Effect of artificial intelligence on novice-performed colonoscopy: a multicenter randomized controlled tandem study

Abstract

Background and aimsThe efficacy and safety of AI-assisted novices performed colonoscopy remain unknown. Here, we aim to compare the lesion detection capability of novices, AI-assisted novices and experts. MethodsThis multi-center, randomized, non-inferiority, tandem study was conducted across three hospitals in China from May 1, 2022, to November 11, 2022. Eligible patients were randomized into one of three groups: CN group (control novice group, withdrawal performed by a novice independently), AN group (AI-assisted novice group, withdrawal performed by a novice with AI assistance), or CE group (control expert group, withdrawal performed by an expert independently). Participants underwent a repeat colonoscopy conducted by an AI-assisted expert to evaluate the lesion miss rate and ensure lesion detection. The primary outcome is the adenoma miss rate (AMR). Results685 eligible patients were analyzed with 229 in the CN group, 227 in the AN group, 229 in the CE group. Both AMR and polyps miss rate (PMR) were lower in the AN than in the CN group (18.82% vs 43.69%, P＜0.001; 21.23% vs 35.38%, P＜0.001, respectively). The non-inferiority margin was met between AN and CE group of both AMR and PMR (18.82% vs 26.97%, P=0.202; 21.23% vs 24.10%, P＜0.249, respectively). ConclusionAI-assisted colonoscopy lowered the AMR of novices, making them non-inferior to experts. Withdrawal technique of new endoscopists can be enhanced by AI-assisted colonoscopy; ClincialTrials.gov, NCT05323279.