183 320
reports based on PI-RADSv1 or an in-house scoring system
[29]. Regarding the reference standard domain, eight
studies had a high risk of bias. Seven were based on
either only systematic biopsy or targeted biopsy
[18,19,21,28,30,32,33]; in one study, targeted biopsy was
performed, but on lesions that were suspicious on
ultrasonography, and not on MRI
[31]. Those in which
radical prostatectomy or a systematic plus targeted biopsy
(MRI guided, MRI–transrectal ultrasound fusion, or cogni-
tive) was used as the reference standard were considered to
have a low risk of bias. In 10 studies, the definition of
clinically significant cancer did not abide by those described
in the PI-RADSv2 guidelines, and therefore showed high
concern for applicability
[16,21,22,24,29,30,32,34–36]. Re-
garding the flow and timing domain, two studies had a high
risk of bias as patients did not receive the same reference
standard
[17,18].
3.4.
Diagnostic accuracy of PI-RADSv2
The sensitivity and specificity of individual studies were
73–100% and 8–100%, respectively. The Q-test revealed that
substantial heterogeneity was present (
p
<
0.001). The
Higgins
I
2
[3_TD$DIFF]
statistics demonstrated substantial heterogeneity
in terms of the sensitivity (
I
2
= 85.55%) and considerable
heterogeneity in terms of the specificity (
I
2
= 95.30%). The
coupled forest plot of the sensitivity and specificity
demonstrated the absence of a threshold effect
( Fig. 3 ).
The Spearman correlation coefficient between the sensitiv-
ity and the false-positive rate was 0.45 (95% confidence
interval [CI] 0.023–0.738), also indicating the lack of a
threshold effect.
For all 21 studies combined, the pooled sensitivity was
0.89 (95% CI 0.86–0.92) with specificity of 0.73 (95% CI 0.60–
0.83;
Fig. 3). In the HSROC curve, there was a large
difference between the 95% confidence region and the 95%
prediction region, thus indicating heterogeneity between
the studies
( Fig. 4). The area under the HSROC curve was
0.91 (95% CI 0.88–0.93). According to the Deeks’ funnel plot,
the likelihood of publication bias was low, with a
p
value of
0.75 for the slope coefficient
( Fig. 5 ).
For the six studies that provided a head-to-head
comparison between PI-RADSv1 and PI-RADSv2, PI-RADSv2
demonstrated higher pooled sensitivity of 0.95 (95% CI
0.85–0.98) compared with 0.88 (95% CI 0.80–0.93) for PI-
RADSv1 (
p
= 0.04). However, the pooled specificity did not
show a significant difference between the two versions of
PI-RADS: 0.73 (95% CI 0.47–0.89) for v2 and 0.75 (95% CI
0.36–0.94) for v1 (
p
= 0.90).
3.5.
Heterogeneity exploration
As we found a considerable heterogeneity among the
included studies,
[3_TD$DIFF]
meta-regression analyses were performed
(Supplementary
Table 1). Among several potential vari-
ables, proportion of patients with PCa, magnetic field
strength, and reference standard were significant factors
affecting the heterogeneity (
p
<
0.01 for all). However,
among these three, only specificity according to the
Woo
(2016)
[35]
Retrospective
NR RP
6 mo
2
Consensus
22/10
Yes
a3
Philips,
Siemens
Ingenia,
Verio/Trio
No Strict
4
Whole Patient csPCa GS 7
Zhao
(2016)
[36]
Retrospective
NR STRUSGB + targeted
MRI–TRUS biopsy
3 mo
2
Independent
NR
NR 3
NR
NR
No Strict
3
Whole Patient csPCa GS 7
csPCa = clinically significant prostate cancer; GS = Gleason score; EPE = extraprostatic extension; MRGB = magnetic resonance imaging-guided biopsy; MRI = magnetic resonance imaging; MRI–TRUS = fusion of magnetic
resonance imaging and transrectal ultrasound images; NR = not reported; PCa = prostate cancer; PI-RADSv2 = Prostate Imaging Reporting and Data System version 2; PZ = peripheral zone; RP = radical prostatectomy;
STRUSGB = systematic transrectal ultrasound-guided biopsy; TTB = transperineal template biopsy; TZ = transition zone.
a
Blinded but aware that patients had biopsy-proven PCa.
b
PI-RADSv2 scores were generated from reports based on PI-RADSv1 and simplified qualitative system.
c
Epstein’s criteria = Gleason pattern 4, or Gleason 3 + 3 disease with core length 50% and/or
>
2 cores positive.
d
For transition zone, cutoff value = 4.
E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 1 7 7 – 1 8 8
183