Salivary gland neoplasms with basaloid features in the era of the Milan system for reporting salivary gland cytology: Classification and interobserver agreement.

BACKGROUND: The Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) has been shown to have moderate to good reproducibility for categorization of salivary gland fine-needle aspiration (FNA) specimens. Less is known of its accuracy and interobserver reproducibility for categorization of the diagnostically difficult group of basaloid neoplasms. METHODS: Forty-five salivary gland specimens with a basaloid morphology (pleomorphic and monomorphic adenomas and adenoid cystic carcinomas) were independently assigned by seven cytopathologists to one of the MSRSGC categories. Interobserver agreement was assessed for average agreement, chance expected agreement and by Cohen's kappa and diagnostic accuracy. Correlation of the salivary gland neoplasm of unknown malignant potential (SUMP) category with histologic diagnosis and benign or malignant designation along with interobserver reproducibility were calculated. RESULTS: Average observed agreement for assignment to the MSRSGC was 46% and Cohen's kappa = 0.2%. The SUMP category did not correlate with tumor type or with the benign or malignant nature of the neoplasm. Diagnostic specificity and sensitivity were 92% and 100% for consensus diagnosis, but were 76% and 77% for individual diagnoses. CONCLUSION: The interobserver agreement in categorizing basaloid neoplasms by the MSRSGC is poorer than for salivary gland lesions overall. This reflects the difficulty in diagnosing basaloid neoplasms. Nonetheless, diagnostic accuracy appears similar to that of salivary gland neoplasms as a whole.