TITLE:EvaluatingMulti-ParametricMRIandPI-RADSv2forPredictingIndolentProstateCanceronPostOperativePathology
Faina Shtern1, Fiona Fennessy2, Mukesh Harisinghani3, Elmira Hassenzadeh2, Adam S. Kibel2, Quoc-Dien Trinh2,
MichelleS.Hirsch2,ZhengZhang4,DavidZurakowski5,andClareM.Tempany2
1AdMeTechFoundation,2BrighamandWomen’sHospital,3MassachusettsGeneralHospital,4BrownUniversity,and
5BostonChildren’sHospital
INTRODUCTION. Prostate cancer (PC) is the most common and the second most lethal malignancy in American
men.1-4 Extensive studies have demonstrated the value of observation strategy (i.e., active surveillance) for lowrisk,low-volume,indolentPC(GleasonScore,orGS≤6),andtheimportanceofdefinitivetreatmentforclinically
significantPC,asdefinedbymoreaggressive(GS≥7)disease,whichismorelikelytoprogress,causemetastases
anddeath.1-8Theaccuratedifferentiationofaggressivevs.indolentdiseaseisofparticularimportanceintheeraof
prostate specific antigen (PSA), when many men present with GS ≤ 6 and can be safely managed with active
surveillance, but unfortunately, are often treated.1-8 While observation strategies are increasingly adopted, postoperative pathology (POP) GS ≤ 6 has been reported in as many as 42% to 62% of men undergoing radical
prostatectomy(RP).4,9TheneedtoadvancediagnostictoolsforimprovingpredictionofPOPGSandreducingovertreatmentofindolentdiseasehasbeenwidelyrecognized.1-4
SPECIFIC AIM: The goal of this clinical study is to determine whether multi-parametric (mp) MRI and related
standardized Prostate Imaging–Reporting and Data System Version 2 (PI-RADS™ v2) can improve prediction of
POPGS≤6andhelpselectmenforactivesurveillancevs.treatment.
BACKGROUND AND RATIONALE: Over the last ten years, mpMRI has emerged as one of the most promising
diagnostic tools for the detection (and exclusion) of aggressive PC.10,11 The goal of PI-RADS™ v2 scoring is to
improveriskstratificationevenfurther,withstandardizedtechniquesforimageacquisitionandreportingcriteria,
including qualitative assessment that assigns a score of 1 to 5 based on the degree of clinical suspicion.12 While
developed based primarily on international expert consensus, PI-RADS v2 scoring requires validation, including
correlationwithPOPoutcomes.Ourstudywasconductedtoaddressthisunmetneed.
METHODS AND MATERIALS: This institutional review board-approved, retrospective study accrued 366 men
aged40to80yearsoldwhohadabnormalPSA(>4ng/mL)and/orabnormaldigitalrectalexamandunderwent
TRUS biopsy and RP at BWH between 2008 and 2014. The two study cohorts included: 1) Patients who had
mpMRI,whichwasperformedatleast6weeksafterTRUSbiopsyandwithin6monthspriortoRP(N=190);and2)
PatientswhodidnothavempMRIpriortoRP(N=176).MpMRIwasdefinedas“negative”forGP≥4ifPI-RADSv2
scores were 1-3, and “positive” if PI-RADS v2 scores were 4-5. POP served as a reference standard. Odds of POP
GS< 6 were analyzed using multivariable logistic regression withSAS software (version 9.3,SAS Institute, Cary,
NC).13
Availability of a complete standard diagnostic clinical information (including PSA level, POP reports and
TRUS biopsy results) was required for both cohorts. Lesion volume was measured on POP. All mpMRI exams
includedinthestudywereacquiredat3Tesla(3T)andmettechnicalstandardsforimageacquisition14andclinical
criteria for acceptable image quality. Two readers independently reviewed all mpMRI de-identified studies and
assigned PI-RADS v2 score at a patient level, based on the dominant lesion. In cases of disagreement between
readers,athirdradiologistservedasanadjudicatoranddeterminedafinaloverallassessment.Allreviewerswere
blindedtoclinicalinformation.
