Early clinical experience with digital breast tomosynthesis for screening mammography.

PURPOSE To examine recall rates from screening mammography and the mammographic findings that caused recall in women who underwent digital breast tomosynthesis with conventional mammography (referred to as two-dimensional [ 2D two-dimensional ] with three-dimensional [ 3D three-dimensional ] imaging [ 2D two-dimensional + 3D three-dimensional ]) and in women who underwent conventional mammography alone (referred to as 2D two-dimensional ). MATERIALS AND METHODS This was an institutional review board-approved, HIPAA-compliant study with waivers of informed consent. A retrospective review of 2D two-dimensional + 3D three-dimensional and 2D two-dimensional screening mammograms from August 1, 2011, to December 31, 2012, was performed. Recall rates and abnormalities that caused recall were compared by controlling for differences in patient age, breast density, and risk factors. Cancer detection rate was assessed from this time period and from 1 year before the introduction of tomosynthesis for a historic control. RESULTS This study included 17 955 screening mammograms; of the total, there were 8591 (47.8%) 2D two-dimensional + 3D three-dimensional screening examinations and 9364 (52.2%) 2D two-dimensional examinations. The recall rate was 7.8% (671 of 8592) for 2D two-dimensional + 3D three-dimensional and 12.3% (1154 of 9364) for 2D two-dimensional (P < .0001); the rate of recall was 36.6% lower in the 2D two-dimensional + 3D three-dimensional group than in the 2D two-dimensional group. Recall rates for the 2D two-dimensional + 3D three-dimensional group were significantly lower for patients with asymmetries, ( 2D two-dimensional + 3D three-dimensional vs 2D two-dimensional , 3.1% [267 of 8591] vs 7.4% [689 of 9364], respectively; P < .0001) and calcifications ( 2D two-dimensional + 3D three-dimensional vs 2D two-dimensional , 2.4% [205 of 8591] vs 3.2% [297 of 9364], respectively; P = .0014). For patients with masses and architectural distortion, the difference in recall rates was not significant (masses: 2D two-dimensional + 3D three-dimensional vs 2D two-dimensional , 2.5% [215 of 8591] vs 2.5% [237 of 9364], respectively; P = .90; architectural distortion: 2D two-dimensional + 3D three-dimensional vs 2D two-dimensional , 0.68% [58 of 8591] vs 0.69% [65 of 9364]; P = .88). Cancer detection was highest in the 2D two-dimensional + 3D three-dimensional group at 5.9 cancers per 1000 examinations, with 5.7 cancers per 1000 examinations in the concurrent 2D two-dimensional group, and 4.4 cancers per 1000 examinations in the historic control. CONCLUSION Use of tomosynthesis ( 2D two-dimensional + 3D three-dimensional ) compared with conventional mammography ( 2D two-dimensional ) is associated with a lower recall rate of screening mammography, most often for asymmetries.