Supervised clustering of immunohistochemical markers to distinguish atypical and non-atypical endometrial hyperplasia

Abstract

The risk of endometrial hyperplasia (EH) progressing into endometrioid endometrial cancer ranges from 1% for simple EH without atypia (EHWA) to 46.2% for atypical EH (AEH). Differentiation between both entities is crucial to determine optimal management. As preoperative diagnosis of AEH can be difficult, we aimed to establish clusters of immunohistochemical markers to distinguish EHWA from AEH. We studied 13 immunohistochemical markers (steroid receptors, pro/anti-apoptotic proteins, metalloproteinases (MMP), tissue inhibitor of metalloproteinase (TIMP), CD44 isoforms) known for their role in endometrial pathology. Using supervised clustering, we determined clusters of co-expressed proteins which contributed the most in differentiating EHWA from AEH. From 39 tissue samples (17 EHWA and 22 AEH), we found three clusters of co-expressed proteins: Cluster 1 included two proteins (over-expression of estrogen receptor (ER) and under-expression of progesterone receptor (PR) B in AEH compared to EHWA); Cluster 2: an ER, PR A, MMP-2 and TIMP-1 over-expression and a PR B and TIMP-2 under-expression; Cluster 3: over-expression of ER and MMP-7 and under-expression of PR B and TIMP-2. AEH can be accurately distinguished from EHWA using a supervised clustering of immunohistochemical markers. This promising approach could be useful to improve the preoperative diagnosis of EH.