Interobserver Reproducibility in Assessing Eosinophilic Cells in Ovarian Serous Borderline Tumors to Predict BRAF Mutational Status

Abstract

Ovarian serous borderline tumors (SBTs) harboring the BRAF V600E mutation are associated with decreased risk of progression to low-grade serous carcinoma, and often prominently feature tumor cells with abundant eosinophilic cytoplasm. Since eosinophilic cells (ECs) may be a marker of the underlying genetic driver, we proposed morphologic criteria and evaluated the interobserver reproducibility for assessing this histologic feature. Following the completion of an online training module, representative tumor slides from 40 SBTs (BRAF V600E-mutated, n=18, BRAF-wildtype, n=22) were independently reviewed by 5 pathologists. For each case, reviewers provided a semiquantitative assessment of the extent of ECs (0: absent, 1: <10%, 2: 10%–50%, or 3: >50%, of tumor area). Interobserver reproducibility for estimating the extent of ECs was moderate (κ=0.41). Applying a cut-off score of ≥2, the median sensitivity and specificity for predicting BRAF V600E mutation were 67% and 95%, respectively. With a cut-off score of ≥1, median sensitivity and specificity were 100% and 82%, respectively. Morphologic mimics of ECs, including tumor cells with tufting or hobnail change and detached cell clusters in micropapillary SBTs, were possible contributing factors for discordant interobserver interpretations. BRAF V600E immunohistochemistry showed diffuse staining in BRAF-mutated tumors, including those with few ECs. In conclusion, the finding of extensive ECs in SBT is highly specific for BRAF V600E mutation. However, in some BRAF-mutated SBTs, ECs may be focal and/or difficult to distinguish from other tumor cells with overlapping cytologic features. The morphologic finding of definitive ECs, even when scarce, should therefore prompt consideration for BRAF V600E mutation testing.