Breast Imaging Reporting and Data System v2025 Auditing: Ensuring Outcome Ascertainment for More Reliable Global Benchmarking

We read with great interest the review by Eby and Cohen [1] describing the new auditing and outcomes-monitoring framework introduced in the Breast Imaging Reporting and Data System (BI-RADS) v2025 Manual. The recommendations to audit initial BI-RADS category 3 assessments, monitor preoperative breast magnetic resonance imaging (MRI) performed to determine the extent of disease, and prospectively document the initial method of detection represent important advances toward more comprehensive quality assurance. However, the reliability of these metrics depends on an additional element that deserves explicit consideration: the completeness of longitudinal outcome ascertainment.

A standardized ascertainment interval harmonizes the period during which breast cancer outcomes are evaluated, but it does not guarantee that all diagnoses occurring during that interval are captured. Patients may undergo follow-up imaging, biopsy, surgery, or treatment at another institution, particularly in fragmented healthcare systems without linkage to population-based cancer registries. If the absence of a locally recorded cancer diagnosis is interpreted as the absence of cancer, examinations may be incorrectly classified as true-negative or false-positive, thereby affecting sensitivity, specificity, positive predictive values, and cancer detection rates. Previous work has demonstrated that changes in follow-up attribution alone can materially alter false-negative classification and calculated sensitivity [2]. Differences between institutions may therefore reflect the completeness of data capture rather than true differences in radiologist performance.

This concern is particularly relevant to the proposed BI-RADS category 3 audit. The proportion of initial category 3 assessments resulting in malignancy is meaningful only when follow-up is sufficiently complete. In one study, only 60% of patients completed appropriate early follow-up, whereas 21% were lost to follow-up; adherence also varied according to sociodemographic characteristics [3]. Consequently, the observed malignancy rate may reflect not only appropriate category assignment but also access to care, patient adherence, and institutional tracking capacity. We suggest that category 3 audits additionally report follow-up completion at the recommended milestones, time to diagnostic resolution, the proportion of outcomes verified through external sources, and the proportion of patients with unknown outcomes. Patients with incomplete follow-up should neither be silently classified as not having cancer nor excluded without sensitivity analyses.

Similar considerations apply to auditing MRI for extent-of-disease evaluation. Positive predictive values and the additional cancer detection rate require reliable finding-level correlation with pathology. Selective biopsy, surgery performed elsewhere, or missing histologic outcomes may distort these measures. In addition to the proposed metrics, institutions could report the proportion of BI-RADS category 4 or 5 findings with definitive tissue correlation and selected downstream outcomes, including additional biopsies, changes in surgical management, delays to treatment, and reoperation. This distinction is important because identifying additional abnormalities does not necessarily demonstrate that imaging improved patient-centered outcomes, and preoperative MRI has been associated with changes in surgical treatment patterns [4].

Prospective assignment of the initial method of detection is also promising, and standardized templates have achieved high reporting rates in multisite healthcare systems [5,6]. Nevertheless, correct assignment may be difficult when prior imaging examinations or clinical histories are unavailable. Including an “unknown or insufficient information” quality flag and routinely reporting its frequency could prevent forced classification while identifying deficiencies in information exchange.

To support more reliable global benchmarking, we believe BI-RADS auditing would benefit from a minimum data-quality statement documenting outcome sources, linkage with pathology or cancer registries, completeness of follow-up, losses to follow-up, external care, and missing method-of-detection information. Such reporting would ensure that comparisons reflect imaging performance rather than differences in healthcare integration or data availability. By combining standardized metrics with transparent assessment of data completeness, BI-RADS v2025 can provide a stronger and more equitable foundation for quality improvement across diverse healthcare systems.