EURURO Vol. 71 No. 4

519 692

Published primary diagnostic studies reporting on PCa
detection rates among patients at risk of PCa using MRI-TB,
or FUS-TB, or COG-TB were included. A direct comparison of
MRI-GB techniques was not obligatory. Studies were
excluded if they reported detection rates of PCa among
patients with prior diagnosed PCa (including active surveil-
lance populations, andmixed populations if data for patients
with no or negative prior biopsies was not separately
reported upon); if the MRI acquisition was not in accordance
to the 2012 ESUR guidelines if the language was other
than English, and if studies used alterative target biopsy
strategies (such as contrast-enhanced TRUS).
Since the interval between data presentation and initial
search was significant, a cursory repeat search was
performed on December 15, 2015. This search identified
an additional four studies which were not included in the
meta-analysis, but are incorporated in the discussion section
of this paper.
2.2. Selection procedure
Following initial identification of studies, duplicates were
removed by a single reviewer (OW). Titles and abstract of all
studies were screened for relevance by two reviewers (OW,
RS). Full text review of eligible studies was performed by
three reviewers (OW, RS, and HM). Any disagreement was
handled by consensus, refereed by a fourth reviewer (RB).
The selection procedure followed the Preferred Reporting
Items for Systematic Reviews and Meta-analysis (PRISMA)
principles and is presented using a PRISMA flow chart
2.3. Quality assessment
The methodological quality of studies was assessed using
the Quality Assessment of Diagnostic Accuracy Studies-2
checklist by two reviewers in consensus (OW, LH)
. Using the Quality Assessment of Diagnostic Accuracy
Studies-2 checklist the risk of bias and concerns of
applicability to the review questions was assessed. A
sensitivity analysis was performed excluding the studies
assessed to have high risk of bias or high concerns regarding
applicability to the review questions.
2.4. Data extraction
The data for quantitative assessment was extracted by a
single reviewer (OW) in accordance to the START recom-
mendations . Data was collected on the method of
recruitment; population investigated; methods of MRI
acquisition and evaluation; MRI findings and/or PI-RADS
score; threshold applied for MRI positivity; methods of
biopsy procedure; number of (systematic and target) cores
taken; detection rates of csPCa (per patient and per core);
and the applied definition of csPCa.
2.5. Data analysis
For the first review question on the difference in accuracy
between TRUS-GB andMRI-GB, we combined the data of the
three MRI-GB techniques. For this analysis, we focused on
paired studies reporting results of both TRUS-GB and MRI-
GB separately. The main accuracy measure was the
sensitivity of each technique, which was defined as the
number of patients with detected cancer by TRUS-GB
(or MRI-GB), divided by the total number of patients with
detected cancer by the combination of TRUS-GB and MRI-
GB. In other words, 1 minus the sensitivity of a technique is
the percentage of patients with a cancer missed by this
technique. We calculated the relative sensitivity for each
study by dividing the sensitivity of MRI-GB by the
sensitivity of TRUS-GB. We used the formula for the
standard error of a relative risk without taking the paired
nature into account because not all studies reported their
data in a paired format . A random effects pooled
estimate of this relative sensitivity was calculated using the
generic inverse variance method All sensitivity
analyses were done twice: once for all PCa detected as
the condition of interest and once focussing on csPCa only.
For the per core analysis and detection of insignificant PCa
we performed a yield analysis as accuracy measure, which
was defined as the number of patient with detected cancer,
divided by the total number of patient that underwent
biopsy. We calculated the relative yield for each study by
dividing the yield of MRI-GB by the yield of TRUS-GB.
For the second review question on the difference in
accuracy between the various techniques of MRI-GB, we
used studies reporting on at least one of the MRI-GB
techniques (MRI-TB or FUS-TB or COG-TB). The applied
accuracy measurement was the sensitivity of each MRI-GB
technique as defined earlier. These proportions were meta-
analysed using a random effects model, incorporating
heterogeneity beyond chance due to clinical and methodo-
logical differences between studies. The within-study
variances (ie, the precision by which yield has been
measured in each study) was modelled using the exact
binomial distribution. Differences in sensitivity between
MRI-GB techniques were assessed by adding the type of
MRI-GB technique as covariate to the random effects meta-
regression model. These analyses were performed for all
PCa and csPCa. Extracted data was analysed using SPSS
version 22.0 (SPSS Inc., IBM, Chicago, IL, USA), and the
random effects models were analysed in SAS version 9.2
(SAS Institute Inc., Cary, NC, USA).
3. Evidence synthesis
3.1. Search and selection
Using the three databases 2562 studies were identified.
Following removal of duplicates, abstract and title screen-
ing, and full text assessment a total of 43 articles were
deemed relevant for the current review question. For an
overview of the selection procedure and reason for
exclusion see the PRISMA flow chart .
3.2. Quality assessment
Of the 43 studies subjected to quality assessment 54%
(n = 23) were estimated to have a low risk of bias, 40%
E U R O P E A N U R O L O G Y 7 1 ( 2 0 1 7 ) 5 1 7 – 5 3 1 519