MRI of the prostate: interobserver agreement compared with histopathologic outcome after radical prostatectomy

The purpose of this study was to evaluate interobserver agreement of prostatic MRI in assessing the performance of staging prostate carcinoma in comparison with histopathologic step section prostate specimens.

We retrospectively evaluated 46 patients who underwent prostatic MRI examination at 1.5 T MRI and "subsequently" radical prostatectomy. All MR-images were reevaluated by two different experienced radiologists (15 and 1.5 years of experience) with special focus on T2/T3 differentiation. Both radiologists were not aware of the patient's clinical data, except that the patient had prostate cancer. These findings were compared with histopathologic whole mount step section prostate specimens, which served as the "gold standard". Fourfold tables were created to calculate sensitivity, specificity, positive and negative predictive values and efficiency for T2/T3 differentiation. Cohen's kappa was calculated to measure inter-rater agreement.

Twenty-eight patients were diagnosed with organ defined cancer (T2), 18 patients were staged with extracapsular extension (T3), and thereof 7 patients were staged with seminal vesicle invasion (T3b) by the pathologists. The experienced reader reached a sensitivity of 77.78% (95%-CI 52.36%; 93.59%) and specificity of 92.86% (95%-CI 76.50%; 99.12%) for T2/T3 differentiation, the less experienced reader however achieved a sensitivity of 33.33% (95%-CI 13.34%; 59.01%) and specificity of 71.43% (95%-CI 51.33%; 86.78%). The Cohen's kappa for inter-rater reliability for differentiation between T2 and T3 stage was κ=0.0129.

Evaluation of prostatic MR imaging requires lengthy experience for accurate interpretation and staging. While a highly experienced reader can achieve good correlation with histopathology even without utilization of functional MR imaging, a less experienced reader with theoretical knowledge falls short of expectation.