Publications

Abstract

PURPOSE: Multiparametric magnetic resonance imaging appears to improve prostate cancer detection but prospective studies are lacking. We determined the accuracy of multiparametric magnetic resonance imaging for detecting significant prostate cancer before diagnostic biopsy in men with abnormal prostate specific antigen/digital rectal examination. MATERIALS AND METHODS: In this single center, prospective study men older than 40 years with abnormal prostate specific antigen/digital rectal examination and no previous multiparametric magnetic resonance imaging underwent T2-weighted, diffusion-weighted and dynamic contrast enhanced imaging without an endorectal coil. Imaging was allocated alternately to 1.5/3.0 Tesla. Imaging was double reported independently using PI-RADS (Prostate Imaging Reporting and Data System) by specialist radiologists. Transperineal grid directed 30-core biopsy was performed with additional magnetic resonance imaging directed cores for regions of interest outside template locations. Four significant cancer definitions were tested. Chi-square and logistic regression analysis was done. Men undergoing prostatectomy were analyzed. RESULTS: Of the 165 men who enrolled in the study 150 were analyzed. Median age was 62.4 years, median prostate specific antigen was 5.6 ng/ml, 29% of patients had an abnormal digital rectal examination and 88% underwent initial biopsy. Multiparametric magnetic resonance imaging was positive (PI-RADS 3 to 5) in 66% of patients, 61% had prostate cancer and 30% to 41% had significant prostate cancer (definitions 1 to 4). For significant cancer sensitivity was 93% to 96%, specificity was 47% to 53%, and negative and positive predictive values were 92% to 96% and 43% to 57%, respectively (definitions 1 to 4). Radical prostatectomy results in 48 men were similar. Aggregate PI-RADS (4 to 20) performed similarly to overall PI-RADS (1 to 5). Negative and positive predictive values (100% and 71%, respectively) were similar in men at higher risk, defined as prostate specific antigen greater than 10 ng/ml with abnormal digital rectal examination. On multivariate analysis PI-RADS score was associated with significant prostate cancer (p <0.001) but magnet strength was not. Adding PI-RADS to the multivariate model improved the AUC from 0.810 to 0.913 (95% CI 0.038-0.166, p = 0.002). Radiologist agreement was substantial (weighted kappa = 0.626). CONCLUSIONS: Multiparametric magnetic resonance imaging reported by expert radiologists achieved an excellent negative predictive value and a moderate positive predictive value for significant prostate cancer at 1.5 and 3.0 Tesla.