Diagnostic performance of digital breast tomosynthesis (DBT) versus digital mammography (DM) in a population clinically referred for breast imaging – a retrospective cohort study

Purpose

To compare the performance of Digital Breast Tomosynthesis (DBT) with Digital Mammography (DM) in patients clinically referred for breast imaging.

Methods

Diagnostic performance of DBT (in 2016, after transition to DBT) was compared with DM (in 2011, before the transition) in consecutively referred patients (N = 10,742 exams). Reference standard was outcomes from all tests including histopathology and clinical review within the same year. Primary outcome was area under receiver operating characteristic curve (AUC-ROC).

Results

Cancer rates did not differ between DBT (1.72 % (CI 1.38–2.15 %)) and DM (1.71 % (CI 1.40–2.08 %)). AUC-ROC was similar for DBT (0.91 (CI 0.87–0.95)) and DM (CI 0.91 (0.88–0.95)). Abnormal interpretation rate for DBT was 2.83 % (CI 2.38–3.36 %) and for DM it was 2.17 % (CI 1.82–2.58 %), and the biopsy rate for DBT was 8.2 % (CI 7.4–9.0 %) and for DM it was 9.9 % (CI 9.1–10.6 %)).

In patients with dense breasts (vs overall cohort) AUC-ROC and sensitivity were lower for both DM and DBT. Within this subgroup, AUC-ROC for DBT was 0.90 (CI 0.84–0.95) and for DM it was 0.85 (CI 0.79–0.92), sensitivity was 78.2 % (CI 65.0–88.2 %) for DBT and 64.8 % (CI 50.6–77.3 %) for DM, and ultrasound was accurate whether it was used with DBT (AUC-ROC 0.95 (CI 0.91 – 0.99)) or DM (AUC-ROC 0.95 (CI 0.90 – 0.99))

Conclusion

In clinically referred patients, diagnostic accuracy and diagnostic yield were similar between DBT and DM. DBT may have a higher abnormal interpretation rate but a lower biopsy rate. DBT may be more accurate than DM in dense breasts.