Objective: Better tools are needed for early diagnosis and classification of pancreatic cystic lesions (PCL) to trigger intervention before neoplastic precursor lesions progress to adenocarcinoma. We evaluated the capacity of molecular analysis to improve the accuracy of cytologic diagnosis for PCL with an emphasis on non-diagnostic/negative specimens. Design: In a span of 7 years, at a tertiary care hospital, 318 PCL endoscopic ultrasound-guided fine needle aspirations (EUS-FNA) were evaluated by cytologic examination and molecular analysis. Mucinous PCL were identified based on a clinical algorithm and 46 surgical resections were used to verify this approach. The mutation allele frequency (MAF) of commonly altered genes (BRAF, CDKN2A, CTNNB1, GNAS, RAS, PIK3CA, PTEN, SMAD4, TP53 and VHL) was evaluated for their ability to identify and grade mucinous PCL. Results: Cytology showed a diagnostic sensitivity of 43.5% for mucinous PCL due in part to the impact of non-diagnostic (28.8%) and negative (50.5%) specimens. Incorporating an algorithmic approach or molecular analysis markedly increased the accuracy of cytologic evaluation. Detection of mucinous PCL by molecular analysis was 93.3% based on the detection of KRAS and/or GNAS gene mutations (p = 0.0001). Additional genes provided a marginal improvement in sensitivity but were associated with cyst type (e.g. VHL) and grade (e.g. SMAD4). In the surgical cohort, molecular analysis and the proposed algorithm showed comparable sensitivity (88.9% vs. 100%). Conclusions: Incorporating somatic molecular analysis in the cytologic evaluation of EUS-FNA increases diagnostic accuracy for detection, classification and grading of PCL. This approach has the potential to improve patient management.
Keywords: molecular next generation sequencing; non-diagnostic cytology; pancreatic cyst classification.