Computer-aided Detection Said To Fail in Randomized Controlled Trial, but It Actually Performed Well
Douglas K. Rex, MD, MASGE, reviewing Ahmad A, et al. Endoscopy 2022 Dec 12.
Artificial intelligence programs for highlighting or detecting polyps during colonoscopy are designated computer-aided detection (CADe) programs. In general, randomized controlled trials of CADe have been positive for adenoma detection rate (ADR) improvement, with average higher gains than with any other device.
In a study performed by 8 experienced endoscopists in the UK National Health Service Bowel Cancer Screening Programme, 614 patients were randomized to CADe or a control group (standard colonoscopy). Endocuff Vision (Olympus) was used in about 70% of patients in each group. The polyp detection rate was borderline higher in the CADe versus control group (85.7% vs 79.7%; P=.05), and the ADR was not significantly higher with CADe (71.4% vs 65%; P=.09). Adenomas per colonoscopy was 2.4 with CADe versus 2.1 in controls (P=.25). The adenoma plus sessile serrated lesion (SSL) detection rate was higher with CADe (79.2% vs 71.6%; P=.03).
COMMENTThe authors of this study interpreted the outcomes as being negative, but considering the very high prevalence of polyps in this population with positive fecal immunochemical test results, the predominant use of Endocuff Vision, the trend toward significance for ADR, and the significance for adenoma plus SSL detection, the study results could easily be viewed as positive for CADe.
Note to readers: At the time we reviewed this paper, its publisher noted that it was not in final form and that subsequent changes might be made.
CITATION(S)
Ahmad A, Wilson A, Haycock A, et al. Evaluation of a real-time computer-aided polyp detection system during screening colonoscopy: AI-DETECT study. Endoscopy 2022 Dec 12. (Epub ahead of print) (https://doi.org/10.1055/a-1966-0661)
