Moderated Poster

Poster, Podium & Video Sessions

MP08-11: Radiologist Experience Level Does Not Predict the Accuracy of Prostate MRI Interpretation for Clinically Significant Prostate Cancer: Are Consensus Reads the Answer?

Friday, May 12
9:30 AM - 11:30 AM
Location: BCEC: Room 160

Presentation Authors: Eric Kim*, Joel Vetter, Anup Shetty, Kathryn Fowler, Aaron Mintz, Cary Siegel, Gerald Andriole, Robert Grubb III, St. Louis, MO

Introduction: To provide standardization as prostate MRI becomes increasingly utilized, the Prostate Imaging-Reporting and Data System (PIRADS) was developed and has been modified to its latest version (v2). Using biopsy outcome as the standard, we examined the predictive accuracy of a PIRADS 4 or 5 read for clinically significant (Gleason 7+) PCa in a blinded fashion.

Methods: We reviewed our prospectively maintained database of consecutive men who underwent prostate MRI prior to biopsy between September 2014 and December 2015. A proportionally representative sample (based on the original clinical PIRADS v2 interpretation) was selected for re-examination (n=32). The prostate MRIs for these patients were de-identified and were loaded by a blinded third party. Four radiologists of varying levels of experience independently interpreted all prostate MRI, blinded to all clinical information. An &[Prime]over-read&[Prime] was defined as a PIRADS 4 or 5 read with biopsy result of benign prostate or Gleason 6 PCa. An &[Prime]under-read&[Prime] was defined as a PIRADS 1-3 read with resulting biopsy result of Gleason 7+ PCa.

Results: The distribution of accuracy is provided in Table 1. Accurate interpretation ranged from 56% (18/32) to 75% (24/32), and the differences among the radiologists were not significant (p=0.48). The improvement of accuracy with a &[Prime]majority read&[Prime], as defined by two or more accurate radiologists&[prime] blinded interpretations, over the original clinical read trends toward significance (p=0.16). No clinical variable was predictive of an incorrect &[Prime]majority read&[Prime], including age, PSA, family history, use of 5-alpha reductase inhibitors, prostate volume, or previous biopsy history.

Conclusions: In a blinded assessment of radiologists at our institution, we find that the predictive accuracy of PIRADS 4 or 5 for clinically significant PCa varies among radiologists independent of experience level. A &[Prime]majority read&[Prime] performed better than the original clinical interpretation, suggesting that consensus interpretation of prostate MRI may improve predictive accuracy.

Source Of Funding: None

Send Email for Eric Kim


Assets

MP08-11: Radiologist Experience Level Does Not Predict the Accuracy of Prostate MRI Interpretation for Clinically Significant Prostate Cancer: Are Consensus Reads the Answer?



Attendees who have favorited this

Send Email for Radiologist Experience Level Does Not Predict the Accuracy of Prostate MRI Interpretation for Clinically Significant Prostate Cancer: Are Consensus Reads the Answer?