RESULTS: We compared overall numbers/percentages of POP GS≤6 in the mpMRI cohort (all, “positive” and
“negative” PI-RADS™ v2 scores) vs. the control group (no mpMRI). We have also examined the effect of lesion
volume,asdefinedbyPOP(≤0.5ccvs.>0.5cc),onmpMRI.Table1showspatientpopulationinbothstudycohorts.
Variable
MRI(AnyPI-RADS™)
(N=190)
Age,years,mean±SD
58.7±6.9
Race,no.(%)
Caucasian
164(86)
AfricanAmerican&Hispanic
20(11)
Other/declinedtoanswer
6(3)
TNMStage,median(range)
2(1-3)
PSA,ng/mL,median(IQR)
5.3(4.2-8.1)
*Statisticallysignificant.SD=standarddeviation,IQR=interquartilerange.
NoMRI–Controls
(N=176)
59.8±6.5
143(81)
10(6)
23(13)
2(1-3)
5.0(4.0-6.8)
PValue
0.123
0.002*
0.251
0.069
TABLE 1: Patient populations in the two study cohorts. More patients in the MRI group were African American,
andmoremeninthecontrolgroupdeclinedtoidentifytheirraceand/orethnicity.
I.BaselineComparisonoftheTwoCohortsandRatesofPOPGS<6With/WithoutmpMRI(Irrespectiveof
PI-RADS™v2Scores)
MRICohort(AllPI-RADS™v2)
Variable
(N=190)
Age,years,mean±SD
58.7±6.9
TNMStage,median(range)
2(1-3)
PSA,ng/mL,median(IQR)
5.3(4.2-8.1)
POPGS<6
32(17%)
*Statisticallysignificant.SD=standarddeviation,IQR=interquartilerange.
NoMRI–Controls
(N=176)
59.8±6.5
2(1-3)
5.0(4.0-6.8)
74(42%)
PValue
0.123
0.251
0.069
<0.0001*
TABLE2.RatesofPOPGS<6were17%withmpMRIand42%withoutmpMRI,resultinginanabsolutereduction
of 25% based on the studied cohorts (95% confidence interval (CI): 16% to 34% reduction). Logistic regression
indicatedthattheunadjustedoddsofPOPGS<6arereducedby72%withMRI(oddsratio:0.28,95%CI:0.170.45,likelihoodratiotest=28.7,P<0.0001).Multivariablelogisticregressionadjustingforcovariatesincluding
age,race,clinicalTNMstage,andPSAlevelconfirmedthattheoddsofPOPGS<6arereducedby74%withMRI
(adjustedoddsratio:0.26,95%CI:0.15-0.43,likelihoodratiotest=28.6,P<0.0001).Therefore,afteraccounting
for the influence of the other four variables, the effect of MRI (any PI-RADS™) had a significant independent
reductionof74%intheoddsofPOPGS<6witha95%CIfrom57%to85%.
II. Baseline Comparison of the Two Cohorts and Rates of POP GS < 6 With/Without MRI with Positive PIRADS™v2Scores4-5:
PositiveMRI(PI-RADS™4-5)
Variable
(N=150)
Age,years,mean±SD
59.1±6.8
TNMStage,median(range)
2(1-3)
PSA,ng/mL,median(IQR)
5.3(4.2-8.5)
POPGS<6
13(9%)
*Statisticallysignificant.SD=standarddeviation,IQR=interquartilerange.
NoMRI–Controls
(N=176)
59.8±6.5
2(1-3)
5.0(4.0-6.8)
74(42%)
PValue
0.357
0.091
0.034*
<0.0001*
TABLE 3: Rates of POP GS < 6 were 9% with “positive” MRI and 42% without MRI, resulting in an absolute
reduction in the percentage of patients with POP GS < 6of33% based onthestudied cohorts(95% confidence
interval (CI): 24% to 42% reduction). Logistic regression indicated that the unadjusted odds ofPOP GS < 6 are
reduced by over 87% with MRI (odds ratio: 0.13, 95% CI: 0.07-0.25, likelihood ratio test = 50.3, P < 0.0001).
