Breast ImagingE2981. Decreasing Screening Mammography Recall Rates With Consensus Recall: The Effect of Breast Density
Panigrahi B, Mullen L, Harvey S. Johns Hopkins Hospital, Baltimore, MD
Address correspondence to B. Panigrahi (email@example.com)
Objective: Standard recall rates in the United States for screening mammography are 10% or less. Recalls from both full field digital mammography (FFDM) and digital breast tomosynthesis (DBT) screening studies are costly, time consuming, and stressful for the patient. Consensus recall has been proposed as a method to decrease recall rates. However, the effect of breast density on the effectiveness on consensus recall has not been described.
Materials and Methods: We retrospectively reviewed all screening FFDM and DBT studies from January 2012 to April 2016, stratified by breast density. After a baseline period, imagers implemented consensus recalls in which two radiologists were required to make final decisions on whether or not to recall patients from screening mammography. Recall rates, cancer detection rate, and positive predictive values (PPV3) were obtained and stratified by breast density before and after the implementation of consensus recall. The two-tailed t test was used to distinguish differences between groups.
Results: During the baseline period from January 3, 2012, to February 3, 2015, 47,773 FFDM and 7780 DBT screening examinations were included. During the consensus recall period from September 4, 2015, to April 3, 2016, 2642 FFDM and 8240 DBT screening examinations were included. For FFDM studies, baseline recall rates were the lowest in fatty breasts (7.0%) and highest in heterogeneously dense breasts (14.0%). Implementation of consensus recalls decreased FFDM recall rates in all breast densities, though only statistically significant in scattered density breasts (10.61% decreased to 7.14%, p = 0.00044). For DBT studies, baseline recall rates were lowest in scattered density breasts (6.0%) and highest in extremely dense breasts (10.2%). Implementation of consensus recalls significantly decreased DBT screening recall rates in fatty breasts (6.9% decreased to 3.15%, p = 0.0029) and extremely dense breasts (10.22% decreased to 6.48%p, = 0.049). Notably, implementation of consensus recalls showed a trend of increased recall rates for scattered and heterogeneously dense breasts being evaluated with DBT. PPV3 increased in all breast density groups after the implementation of consensus recalls in both FFDM and DBT studies, implying more accurate criteria for biopsy were used during the screening process. Cancer detection rate did not differ significantly before and after implementation of consensus recall.
Conclusion: Awareness of how breast density and screening modality affects recall rates may guide the selective implementation of consensus recall. After implementation of consensus recall with FFDM, the recall rates for all breast densities declined. However, recall rates declined only for fatty and extremely dense breasts with DBT. This implies that consensus recall is an effective method to decrease recall rates in all breast densities using FFDM, but may not be as effective in decreasing recall rates in scattered or heterogeneously dense breasts using DBT.