While numerous studies have reported outstanding accuracy of mpMRI, most have come from high-volume institutions using radiologists with particular expertise in prostate MRI. As mpMRI becomes more widely adopted, we sought to evaluate the performance of mpMRI when used in routine clinical care by radiologists with varying experience in prostate MRI.
We identified all men at our institution who underwent mpMRI and targeted biopsy over a 2-year period and evaluated the distribution of PIRADS scores and cancer yield. MRIs were read by one of nine radiologists who all had experience in body imaging. Radiologists varied in years of prostate MRI experience (median =6, range 1-25). Four of the nine had specific prostate MRI training. All MRIs were read as part of routine clinical workflow and not re-read prior to biopsy.
We found striking differences in the distribution of PIRADS scores (e.g. certain radiologists classified a higher proportion of lesions as PIRADS 5) and cancer yield according to PIRADS score (Figure 1) across the 9 radiologists. After adjusting for known prognostic variables (e.g. age, PSA, biopsy indication) both the PIRADS score and the radiologist who read the MRI remained significant. Interestingly, we found that radiologist performance did not significantly improve over time, despite the increase in experience gained over 2 years. Finally, we found no significant difference in performance between the higher-volume radiologists (>60 cases) compared to the lower-volume radiologists (<25 cases).
Our study results suggest that mpMRI interpretation in routine clinical care may vary significantly amongst radiologists, and that radiologist performance might not improve with increased experience alone. Our study highlights the importance of internal validation of published results at centers using mpMRI for clinical decision-making and suggests that specific mpMRI training including structured feedback may be necessary to maximize the success of mpMRI interpretation.
Figure 1. This figure displays the per-lesion cancer yield for each radiologist by PIRADS score for clinically significant cancer (A) and any prostate cancer (B). Each dash represents the performance of an individual radiologist. For example, the clinically significant cancer yield for PIRADS 5 lesions ranges from 40-80% across radiologists. The width of the gray shading associated with each dash reflects the number of lesions classified as that PIRADS score by that radiologist.
Written by: Nancy N. Wang and Geoffrey A. Sonn
References:
- Ahmed HU, Bosaily AES, Brown LC, Gabe R, Kaplan R, Parmar MK, Collaco-Moraes Y, Ward K, Hindley RG, Freeman A, Kirkham AP, Oldroyd R, Parker C, Emberton M. Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study. Lancet. 2017 Feb; 389(10071): 815-22.
Read the Abstract