Diagnosis of pancreatic cystic lesions (PCLs) remains challenging. EUS with FNA is limited by sampling error and nondiagnostic cytology. Needle-based confocal laser endomicroscopy (nCLE) performed during EUS can be used to improve diagnostic yield via FNA by providing in vivo histology of PCLs. However, the interobserver agreement (IOA) of nCLE of PCLs has yet to be studied.

Fifteen deidentified nCLE video clips of PCLs were sent to 6 interventional endoscopists at 5 institutions. Six variables were assessed for IOA: presence or absence of (1) vessels, (2) villi, (3) dark clumps, (4) reticular pattern, (5) acinar cells pattern, and (6) debris. PCL interpretation was categorized as mucinous, serous, pseudocyst, malignant, or indeterminate and final diagnosis as benign, malignant, or indeterminate.

IOA ranged from “poor” to “fair.” The K statistics were –.04 (SE = .05) for vessels, .16 (SE = .07) for villi, .22 (SE = .06) for dark clumps, .13 (SE = .06) for reticular pattern, .14 (SE = .06) for acinar cells pattern, .06 (SE = .06) for debris, .15 (SE = .03) for interpretation, .13 (SE = .05) for final diagnosis, and .19 (SE = .05) for image quality. The final diagnosis was malignant (10), benign (13), and indeterminate (2). The mean accuracy of the observers was 46%, with the lowest being 20% and highest being 67%.

The IOA and accuracy for PCL diagnosis were low. The results of this study support the need to identify and validate imaging criteria to determine whether nCLE has diagnostic value for pancreatic pathology. (Clinical trial registration number: NCT02166086.)