Abstract: PD61-06
Sources of Funding: This investigation was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under award number UL1TR001105.

Introduction

_x000D_ Preoperative detection of extraprostatic extension (EPE) in prostate cancer (PCa) may affect clinical management and surgical planning. Qualitative descriptors commonly used in multiparametric magnetic resonance Imaging (mpMRI) reports have not been systematically validated and are potentially difficult to reproduce. This study aimed to determine the inter-reader agreement and diagnostic performance of qualitative descriptors versus a 5-point Likert scale for determination of EPE._x000D_

Methods

_x000D_ This was an IRB-approved, HIPAA-compliant, single-center, retrospective study with 3 experienced and 2 non-experienced readers. Men who underwent mpMRI of the prostate followed by radical prostatectomy between Nov/2015 and Jul/2016 were eligible. Whole-mount prostatectomy specimen processed with a 3D-printed, patient-specific mold for precise anatomical registration was the standard of reference. Reviewers chose one or more of 8 qualitative descriptors (e.g., capsular bulging, irregular margin) and, after a washout period, assigned a Likert score for the likelihood of EPE: 1, highly unlikely; 2, unlikely; 3, indeterminate; 4, likely; 5, highly likely. Reproducibility among reviewers was assessed with weighted kappa statistics (<0,no agreement; 0-0.20 slight, 0.21-0.40 fair, 0.41-0.60 moderate, 0.61-0.80 substantial, and 0.811 almost perfect). Cochran-Armitage Trend test was used to test the association bewteen pathology-proven EPE and MRI-based Likert score._x000D_

Results

_x000D_ Eighty men met eligibility criteria; mean age: 64 years, PSA: 8.0 ng/mL; prostate volume: 39.9 cc; Histologic index lesion size: 22±9 mm; Gleason score≤ 3+4, 62.5%; ≥4+3, 37.5%. EPE was present in 40(50%) men. Qualitative descriptors had inconsistent reproducibility (kappa as low as 0.33 for experienced and -0.02 for inexperienced readers) and poor accuracy (as low as 0.41). Agreement was moderate for experienced (k=0.56) and non-experienced (0.46) readers with the Likert scale. There was significant increase in the likelihood of EPE with higher Likert scores (Figure 1). _x000D_

Conclusions

_x000D_ A 5-point Likert scale improves inter-reader reproducibility and the diagnostic performance of mpMRI compared to qualitative descriptors of EPE, facilitating informed decision making, treatment planning and patient counseling._x000D_ _x000D_

Funding

This investigation was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under award number UL1TR001105.