Identifying high-grade features in patients with pancreatic mucinous neoplasms (MNs) is important for patient management. The reproducibility of MN cytology grading has been evaluated to a limited extent. In the current study, the authors evaluated interobserver variability in grading MNs and the identification of neoplastic mucin in endoscopic ultrasound-guided fine-needle aspiration specimens.
A 54-case grading set was created from histologically confirmed MNs (44 MNs) and nonmucinous lesions with abundant gastrointestinal contamination (10 nonmucinous lesions). Six observers received a tutorial, reviewed prescreened slides, and recorded: 1) a diagnosis according to a 6-tiered system (TS) (nondiagnostic, atypical [ATP], mucinous cyst low grade [MCLG], mucinous cyst high grade, suspicious for adenocarcinoma, and positive for adenocarcinoma); 2) the cyst fluid carcinoembryonic antigen diagnosis (CEADX); and 3) the presence of neoplastic musin. Interobserver agreement (IOA) was evaluated by calculation of kappa coefficients (Kappa). Diagnostic accuracy was not evaluated.
The IOA was lowest for the 6-TS (Kappa, 0.13; P<.001). The CEADX was available for 18 cases (33%), including 6 of 24 MCLG cases (25%). CEADX modestly improved IOA for combined tiers of the 6-TS with ATP and MCLG as separate categories. The highest IOA was noted with a 3-TS (nondiagnostic, ATP/MCLG, and mucinous cyst high grade/suspicious for adenocarcinoma/positive for adenocarcinoma [Kappa, 0.28; P<.001]) and various 4-TS (Kappa, 0.22-0.23). IOA was found to be low for neoplastic mucin (Kappa = 0.15; P<.001).
In a study using simulated cytology practice, observers demonstrated fair IOA for grading MNs and low IOA for identifying neoplastic mucin. Knowledge of the cyst fluid CEA level was found to modestly improve the IOA for low-grade lesions.