High interobserver variability exists in grading appendiceal goblet cell adenocarcinoma using World Health Organization 5th edition criteria.

Appendiceal goblet cell adenocarcinoma (GCA) is an uncommon malignancy that has been described under various names and grading schemes. The 5th Edition of the World Health Organization (WHO) Classification of Digestive System Tumours provides a three-tiered system for grading these neoplasms, but the reproducibility of this classification scheme has not been studied. We scanned 58 H&E-stained slides from 20 GCA and circulated the whole-slide images among seven pathologists with interest in appendiceal pathology. They evaluated each slide for the presence of 15 histological patterns defined by the WHO as criteria for low-grade (n = 5) and high-grade (n = 10) GCA. Cases were also evaluated for the presence of extracellular mucin. Participants also reported the percentage of high-grade features in each whole slide image and each case. Interobserver variability was assessed statistically. All seven observers agreed on the WHO grade for four of the 20 cases (20%; one grade 1, three grade 3). Using Fleiss's kappa statistic, overall agreement for cases was fair at 0.29 (95% confidence interval [CI]: 0.14-0.44), and pairwise agreement between observers ranged from 0.00 to 0.82 (median = 0.13). Gwet's agreement coefficient ranged from 0.10 to 0.87 (median = 0.28), while overall agreement was 0.36 (95% CI: 0.17-0.54). There was significant variability with respect to assessing the presence of individual features. The best agreement was seen for extracellular mucin (neutral feature, κ = 0.43) and tumour sheets (high-grade feature, κ = 0.41), whereas the worst agreement was seen for mild architectural disarray/tubular fusion (low-grade feature, κ = 0.05) and necrosis (high-grade feature, κ = 0.07). We conclude that interobserver agreement for grading GCA using the three-tiered WHO 5th Edition classification system is fair at best. A validated two-tiered system (i.e. low- versus high-grade) may be more reproducible.