Additional Breast Cancer Detection at Digital Screening Mammography through Quality Assurance Sessions between Technologists and Radiologists

Background: Screening technologists may function as readers in breast cancer screening programs. In the Netherlands, they attend quality assurance sessions. The frequency and characteristics of additional breast cancers detected through these sessions have not been reported.

Purpose: To determine the frequency and characteristics of cancers detected through quality assurance sessions.

Materials and Methods: This secondary analysis of a prospective cohort included 466 647 screening mammograms obtained between January 1, 2009, and January 1, 2017. Mammograms were single read by certified screening technologists before being double read by two certified screening radiologists who were not blinded to the technologists' reading. The technologists and a coordinating screening radiologist regularly discussed mammograms that the technologists considered suspicious but that did not prompt recall at radiologist double reading. The coordinating radiologist decided whether secondary recall was indicated. During a 2-year follow-up, radiologic and pathologic outcome data for all recalled women were obtained. Characteristics of cancers detected at radiologist double reading and those detected through quality assurance sessions were compared by using chi(2) and Fisher exact tests.

Results: A total of 14 142 women (mean age, 59 years +/- 7.8 [standard deviation]; range, 49-75 years) were recalled (recall rate, 3.0% [14 142 of 466 647]): 14 057 after radiologist double reading and 85 by the coordinating radiologists after quality assurance sessions. This resulted in 3156 screening-detected cancers (6.8 cancers detected per 1000 screenings), of which 26 (0.8% of screening-detected cancers [26 of 3156]) were detected after secondary recall through quality assurance sessions. The latter comprised eight ductal carcinomas in situ (88% intermediate or high grade [seven of eight]) and 18 invasive cancers (14 T1a-c and four T2+ cancers, 89% Nottingham grade I or II [16 of 18]). No significant differences in tumor characteristics were found (P values ranging from.22 to.95). Sensitivity of quality assurance sessions for additional cancer detection was 52% (26 of 50; 95% confidence interval: 38%, 66%).

Conclusion: The role of quality assurance sessions in additional cancer detection is limited. Tumor characteristics did not differ significantly from those of cancers detected at radiologist double reading. (C) RSNA, 2020.