Feasibility of Auditing Preoperative Breast MRI for Extent-of-Disease Evaluation Using the BI-RADS v2025 Manual

Background The forthcoming Breast Imaging Reporting and Data System (BI-RADS) version 2025 (hereafter, v2025) manual recommends auditing breast MRI examinations performed to evaluate the extent of disease in women newly diagnosed with breast cancer. Purpose To investigate the feasibility of this recommended audit using preoperative breast MRI examinations performed over 2 years, report the calculated metrics, and identify any clinical or imaging characteristics associated with improved MRI performance metrics. Materials and Methods Consecutive breast MRI examinations performed from January 2021 to December 2022 at The University of Texas MD Anderson Cancer Center were retrospectively reviewed, excluding screening, diagnostic, neoadjuvant chemotherapy response, and nondiagnostic or incomplete preoperative MRI examinations. Recommended audit calculations were performed as described in the BI-RADS v2025 manual: abnormal interpretation rate (AIR) (percentage of positive preoperative MRI examinations), positive predictive value of biopsies recommended (PPV2) (percentage of all distinct BI-RADS category 4 or 5 findings, not examinations, with a tissue diagnosis of cancer at image-guided biopsy), positive predictive value of biopsies performed (PPV3) (percentage of BI-RADS category 4 or 5 findings, not examinations, with preoperative image-guided biopsy revealing a tissue diagnosis of cancer), and additional cancer detection rate (ACDR) (number of additional malignant lesions diagnosed beyond the primary cancer per 1000 MRI examinations). Total time collecting data and performing calculations was recorded. Characteristics of benign and malignant BI-RADS 4 or 5 lesions were compared using the Wilcoxon rank sum test and Fisher exact test. Logistic regression was used to determine characteristics associated with malignancy. Results A total of 1533 preoperative MRI examinations were included (median patient age, 54.6 years; IQR, 45.5-65.2 years; all female), with 464 MRI examinations revealing 609 suspicious breast findings. Subsequent image-guided biopsy for 432 of these findings revealed 139 malignancies. The overall AIR was 30.3% (95% CI: 28.0, 32.6), overall PPV2 was 22.8% (95% CI: 19.5, 26.1), overall PPV3 was 32.2% (95% CI: 27.8, 36.6), and overall ACDR was 90.7 per 1000 examinations (95% CI: 74.3, 107.1). Seven radiologists spent a combined total of 141 hours collecting data and performing audit calculations. Older patient age (odds ratio, 1.04; 95% CI: 1.02, 1.05; P < .001) and presence of lesions ipsilateral to the index malignancy (odds ratio, 1.81; 95% CI: 1.20, 2.75; P = .005) were associated with higher odds of malignancy. Conclusion Auditing of preoperative breast MRI examinations performed to evaluate disease extent, as recommended in the forthcoming BI-RADS v2025 manual, was feasible and yielded metrics within expected ranges based on limited prior data. © RSNA, 2025.