Multivariable logistic regression adjusting for covariates including age, race, clinical TNM stage, and PSA level
confirmedthattheoddsofPOPGS<6arereducedby89%withMRI(adjustedoddsratio:0.11,95%CI:0.05-0.23,
likelihoodratiotest=44.4,P<0.0001).Therefore,afteraccountingfortheinfluenceoftheotherfourvariables,
the effect of positive MRI (PI-RADS™ 4-5) had a significant independent reduction in the odds of POP GS < 6 of
89%witha95%CIfrom77%to95%.
III.BaselineComparisonoftheTwoCohortsandRatesofPOPGS<6With/WithoutmpMRIInaSub-Group
ofMenwithNegativePI-RADS™Scores1-3:
NegativeMRI(PI-RADS™1-3)
Variable
(N=40)
Age,years,mean±SD
57.2±7.0
TNMStage,median(range)
2(1-3)
PSA,ng/mL,median(IQR)
5.4(4.0-7.4)
POPGS<6
19(48%)
*Statisticallysignificant.SD=standarddeviation,IQR=interquartilerange.
NoMRI-Controls
(N=176)
59.8±6.5
2(1-3)
5.0(4.0-6.8)
74(42%)
PValue
0.026*
0.297
0.912
0.597
TABLE4:RatesofPOPGS<6weresimilarinbothcohorts-48%withMRIand42%withoutMRI,resultingina
non-significant absolute increase of 6% based on MRI (95% confidence interval: -11% to 22%). Logistic
regression indicated that the unadjusted odds of POP GS < 6 are increased by 25% with MRI, which is not
statisticallysignificant(oddsratio:1.25,95%CI:0.63-2.48,likelihoodratiotest=0.39,P=0.531).Multivariable
logistic regression adjusting for covariates including age, race, clinical TNM stage, and PSA level indicated that
theoddsofPOPGS<6arenon-significantlyincreasedby15%withMRI(adjustedoddsratio:1.15,95%CI:0.542.44, likelihood ratio test = 0.13, P = 0.715). Therefore, after accounting for the influence of the other four
variables,theimpactofnegativeMRIhadnosignificantmeasurableeffectontheoddsofpost-opGS<6.
IV. Baseline Comparison of the Two Cohorts and Rates of POP GS < 6 With/Without mpMRI in a SubGroupofMenwithPositivePI-RADS™ScorresandPOPLesionVolumeGreaterThan0.5cc(Table5).Rates
of POP GS < 6 were 2% with MRI and 42% (Table 5) without MRI resulting in an absolute reduction of 40%
based on the studied cohorts (95% confidence interval (CI): 30% to 47% reduction). Logistic regression
indicatedthattheunadjustedoddsofPOPGS<6arereducedbyover96%withMRI(oddsratio:0.04,95%CI:
0.01-0.14, likelihood ratio test = 55.8, P < 0.0001). Multivariable logistic regression adjusting for covariates
including age, race, clinical TNM stage, and PSA level confirmed that the odds of POP confirmed GS < 6 are
reducedby97%withMRI(adjustedoddsratio:0.03,95%CI:0.01-0.15,likelihoodratiotest=45.2,P<0.0001).
Therefore,afteraccountingfortheinfluenceoftheotherfourvariables,theeffectofpositiveMRI(PI-RADS™4-5)
amongpatientswithlesionvolume>5cchadasignificantindependentreductionof97%intheoddsofPOPGS<
6witha95%CIfrom85%to99%.
Positive MRI (PI-RADS™ 4-5) NoMRI-Controls
andLesionVolume>0.5cc
Variable
(N=84)
(N=176)
Age,years,mean±SD
59.0±6.2
59.8±6.5
TNMStage,median(range)
2(1-3)
2(1-3)
PSA,ng/mL,median(IQR)
6.0(4.5-9.6)
5.0(4.0-6.8)
POPGS<6
2(2%)
74(42%)
•
Statisticallysignificant.SD=standarddeviation,IQR=interquartilerange.
