AI-Assisted Colonoscopy Increases ADR by 20%

TOPLINE:

Artificial intelligence (AI)–assisted colonoscopy increases the adenoma detection rate (ADR) and reduces the adenoma miss rate, regardless of the endoscopist’s experience, technology system, or healthcare setting.

METHODOLOGY:

• Missed lesions during colonoscopy significantly contribute to cases of post-colonoscopy colorectal cancer, with up to 25% of adenomatous lesions potentially being missed.
• Researchers conducted a systematic review and meta-analysis to evaluate whether AI-driven computer-aided detection (CADe) systems improve adenoma detection compared with routine colonoscopy.
• It included 28 randomized controlled trials published between 2019 and 2024 involving 23,861 participants who underwent colonoscopy for screening, post-polypectomy surveillance, or symptomatic reasons using a high-definition endoscope.
• The primary outcomes were the ADR and the adenoma miss rate, and secondary outcomes were adenomas per colonoscopy, sessile serrated lesion detection rates and miss rates, and withdrawal time.

TAKEAWAY:

• AI-assisted colonoscopy increased the ADR by 20% (P P
• In studies involving only expert endoscopists, AI-assisted colonoscopy showed a 19% increase in the ADR compared with routine colonoscopy (P
• The use of AI led to an increase in withdrawal time of just 9 seconds on average (P = .01).
• AI use increased adenomas per colonoscopy (weighted mean difference, 0.21; P

IN PRACTICE:

“Overall, this study supports the use of CADe systems for improved performance of colonoscopy,” the authors wrote.

SOURCE:

This study, led by Jonathan Makar, Department of Medicine, University of Melbourne, Melbourne, Australia, was published online in Gastrointestinal Endoscopy.

LIMITATIONS:

The fact that only one of the control arms in the included randomized controlled trials used a placebo raises the possibility of performance bias or the Hawthorne effect. Significant heterogeneity in patient ages, bowel preparation scores, and colonoscopy indications were noted across the studies and could confound the results. The omission of per-protocol or intention-to-treat analyses in some studies raises the potential for attrition bias.

DISCLOSURES:

The study did not receive any specific funding. The authors declared no conflicts of interest.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.