AI can offset dwindling adenoma detection for colonoscopies performed later in the day
Click Here to Manage Email Alerts
Although colonoscopies performed later in the day correlated with a decreased adenoma detection rate, researchers reported that artificial intelligence systems may offset time-related decline in colonoscopy quality.
“Time of day has been identified as an indispensable factor related to suboptimal ADR. ... Continuous and repetitive visual stimuli may lead to weaker reliable response and poorer judgement, which jeopardize diagnostic abilities and efficiency,” Zihua Lu, MD, of the department of gastroenterology at Renmin Hospital of Wuhan University, and colleagues wrote in JAMA Network Open. “According to available evidence, the incorporation of AI as an aid for colonoscopy results in a significant increase in ADR. ... However, the extra benefit of AI systems in eliminating the time-related decline of ADR remains unknown.”
In a secondary analysis of two prospective trials, Lu and colleagues sought to determine whether AI assistance could improve time-related decline in ADR during colonoscopy. They enrolled 1,780 patients (mean age, 48.61 years; 47.02% women), who were randomly assigned to AI-assisted or standard colonoscopy groups, and compared ADRs of early vs. late colonoscopy sessions per half day before and after AI intervention. Researchers defined a half day as morning (earlier than 1 p.m.) and afternoon (1 p.m. or later).
A total of 1,041 colonoscopies were performed during early sessions (65.71% AI-assisted, 34.29% standard) and 739 during late sessions (64.41% and 35.59%, respectively).
According to results, ADR during early sessions was “significantly higher” compared with late sessions (13.73% vs. 5.7%; OR = 2.42; 95% CI, 1.31-4.47), with no significant difference between sessions on ADR (22.95% vs. 22.06%; OR = 0.96; 95% CI, 0.71-1.29) or polyp detection rate (55.99% vs. 52.31%; OR = 1.08; 95% CI, 0.83-1.4) after AI intervention.
However, AI assistance improved ADR in both early (22.95% vs. 13.73%; OR = 1.6; 95% CI, 1.1-2.34) and late sessions (22.06% vs. 5.7%; OR = 3.81; 95% CI, 2.1-6.91) with higher assistance capability in the late sessions (OR = 3.81; 95% CI, 2.1-6.91) compared with early sessions (OR = 1.6; 95% CI, 1.1-2.34).
“Our results suggest that later sessions per half day were associated with a decline adenoma detection. Furthermore, AI systems could eliminate the time-related degradation of colonoscopy quality,” Lu and colleagues concluded. “In the future, the application of AI systems has the potential to maintain high quality and homogeneity of colonoscopies and further improve endoscopist performance in large screening programs and centers with high workloads.”
Perspective
Back to Top
Carol A. Burke, MD, FACG, FASGE, AGAF, FACP
The efficacy of colonoscopy is affected by the quality of the examination, and one of the key quality metrics is ADR. Data demonstrates wide variability in ADR between endoscopists and on the impact of time of day on colonoscopy and ADR.
In 2009, Madhusudhan Sanaka, MD, and I published one of the earliest studies showing that ADR decreases for each subsequent hour of day, such that ADR is highest in colonoscopy started at the first hour of the day and progressively wanes over the course of the day. The exact reasons are not known, but endoscopist fatigue or lack of concentration are likely contributory.
In this study, the authors posit that AI could overcome the time-related decline in ADR during colonoscopy as the day progresses. This makes perfect sense. Colonoscopy with AI has been shown to be more reliable and effective for ADR enhancement than when colonoscopy is performed without AI.
The data for this secondary analysis was derived from two prospective, randomized controlled trials comparing polyp detection in consecutive patients undergoing colonoscopy with AI vs. non-AI from a single center in China. A few findings were no surprise to me: First, ADR was higher with AI than without for both morning and afternoon exams (23% vs. 14% and 22% vs. 6%, respectively). Second, the difference in ADR between morning and afternoon exams with AI was not significantly different (23% vs. 22%), but was so in the non-AI arms (14% vs. 6%).
I have a few major concerns about the generalizability of this study. ADR — regardless of time of day — in the control group was substandard compared with established quality metrics. ADR in men and women undergoing screening colonoscopy should be a minimum of 25% or higher. Additionally, even when AI was used, the ADR in these patients (at about 23%, regardless of time of day) is much lower than has been found in other AI colonoscopy studies, in which ADRs have been reported as high as 55%.
I agree with the authors that AI colonoscopy can maintain ADR at a constant level throughout the day. It also shows that AI assistance could boost this center’s quality of colonoscopies to national and international ADR minimum benchmarks, as well as in other centers where ADR targets are not met.
Reference:
Sanaka M, et al. Am J Gastroenterol. 2009;doi:10.1038/ajg.2009.249.
Collapse
Read more about
Click Here to Manage Email Alerts