Macroscopic versus rapid on-site evaluation for EUS-guided tissue acquisition of solid pancreatic masses: a systematic review, diagnostic test accuracy meta-analysis, and network meta-analysis

Why clinicians should know about it

• Picked for Pathology and Forensic Medicine (top studies of the week, 8 November 2026): Meta-analysis compares methods for pancreatic tissue diagnostic accuracy

Abstract

BACKGROUND: Endoscopic ultrasound-guided tissue acquisition (EUS-TA) is central to the pathological diagnosis of solid pancreatic masses, but the comparative value of rapid on-site cytologic evaluation (ROSE), macroscopic on-site evaluation (MOSE), and no on-site evaluation (NONE) may vary by needle type and remains uncertain. METHODS: We performed a PRISMA-DTA systematic review registered in PROSPERO (CRD420251241446). Potential cohort overlap and data dependency were assessed before quantitative synthesis. Diagnostic 2 × 2 data were synthesized with a bivariate binomial-normal random-effects model. Independent direct comparisons were pooled by random-effects meta-analysis for correct versus incorrect diagnosis, with complementary absolute differences in sensitivity and specificity. An exploratory three-node network meta-analysis (NMA) compared ROSE, MOSE, and NONE. RESULTS: Thirty-one eligible reports contributed 67 strategy-specific diagnostic datasets. After excluding 21 duplicate, nested, paired, or otherwise dependent datasets, 46 independent datasets representing 4,901 arm-level observations were included in the primary DTA. Pooled sensitivity was 0.94 (95% CI 0.92-0.95) and specificity was 1.00 (0.98-1.00); Deeks' asymmetry test was positive (P <0.001). In direct comparisons, ROSE improved composite diagnostic accuracy versus NONE in FNA studies (OR 5.56, 95% CI 1.76-17.62; k = 4), but not in FNB studies (OR 1.08, 0.42-2.79; k = 4); the FNA-versus-FNB interaction was exploratory (P = 0.012). MOSE versus NONE in FNB was uncertain (OR 1.31, 0.60-2.87; k = 4), as was MOSE versus ROSE (OR 1.53, 0.05-49.13; k = 3). The exploratory NMA did not identify a statistically clear difference for ROSE versus NONE (OR 1.67, 0.94-2.95), MOSE versus NONE (OR 1.89, 0.86-4.20), or MOSE versus ROSE (OR 1.14, 0.48-2.71). CONCLUSIONS: EUS-TA showed high overall diagnostic performance. ROSE may add value in FNA-era workflows, whereas a clear incremental benefit was not detected with FNB. MOSE remains a plausible resource-sparing strategy, but comparative estimates are imprecise. The NMA should be regarded as hypothesis-generating because of sparse comparisons, methodological heterogeneity, small-study effects, and low-to-very-low certainty. SYSTEMATIC REVIEW REGISTRATION: https://www.crd.york.ac.uk/PROSPERO/, identifier CRD420251241446.

Abstract as published, via PubMed.