OBJECTIVE:This study compared breast cancer screening outcomes between high resolution (HR; 70-micron) and standard resolution (SR; 100-micron) digital breast tomosynthesis (DBT) systems in real-world practice.
METHODS:This retrospective, observational cohort study included women ages 40-79 screened for breast cancer at a U.S. health system from 2013 to 2023. Cancer detection rate (CDR), recall rate (RR), and positive predictive value of recall (PPV1) were reported. The odds of each outcome following HR versus SR DBT, adjusted for age, race, density, risk status, prior resolution, facility, and radiologist, were estimated using logistic regression. Additional analyses assessed the impact of having prior mammograms on outcome measures.
RESULTS:A total of 184,006 mammograms were included (95,633 SR, 88,373 HR). The CDR was 5.38/1000 (HR) and 4.87/1000 (SR) (p = 0.1296). The increase in cancer detection with HR was statistically significant after adjusting for potential confounders (OR = 1.370, 95 % CI:1.117, 1.681). The RR was 9.80 % (HR) and 9.07 % (SR) (p < 0.0001), with an adjusted OR of 1.392 (95 % CI:1.327, 1.460). PPV1 was similar: 5.57 % (HR), 5.45 % (SR) (p = 0.2730). For exams with a known prior, the CDR was 5.38/1000 (HR) and 4.22/1000 (SR) (p = 0.0020), the RR was 9.39 % (HR) and 7.80 % (SR) (p < 0.0001), and the PPV1 was 5.74 % (HR) and 5.49 % (SR) (p = 0.0437). For HR exams with a known prior, the RR was 10.00 % (SR prior) and 9.14 % (HR prior) (p = 0.0001).
CONCLUSIONS:This large, real-world study demonstrated that HR DBT is associated with a higher CDR than SR DBT, with a greater increase in CDR for exams with a known prior.