AI-assisted colonoscopy boosts polyp removal, but ‘only benign, non-adenomatous lesions’
Click Here to Manage Email Alerts
Key takeaways:
- AI-assisted colonoscopy induced more removals of benign lesions without adenomatous polyps.
- Additional “procedural risks as well as costs” may offset the value of preventive CRC screening.
PHILADELPHIA— AI-assisted-colonoscopy significantly increased the number of colonoscopy procedures in which only benign lesions were removed without additional adenomatous polyps, according to data at the ACG Annual Scientific Meeting.
“Artificial intelligence-assisted colonoscopy with computer aided detection, or CADe, is designed to improve colonoscopy quality, primarily through improvement in adenoma detection and adenoma misreads,” Tessa Herman, MD, chief resident of internal medicine at the University of Minnesota Medical Center, told attendees. “AI-assisted colonoscopy also appears to increase the resection of non-neoplastic lesions.”
She added: “[However,] the rate at which colonoscopy is performed for colorectal cancer prevention that results in polypectomy exclusively for resection of benign lesions, without adenoma removal, is currently unknown.”
To assess the effect of AI-assisted colonoscopy on adenoma detection rate compared with unassisted colonoscopy, Herman and colleagues performed an ad-hoc analysis using data from a previous study of adults who underwent colonoscopy for screening, surveillance or FIT-positive testing at the Minneapolis Veteran Affairs Medical Center. The researchers identified 441 unassisted and 599 AI-assisted colonoscopies, with identical demographics between the two cohorts and the same endoscopists involved throughout the study.
In the original study, Herman noted an “increased resection rate of benign, non-adenomatous lesions with AI at a rate of 30.9% vs. 22.7%, which is a statistically significant result.”
The primary outcome of the ad hoc analysis was the proportion of AI-assisted vs. unassisted colonoscopies that led to the removal of only benign lesions without removing adenomas.
According to the study results, 12.4% of AI-assisted colonoscopies led to polypectomies that only showed benign lesions compared with just 8.4% among unassisted colonoscopies (P =.04).
The researchers noted that the most common resected benign lesions were benign colonic mucosa, lymphoid aggregates and hyperplastic polyps.
“Our analysis found that AI-assisted colonoscopy resulted in a statically, and potentially clinically, significant increase in the proportion of colonoscopies where only benign, non-adenomatous lesions were removed without synchronous removal of adenomas,” Herman, told attendees. “This [represents] an absolute increase of approximately 4%.”
She added: “The potential implications of this work are that there could be increased procedural risks as well as costs, such as pathology costs and health care expenditures, without additional colorectal cancer prevention benefit.”
Collapse
Herman T, et al. Artificial intelligence-assisted colonoscopy is associated with higher conversions from screening to therapeutic exams; Oct. 25-30, 2024; Philadelphia (hybrid meeting).
Click Here to Manage Email Alerts