Abstract

Reproducible diagnosis of ovarian carcinoma cell types is critical for cell type-specific treatment. The purpose of this study was to test the reproducibility of cell type diagnosis across Canada. Analysis of the interobserver reproducibility of histologic tumor type was performed among 6 pathologists after brief training in the use of modified World Health Organization criteria to classify ovarian carcinomas into 1 of 6 categories: high-grade serous, endometrioid, clear cell, mucinous, low-grade serous, and other. These 6 pathologists independently reviewed a test set of 40 ovarian carcinomas. A validation set of 88 consecutive ovarian carcinomas drawn from 5 centers was subject to local review by 1 of the 6 study pathologists, and central review by a single observer. Interobserver agreement was assessed through calculation of concordance and kappa values for pair-wise comparison. For the test set, the paired concordance between pathologists in cell type diagnosis ranged from 85.0% to 97.5% (average 92.3%), and the kappa values were 0.80 to 0.97 (average 0.89). Inclusion of immunostaining results did not significantly improve reproducibility (P=0.69). For the validation set, the concordance between original diagnosis and local review was 84% and between local review and central review was 94%. The kappa values were 0.73 and 0.89, respectively. With a brief training exercise and the use of defined criteria for ovarian carcinoma subtyping, there is excellent interobserver reproducibility in diagnosis of cell type. This has implications for clinical trials of subtype-specific ovarian carcinoma treatments.