Los Angeles General Hospital, University of Southern California Los Angeles, CA
Linda Huang, MD1, Patrick Chang, MD2, Sarah Wang, MD, MPH3, Denis Nguyen, MD3, Frederick W. Chang, DO4, Helen Lee, NP2, Jennifer Phan, MD3, Ara B. Sahakian, MD3, James Buxbaum, MD, MS3 1Los Angeles General Hospital, University of Southern California, Los Angeles, CA; 2University of Southern California, Los Angeles, CA; 3Keck School of Medicine of USC, Los Angeles, CA; 4University of Arizona, Oxnard, CA
Introduction: Colonoscopy is a widely used screening and diagnostic tool for colorectal neoplasms. The application of artificial intelligence (AI) has the potential to enhance adenoma detection rate (ADR) during colonoscopy. However, the impact of AI-assisted colonoscopy on ADR among different endoscopist proficiencies remains unclear. This meta-analysis aims to determine the impact on ADR by AI-assisted colonoscopy compared to standard colonoscopy among high and low ADR endoscopists.
Methods: A comprehensive literature search was conducted using PubMed, Embase, and Cochrane Library databases to identify randomized controlled trials (RCTs) comparing AI-assisted colonoscopy with standard colonoscopy. Additional inclusion criteria included discrete reported ADR for both control and AI-assisted arms. Groups were stratified by baseline ADR extracted from the control arms into high detection (ADR ≥25%) and low detection (ADR < 25%) endoscopists based on the latest American Gastroenterology Association (AGA) Clinical Practice Update for clinically concerning low ADR rates (< 25%). The primary outcome was difference of ADR comparing high and low ADR groups. The pooled odds ratio (OR) with a 95% confidence interval (CI) was calculated using a random-effects model.
Results: From January 2019 to April 2023, a total of twenty-two RCTs were identified to meet criteria, aggregating a total of 17,404 patients. The pooled analysis demonstrated significantly more ADR improvement in low ADR endoscopists compared to the high ADR endoscopists (Table 1, Figure 1; ADR improvement for Low ADR vs. High ADR, OR 1.48, 95% CI 1.35-1.63).
Within group analysis by stratified by ADR proficiencies showed ADR improvement with AI-assisted colonoscopy compared to the standard colonoscopy in both low ADR endoscopists (Low ADR: No AI: 17.8 +/- 5.1% vs. AI: 27.81+/-7.0%) and high ADR endoscopists (High ADR: No AI: 43.3+/- 11.0% vs. AI: 50.7+/- 11.3%). The magnitude of improvement in ADR was greater among low ADR endoscopists compared to high ADR endoscopists.
Discussion: This meta-analysis shows that AI-assisted colonoscopy may lead to greater ADR improvement in low ADR endoscopists (< 25%). Thus, we further highlight a use case of AI assisted colonoscopy for detectors with low ADR given its relevance around an AGA identified cut point.
Figure: Figure 1: A forest plot comparison of the odds ratio of improvement by AI-assisted colonoscopy comparing low and high ADR endoscopists
Disclosures:
Linda Huang indicated no relevant financial relationships.
Patrick Chang indicated no relevant financial relationships.
Sarah Wang indicated no relevant financial relationships.
Denis Nguyen indicated no relevant financial relationships.
Frederick Chang indicated no relevant financial relationships.
Helen Lee indicated no relevant financial relationships.
Jennifer Phan: Boston Scientific – Consultant. Cook Medical – Consultant. Olympus – Consultant.
Ara Sahakian indicated no relevant financial relationships.
James Buxbaum: Boston Scientific – Consultant. Olympus – Consultant.
Linda Huang, MD1, Patrick Chang, MD2, Sarah Wang, MD, MPH3, Denis Nguyen, MD3, Frederick W. Chang, DO4, Helen Lee, NP2, Jennifer Phan, MD3, Ara B. Sahakian, MD3, James Buxbaum, MD, MS3. P0554 - Artificial Intelligence-Assisted Colonoscopy Improves Adenoma Detection Rate (ADR) in Both Low and High ADR Endoscopists: A Meta-Analysis, ACG 2023 Annual Scientific Meeting Abstracts. Vancouver, BC, Canada: American College of Gastroenterology.
