Reply to Henry and East

10.1055/a-2749-8345

My colleagues and I appreciate the interest of Dr. Henry and Prof. East in our study [1] and the opportunity to address their concerns regarding the detection of clinically significant lesions.

The group who did not undergo computer-aided detection (non-CADe) did indeed have more high grade adenomas (4.6% vs. 3.0%; P = 0.009) and invasive cancers (1.8% vs. 0.7%; P = 0.001) detected; however, several methodological considerations warrant attention. First, the absolute numbers were small, thereby limiting statistical power: high grade adenomas, CADe, n = 72 and non-CADe, n = 79; invasive cancer, CADe, n = 16 and non-CADe, n = 31. The overall advanced neoplasm detection rate showed no significant difference (4.1% vs. 4.7%; P = 0.39), suggesting the observed differences may reflect sampling variation rather than systematic detection failure. Critically, these lesions were substantial in size (mean diameters: for high grade adenomas, 10.3 mm; for invasive cancers, 20.6 mm). Given CADe's demonstrated benefit for primarily smaller, subtle lesions, these large lesions would be unlikely to benefit from CADe assistance. Therefore, the observed difference more likely reflects unmeasured confounding factors – an inherent limitation of observational studies using propensity score matching.

Second, the non-CADe group's performance may reflect skill acquisition from CADe exposure. Our cumulative summation (CUSUM) analysis showed that high detectors maintained their improved performance during subsequent non-CADe procedures, suggesting possible skill transfer. Endoscopists performed both CADe and non-CADe procedures throughout the study, making it difficult to separate direct CADe effects from potential skill enhancement acquired through CADe exposure. Recent studies examining deskilling associated with CADe use have varied considerably in their study periods, populations, and methodological approaches, precluding definitive conclusions [2] [3] [4].

Most importantly, we agree with your fundamental assertion that whether AI-supported adenoma detection reduces post-colonoscopy colorectal cancer remains unclear. The ultimate measure – prevention of interval cancers and cancer-related mortality – will require substantially longer follow-up and larger multicenter cohorts [5].

In conclusion, while our study demonstrates that detection performance was maintained without evidence of deskilling, we concur that cautious interpretation is warranted. Only large-scale, long-term studies can definitively answer whether AI-supported colonoscopy impacts patient outcomes.

Publication History

Article published online:
20 February 2026

• References

• 1 Okumura T, Kudo SE, Ide Y. et al. Long-term impact of computer-aided adenoma detection: a prospective observational study. Endoscopy 2025;
• 2 Budzyń K, Romańczyk M, Kitala D. et al. Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study. Lancet Gastroenterol Hepatol 2025; 10: 896-903
• 3 Takasu A, Kogure H, Dai Z. et al. Impact of introducing artificial intelligence on colonoscopy: a retrospective study on potential benefits and drawbacks. J Gastroenterol Hepatol 2025; 40: 2258-2266
• 4 Orzeszko Z, Gach T, Necka S. et al. The implementation of computer-aided detection in an initial endoscopy training improves the quality measures of trainees' future colonoscopies: a retrospective cohort study. Surg Endosc 2025; 39: 5276-5286
• 5 Sultan S, Shung DL, Kolb JM. et al. AGA Living Clinical Practice Guideline on computer-aided detection-assisted colonoscopy. Gastroenterology 2025; 168: 691-700