Cancers found at breast screening: is the radiologist giving the surgeon all the information he needs in the referral letter?

Abstract

Introduction Digital breast tomosynthesis (DBT) may improve the accuracy of mammography by enabling visual separation of overlapping tissues (Andersson et al. 2008, Poplack et al. 2007). Methods Following local research ethics approval, all women attending the assessment clinic for evaluation of a mammographic abnormality found on routine screening (film-screen) were invited to take part in the study subject to informed consent. Participants underwent bilateral two-view two-dimensional (2D) digital mammography and bilateral twoview DBT. Mammography scores using the RCR Breast Group classification were sequentially obtained for the screening mammogram, 2D digital and DBT, and these were each compared with the final assessment outcome. Results Ninety-one percent of eligible women participated. Results from the first 300 participants are presented in Table 1 below. Kappa coefficients for agreement of each imaging method with final assessment outcome were calculated. Screening mammograms had the lowest agreement with the final outcome (Kappa = 0.02; P = 0.22), 2D digital mammography was better (Kappa = 0.26; P = 0.0000) and DBT had the highest score (Kappa = 0.37; P = 0.0000). Conclusion The preliminary results of this ongoing study show that DBT increases the diagnostic confidence of the radiologist. This supports the need for a larger multicentre study. Markers of screen reading performance (overall and first reader cancer detection rates, recall rates, positive predictive value of recall and missed cancers) were compared with volume of films read. Readers were categorised into four groups, according to film reading volume over the 3-year period: <15,000 (that is, on average less than the recommended 5,000/year); 15,000 to <20,000; 20,000 to <25,000; and ≥25,000. Statistical analysis was undertaken using SPSS for Windows version 13. Results The recall rate in low volume readers (<5,000/year) was 6.9% and was significantly higher than in the other groups combined (4.8%; P ≤ 0.001). These readers also had a lower positive predictive value than higher volume readers (11.7% versus 15.7%, P ≤ 0.001).The cancer detection rate at first read was significantly lower in the higher volume readers (≥25,000) in comparison to the other groups combined (6.6 per 1,000 versus 8.2 per 1000, P ≤ 0.001). Conclusion These data support the recommendation that readers should read a minimum of 5,000 mammograms/year. They also suggest that there is an upper limit above which reader performance deteriorates (in terms of cancer detection). With the imminent programme expansion this has implications for service quality. Consideration should be given to the introduction of an upper limit of mammographic reads. O2 Breast Cancer Research