Unravelling interobserver variability in gastrointestinal glandular neoplasia: a contemporary study of US and Korean pathologists

Abstract

Aims: Interobserver variability in the assessment of gastric neoplasia biopsies between most Western and Eastern (predominantly represented by Japanese in the literature) pathologists has been documented. It is unknown if such variability exists between the US and Korean pathologists in the current era.

Methods: Ten gastrointestinal (GI) pathologists from the USA (n=5) and South Korea (n=5) evaluated 100 scanned images of gastric (n=50) and colorectal (n=50) neoplasia biopsies and answered multiple questionnaires. Consensus was defined as the answer chosen by the majority. Cohen's (κc) and Fleiss' kappa (κf) values were calculated between the consensus of the two groups and among the raters, respectively.

Results: Both groups reached a consensus in the majority of cases (74%-100%) with slight to perfect intergroup (κc=0.049-1.000) and no to substantial intragroup (κf=-0.083 to 0.660) agreements. For gastric neoplasia, Korean pathologists relied heavily on cytoarchitectural atypia, whereas the US pathologists focused on stromal invasion when diagnosing adenocarcinoma. For colorectal neoplasia, the Korean pathologists identified concurrent intramucosal carcinoma when diagnosing invasive adenocarcinoma, while the presence of desmoplasia was a prerequisite for the diagnosis of invasive adenocarcinoma for the US pathologists.

Conclusions: For GI neoplasia biopsy interpretation, the diagnostic approach of Korean pathologists is similar to that of Eastern/Japanese pathologists. Consensus outperformed kappa statistics in capturing the magnitude of inter-rater and intergroup reliability, highlighting the potential benefit of consensus meetings to decrease the gap between Western and Eastern diagnostic approaches.