Diagnostic accuracy of limited gastric biopsy strategies for detecting premalignant gastric lesions: A prospective paired-design diagnostic accuracy study using the Updated Sydney protocol as a reference standard

Although the Updated Sydney System recommends systematic biopsy mapping of the antrum, incisura angularis, and corpus, single-site sampling remains common in routine endoscopy. This study quantified the diagnostic loss of single-site biopsy strategies and the residual probability of a premalignant gastric lesion after a negative limited biopsy. In this prospective, paired-design diagnostic accuracy study, 300 adults undergoing upper gastrointestinal endoscopy for dyspeptic symptoms at a single tertiary center between November 2024 and April 2025 underwent the complete Updated Sydney protocol. Antral-only and corpus-only strategies were simulated within each patient and compared with the complete protocol (reference standard). The primary outcome was premalignant gastric lesions (atrophy and/or intestinal metaplasia). Paired proportions were compared using the exact McNemar test. A second gastrointestinal pathologist independently reassessed a random 15% subsample to estimate interobserver agreement. The study was reported according to Standards for Reporting of Diagnostic Accuracy Studies 2015. The study was not registered in a public registry. Premalignant gastric lesions were present in 74 patients (24.7%) and Helicobacter pylori in 118 (39.3%). Sensitivity for premalignant lesions was 86.5% (95% confidence interval [CI]: 76.9–92.5) for antral-only and 67.6% (95% CI: 56.3–77.1) for corpus-only biopsy (paired difference 18.9 percentage points; McNemar P =.009). Antral-only biopsy missed 10 of 74 patients with lesions (13.5%; 3.3 per 100 endoscopies) and corpus-only missed 24 (32.4%; 8.0 per 100 endoscopies). Negative likelihood ratios were 0.15 and 0.34, and diagnostic odds ratios were 69.7 and 50.2, respectively. After a negative antral-only biopsy, the residual probability of a premalignant lesion was 4.6% at the observed prevalence and ≤2.5% at pretest probabilities ≤15%, compared with 10.0% and ≤5.6% after a negative corpus-only biopsy. Interobserver agreement was κ = 0.78 (95% CI: 0.62–0.91) for intestinal metaplasia and κ = 0.58 (95% CI: 0.39–0.75) for atrophy. Antral-only sampling lost less diagnostic information than corpus-only sampling, but neither strategy reliably excluded premalignant lesions at moderate-to-high pretest probability. When complete mapping is not feasible, the antrum should be sampled preferentially, and complete mapping should be prioritized for patients at higher pretest risk.