Misclassification rates in breast histopathology biopsies: how can they be lowered?
13 February, 2017:
Because misclassification of breast biopsies is relatively common, and no prior study had analyzed strategies for reducing error, the recent Elmore study is timely [1]. Here, 12 different strategies for acquiring second opinions were compared in order to help define which strategies worked best to reduce misclassification errors. The authors systematically tested whether and which pathology classification affected the best strategic choice for: invasive breast cancer, ductal carcinoma in situ (DCIS), atypia, proliferative without atypia, or benign without atypia. Also analyzed was the influence of the perceived case difficulty, the pathologists’ clinical volumes, and local institutional policy.
The test sample consisted of a set of 240 histopathology slides (one per case) that had previously undergone expert consensus to form a reference diagnosis. A total of 115 pathologists were assigned to one of four groups of 60 slides distributed according to the range of outcomes usually found. Each pathologist independently interpreted all 60 slides from one of the four sets of 60 breast biopsy specimens. The key findings were: (1) Over-interpretation of benign cases without atypia was cut in half (12.9% to 6.0%) by second opinions when initial diagnosis of atypia, DCIS and invasive cancer always included a second opinion; (2) atypia cases had the highest misclassification rate after a single interpretation (52.2%) which remained at more than 34% in every second opinion strategy tested; (3) excluding invasive breast cancer slides, the misclassification rates decreased (p < 0.001) from 24.7% to 18% when all of these received a second opinion; (4) high-volume pathologists (> nine cases per week) consistently delivered fewer misclassifications; (5) accuracy of diagnosis improved by the second opinion regardless of the pathologists’ confidence in their experience or diagnosis.
Read more …Misclassification rates in breast histopathology biopsies: how can they be lowered?