Background/aims: Among various specimen preparation techniques for rapid on-site evaluation (ROSE), touch imprint cytology (TIC) is widely used; however, its utility for pancreatic lesions has not been fully validated. This study aimed to evaluate the diagnostic performance of TIC-based ROSE during endoscopic ultrasound-guided tissue acquisition (EUS-TA) for pancreatic lesions.
Methods: This retrospective study included 385 patients who underwent EUS-TA for pancreatic lesions at a single center. ROSE using TIC was performed in 370 cases (96.1%). Diagnostic concordance between TIC-based ROSE and histology, as well as diagnostic accuracy relative to the final diagnosis, were assessed. The diagnostic performance of TIC was compared with that of spray cytology (SC) and instrument rinse cytology (IRC). Multivariate analysis was conducted to identify factors associated with discordance between TIC-based ROSE and histology.
Results: The concordance rate between TIC-based ROSE and histology was 85.1% (κ=0.575). For the final diagnosis, TIC-based ROSE demonstrated a sensitivity of 92.3%, specificity of 96.9%, and accuracy of 92.7%, showing superior diagnostic performance within the same workflow compared with SC and IRC. Smaller lesion size and the use of a 25-gauge needle were identified as independent factors for discordance.
Conclusions: TIC is a reliable ROSE technique in EUS-TA for pancreatic lesions, providing high diagnostic accuracy and a low inadequacy rate.
Keywords: Endoscopic ultrasonography; Fine-needle aspiration; Pancreatic neoplasms cytodiagnosis; Rapid on-site evaluation.