Prostate Imaging Reporting and Data System and Likert Scoring System: Multiparametric MR Imaging Validation Study to Screen Patients for Initial Biopsy.

Abstract

To compare the diagnostic performance of the magnetic resonance (MR) imaging-based Prostate Imaging Reporting and Data System (PI-RADS) and a Likert scale in the detection of prostate cancer in a cohort of patients undergoing initial prostate biopsy. This institutional review board-approved two-center prospective study included 118 patients with normal digital rectal examination (DRE) results but elevated prostate-specific antigen (PSA) levels (4-20 ng/mL) who were referred for initial prostate biopsies and had one suspicious (Likert scale score, ≥3) focus at prebiopsy 1.5-T multiparametric MR imaging performed with T2-weighted, diffusion-weighted [DW], and dynamic contrast material-enhanced imaging. Targeted core biopsies and random systematic core biopsies were performed. The elementary unit for analysis was the core. Relationships were assessed by using the Mann-Whitney U test. Yates corrected and Pearson χ(2) tests were used to evaluate categoric variables. A training set was randomly drawn to construct the receiver operating characteristic curves for the summed PI-RADS scores and for the Likert scale scores. The thresholds to recommend biopsy were obtained from the Youden J statistics and were tested in the remaining validation set in terms of predictive characteristics. Interobserver variability was analyzed by using weighed κ statistics in a random set of 50 patients. Higher T2-weighted, DW, and dynamic contrast-enhanced imaging PI-RADS scores were observed in areas that yielded cancer-positive cores. The percentage of positive cores increased with the sum of scores aggregated in five classes as follows: For summed PI-RADS scores of 3-5, the percentage of positive cores was 2.3%; for scores of 6-8, it was 5.8%; for scores of 9 or 10, it was 24.7%; for scores of 11 or 12, it was 51.8%; and for scores of 13-15, it was 72.1% (P for trend, <.0001). For the threshold of summed PI-RADS scores of 9 or greater, sensitivity was 86.6%, specificity was 82.4%, the positive predictive value was 52.4%, the negative predictive value was 96.5%, and accuracy was 83.2%. The respective data for Likert scale scores of 3 or greater were 93.8%, 73.6%, 44.3%, 98.1%, and 73.3%. Good interobserver agreement was observed for the Likert scale (κ = 0.80) and the summed PI-RADS (κ = 0.73) scoring systems. PI-RADS provided the site-specific stratified risk of cancer-positive cores in biopsy-naive men with normal DRE results and elevated PSA levels. There was no significant difference between summed PI-RADS scores of 9 or greater and Likert scale scores of 3 or greater in the detection of cancer in the peripheral zone.