Assessing the Positive Predictive Value of Architectural Distortion Detected with Digital Breast Tomosynthesis in BI-RADS 4 Cases

Abstract

Objective The purpose of this study was to evaluate the positive predictive value of biopsy (PPV3) of architectural distortion (AD) detected on digital breast tomosynthesis (DBT) in BI-RADS 4 cases, where suspicion for malignancy remains broad. Methods This Institutional Review Board-approved, retrospective study included screening and diagnostic mammograms performed from August 2015 to December 2017 with DBT and digital mammography (DM) revealing suspicious AD with a BI-RADS 4 assessment. Medical records were reviewed for clinical data, imaging, and pathology results. Malignancy rate was assessed by lesion visibility on DM and DBT. Multivariate analysis was performed to assess the odds ratio (OR) of malignancy. Results A total of 63/179 cases were malignant, yielding a PPV3 of 35%. No significant difference in PPV3 was found by race, personal or family history of breast cancer, presence of microcalcifications, or mammogram type. Architectural distortion was more likely to be malignant when an US correlate was present (PPV3 49% vs 19%; P < 0.0001). Multivariate analysis demonstrated a 3-fold increased OR for malignancy with an US correlate present (P = 0.005). Lesion visibility analysis revealed a higher PPV3 for AD visible on DM-DBT compared with DBT alone (44% vs 26%; P = 0.01) and when an US correlate was present (DM-DBT 54% vs 30%, P = 0.02; DBT-only 43% vs 11%, P < 0.001). Conclusions Tomosynthesis-detected BI-RADS 4 AD are malignant in 35% of cases and are more likely to be malignant if an US correlate is present and if visible on both DM and DBT.