MP13-12 IMPROVING THE IDENTIFICATION OF CLINICALLY SIGNIFICANT PROSTATE CANCER USING MULTIPARAMETRIC MAGNETIC RESONANCE IMAGING (MPMRI)

Abstract

INTRODUCTION AND OBJECTIVE: The aim of this study is to validate PIRADSv2 on biopsy-naive – mpMRI in detecting all prostate cancer lesions and clinically significant prostate cancer by correlation with whole-mount pathology after radical prostatectomy, using a per-lesion analysis. And using the correlation study, we will attempt to establish a new model with additional clinical variables to improve the identification of clinically significant prostate cancer using mpMRI. METHODS: Retrospective review of our institutional database of men who have radical prostatectomies operated between 1st Jan 2015 to 31st Dec 2017. The inclusion criteria were pre-operative mp-MRI performed in our institution, either pre-biopsy or >1 year from biopsy, and available whole-mount histology. PIRADSv2 was used to assess the imaging. Each lesion was outlined on wholemount histology, with the low grade (LG) (Gleason 3) and high-grade (HG) (Gleason 4 and 5) colour coded differently. Lesions on imaging were matched to wholemount histology and measured at the same axial plane taking the mean of 2 readings. For the model derivation, data from PIRADSv2 and other clinical parameters were used. Receiving operator characteristic curves was generated and the area under curves (AUC) were compared. RESULTS: A total of 70 patients met criteria. mpMRI had a sensitivity of 76% (95% CI: 0.69 – 0.83) and a positive predictive value of 72% (95% CI: 0.64 – 0.78) for prostate cancer detection. Patients with at least 1 HG lesion has median PSA density of 0.24 ng/ml2 which is significantly higher than patients with no HG lesion which have a median PSA density of 0.17 ng/ml2 (p = 0.008). Most missed lesions by mpMRI were LG. Missed HG lesions tended to be at least 50% smaller compared to HG lesions detected by mpMRI with a median area of 21.7mm2 versus 45.8 mm2 (p < 0.001). Our model using a combination of PIRADSv2, PSA density and MRI Volume of lesion showed improved predictive performance of AUC=0.799, superior to that of mpMRI using PIRADSv2 alone (AUC=0.769, p=0.001). A cut-off point of predicted probability of HG, P(HG) ≥ 20% corresponded to a negative predictive value of 87.5% and a positive predictive value (PPV) of 59.7%. Which means that for P(HG)< 20%, the lesion is most likely to be a LG. Whereas a cut-off point of P(HG) ≥ 50% corresponded to a NPV of 70.8% and a PPV of 74.0%. CONCLUSIONS: Our study derived a risk model using PIRADSv2 score and PSA density and mpMRI volume of lesion to better differentiate a LG or HG lesion compared to PIRADSv2 alone. This could help clinicians better select for gland sparing strategies and/or, avoid unnecessary biopsies. Source of Funding: N.A.