TOPLINE: 

A meta-analysis of 11 randomized controlled trials involving more than 50,000 participants showed that AI– assisted esophagogastroduodenoscopy outperformed the conventional procedure in detecting upper gastrointestinal neoplasms. Benefits were evident for lesions ≤ 10 mm and for low‑grade intraepithelial neoplasia, high‑grade intraepithelial neoplasia, and carcinoma.

METHODOLOGY:

• Esophagogastroduodenoscopy is a key technique for diagnosing upper gastrointestinal lesions, but operator‑dependent variability limits the detection of early gastric and esophageal cancers despite training and imaging advances. AI-based technologies are being explored to improve early lesion recognition.
• Researchers conducted a systematic review and meta-analysis of 11 randomized controlled trials comparing AI-assisted esophagogastroduodenoscopy with conventional esophagogastroduodenoscopy for detecting upper gastrointestinal neoplasms in adult patients undergoing the procedure.
• Across the trials (10 in China and one in Japan), a total of 57,512 participants were included: 28,772 were allocated to receive AI‑assisted esophagogastroduodenoscopy and 28,740 to receive conventional esophagogastroduodenoscopy.
• The primary outcome was neoplasm detection rate per patient, and secondary outcomes included detection rate per lesion, stratification of detection rate by lesion size, and detection rates by histologic classification, including low-grade intraepithelial neoplasia, high-grade intraepithelial neoplasia, and carcinoma.

TAKEAWAY:

• Compared with the conventional procedure, AI-assisted esophagogastroduodenoscopy demonstrated higher neoplasm detection rates per patient (risk ratio [RR], 1.57; 95% CI, 1.23-2.01) and per lesion (RR, 1.55; 95% CI, 1.33-2.18), according to meta-analyses of six and eight trials, respectively.
• Analysis of five trials showed that AI-assisted procedures showed more than double the detection rate of conventional procedures for lesions ≤ 10 mm (RR, 2.24; 95% CI, 1.72-2.91); detection rates for lesions > 10 mm were comparable between groups.
• AI-assisted esophagogastroduodenoscopy achieved higher detection rates for low‑grade intraepithelial neoplasia (RR, 1.73; 95% CI, 1.30-2.32), high‑grade intraepithelial neoplasia (RR, 1.61; 95% CI, 1.39-1.87), and carcinoma (RR, 1.54; 95% CI, 1.32-1.79) than the conventional procedure in meta-analyses of seven, six, and seven trials, respectively.
• AI-assisted procedures were associated with fewer blind spots during examination, whereas inspection time and biopsy rates showed no difference between AI-assisted and conventional approaches.

IN PRACTICE:

"AI-EGD [esophagogastroduodenoscopy] demonstrated superior performance compared to Co-EGD [conventional EGD] in detecting UGI [upper gastrointestinal] neoplasms, including lesions in both the stomach and esophagus," the authors wrote.

SOURCES:

The study was led by Mohammad Al Hayek, MD, Damascus University, Damascus, Syrian Arab Republic; and Brigida Barberio, MD, PhD, University of Padova, Padova, Italy. It was published online in Gastrointestinal Endoscopy.

LIMITATIONS:

The study's generalizability is limited by its geographic restriction, with 10 of 11 trials conducted in China and one in Japan, which may not reflect performance in Western healthcare settings. Most studies utilized AI systems whose commercial availability and regulatory approval in Western countries remain inconsistent. Endoscopy techniques and operator experience varied across studies or were incompletely reported, limiting the ability to conduct expertise-based subgroup analyses. 

DISCLOSURES:

No funding was received for this study. All authors reported no conflicts of interest related to this work.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.