Histologic grading of breast carcinoma. A reproducibility study

Abstract

Background. A concern with the histologic grading of breast cancer is that tumor grading is a subjective evaluation that may have problems with reproducibility. Methods. A single slide from 10 invasive breast cancers was submitted to 25 pathologists who practice in six separate groups. Pathologists graded the tumors using a modified Bloom‐Richardson (B‐R) scheme, and the results were compared. Results. In 8 of the 10 cases, there was greater than 87% agreement by the pathologists as to the final combined B‐R grade, with complete agreement in 2 cases. Only one case had any discrepant opinions that ranged from low to high grade, and this involved only 3 of the 25 pathologists. With respect to B‐R score, the pathologists tended to score the tumors as either one of two adjacent scores. Due to this clustering, the B‐R scheme appears reproducible into five groups: very low and very high grade tumors and B‐R score “5,6,” “6,7,” and “7,8” tumors. This clustering was especially noticeable in two cases with split decisions, in which the discrepancy in final combined grade was largely due to the tumors being given B‐R scores that straddled and were then condensed into two B‐R grades. A consensus from each pathology group tended to merge with the majority opinion of all 25 pathologists and was correct for outliers. Conclusions. This study indicates that reproducibility of grading breast cancers can be achieved when a histologic grading scheme with specified guidelines is used. Pathologists must be aware of the limits of reproducibility, with appropriate guidelines being followed to help optimize agreement, and there should be an awareness of how pathologists group in their evaluations. Also, it may be advisable to better correlate or link reproducibility data with prognostic data in the design of grading schemes. Cancer 1994; 73:2765–70.