Does AI-Assisted Colonoscopy Actually Improve Colorectal Cancer Prevention?

Why adenoma detection is attractive

Adenoma detection rate (ADR) is strongly associated with post-colonoscopy colorectal cancer risk, so technology that increases ADR is plausible as a cancer-prevention tool. Randomized trials show CADe can increase detection, particularly of small lesions.

The surrogate-outcome problem

More polyps detected is not identical to fewer cancers. Many additional lesions are diminutive and low risk. CADe can increase polypectomy burden, pathology volume and surveillance recommendations. Whether those downstream effects improve net patient outcomes enough to justify universal adoption remains uncertain.

What AGA concluded in 2025

After GRADE review, AGA made no recommendation for or against CADe-assisted colonoscopy. The panel acknowledged improved ADR but emphasized very low certainty for critical long-term outcomes such as colorectal cancer incidence, colorectal cancer mortality and post-colonoscopy colorectal cancer.

AI can amplify both good and poor technique

A fatigued or hurried endoscopist may benefit from a second visual observer, but CADe cannot compensate for an unclean colon, failure to reach the cecum, inadequate withdrawal time or poor resection technique. False-positive prompts can also create distraction. The technology is best considered one component of an endoscopy quality system.

What evidence would change practice?

The field needs pragmatic studies with long-term outcomes, performance in diverse community settings, cost-effectiveness, effect on clinically significant lesions such as advanced adenomas and sessile serrated lesions, and impact on surveillance burden.

A practical clinical approach

1. Maintain core quality standards regardless of whether CADe is used.
2. Use CADe as an adjunctive detector, not an autonomous decision-maker.
3. Track ADR, sessile serrated lesion detection, bowel-prep quality and appropriate surveillance recommendations.
4. Avoid shortening surveillance intervals solely because AI finds additional low-risk diminutive lesions beyond guideline logic.
5. Reassess adoption as long-term outcome data mature.

Common errors to avoid

• Claiming that improved ADR already proves reduced cancer mortality.
• Allowing AI to distract from withdrawal technique.
• Using CADe without measuring baseline endoscopist quality.
• Treating every alert as a lesion that requires removal without clinical inspection.

What should trainees remember?

AI-assisted colonoscopy improves detection, but the cancer-prevention claim remains unproven. The right standard is patient-important outcomes plus high-quality colonoscopy, not technology adoption by itself.

Free further reading from Dr. Alan B. R. Thomson

See Dr. Thomson's Endoscopy and Diagnostic Imaging and Images in Gastroenterology and Hepatology for visual/endoscopic foundations.

Frequently asked questions

Does CADe increase adenoma detection?

Yes, in many trials.

Has CADe been proven to reduce colorectal cancer mortality?

No.

Does AGA recommend routine CADe for all colonoscopies?

The 2025 AGA guideline made no recommendation for or against routine use because evidence for long-term outcomes was very uncertain.

References

1. Thomson ABR. Endoscopy and Diagnostic Imaging, Parts I-II. CAPstone Academic Publishers; 2012. ISBN 978-1477400579 and 978-1477400654.

2. Thomson ABR. Images in Gastroenterology and Hepatology, Parts 1-2. CAPstone Academic Publishers; 2021. ISBN 979-8719829074 and 979-8743669325.

3. Sultan S, et al. AGA Living Clinical Practice Guideline on Computer-Aided Detection-Assisted Colonoscopy. Gastroenterology. 2025. doi:10.1053/j.gastro.2025.01.002.

4. American Gastroenterological Association. Use of Computer-Aided Detection Systems (CADe) in Colonoscopy. Published March 20, 2025.

5. Rex DK, et al. Quality Indicators for Colonoscopy. ACG/ASGE Position Statement. Am J Gastroenterol. 2024.