Comparison of two-view digital breast tomosynthesis to three-view digital mammography in a simulated screening setting

Background: Although digital breast tomosynthesis (DBT) may be an effective breast cancer screening modality, alternatives are important to explore as this technique presents some challenges for population screening implementation. Purpose: To compare the detection performance of three-view (craniocaudal [CC] + mediolateral oblique [MLO] + mediolateral/lateromedial [ML/LM]) digital mammography (DM) to two-view (CC + MLO) DBT. Material and Methods: In this institutional review board-approved study, recalled screening DBT exams from 3 March 2014 to 31 March 2015 with a ML/LM view acquired within 30 days were retrospectively reviewed for detection of breast cancer. Six radiologists blinded to outcomes evaluated two-view DBT, which included DM and tomosynthesis images, and three-view DM. Probability of malignancy and a BI-RADS score were assigned. Primary outcomes were area under the receiver operating characteristic curve (AUC) and benign and malignant recall rates. Results: The cohort included 121 cases: 37% were malignant and 63% were benign. Reader-averaged AUCs were 0.846 (95% confidence interval [CI] = 0.787–0.905) and 0.853 (95% CI = 0.802–0.904) for three-view DM and DBT, respectively. Benign and malignant recall rates were 53% and 74% for three-view DM and 52% and 75% for DBT, respectively. There was no difference in AUCs (P = 0.692), and benign (P = 1) and malignant (P = 1) recall rates. Conclusion: In this retrospective initial study, the detection performance of two-view DBT and three-view DM were statistically equivalent. Future studies should determine whether three-view DM is a viable alternative to DBT for screening, given its cost and work-flow.