The diagnosis of endometrial carcinomas with clear cells by gynecologic pathologists: An assessment of interobserver variability and associated morphologic features

The purposes of this study are to assess the level of interobserver variability in the diagnosis of endometrial carcinomas with clear cells by gynecologic pathologists based purely on their morphologic features and to comparatively describe the cases of putative clear cell carcinoma (CCC) with and without significant interobserver variability. A total of 35 endometrial carcinomas (1 slide per case) were reviewed by 11 gynecologic pathologists (median experience: 10 y) from 11 North American institutions. The cases were selected from the files of 3 institutions on the basis of the presence of at least focal clear cells and had previously been classified as a variety of histotypes at these institutions. Diagnoses were rendered in a blinded manner and without predetermined diagnostic criteria or categories. The κ values between any pair of observers ranged from 0.18 to 0.69 (combined 0.46), which was indicative of a "moderate" level of interobserver agreement for the group. Subgroups of "confirmed CCC" [cases diagnosed as such by at least 8 (73%) of the 11 observers, n=14] and "possible CCC" (cases diagnosed as CCC by 1 but <8 observers, n=13) were compared with regard to a variety of semiquantified morphologic features. By combining selected morphologic features that displayed statistically significant differences between the 2 groups on univariate analyses, the following approximate morphologic profile emerged for the confirmed CCC group: papillae with hyalinized cores in 33% of the lesion, clear cells in 33% of the lesion, hyperchromasia in 33% of the lesion, the absence of nuclear pseudostratification in >3 cells on the papillae, the absence of nuclear pseudostratification in glands in 33% of the lesion, the absence of diffuse grade 3 nuclei, the absence of long and slender papillae in 33% of the lesion, and glands and papillae lined by cuboidal to flat, noncolumnar cells. In a backward stepwise logistic regression analysis, features from the profile that predicted the confirmed CCC group included: (1) absence or minimality of diffuse sheets of grade 3 nuclei [P=0.025; 95% confidence interval (CI), 0.0266-0.363]; (2) absence or minimality of nuclear stratification in glands and papillae (P=0.040; 95% CI,-0.228 to-0.0054); and (3) glands and papillae lined by cuboidal to flat, noncolumnar cells (P=0.008; 95% CI, 0.0911-0.566). The 2 groups displayed significant overlap regarding a wide variety of features, and no single case displayed a full complement of potentially diagnostic features. Morphologic patterns associated with cases with very high levels of interobserver variability (defined as cases with 4 different diagnoses rendered for them, n=9) included the near-exclusive or exclusively solid pattern of clear cells (3/9) and glandular/papillary proliferations whose only CCC-like feature was the presence of clear cells (2/9). In conclusion, the diagnosis of endometrial carcinomas with clear cells by gynecologic pathologists is associated with a moderate level of interobserver variability. However, there is a morphologic profile that characterizes cases that gynecologic pathologists more uniformly classify as CCC, and the presence of these features is supportive of a CCC diagnosis in an endometrial carcinoma with clear cells. Cases that display broad and significant qualitative deviations from the aforementioned profile should prompt the consideration of a diagnosis other than CCC.