Peripheral gaze guidance improves adenoma detection rate

Share

Real-time guidance that prompts endoscopists to look toward the periphery of the colonoscopy screen significantly increased adenoma detection rate (ADR) compared with standard practice, without lengthening withdrawal time, according to a multicenter randomized controlled trial conducted in Japan.
 
“Ensuring adequate withdrawal time is an important surrogate marker for improving ADR and  APC [adenomas per colonoscopy],” noted corresponding author Satoshi Ono, PhD, of the department of gastroenterology and gastrointestinal endoscopy at Tokyo Metropolitan Institute for Geriatrics and Gerongology, and colleagues.
 
“Feedback on withdrawal time has also been shown to support higher ADRs. Despite adequate withdrawal time, substantial variability in ADR and APC persists across endoscopists, highlighting the need for additional quality metrics.”

In the EYE-SIGHT trial, 400 patients aged 40 to 90 years undergoing colonoscopy at four institutions were randomized to eye-tracking and feedback (ETF)–guided withdrawal or usual colonoscopy. The primary endpoint, APC, was higher in the intervention group than in controls (1.34 vs 0.95), according to the research published in Clinical Gastroenterology and Hepatology

ADR also improved, rising to 53.3% with gaze guidance compared with 39.2% in the control group. Polyp detection rate was 66.5% vs 47.2%. Observation time did not differ significantly between groups (8.4 vs 8 minutes). Advanced adenoma and sessile serrated lesion detection rates were similar.
 
The researchers used a web-based system to stratify patients by age, sex and institution. Patients were blinded to allocation; endoscopists were not. Eighteen endoscopists, including experts and non-experts, performed procedures. Baseline characteristics, including age, sex distribution, bowel preparation quality and sedation use, were balanced between groups.
 
The ETF system used a screen-mounted eye tracker that recorded gaze coordinates in real time. The endoscopy monitor was divided into a 5×5 screen grid, and the 16 outer segments were designated as the areas of interest. During withdrawal in the intervention group, a high-pitched tone sounded when gaze fell within peripheral segments and a low-pitched warning tone when it did not. Controls received a continuous tone without gaze-dependent feedback.
 
Lesion-level analyses showed that the intervention increased detection of 5- to 9-mm adenomas, flat-type adenomas and left-sided adenomas. In per-protocol analyses, right-sided detection was also higher. Detection of lesions 10 mm or more did not differ between groups.
 
Gaze analysis in 340 patients confirmed that peripheral gaze rate was significantly higher in the intervention arm (33.7% vs 25.8%). In multivariable analysis, a peripheral gaze rate above 26% independently predicted adenoma detection (odds ratio 2.07), along with older age and male sex.
 
The benefit was concentrated among endoscopists with lower baseline peripheral gaze rates. Those starting below 26% showed significant gains in both gaze rate and ADR, while those already at or above that threshold did not improve.
 
The authors noted limitations of the analysis, including use of a proprietary ETF system, lack of improvement in advanced adenoma detection, absence of endoscopist blinding, and enrollment of endoscopists with baseline ADRs of at least 25%, which limits generalizability to low detectors.
 
“Peripheral gaze guidance significantly improved adenoma detection — particularly for flat lesions — regardless of endoscopist experience level,” the authors reported. “Importantly, this intervention did not increase procedure time, suggesting no additional procedural burden. However, no improvement in AADR [advanced adenoma detection rate] or SSLDR [sessile serrated lesion detection rate] was observed. A subsequent large-scale trial involving endoscopists across diverse settings, particularly low-detector endoscopists, is needed to evaluate user ergonomics and validate improvements in clinically relevant metric.”
 
The work was supported by Japanese Society for the Promotion of Science and the Kowa Life Science Foundation. Two of the 12 study authors reported receiving honoraria for lectures from Fujifilm Corporation and Olympus Corporation. The other authors declared no conflicts of interest.

Summary content