Interobserver Agreement in Dysplasia Grading of Intraductal Papillary Mucinous Neoplasms: Performance of Kyoto Guidelines and Optimization of Endomicroscopy Biomarkers through Pathology Reclassification.

Background and aims: Interobserver agreement (IOA) among pancreaticobiliary (PB) pathologists in evaluating high-grade dysplasia and/or invasive carcinoma (HGD-IC) of IPMNs remains understudied. EUS-guided needle-based confocal endomicroscopy (nCLE) can evaluate papillary architecture in branch-duct (BD)-IPMNs. We assessed IOA among PB pathologists in classifying dysplasia in resected IPMNs and compared the performance of the Kyoto guidelines' high-risk stigmata (HRS) and pre-surgical EUS-nCLE against reclassified pathology.

Methods: Subjects in prospective clinical trials (2015-2023) with resected IPMNs were included. Blinded PB-pathologists independently reviewed histopathology, achieving a consensus diagnosis. The accuracy of cyst fluid next-generation sequencing (NGS) analysis, EUS-nCLE, and Kyoto-HRS in predicting HGD-IC was compared with the reclassified pathology.

Results: Among 64 subjects, 25 (39%) exhibited HGD-IC (17 HGD, 8 IC). Disagreements occurred in 14% of cases with substantial IOA (kappa=0.70; 95%CI: 0.53-0.88) between two PB-pathologists for differentiating HGD-IC vs. low-grade dysplasia. To detect HGD-IC, the sensitivity, specificity, and accuracy of Kyoto-HRS and EUS-nCLE were 52%, 95%, 78%, and 68%, 87%, 80%, respectively. Integrating nCLE with Kyoto-HRS improved sensitivity to 80%, with specificity and accuracy at 82% and 81%, respectively. The sensitivity, specificity, and accuracy of NGS (n=47) to detect HGD-IC was 6.3%, 100%, and 68%, respectively. A unique subset of IPMNs were identified in all (n=8, p=0.01) cases where pre-surgical EUS-nCLE underestimated dysplasia revealing a distinct micropapillary architecture on post-surgical histopathology.

Conclusions: Despite substantial IOA among experienced PB-pathologists, a second pathologist's review may be warranted for dysplasia classification in IPMNs under certain circumstances. Incorporating an imaging biomarker such as EUS-nCLE with Kyoto-HRS improves sensitivity for HGD-IC without sacrificing accuracy.