatectomy: incidence and
predictive factors using the modified Gleason
grading system and factoring in tertiary grades.
Eur Urol 2012; 61:10191024
33. Rosenkrantz AB, Triolo MJ, Melamed J, Rusinek
H, Taneja SS, Deng FM. Whole-lesion apparent
diffusion coefficient metrics as a marker of per-
centage Gleason 4 component within Gleason 7
prostate cancer at radical prostatectomy. JMagn
Reson Imaging 2015; 41:708714
34. Rozenberg R, Thornhill RE, Flood TA, Hakim
SW, Lim C, Schieda N. Whole-tumor quantitative
apparent diffusion coefficient histogram and tex-
ture analysis to predict Gleason score upgrading
in intermediate-risk 3+ 4= 7 prostate cancer. AJR
2016; 206:775782
35. Junker D, Schafer G, Edlinger M, et al.
Evaluation of the PI-RADS scoring system for
classifying mpMRI findings in men with suspi-
cion of prostate cancer. Biomed Res Int 2013;
2013:252939
36. Roethke MC, Kuru TH, Schultze S, et al. Evalua-
tion of the ESUR PI-RADS scoring system for
multiparametric MRI of the prostate with targeted
MR/TRUS fusion-guided biopsy at 3.0 Tesla. Eur
Radiol 2014; 24:344352
37. Reister LA, Ftterer JJ, Halvorsen OJ, et al.
1.5-T multiparametric MRI using PI-RADS: a re-
gion by region analysis to localize the index-tu-
mor of prostate cancer in patients undergoing
prostatectomy. Acta Radiol 2015; 56:500511
38. Muller BG, Shih JH, Sankineni S, et al. Prostate
cancer: interobserver agreement and accuracy
with the revised Prostate Imaging Reporting and
Data System at multiparametric MR imaging.
Radiology 2015; 277:741750
39. Rosenkrantz AB, Ginocchio LA, Cornfeld D, et
al. Interobserver reproducibility of the PI-RADS
version 2 lexicon: a multicenter study of six expe-
rienced prostate radiologists. Radiology 2016;
280:793-804
40. Rosenkrantz AB, Kim S, Lim RP, et al. Prostate
cancer localization using multiparametric MR
imaging: comparison of Prostate Imaging Report-
ing and Data System (PI-RADS) and Likert
scales. Radiology 2013; 269:482492
41. Langer DL, van der Kwast TH, Evans AJ, et al.
Prostate tissue composition and MR measure-
ments: investigating the relationships between
ADC, T2, Ktrans, ve, and corresponding histologic
features. Radiology 2010; 255:485494
42. Langer DL, van der Kwast TH, Evans AJ, et al.
Intermixed normal tissue within prostate cancer:
effect on MR imaging measurements of apparent
diffusion coefficient and T2sparse versus dense
cancers. Radiology 2008; 249:900908
43. de Rooij M, Hamoen EH, Witjes JA, Barentsz JO,
Rovers MM. Accuracy of magnetic resonance im-
aging for local staging of prostate cancer: a diag-
nostic meta-analysis. Eur Urol 2016; 70:233245
44. Lim C, Flood TA, Hakim SW, et al. Evaluation of
apparent diffusion coefficient and MR volumetry
as independent associative factors for extra-pros-
tatic extension (EPE) in prostatic carcinoma.
JMagn Reson Imaging 2016; 43:726736
45. Ruprecht O, Weisser P, Bodelle B, Ackermann H,
Vogl TJ. MRI of the prostate: interobserver agree-
ment compared with histopathologic outcome af-
ter radical prostatectomy. EurJ Radiol 2012;
81:456460
46. Engelbrecht MR, Jager GJ, Laheij RJ, Verbeek AL,

van Lier HJ, Barentsz JO. Local staging of prostate
cancer using magnetic resonance imaging: a meta-
analysis. Eur Radiol 2002; 12:22942302
47. Hoeks CM, Barentsz JO, Hambrock T, et al. Pros-
tate cancer: multiparametric MR imaging for de-
tection, localization, and staging. Radiology 2011;
Downloaded from www.ajronline.org by 114.121.238.76 on 05/02/17 from IP address 114.121.238.76. Copyright ARRS. For personal use only; all rights reserved
261:4666
48. Baco E, Rud E, Vlatkovic L, et al. Predictive value
of magnetic resonance imaging determined tumor
contact length for extracapsular extension of pros-
tate cancer. JUrol 2015; 193:466472
49. Rosenkrantz AB, Shanbhogue AK, Wang A,
Kong MX, Babb JS, Taneja SS. Length of capsular
1044 AJR:208, May 2017
Lim et al.
contact for diagnosing extraprostatic extension on
prostate MRI: assessment at an optimal threshold.
JMagn Reson Imaging 2016; 43:990997
50. Chong Y, Kim CK, Park SY, Park BK, Kwon GY,
Park JJ. Value of diffusion-weighted imaging at
3 T for prediction of extracapsular extension in
patients with prostate cancer: a preliminary study.
AJR 2014; 202:772777
51. Kayat Bittencourt L, Litjens G, Hulsbergen-van de
Kaa CA, Turkbey B, Gasparetto EL, Barentsz JO.
Prostate cancer: The European Society of Uro-
genital Radiology Prostate Imaging Reporting
and Data System criteria for predicting extrapros-
tatic extension by using 3-T multiparametric MR
imaging. Radiology 2015; 276:479489
52. Schieda N, Quon JS, Lim C, et al. Evaluation of
the European Society of Urogenital Radiology
(ESUR) PI-RADS scoring system for assessment
of extra-prostatic extension in prostatic carcino-
ma. EurJ Radiol 2015; 84:18431848
53. Flood TA, Schieda N, Keefe DT, et al. Utility of
Gleason pattern 4 morphologies detected on
transrectal ultrasound (TRUS)-guided biopsies
for prediction of upgrading or upstaging in Glea-
son score 3+ 4= 7 prostate cancer. Virchows Arch
2016; 469:313319