Radiologic–Pathologic Discordance After Image-Guided Breast Biopsy: A Systematic Review of Prevalence and Outcomes

Background: Radiologic–pathologic discordance remains an important concern owing to the absence of standardized guidelines. This systematic review aimed to summarize the prevalence of discordant benign outcomes, defined as suspicious imaging findings with benign biopsy histology insufficient to explain the imaging abnormality, and to quantify malignancy upgrades subsequent to additional tissue assessment. Methods: This review was conducted in accordance with PRISMA 2020 guidelines and was prospectively registered. Eligible studies reported primary patient-level or aggregated data on radiologic–pathologic correlations post-image-guided breast biopsy and provided extractable data on discordant benign prevalence and/or subsequent malignancy upgrades. Results: Twenty-three studies were included. Lesion-/biopsy-based cohorts focused on biopsied abnormalities for analysis. Twelve studies directly estimated discordant-benign prevalence, whereas 11 studies did not, as study designs were discrepant-only, lesion-defined, or excision-restricted. Unselected cohorts with a cohort-wide correlation reported 1.2–5.3% discordant benign prevalence for all biopsies. When restricted to excised lesions, the discordant benign ascertainment rate was 7.4%, representing an excision-ascertained subset rather than the cohort-wide prevalence. Using benign-biopsy denominators, the discordance rate was 1.5–19.2%. Malignancy upgrades among discordant benign lesions ranged from 0 to 100% in selected subsets; however, several clinically relevant cohorts reported representative rates of approximately 20–40%, with some high-risk cohorts exceeding 50%. Conclusions: Discordant benign biopsy results are rare in unselected biopsy populations but carry a clinically meaningful upgrade risk, which warrants structured radiologic–pathologic correlation and prompt diagnostic resolution through repeat sampling or excision. Improvements in comparability and management algorithms require standardized definitions, uniform denominators aligned with all biopsied lesions, and prospective multicenter designs.