Artificial intelligence (AI) holds the promise of identifying premalignant and advanced malignant lesions during colonoscopy that might otherwise be missed.
Is it living up to that promise?
It seems that depends on where, how, and by whom it’s being implemented.
Clinical Trials vs the Real World
The majority of randomized clinical trials of AI use conducted worldwide “clearly show an increase in the adenoma detection rate (ADR) during colonoscopy,” Prateek Sharma, MD, a gastroenterologist at The University of Kansas Cancer Center, Kansas City, told this news. “But the real-world results have been quite varied; some show improvement, and others don’t.”
Dr. Sharma is coauthor of a recent pooled analysis of nine randomized controlled trials on the impact of AI on colonoscopy surveillance after polyp removal. It found that AI use increased the proportion of patients requiring intensive surveillance by approximately 35% in the United States and 20% in Europe (absolute increases of 2.9% and 1.3%, respectively).
“While this may contribute to improved cancer prevention, it significantly adds patient burden and healthcare costs,” the authors concluded.
A recent retrospective analysis of staggered implementation of a computer-aided detection (CADe) system at a single academic center in Chicago found that for screening and surveillance colonoscopy combined, endoscopists using CADe identified more adenomas and serrated polyps — but only endoscopists who used CADe regularly (“majority” users).
A systematic review and meta-analysis of 21 randomized controlled trials comparing CADe with standard colonoscopy found increased detection of adenomas, but not of advanced adenomas, as well as higher rates of unnecessary removal of non-neoplastic polyps.
Adding to the mix, a multicenter randomized controlled trial of patients with a positive fecal immunochemical test found that AI use was not associated with better detection of advanced neoplasias. Lead author Carolina Mangas Sanjuán, MD, PhD, Hospital General Universitario Dr. Balmis, Alicante, Spain, told this news organization the results were “surprising,” given previous studies showing benefit.
Similarly, a pragmatic implementation trial conducted by Stanford, California, researchers showed no significant effect of CADe on ADR, adenomas per colonoscopy, or any other detection metric. Furthermore, CADe had no effect on procedure times or non-neoplastic detection rates.
The authors cautioned against viewing their study as an “outlier,” however, and pointed to an Israeli study comparing adenoma and polyp detection rates 6 months before and after the introduction of AI-aided colonoscopy. Those authors reported no performance improvement with the AI device and concluded that it was not useful in routine practice.
A ‘Mishmash’ of Methods
“It’s not clear why some studies are positive, and some are negative,” Dr. Sharma acknowledged.
Study design is a factor, particularly in real-world studies, he said. Some researchers use the before/after approach, as in the Israeli study; others compare use in different rooms — that is, one with a CADe device and one without. Like the Chicago analysis, findings from such studies probably depend on whether the colonoscopists with the CADe device in the room actually use it.
Other real-world studies look at detection by time, Dr. Sharma said.
For example, a study of 1780 colonoscopies in China found that AI systems showed higher assistance ability among colonoscopies performed later in the day, when adenoma detection rates typically declined, perhaps owing to fatigue.
These authors suggest that AI may have the potential to maintain high quality and homogeneity of colonoscopies and improve endoscopist performance in large screening programs and centers with high workloads.
“There’s a mishmash of different kinds of real-world studies coming in, and it’s very difficult to figure it all out,” Dr. Sharma said. “We just have to look at these devices as innovations and embrace them and work with them to see how it fits it in our practice.”