Implications of pathologist concordance for breast cancer assessments in mammography screening from age 40 years.

Three pathologists reviewed slides and reports of cancers arising in both the study and control populations of the U.K. trial of annual mammography screening from age 40 years. A total of 875 cases were scored independently as noninvasive, microinvasive, or invasive cancer, with the last also evaluated for histology grade, type, and lymphatic vascular invasion. Of these, 870 (99.2%) were confirmed malignant, 1 case had cytology only, and 5 were judged by all reviewers as benign. Reviewer complete concordance for the three classes of malignancy was achieved in 826 (95%) and majority agreement in 31 (3.6%) of 870 with complete data. All three readers recorded grade in 736 cancers, giving a kappa statistic of 0.69, 0.52, and 0.66 for grades I, II, and III, respectively, and 0.61 overall. Agreement that the cancer was special type or not was obtained in 671 (89.0%) with complete concordance in the nature of the type in 504 and majority view in 167; another 58 (7.7%) were characterised as "part special" pattern, with type disagreement in 23 (3%). The kappa statistic for single type subcategories in those cancers was substantial, at 0.68 overall. This improved to 0.76 for the last 230 invasive cancers after the pathologists agreed more explicit criteria for type discrimination. There was almost perfect agreement between original and review diagnosis of breast malignancy for both noninvasive/microinvasive and invasive cancer (kappa 0.84 and 0.91, respectively), justifying confidence in the diagnosis of breast cancer by U.K. pathologists. The specialists agreed substantially on qualitative histology features of type and grade of cancers, and improved further for typing by defining criteria. These consensus data, along with invasive size and node status, are reliable for use as surrogate measures of outcome, and to enhance interpretation of effect, when the trial case population sources are disclosed. Copyright 2002, Elsevier Science (USA). All rights reserved.