This study aimed to evaluate the diagnostic accuracy of Endoscopic Ultrasound-guided Tissue Acquisition (EUS-TA) for pancreatic lesions in a tertiary referral center without Rapid On-Site Evaluation (ROSE). Secondary objectives pointed to assess the influence of specific variables (lesion size, location, needle gauge, number of passes, and needle type) on the diagnostic yield.

We conducted a retrospective single-center study on 129 patients with 134 pancreatic lesions undergoing EUS-TA between 2020 and 2024. All procedures were performed by experienced endoscopists. Diagnostic yield was calculated based on the definitive result of the most recent procedure; indeterminate cases (Papanicolaou C3/H3) were reclassified based on KRAS mutation analysis. A composite gold-standard (GS) was defined as histological confirmation from surgery or minimum 6-month clinical-radiological follow-up consistent with disease evolution. Univariate analysis (Chi-square test) was performed to evaluate the impact of lesion size, location, needle gauge, number of passes, and needle type (FNA vs FNB) on diagnostic accuracy. McNemar’s test was used to analyse the efficacy of repeat biopsies.

Out of 134 included lesions, 130 yielded a conclusive result and were included in the performance analysis. Comparison with the GS revealed 107 True Positives, 15 False Negatives, 7 True Negatives, and 1 False Positive. Overall accuracy was 87.7% (95% CI: 80.9–92.3). Sensitivity was 87.7% (95% CI: 82.1–93.3), Specificity 87.5% (95% CI: 65–100) and Positive Predictive Value (PPV) 99.1% (95% CI: 94.9–99.8). The Negative Predictive Value (NPV) was 32.0% (95% CI: 12.4–52.1). Univariate analysis revealed no significant correlations with lesion location, size, needle gauge, or number of passes. Specifically, no significant difference in failure rates was observed between FNA (9.6%) and FNB (14.1%) (p=0.43). However, repeat biopsy significantly improved the diagnostic yield in initially inconclusive cases (p<0.001).

EUS-TA without ROSE demonstrates high diagnostic accuracy and an excellent PPV (99.1%) for pancreatic lesions. Our analysis confirms that diagnostic performance is not significantly influenced by needle type (FNA vs FNB) or other technical variables in this setting. However, in a tertiary setting with a high pre-test probability of disease, the low NPV highlights that EUS-TA does not reliably exclude malignancy, therefore requiring careful evaluation and repeated EUS-TA to resolve diagnostic uncertainty.