Poor Interobserver Reproducibility in the Diagnosis of High-grade Endometrial Carcinoma

Patients with high-grade subtypes of endometrial carcinoma (grade 3 endometrioid, serous, clear cell, or carcinosarcoma) have a relatively poor prognosis. The specific subtype may be used to guide patient management, but there is little information on the reproducibility of subtype diagnosis in cases of high-grade endometrial carcinoma. Fifty-six cases diagnosed as a high-grade subtype of endometrial carcinoma were identified from the pathology archives of Vancouver General Hospital. All slides for each case were reviewed independently by 3 pathologists, who diagnosed the specific tumor subtype(s) and assigned the percentage of each subtype for mixed tumors. Agreement between observers was categorized as follows: major disagreement: (A) no consensus for low-grade endometrioid versus high- grade carcinoma (any subtype), or (B) no consensus with respect to the predominant high- grade subtype present; minor disagreement: consensus was reached about the cell type of the predominant component of a mixed tumor, but there was disagreement about the subtype of the minor component. A tissue microarray was constructed from these cases and immunostained for p16, ER, PR, PTEN, and p53. In 35 of 56 (62.5%) cases, there was agreement between all 3 reviewers regarding the subtype diagnosis of the exclusive (in pure tumors) or predominant (in mixed tumors) high- grade component. Of these cases, there was a minor disagreement (ie, disagreement about the minor high- grade component subtype in a mixed tumor) in 4 cases (4/56, 7.1%). In 20 of 56 (35.8%) cases there was a major disagreement; in 17 (30.4%) of these cases there was no consensus about the major subtype diagnosis, whereas in 3 (5.4%) cases there was disagreement about whether a component of high- grade endometrial carcinoma was present. In the final case, all 3 reviewers diagnosed the case as low-grade endometrioid carcinoma, disagreeing with the original diagnosis of high- grade carcinoma. The most frequent areas of disagreement were serous versus clear cell (7 cases) and serous versus grade 3 endometrioid (6 cases). Immunostaining results using the 5-marker immunopanel were then used to adjudicate in the 6 cases in which there was disagreement between reviewers with respect to serous versus endometrioid carcinoma, and these supported a diagnosis of serous carcinoma in 4 of 6 cases and endometrioid carcinoma in 2 of 6 cases. Pairwise comparison between the reviewers for the 20 cases classified as showing major disagreement was as follows: reviewer 1 and reviewer 2 agreed in 5/20 cases, reviewer 1 and reviewer 3 agreed in 7/20 cases, and reviewer 2 and reviewer 3 agreed in 8/20 cases, indicating that disagreements were not because of a single reviewer holding outlier opinions. Diagnostic consensus among 3 reviewers about the exclusive or major subtype of high- grade endometrial carcinoma was reached in only 35/56 (62.5%) cases, and in 4 of these cases there was disagreement about the minor component present. This poor reproducibility did not reflect systematic bias on the part of any 1 reviewer. There is a need for molecular tools to aid in the accurate and reproducible diagnosis of high- grade endometrial carcinoma subtype.