P0351Reliability and Validity of Endoscopic Scoring in Pouchitis: Agreement Among Reviewers and Correlation with Histology

Abstract Background The Pouchitis Disease Activity Index (PDAI) is widely used to assess pouchitis but is limited by clinical subjectivity and suboptimal endoscopic reliability.1-2 The recently developed Atlantic Pouchitis Index (API) offers a more objective evaluation of pouch inflammation by excluding clinical parameters.3 Central reading remains the gold standard in clinical trials to enhance endoscopic scoring reliability.4 However, discrepancies between endoscopic and histologic findings continue to complicate pouchitis assessment. This study aimed to evaluate inter-rater variability in endoscopic scoring using video and still images, examine correlations with histologic scores, and validate the API. Methods Forty pouchoscopy procedures with corresponding histology were evaluated. Three blinded gastroenterologists independently scored video and still images using ePDAI, SES-CD, UCEIS, and Monash Endoscopic Score. Two blinded gastrointestinal pathologists assessed histologic inflammation using hPDAI, Nancy Index, RHI, Geboes Score, and IBD-DCA. Composite assessments included PDAI and API. Agreement between video and still-image scores was analyzed with a modified Bland–Altman method accounting for repeated measures. Interobserver agreement was evaluated using intraclass correlation coefficients (ICC). Results The UCEIS endoscopic scoring system showed the closest agreement between measures recorded by still-images and videos (mean difference -0.09, 95% limits of agreement: -2.45 to 2.28) (Table 1a). The ICC was higher when using video scoring for ePDAI and segmental SES-CD, with total SES-CD having the highest ICC when using still-images (0.765, 95% CI 0.627-0.862) and videos (0.766, 95% CI 0.642-0.860) (Table 1b). The highest ICCs among histology scoring system were the Geboes (0.774, 95% CI 0.611-0.874) and RHI (0.770, 95% CI 0.607-0.871) (Table 1c). The correlations between endoscopy and histology for pouch measures ranged from 0.46 to 0.71, with Spearman correlations more often higher with video endoscopy measures (Table 2). The highest correlation between both still-image and video scores and histology was between hPDAI and UCEIS. The Pearson correlation coefficient between the API and PDAI are strong for both still-image and video scoring (r = 0.802 and r = 0.844 respectively). Conclusion Video-based assessments showed higher interobserver agreement and stronger endoscopic–histologic correlations than still images. UCEIS demonstrated the closest agreement between modalities, and SES-CD showed the highest reliability. The hPDAI correlated most strongly with ePDAI and UCEIS, while Geboes and RHI had the best reproducibility. Composite PDAI and API scores were strongly correlated, supporting the use of API for pouchitis. References: 1. Shen B, Achkar JP, Connor JT, et al. Modified pouchitis disease activity index: a simplified approach to the diagnosis of pouchitis. Dis Colon Rectum. 2003;46(6):748-53. 2. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33(1):159-74. 3. Sedano R, Ma C, Hogan M, et al. Development and external validation of an endoscopic and histologic pouchitis assessment tool: the Atlantic Pouchitis Index. Clin Gastroenterol Hepatol.2025. 4. Samaan MA, Shen B, Mosli MH, et al. Reliability among central readers in the evaulation of endoscopic disease activity in pouchitis. Gastroint Endoscopy. 2018;88(2). Conflict of interest: Dr. Khoo, Emi: Speaker honorarium from CSL Seqirus, Menarini, Research review, Ferring Education sponsorship from Arrotex, Takeda, Abbvie, Janssen, Pfizer, Falk Pharma Research grant from Pfizer, Celltrion, Gastroenterological Society of Australia (GESA) Fernandes, Richard: Grant support from Gastroenterology Network of Intestinal Ultrasound (GENIUS) A research scholarship from Crohn’s and Colitis Australia (CCA) Speaker, advisory board or consultation honoraria from Janssen, Abbvie, Takeda and Ferring Conference support from Ferring, Pfizer, Sandoz, AbbVie, Janssen and Takeda Walker, Nicole: No conflict of interest Howlett, Mariko: No conflict of interest Lourie, Rohan: No conflict of interest Liu, John: No conflict of interest Webb, Lachlan: No conflict of interest Begun, Jakob: Consulting fee from AbbVie, Bristol Myers Squibb, Dr. Falk Pharma, Eli Lilly, Janssen, Pfizer, and Takeda Speaker fees from AbbVie, Bristol Myers Squibb, Dr. Falk Pharma, Eli Lilly, Ferring, GlaxoSmithKline, Janssen, Pfizer, and Takeda Grants from Janssen, Abbvie and Pfizer.