IOlBD evidence-based consensus on the use of artificial intelligence for assessment of endoscopic and histologic endpoints in clinical trials of inflammatory bowel disease

Background and Aims Central reading of endoscopy and histopathology is the current standard for disease activity assessment in inflammatory bowel disease (IBD) clinical trials but is limited by interreader and intrareader variability, operational delays, and cost. Artificial intelligence (AI) and machine learning (ML) offer the potential to improve accuracy, efficiency, and reproducibility. The International Organization for the Study of IBD (IOIBD) developed consensus statements on AI/ML use for endoscopic and histologic endpoint assessment in IBD trials. Methods As part of the IOIBD endpoints cluster initiative, a narrative, evidence-informed review with literature searches of Medline and Embase (January 2018-February 2025) identified studies applying AI/ML to endoscopy or histology in IBD. Relevant evidence informed 36 survey statements formulated by a steering committee. Seventy-two IOIBD members were invited to vote online; consensus required ≥80% agreement (score: 7-10 on a 10-point scale). Results Forty-five members completed the survey. Consensus was reached for 28 statements related to endoscopy, pathology, and trial design. Experts agreed that AI-based central reading could improve diagnostic accuracy, expedite processes, reduce costs, and enhance reproducibility. Combining human and AI assessments was favored over AI replacement. The key limitations identified included insufficient validation, generalizability concerns, and dependence on human-annotated training datasets. Conclusions This IOIBD consensus supports the integration of AI/ML into central reading for IBD clinical trials to improve objectivity, efficiency, and consistency, while maintaining human oversight. Further research should address validation, regulatory frameworks, and multimodal integration to enable broader adoption in both trials and clinical practice.