This nationwide observational cohort study aimed to assess the reliability of optical diagnosis (based on Kudo’s pit pattern) compared to histologic results of colorectal polyps removed during screening colonoscopies. 

Data regarding Kudo’s pit pattern (Kudo I-V), and histologic results were obtained from the prospectively collected registry of Hungarian National Public Health Institute. The correlation between Kudo I-II and non-neoplastic lesions, Kudo III-IV and neoplastic lesions, and Kudo V and submucosally invasive cancer (SMIC) was characterized by Cohen’s Kappa coefficient, respectively. The potential effects of polyp size and location, and operator characteristics (adenoma detection rate [ADR], annual screening colonoscopy rate and experience in years) were also evaluated.

A total of 8560 lesions (7423 neoplastic [785 SMIC], 1137 non-neoplastic) were registered between 2019 and 2023. Of the 979 serrated polyps 100 contained dysplasia. Kudo I-II and Kudo III-IV demonstrated moderate agreement with non-neoplastic (κ=0.449) and neoplastic lesions (κ=0.528), respectively, while Kudo V demonstrated substantial agreement (κ=0.782) with SMIC. For serrated lesions, Kudo I-II correlated only slightly (κ=0.143) with non-neoplastic lesions. Left-sided location resulted in higher levels of agreement for non-neoplastic (κ=0.489), and neoplastic lesions (κ=0.556), respectively. For lesions larger than 2 cm, levels of agreement were substantial for neoplastic lesions (κ=0.758), and near perfect for SMIC (κ=0.803). ADR reaching 25%, and annual screening colonoscopy rate ≥100 was associated with slightly better results for both non-neoplastic (κ=0.454 and κ=0.530), neoplastic lesions (κ=0.535 and κ=0.585), and SMIC (κ=0.802 and κ=0.804). This was not observed for ADR reaching 35%. Operators with less than 10 years of prior endoscopic experience had better results: κ=0.581 for non-neoplastic, κ=0.651 for neoplastic lesions, and κ=0.827 for SMIC. 

This nationwide cohort demonstrates a moderate and substantial reliability of surface pattern-based identification of neoplasia and SMIC during screening colonoscopies, with better results for left-sided lesions and those larger than 2 cm. Endoscopists with ADR reaching 25% (but not that reaching 35%), higher annual rate of screening colonoscopies and those early in their career perform better in optical diagnosis based histologic prediction.