Mangas-Sanjuan CADe (advanced neoplasia)
2023
Role of Artificial Intelligence in Colonoscopy Detection of Advanced Neoplasias
In 3,213 FIT-positive adults in a colorectal cancer screening programme, computer-aided polyp detection did not improve detection of advanced colorectal neoplasia compared with standard colonoscopy.
Result
Efficacy: Advanced colorectal neoplasia detection rateaRR 1.01 (95% CI 0.92-1.10)
Below 1 favours standard colonoscopy, above 1 favours computer-aided polyp detection.
- Absolute difference
- 0.2 points
The absolute difference is derived from the event rates.
Also reported: Advanced neoplasias per colonoscopy 0.54 vs 0.52 (adjusted rate ratio 1.04, 99.9% CI 0.88–1.22); ADR 64.2% vs 62.0% (aRR 1.06, 99.9% CI 0.91–1.23). CADe did increase nonpolypoid lesions (0.56 vs 0.47) and proximal adenomas (0.94 vs 0.81).
- Subtopic
- Adenoma detection
- Design
- Multicentre RCT, Spain
- Relevance
- Practice-defining: a trial the field needs to know.
- Source
- Mangas-Sanjuan C, et al. Ann Intern Med. 2023;176(9):1145-1152.doi 10.7326/M22-2619
Discussion & critique
The landmark negative of the cluster: in a FIT-positive screening population with high baseline detection, CADe found more small, flat and proximal lesions but no more advanced neoplasia, the lesions that matter for cancer prevention. The authors note that the high control-group ADR (62.0%) may limit generalisability to low-detector endoscopists; the trial was funded by Medtronic. With the deskilling signal (Budzyń) it reframes the ADR gains of Wang, Repici and COLO-DETECT as surrogate-endpoint wins of uncertain clinical value.