83 148
MRI, T1-weighted imaging (T1WI), T2-weighted imaging
(T2WI), or short tau inversion recovery (STIR), and their
corresponding technical parameters; and criteria for bone
metastasis.
The methodological quality of the included studies was
assessed using the Quality Assessment of Diagnostic
Accuracy Studies-2 (QUADAS-2) tool
[8] .Data extraction
and quality assessment were performed independently by
two reviewers (S.W. and C.H.S.), and consensus was reached
via discussion with a third reviewer (S.Y.K.).
2.4.
Data synthesis and analysis
The primary outcome of this meta-analysis was the per-
patient diagnostic performance of MRI for the detection of
bone metastasis in patients with prostate cancer. As a
secondary outcome, we aimed to assess the presence of
heterogeneity among the included studies and explore
potential causes.
Two by two tables were tabulated for the included
studies to calculate their sensitivity and specificity. If
diagnostic performance of several MRI sequences were
separately assessed, we selected the results using the most
advanced MRI sequence or most comprehensive MRI
protocol (ie, DWI + conventional sequences
>
DWI
>
conventional sequences). When results from multiple
independent readers were given, the result with the highest
accuracy was used. Of note, although one study included in
our meta-analysis
[9]assessed bone metastasis on a per-
lesion basis, it was analyzed as per-patient for the following
reasons: (1) this was the only study not providing per-
patient diagnostic performance and (2) the mean number of
lesions per patient in this study was 1.1 (26/24).
Summary estimates of sensitivity and specificity were
calculated using hierarchical logistic regression modeling
including bivariate and hierarchical summary receiver
operating characteristic (HSROC) modeling
[10–12]. These
results were plotted using HSROC curves with 95%
confidence and prediction regions. Publication bias was
evaluated using visual analysis of the Deeks et al’s funnel
plot and calculating the
p
value using Deeks et al’s
asymmetry test
[[4_TD$DIFF]
13].
Heterogeneity was determined using the following: (1)
Cochran’s
Q
test with
p
<
0.05 indicating the presence of
heterogeneity; (2) Higgins
I
2
[15_TD$DIFF]
test with the following criteria
for the interpretation of the degree of heterogeneity:
inconsistency index (
I
2
) = 0–40%, heterogeneity might not
be important; 30–60%, moderate heterogeneity may be
present; 50–90%, substantial heterogeneity may be present;
and 75–100%, considerable heterogeneity
[14]; and (3)
testing for the presence of a threshold effect (a positive
correlation between sensitivity and false positive rate)
among the selected studies.
[16_TD$DIFF]
Meta-regression analyses using several covariates were
performed to explore the cause of heterogeneity as follows:
(1) clinical setting (newly diagnosed vs treated), (2)
reference standard (BVC only vs inclusion of histopatholo-
gy), (3) magnet field strength (1.5 vs 3 T), (4) MRI coverage
(pelvis vs axial skeleton/whole body), (5) MRI sequence
(only conventional sequences vs DWI included), (6) number
of imaging planes (1 vs 2), and (7) minimum slice
thickness among sequences used ( 4 vs
>
4 mm). In
addition, sensitivity analyses for the various settings
stratified to the covariates described above were performed.
The ‘‘midas’’ module in Stata 10.0 (StataCorp LP, College
Station, TX, USA) and ‘‘mada’’ package in R software version
3.2.1 (R Foundation for Statistical Computing, Vienna,
Austria) were used for statistical analyses,
with
p
<
0.05 indicating statistical significance.
3.
Evidence synthesis
3.1.
Literature search
The systematic literature search initially yielded 1689 arti-
cles. After removing 697 duplicates, screening of the
992 titles and abstracts yielded 46 potentially eligible
original articles. Full-text reviews were considered and
36 studies were excluded due to the following reasons: not
in the field of interest (
n
= 21), insufficient data to
reconstruct 2 2 tables (
n
= 9),
<
10 patients (
n
= 2), study
population shared with other studies (
n
= 3), and non-
English publication (
n
= 1). Ultimately, 10 studies including
1031 patients evaluating the diagnostic performance of MRI
for the detection of bone metastasis in patients with
prostate cancer were included in this meta-analysis
[4–[22_TD$DIFF]
6,9,15–[11_TD$DIFF]
20]. The study selection process is summarized in
Figure 1.
3.2.
Characteristics of included studies
The patient characteristics are described in
Table 1. The size
of the study population ranged from 21 to 308 patients,
with the percentage of patients with bone metastasis
ranging from 6.8% to 71.4%. Four studies included only
patients with newly diagnosed prostate cancer, three
included only those with treated prostate cancer, and three
included a mixed population of newly diagnosed and
treated prostate cancer. The patients had a median age of
63–78 yr. Six studies were based on patients with a
clinically ‘‘high risk’’ of bone metastasis, two studies
included patients with any risk, and two studies were
unclear regarding this risk. The median PSA and Gleason
scores were 2.7–31 ng/ml and 7–9, respectively.
The study characteristics are summarized in
Table 2 .The
study design was prospective in six studies and retrospec-
tive in four. All but two studies were single-center studies.
Patient recruitment was consecutive in all but two studies
(nonconsecutive 1:3 matching for normal and metastasis in
one study and not explicit in another). Four studies used
either histopathology or BVC as the reference standard;
while the other six used only BVC. The imaging modalities
used in the included studies for BVC included BS, targeted x-
ray, computed tomography (CT), MRI, and positron emis-
sion tomography/CT. The interval between MRI and the
reference standard was not provided in three studies. MRI
was interpreted blinded to the reference standard in all but
one study, which was not explicit.
E U R O P E A N U R O L O G Y 7 3 ( 2 0 1 8 ) 8 1 – 9 1
83