Despite simplified diagnostic criteria, intraobserver and interobserver variability remain in the interpretation of colorectal serrated polyps

AIMS: Many studies have highlighted interobserver variability in histologic distinction between colorectal sessile serrated lesion (SSL) and hyperplastic polyp (HP). In 2019, the WHO updated their criteria for the diagnosis of SSL, requiring 'at least 1 unequivocal architecturally distorted serrated crypt'. Even with this simplified criterion, as well as experience accumulated in recognizing SSL over 25 years, SSL and HP remain difficult to distinguish in some instances. This study aimed to assess observer variability and preferred criteria in diagnosing SSL among gastrointestinal pathologists. METHODS AND RESULTS: We retrospectively identified 60 serrated colorectal polyps, produced uniform H&E recuts, and created whole-slide images for each case. Cases were selected to cover a spectrum of non-dysplastic serrated lesions, as confirmed via review by four pathologists who individually interpreted each as SSL, HP, or serrated polyp NOS (SP-NOS). The cases were then reviewed by nine additional pathologists. A second round of review followed after a 5-month washout period. A third round of reviews was completed after 5 additional months, at which time reviewers were provided information regarding polyp size and site. Fleiss and Cohen kappa values were calculated to determine overall inter- and intra-observer agreement. The top three criteria pathologists used to favour a diagnosis of SSL over HP included crypt distortion (13/13), polyp location (8/13) and size (4/13). There was moderate agreement among all 13 pathologists when classifying the 60 cases for rounds 1 (κ = 0.50) and 2 (κ = 0.46), and good agreement for round 3 (κ = 0.63), the round using criteria beyond those of the WHO; stratification by location showed agreement was worst for transverse polyps. Twenty-one (35%) cases were called SSL >80% of the time, and 16 (27%) cases were classified as HP >80% of the time. Agreement was moderate (κ = 0.43) for polyps measuring ≥1.0 cm and was good (κ = 0.63) for polyps measuring ≤0.4 cm. In keeping with current WHO criteria, crypt distortion was the only feature all pathologists considered useful to diagnose SSL. Overall interobserver agreement improved from moderate to good when pathologists were aware of the polyp size and site. Pathologists had the worst agreement when classifying lesions in the transverse colon or polyps ≥1.0 cm. CONCLUSIONS: Non-histologic criteria (e.g. polyp site and size) may be necessary to accurately and reproducibly distinguish SSL from HP, if properly validated.