Background: Understanding the effect of specific experience in prostate MRI interpretation on diagnostic performance would help inform the minimum interpretation volume to establish proficiency. Objective: To assess for an association between increasing experience in prostate MRI interpretation and change in radiologist-level PPVs for PI-RADS version 2 (v2) categories 3, 4, and 5. Methods: This retrospective study included prostate MRI examinations performed between July 1, 2015 and August 13, 2021 that were assigned PI-RADS v2 category of 3, 4, or 5 and with MRI-ultrasound fusion biopsy available as reference standard. All examinations were among the first 100-200 prostate MRI examinations interpreted using PI-RADS v2 by fellowship-trained abdominal radiologists. Radiologists received feedback through a quality assurance program. Radiologists' experience levels were classified using progressive subsets of 50 interpreted examinations. Change with increasing experience in distribution of individual radiologists' whole-gland PPVs for Gleason sum score ≥7 prostate cancer, stratified by PI-RADS category, was assessed by hierarchical linear mixed models. Results: The study included 1300 prostate MRI examinations in 1,037 patients (mean age, 66±7 years), interpreted by 8 radiologists [median, 13 years of post-fellowship experience (range, 5-22 years)]. Aggregate PPVs were 20% (68/340) for PI-RADS category 3, 49% (318/652) for category 4, and 73% (220/308) for category 5. Interquartile ranges (IQRs) of PPVs overlapped for category 4 (51% [IQR: 42-60%]) and category 5 (70% [IQR: 54-75%]) for radiologists' first 50 examinations. IQRs of PPVs did not overlap between categories of greater experience; for example, at the 101-150 examination level, PPV for category 3 was 24% [IQR: 20-29%], category 4 was 55% [IQR: 54-63%], and category 5 was 81% [IQR: 77-82%]). Hierarchical modeling showed no change in radiologists' absolute PPV with increasing experience (category 3: p=.27; category 4: p=.71; category 5: p=.38). Conclusion: Absolute PPVs at specific PI-RADS categories did not change during radiologists' first 200 included examinations. However, resolution of initial overlap in IQRs indicates improved precision of PPVs after the first 50 examinations. Clinical Impact: If implementing a minimum training threshold for fellowship-trained abdominal radiologists, 50 prostate MRI examinations may be sufficient in the context of a quality assurance program with feedback.
AJR. American journal of roentgenology. 2022 Mar 23 [Epub ahead of print]
Bassel R Salka, Prasad R Shankar, Jonathan P Troost, Shokoufeh Khalatbari, Matthew S Davenport
University of Michigan School of Medicine, 1500 E. Medical Center Drive, Ann Arbor, MI 48109., Department of Radiology, Michigan Medicine, 1500 E. Medical Center Drive, Ann Arbor, MI 48109., Michigan Institute for Clinical and Health Research (MICHR), Michigan Medicine, 1500 E. Medical Center Drive, Ann Arbor, MI 48109.