Optimizing surveillance in Lynch syndrome: lesion detection and comparative performance of different colonoscopy modalities-a systematic review and network meta-analysis.

Abstract

Purpose: Lynch syndrome patients are at a high risk for developing colorectal cancer; thus, optimal surveillance strategies are required. Although colonoscopic imaging methods differ in diagnostic performance, direct comparisons in this population are not very common. We aimed to evaluate and compare the detection capabilities of white-light endoscopy (WLE), chromoendoscopy, virtual chromoendoscopy (NBI: narrow band imaging, LCI: linked color imaging, I-SCAN), and AI-assisted colonoscopy in detecting neoplastic and non-neoplastic lesions in individuals diagnosed with Lynch syndrome.

Methods: Up until March 2025, PubMed, WOS, and Scopus were searched. Relevant studies included observational or interventional designs that contrasted various forms of colonoscopy in adults with Lynch syndrome. The primary outcomes were the lesion detection rate and number of lesions per colonoscopy. Secondary outcomes included total procedure time and withdrawal time. Credibility of the evidence was assessed employing CINeMA.

Results: Nine studies were included. LCI and chromoendoscopy demonstrated a significantly higher neoplastic lesion detection rate compared to WLE (RD 0.11, 95% CI [0.01, 0.21], P = 0.03) and (RD 0.07, 95% CI [0.01, 0.14], P = 0.03), respectively, and LCI significantly detected more lesions per procedure (MD = 0.23, 95% CI 0.01-0.45, P = 0.04). Chromoendoscopy was better at marking the non-neoplastic lesions (RD 0.16, 95% CI [0.05, 0.27], P = 0.005) but had the longest procedure and withdrawal times. AI-assisted, as well as virtual ones, were better than WLE but were not as effective as LCI or chromoendoscopy.

Conclusion: In terms of efficiency, LCI and chromoendoscopy improved WLE in detecting neoplastic lesions in Lynch syndrome. Chromoendoscopy remains valuable for non-neoplastic detection, but procedural time is a major drawback. AI-assisted technologies are promising, which require additional investigation.