Comparative Accuracy of Digital Breast Tomosynthesis and Digital Mammography for Surveillance After Breast Cancer Treatment

Background: Digital breast tomosynthesis (DBT) has been less extensively evaluated for surveillance after breast cancer treatment than for screening. We compared the performance of DBT and digital mammography (DM) for surveillance of asymptomatic women with a personal history of breast cancer (PHBC), using US clinical practice data adjusted for population differences. Methods: We conducted a retrospective cohort study of 28,042 women with PHBC and posttreatment surveillance mammograms performed between 2012 and 2018 at 37 facilities within the Breast Cancer Surveillance Consortium. We calculated performance measures and absolute differences (ADs) for DBT versus DM, including cancer detection rate (CDR), invasive CDR, interval cancer rate, recall rate, and false-positive biopsy recommendation rate, sensitivity, specificity, and positive predictive value (PPV1), with 95% confidence intervals, adjusting for population characteristics using overlap weighting. Results: Among 80,077 mammograms (DBT, n=27,852; DM, n=52,225) and 1,098 subsequent breast cancer diagnoses, DBT had a higher CDR (AD, 2.3 per 1,000 95% CI, 0.3 to 4.3), invasive CDR (AD, 2.9 per 1,000 95% CI, 1.0 to 4.7), and PPV1 (AD, 5.2% 95% CI, 2.5% to 7.8%) and lower recall rate (AD, –1.0% 95% CI, –1.5% to −0.5%) compared with DM. No significant differences overall were identified for interval cancer rate or other performance measures. Surveillance-detected cancers with DBT or DM had similar proportions of favorable prognostic characteristics (smaller size, node-negative, and estrogen receptor–positive). Conclusions: For women with PHBC, DBT had better detection metrics compared with DM, as demonstrated by higher CDR, higher invasive CDR, and lower recall rate. However, the lack of significant reduction in interval cancer rate suggests that both DBT and DM may be used for posttreatment surveillance. Identifying effective imaging surveillance strategies for reducing interval cancer rates is needed to further improve posttreatment outcomes.