PValue
0.353
0.108
<0.001*
<0.0001*
TABLE5
V.BaselineComparisonoftheTwoCohortsandRatesofPOPGS<6With/WithoutmpMRIinaSub-Group
ofMenwithPositivePI-RADS™ScoresandPOPLesionVolumeLessThan0.5cc(Table6):
Positive MRI (PI-RADS™ 4-5)
andLesionVolume<0.5cc
Variable
(N=66)
Age,years,mean±SD
59.2±7.5
TNMStage,median(range)
2(1-3)
PSA,ng/mL,median(IQR)
4.9(4.0-7.0)
POPGS<6
11(17%)
*Statisticallysignificant.SD=standarddeviation,IQR=interquartilerange.
NoMRI-Controls
(N=176)
59.8±6.5
2(1-3)
5.0(4.0-6.8)
74(42%)
PValue
0.353
0.216
0.845
0.0002*
TABLE6.RatesofPOPGS<6were17%withMRIand42%withoutMRIresultinginanabsolutereductionof25%
(95%confidenceinterval(CI):13%to36%reduction).Logisticregressionindicatedthattheunadjustedoddsof
pathGS<6arereducedbyover72%withMRI(oddsratio:0.28,95%CI:0.14-0.56,likelihoodratiotest=14.7,P=
0.0001).Multivariablelogisticregressionadjustingforcovariatesincludingage,race,clinicalTNMstage,andPSA
levelconfirmedthattheoddsofPOPconfirmedGS<6arereducedby78%withMRI(adjustedoddsratio:0.22,
95% CI: 0.09-0.55, likelihood ratio test = 13.2, P = 0.0003). Therefore, after accounting for confounding and the
influenceoftheotherfourvariables,theeffectofpositiveMRI(PI-RADS™4-5)amongpatientswithlesionvolume
<5cchadasignificantindependentreductionof78%intheoddsofpost-opGS<6witha95%CIfrom45%to
91%.
VI.Summary:TherateofPOPGS≤6inthecontrolgroupis42%andalignedwithpreviousstudies.4,9Inmenwho
hadpositivempMRI(PI-RADS™v24-5scores),therateofindolentPOPGS≤6isaslowas9%,withtheoddsof
yieldpredictedtobereducedby89%.WithmpMRI(irrespectiveofPI-RADS™v2scores),therateofPOPGS≤6is
17%, also significantly lower compared to the control cohort. Based on this data, 74% reduction in the odds of
yieldofindolentPCwithmpMRI(irrespectiveofPI-RADSv2score)onPOPispredictedcomparedtothecontrol
group.TheeffectofnegativePI-RADS™v2scores1-3ontheoddsofyieldofPOPGS≤6isinsignificant.
AmongpositivePI-RADS™v2scores(Tables5-7),theimpactontheoddsofyieldofPOPGS≤6dependson
lesionvolume.Insmaller(≤0.5cc)lesions,theyieldofPOPGS≤6is17%.Inlargerlesions,therateofPOPGS≤6is
aslowas2%(withpredicted97%reductionintheoddsofyield).
DISCUSSION:Themajorityofpatients(83%oftheMRIcohortand58%ofcontrols)inthissurgicalpopulationhad
aggressivedisease(POPGS≥7).Thisisinagreementwithotherrecentstudies.15,16TherateofindolentPOPGS≤6
of42%inthecontrolgroupreflectscurrentstandardsofPCcareinleadingacademiccenters.4.9Insharpcontrast,
with positive mpMRI (PI-RADS™ v2 4-5 scores), the probability of over-estimating PC is low, as indicated by the
rateofindolentPConPOPaslowas9%.WepredictasignificantreductionintheoddsofyieldofindolentPCon
POPwithpositivePI-RADSscores,ifmpMRIweretobeusedfordiagnosticpatientevaluationpriortoRP.
