PurposeBreast magnetic resonance imaging (MRI) can detect some malignant lesions that are not visible on mammography (MX) or ultrasound (US). If a targeted, second-look fails, MRI-guided breast biopsy is the only available tool to obtain a tissue sample and pathological proof of these “MRI-only lesions”. The aim of this study is to report the performance and underestimation rate of 9G MRI-guided vacuum-assisted breast biopsy (VABB) over 12 years at a single center.Material and methodsAll 9G MRI-VABB procedures performed from January 2010 to December 2021 were retrospectively reviewed. Two MRI scanners (1.5 T and 3 T) were used with the same image resolution and contrast media. All suspicious lesions detected only by breast MRI underwent biopsy. Reference standard was histological diagnosis or at least 1-year negative follow-up. All malignant and atypical lesions underwent surgery, which was used as the reference standard.ResultsA total of 293 biopsies were retrospectively reviewed. Histopathological VABB results revealed 142/293 (48.4%) benign lesions, 77/293 (26.2%) high-risk lesions, and 74/293 (25.2%) malignant lesions. No significant complications were observed. Surgical pathology results allowed for the reclassification of n = 7/48 B3b lesions: n = 4 were ductal carcinoma in situ, while n = 3 presented invasive features at surgical histology (2 IDC; 1 ILC). B3b underestimation occurred overall in 14.6% of B3 cases. Breast follow-up was achieved for all benign VABB results, and only one false-negative case was observed.ConclusionOur results confirm that 1.5 T and 3 T MRI-guided VABB is an accurate and safe procedure for histopathologic final diagnosis of MRI-only lesions. Critical issues remain the potential high-risk underestimation rate of B3b VABB results and management of follow-up of benign lesions.