EURURO Vol. 72 No. 2

262 320

Owing to the large variations of NPV induced by differences
in definitions of csPCa, we did not include different
definitions in the meta-analysis since this would have
introduced unacceptable clinical heterogeneity in the
results, possibly resulting in erroneous and biased esti-
mates. We, therefore, a priori restricted the definition of
csPCa to cancers with a Gleason score of 7, given the low
lethal potential of Gleason 6 cancers and the lack of
consensus among pathologists on the best method to
measure biopsy core invasion length .
In this more homogeneous group of studies, the preva-
lence range was still large (31.3–63.7%). As a result, we
modelled the evolution of NPV (and PPV) as a function of
overall PCa prevalence. Unfortunately, we could not duplicate
this for csPCa since only one study reporting NPV for Gleason
7 cancers met the inclusion criteria for meta-analysis.
3.6.2. Reference standard
We included only studies that reported the results of
systematic/standard biopsy in patients with negative
mpMRI and used the systematic/standard biopsy as a
reference standard. It is well known that TRUS-guided
biopsy harbours both random and systematic errors, as
evidenced by the high rates of positivity of immediate
repeat biopsy after a first series of negative biopsies ,
and as confirmed recently by the PROMIS trial . There-
fore, using TRUS-guided biopsy as a reference standard may
have overestimated the NPV of mpMRI. However, studies
using radical prostatectomy specimens as a reference
standard have already reported mpMRI detection rates in
relation to PCa Gleason score and volume In this review,
we intended to address the more pragmatic question as to
whether a negative mpMRI could predict a negative
subsequent biopsy. This is an important question because
if the NPV of mpMRI was sufficiently high in comparison
with the reference standard of systematic/standard biop-
sies, then in practice a negative mpMRI result could indeed
avoid the need for prostate biopsy. Therefore, studies
reporting only biopsy results when the mpMRI was positive
(eg, obtained through MRI-targeted, guided, or fusion
biopsies with added systematic biopsies) were not included
in this review.
3.6.3. Impact on clinical practice and research
It is now well established that mpMRI is a sensitive tool for
detecting aggressive PCa However several reasons
preclude its broad use as a triage test before biopsy.
Firstly, the population referred to prostate biopsy is not
standardised. The large range of reported prevalence for
overall PCa and csPCa suggests substantial heterogeneity in
the way patients are selected for biopsy. Owing to this
heterogeneity, we did not provide a pooled estimate for
mpMRI NPV. The role of mpMRI as a triage test before
prostate biopsy should be evaluated in the broader context
of the selection of patients with a suspicion of (aggressive)
PCa. In a recent retrospective study of 514 patients, mpMRI
NPV for Gleason 7 cancers was 91% when the PSA density
was 0.2 ng/ml/ml, and only 71% when the PSA density
was >0.2 ng/ml/ml (p = 0.003) . In another series of
288 biopsy-naı¨ve patients, no csPCa (Gleason score 7 or
maximum cancer core length 4 mm) was found in
44 patients with a PSA density of <0.15 ng/ml/ml and a
PI-RADS v2 score of <3/5 . We believe that such
prestratification of the risk of csPCa is an interesting way for
rationalising the use of mpMRI before biopsy. Patients
found at very low risk would be spared both mpMRI and
biopsy. Patients at a low risk—for whommpMRI would have
an NPV high enough to be used as a triage test—could avoid
biopsy in case of negative mpMRI. Patients at a higher risk
would need biopsy even in case of negative mpMRI. Many
tools can be used to risk stratify the population of patients
referred to biopsy, ranging from simple parameters such as
PSA density to more complicated risk calculators
. The impact of these tools on the NPV of prebiopsy
mpMRI needs to be carefully evaluated, both in the biopsy-
naı¨ve and in the repeat biopsy setting. For the moment, it is
impossible to make any recommendations on the best way
to risk stratify patients before referring them for mpMRI.
Secondly, the large variability in the definition of csPCa
precludes any definitive conclusion on the ability of mpMRI
to rule out aggressive cancer. The issue of the most
appropriate definition of csPCa on biopsy is complex, since
biopsy results may accurately reflect neither tumour
burden nor aggressiveness. Nonetheless, there is an urgent
need to standardise the histological definition(s) of csPCa, to
allow meaningful comparisons between studies.
Thirdly, the specificity of mpMRI remains moderate, and
there is a substantial proportion of false positives in the
lesions scored 3/5 or 4/5 , even with the new PI-
RADS v2 score In a series of 62 patients with 116 lesions
biopsied under magnetic resonance/ultrasound fusion, the
overall cancer detection rates for PI-RADS v2 scores of 3/5
and 4/5 were only 15.8% and 29.8%, respectively . In
theory, a triage test used to rule out a disease needs to be
highly sensitive for this disease. However, if its specificity is
too low, it will be clinically useless since most patients will
be positive, whether they have the disease or not. Therefore,
if mpMRI is to be used as a triage test in the future, there is a
need to improve its specificity. This could be achieved by a
continuous refinement of scores . Promising results in
characterising csPCa have also been reported with a
quantitative analysis
Table 5 – Positive and negative predictive estimates for prebiopsy
multiparametric MRI as a function of prostate cancer prevalence
(meta-analysis)
PCaPrev PPV NPV
0.30 0.43 (0.34–0.53) 0.88 (0.77–0.99)
0.40 0.54 (0.45–0.64) 0.82 (0.70–0.94)
0.50 0.64 (0.55–0.73) 0.76 (0.64–0.88)
0.60 0.73 (0.65–0.80) 0.67 (0.56–0.79)
0.70 0.81 (0.75–0.87) 0.57 (0.47–0.67)
0.75 0.84 (0.79–0.89) 0.51 (0.42–0.59)
PCaPrev = prevalence of prostate cancer; PPV = positive predictive value;
MRI = magnetic resonance imaging; NPV = negative predictive value.
Intervals in parenthesis are 95% confidence intervals. A score of 3/5 was
used to define positive MRI.
E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 2 5 0 – 2 6 6 262