Related Experiment Video
Updated: Dec 6, 2025
Use of MRI-ultrasound Fusion to Achieve Targeted Prostate Biopsy
Published on: April 9, 2019
Stratifying patients with intermediate-risk prostate cancer: Validation of a new model based on MRI parameters and
Romain Diamand1, Guillaume Ploussard2, Mathieu Roumiguié3
1Urology Department, Hôpital Erasme, University Clinics of Brussels, Brussels, Belgium.
Objective:
Intermediate-risk prostate cancer regroups heterogeneous patients with different oncologic outcomes. Aim of the study is to validate a novel intermediate-risk subclassification ("magnetic resonance imaging [MRI] subclassification") that defines favorable and unfavorable diseases based on multiparametric MRI parameters and compare it to NCCN and AUA intermediate-risk subclassifications.
Methods:
A total of 429 patients treated with radical prostatectomy for NCCN intermediate-risk prostate cancer were identified. Using MRI subclassification, a favorable disease was defined as an organ-confined disease on MRI and international society of urological pathology Grade Group 1 to 2 on targeted biopsy. Remaining was classified as unfavorable. Univariable and multivariable analysis tested MRI subclassification in predicting overall unfavorable disease (OUD: pT3-4 and/or pN1 and/or International Society of Urological Pathology Grade Group ≥ 3), the need for adjuvant therapy and early biochemical recurrence (eBCR). Performance of NCCN, AUA, and MRI models was compared in term of OUD proportion and eBCR prediction using Harrell's c-index, calibrations plots, and decision curve analysis.
Results:
Median (interquartile range) follow-up was 12 months (4-28). In multivariable analysis, MRI subclassification was an independent factor for OUD (odds ratio [OR]: 4.54 [2.85-7.22], P < 0.001), the need for adjuvant therapy (OR: 3.42 [1.36-8.57], P = 0.009), and eBCR (HR: 2.62 [1.18-5.83], P = 0.018). Using this model, the proportion of unfavorable disease decreased from 73.7% and 63.9% to 35.9% (P < 0.001) associated to an increasing proportion of OUD when compared to NCCN and AUA models (63.9% and 67.1%-77.9% respectively, P < 0.001). Performance of the 3 models for eBCR prediction tended to be similar with a poor accuracy ranged from 58.7% to 66.7% (P > 0.05), permanent miscalibration and a net benefit at decision curve analysis.
Conclusions:
We validated an intermediate-risk subclassification based on MRI and targeted biopsy that potentially improves patient selection by reducing the number of patients considered at unfavorable risk while increasing proportion of patients harboring poor oncologic outcomes. Its performance for eBCR detection was comparable to NCCN and AUA models.
