[1130] The Diagnosis of Endometrial Carcinomas with Clear Cells by Gynecologic Pathologists: An Interobserver Variability Study

Oluwole Fadare, Vinita Parkash, William D Dupont, Geza Acs, Kristen A Atkins, Julie A Irving, Edyta C Pirog, Bradley J Quade, M Ruhul Quddus, Joseph T Rabban, Russell Vang, Jonathan L Hecht. Vanderbilt University, TN; Yale University, CT; Moffitt Cancer Center, FL; University of Virginia, VA; Royal Jubilee Hospital, BC, Canada; Cornell University, New York; Harvard University, MA; Brown University, RI; UCSF, CA; Johns Hopkins University, MD

Background: The purposes of this study are to assess the level of interobserver variability in the diagnosis of endometrial carcinomas (ca) with clear cells by gynecologic pathologists, and to describe the cases of clear cell ca that display this variability.
Design: 11 gynecologic pathologists (median experience: 10 years) from 11 institutions in the US and Canada rendered diagnoses on glass slides from 35 endometrial neoplasms in a blinded fashion and without predetermined diagnostic criteria. The cases were selected from the files of 3 institutions by 2 authors based on their having at least focal clear cells, and had previously been diagnosed as clear cell, endometrioid, serous & undifferentiated ca (CC, EC, SC & UC respectively) at those institutions.
Results: The overall study kappa coefficient (κ) was 0.46, indicative of a "moderate" level of interobserver agreement. κ statistics for CC, SC, EC, mixed CC/SC & mixed EC/CC were 0.6, 0.35, 0.58, -.02, & 0.36 respectively. "Confirmed" CC cases (diagnosed as such by ≥80% of observers; n=12; Group 1) tended to be morphologically "classic": clear cells in ≥33% of tumor (10/12), hyalinized papillae (HP, 6/12), hobnail cells (HC, 6/12), small round papillae (SP, 8/12), tubulocystic patterns (TP, 4/12). Cases diagnosed as CC by at least 1 (9%) but <80% of observers (Group 2; n=13) displayed some of the same features but less frequently: clear cells in ≥33% of tumor (6/13), HP (3/13), HC (2/13), SP (6/13), TP (0/13). The groups did not significantly differ in lesional size, mitotic index, or predominant nuclear grade. EC and mixed ca were the most common non-CC diagnoses in Group 2. 44 semi-quantified morphologic variables were analyzed in a logistic regression model, and some traditionally CC-associated features, including hyaline globules and hobnails, were not predictive of confirmed CC from the overall group of 35 cases.
Conclusions: The diagnosis and histotyping of endometrial carcinomas with clear cells, even by experienced gynecologic pathologists, is associated with a moderate level of interobserver variability. For CC, this variabilty is related to the subset of cases displaying morphologic overlap with other histotypes and/or a subtotal composite of diagnostic features.
Category: Gynecologic & Obstetrics

Monday, March 19, 2012 9:00 AM

Platform Session: Section E, Monday Morning


Close Window