What Burden of Prostate Cancer Can Radiologists Rule Out on Multiparametric Magnetic Resonance Imaging? A Sensitivity Analysis Based on Varying the Target Condition in Template Prostate Mapping Biopsies

Abstract

To evaluate the minimum disease burden of prostate cancer at which multiparametric magnetic resonance imaging (MRI) optimally performs. Between 2006 and 2008, 64 men underwent multiparametric MRI imaging (index test) followed by template prostate mapping biopsy (reference test). Three radiologists independently reported each quadrant of every prostate on a scale of 1 to 5: highly likely benign, likely benign, equivocal, likely malignant, highly likely malignant (≥3 or ≥4 was considered positive). There were 256 prostate sectors; bootstrapping adjustment was used to account for nonindependence. The target condition indicating cancer on biopsies was varied by changing the maximum cancer core length (MCCL) and total cancer core length (TCCL) within each sector from 1 mm to 10 mm. The sensitivity, specificity, and positive (PPVs) and negative predictive values (PPVs) were calculated for each MCCL and TCCL. Gleason ≤3+3 and Gleason ≥3+4 cancers were analyzed separately. Mean age was 62 years (range, 40-76 years), and mean prostate-specific antigen level was 8.2 μg/L (range, 2.1-43 μg/L). Fifty percent of quadrants (127 of 256) had prostate cancer, of which 65% (83 of 127) were Gleason ≤3+3. For Gleason ≤3+3, multiparametric MRI had an NPV of ≥95% at an MCCL of ≥5 mm and at a TCCL of ≥7 mm (MRI score ≥3). For Gleason ≥3+4, an NPV of ≥95% was seen at an MCCL of ≥5 mm (MRI score ≥3) and TCCL ≥6 mm. Multiparametric MRI may allow areas of the prostate which test negative to avoid biopsy. Whether multiparametric MRI can be used as a "triage" test before the first biopsy requires results from ongoing prospective validating cohort studies.