AGA: AI-Assisted Colonoscopy Increases Adenoma Detection, But Long-Term Benefits Unclear

©Issara/AdobeStock

A new American Gastroenterological Association (AGA) living guideline published in Gastroenterology evaluates computer-aided detection (CADe) systems during colonoscopy, addressing both potential benefits and associated drawbacks. The multidisciplinary panel—comprising gastroenterologists and methodologists—conducted a systematic review of 44 randomized controlled trials (RCTs) involving more than 30 000 participants, in addition to a microsimulation modeling study and patient and provider surveys.¹

The guideline noted that CADe-assisted colonoscopy was associated with an 8% absolute increase in adenoma detection rate (ADR) (44.8% vs 37.4%; relative risk [RR], 1.22; 95% CI, 1.16–1.29) and a reduction in adenoma miss rate by 19% (RR, 0.47; 95% CI, 0.36–0.60). However, these improvements were largely driven by the detection of diminutive polyps of low malignant potential. The panel also reported a modest increase in detection rates of advanced colorectal neoplasia (2% absolute increase; RR, 1.16; 95% CI, 1.02–1.32).¹

Despite these favorable findings on surrogate markers, the evidence regarding the impact on critical long-term outcomes—such as colorectal cancer (CRC) incidence and CRC-related mortality—remains very low in certainty. Modeling data suggested that CADe might prevent 11 cases of CRC and 2 CRC-related deaths per 10 000 individuals over 10 years, but these benefits are counterbalanced by an estimated 635 additional surveillance colonoscopies per 10 000 individuals. Adverse events such as perforation and bleeding did not differ between CADe-assisted and standard colonoscopy.¹

“We are confident that using AI will lead to more polyps removed and more colonoscopies,” guideline coauthor Benjamin Lebwohl, MD, said in a press release. “We’re less sure about the extent to which it will lead to less colon cancer. AI-assisted colonoscopy technology is promising and exciting. It’s reasonable for practitioners to use the tech now, but we’re not yet at a point where we can recommend universal adoption.”²

Given the close tradeoff between the modest improvements in ADR and the potential for overdiagnosis and increased resource use, the guideline panel concluded with no recommendation for or against the routine use of CADe-assisted colonoscopy. The panel underscored the need for further research into long-term outcomes, cost-effectiveness, and patient-centered perspectives.¹

“If AI is going to be impactful, it needs to be better than the human eye,” guideline coauthor Shahnaz Sultan, MD, MHSc, said in a press release. "Right now, AI is detecting easy-to-detect lesions. This is version 1.0. Before we can recommend everyone use AI, we need version 4.0, where it helps detect polyps that are truly difficult to find.”²

This balanced assessment highlights the evolving nature of artificial intelligence applications in gastroenterology, with an expectation that iterative improvements in CADe technology may eventually shift the risk–benefit profile. For now, the current evidence does not support widespread implementation without additional confirmatory studies.¹

References:

1. Sultan S, Shung DL, Kolb JM, et al. AGA living clinical practice guideline on computer-aided detection-assisted colonoscopy. Gastroenterology. 2025;168:691-700. doi:10.1053/j.gastro.2025.01.002

2. New guideline: AI technology in colonoscopy not yet certain to further prevent colorectal cancer beyond the human eye. News release. American Gastroenterological Association. March 20, 2025. Accessed March 20, 2025. https://gastro.org/news/new-guideline-ai-technology-in-colonoscopy-not-yet-certain-to-further-prevent-colorectal-cancer-beyond-the-human-eye/