Background: In the NSAS-BC protocol, the nuclear atypia and mitotic counts are to be judged by pathologists at each participating hospital for assessing high-risk node-negative breast cancers. Therefore, maintenance of interobserver agreement in diagnosis at a higher level is mandatory during the period of patient entry.
Methods: Individual collaborating pathologists originally evaluated the histological eligibility of 107 cases. Three panel pathologists determined consensus diagnoses and 29-37 collaborating pathologists determined modal diagnoses of these cases at three slide conference sessions. The original diagnoses were compared with the consensus and modal diagnoses to estimate the percentage of erroneous judgments.
Results: The agreement rate in histological type and nuclear atypia score was 69% (74/107) between the original and consensus diagnoses, 76% (81/107) between the original and modal diagnoses and 86% (92/107) between the consensus and modal diagnoses. The strength of interobserver agreement at the slide conference sessions was moderate (0.447-0.535) by kappa statistics. The original, consensus and modal diagnoses were concordant in 71 cases (66%), but were discordant in 36. Of 35 invasive ductal carcinomas with discordant diagnoses, the discordance arose from the intermediate tumor nature in 15, multiple factors in 13 and erroneous diagnosis in seven (6.5%), if the characteristics of the tumor were judged from the percentage interobserver agreement per tumor at the slide conferences.
Conclusion: Nuclear atypia scoring given at individual hospitals on case entry was almost reproducible among the pathologists. Continuous efforts are needed to improve interobserver agreement and to decrease erroneous diagnosis for protocol eligibility.