ThisstudyalsoshowsareductionintheoddsofyieldofindolentPCwithmpMRIirrespectivelyofPI-RADS
v2scores,thoughthismostlikelyreflectsastrongeffectofpositivePI-RADSv2scores.
Performance of positive PI-RADS™ v2 scores appears to depend on POP lesion volume, though dividing
patients into two groups (≤0.5 cc and >0.5 cc) reduces the number and thus lowers power. In smaller (≤0.5cc)
lesions, there was higher probability of over-estimating GS, most likely due to the bias of our readers who were
blindedtoclinicalinformationbutawareofsurgicalpopulation.WeshowamajorbenefitofpositivePI-RADS™v2
scoresinlargerlesions,wheretheyieldofindolentPConPOPisaslowas2%.
CONCLUSION:ThisstudyshowsthatmpMRIalongwithpositivePI-RADS™v2scoresmayplayavaluablerolein
aidingoptimalselectionofcandidatesforsurgeryvs.activesurveillance.Whilepromising,theseresultsneedtobe
considered within the context of this retrospective study, which has an inherent bias of the selected surgical
populationanditsimpactonradiologicreaders.Thisresearchunderscorestheimportanceoffurtherlarger-scale,
prospectiveandmulti-centerclinicaltrial.
REFERENCES
1) GanzPA,et.al.NIHState-of-the-ScienceStatement:Roleofactivesurveillanceinthemanagementofmenwith
localizedprostatecancer.NIHConsensusStateScienceStatements2011Dec5-7;28(1):1-27
2) Dall'Era,M.A.,etal.,Activesurveillanceforearly-stageprostatecancer(PC).Cancer,2008.112(8):1650-9.
3) Stephenson,A.J.,etal.,Prostatecancer-specificmortality(PCSM).JClinOncol,2009.27(26):4300-5.
4) Eggener,S.E.,etal.,Predicting15-yearprostatecancerspecificmortality.JUrol,2011.185(3):869-75.
5) Wilt,T.J.,etal.,Radicalprostatectomy(RP)versusobservation.NEnglJMed,2012.367(3):203-13.
6) Andriole,G.,etal.,ProstatecancerscreeningintherandomizedProstate,Lung,Colorectal,andOvarianCancer
ScreeningTrial:mortalityresultsafter13yearsoffollow-up.JNCI,2012.104(2):125-32.
7) Schroder,F.H.,etal.,PCSMat11yearsoffollow-up.NEnglJMed,2012.366(11):981-90.
8) Freedland,S.J.,etal.,RiskofPCSMfollowingbiochemicalrecurrence.JAMA,2005.294(4):433-9.
9) Epstein,J.,etal.,UpgradinganddowngradingofPCfrombiopsytoRP.EurUrol,2012.61(5):1019-24.
10) Schoots,IG,et.al.MagneticResonanceImaginginActiveSurveillanceofProstateCancer:ASystematicReview.
EuropeanUrology2015.67(4):627–636
11) Barentsz, J.O., et al., Will MRI-Guided Biopsy Replace Systematic Biopsy? European Urology Focus 1 (2015):
152-155
12) Weinreb,JC,PI-RADSProstateImaging–ReportingandDataSystem:2015,Version2.EuropeanUrology2016.
69(1):16-40.
13) Katz MH. Multivariable analysis: A practical guide for clinicians and public health researchers, 3rd ed. New
York:CambridgeUniversityPress;2011:118-39.
14) Hegde, JV, et. al. Multiparametric MRI of prostate cancer: an update on state-of-the-art techniques and their
performance in detecting and localizing prostate cancer. J Magn Reson Imaging. 2013 May; 37(5):1035-54.
PMID:23606141;PMC3741996.
15) Vos,EK,et.al.,MultiparametricMagneticResonanceImagingforDiscriminatingLow-GradeFromHigh-Grade
ProstateCancer.Invest.Radiology2015.50(8):490-7
16) Park,SY,et.al.,ProstateCancer:PI-RADSVersion2HelpsPreoperativelyPredictClinicallySignificantCancers.
Radiology2016.280(1):108-16