Immunohistochemical Classification of the 5 Major Subtypes of Ovarian Carcinoma and the Introduction of the Calculator for Ovarian Subtype Prediction (COSP)
SE Kalloger, M Kobel, S Leung, E Mehl, D Gao, K Masson, DG Huntsman, CB Gilks. Vancouver General Hospital and the British Columbia Cancer Agency, Vancouver, BC, Canada; University of Calgary/CLS, Calgary, AB, Canada
Background: With the emerging evidence that the five major ovarian carcinoma subtypes comprise distinct entities, management of ovarian carcinoma will become subtype specific in the future. Although subtype assignment has been improved recently, it would be desirable to assist subtype assignment with objective molecular markers.
Design: Immunohistochemical expression data of 22 biomarkers were examined using nominal logistic regression in order to produce a subtype prediction model on a set of 322 archival ovarian carcinomas accrued from the British Columbia Cancer Agency between 1984 - 2000. This equation was validated on a cohort of 242 high quality ovarian carcinomas from the Gynaecologic Tissue Bank at Vancouver General Hospital collected between 2001 - 2006.
Results: Using a panel of 9 antibodies (CDKN2A, DKK1, HNF1B, MDM2, PGR, TFF3, TP53, VIM, WT1) morphological type can be predicted with relatively good sensitivity and specificity as shown in the table:
|Subtype||ROC AUC Prediction||Kappa Prediction||ROC AUC Validation||Kappa Validation|
|Clear Cell||0.9944||0.876 +/-0.022||0.9401||0.699 +/- 0.044|