The recent development of microforceps for endoscopic ultrasound (EUS) through-the-needle biopsy (TTNB) of the wall of pancreatic cystic lesions (PCLs) allows to acquire histological specimens, which pathologists never handled and evaluated before. We aimed to estimate the interobserver agreement among pathologists in evaluating such samples.
TTNB slides from 40 PCLs with worrisome features were retrieved and independently evaluated for specimen adequacy, presence of lining epithelium, grade of epithelial dysplasia, presence of ovarian type stroma, and specific diagnosis by 6 expert pathologists from 6 different tertiary care centers. The Gwet's AC1 was used to assess interobserver agreement.
An almost perfect agreement was observed for specimen's adequacy (AC1, 0.82; 95% CI, 0.79;0.98), presence of lesional epithelium (AC1, 0.90; 95% CI, 0.86-0.92), epithelial dysplasia (AC1, 0.97; 95% CI, 0.95-0.99), ovarian-like stroma (AC1, 0.90; 95% CI, 0.86-0.93). When considering all diagnoses separately, a moderate to substantial agreement was observed (AC1, 0.62; 95% CI, 0.57-0.67), similarly to mucinous cysts versus serous adenoma versus other diagnoses (AC1, 0.65; 95% CI, 0.59-0.70) and for mucinous cysts versus all other diagnoses (AC1, 0.74; 95% CI, 0.68-0.84). The agreement for diagnosis of mucinous cystic neoplasm versus intraductal mucinous papillary neoplasm was almost perfect (AC1, 0.88; 95% CI, 0.81-0.95).
Interobserver agreement between expert pathologists in the evaluation of TTNB samples from PCLs with worrisome feature was close to perfection for all the evaluated parameters, except the definitive diagnosis. When mucinous cystic lesions were compared versus all other diagnoses, the agreement became substantial, thus indicating that TTNB specimen can provide important information for PCLs management decisions.