Objective Elevated pancreatic cyst fluid carcinoembryonic antigen (CEA) has been routinely used to classify mucinous cysts. This study incorporates original data that established the CEA ≥192 ng/mL threshold with over 20 years of additional data and reassesses the diagnostic performance of CEA for differentiating mucinous from non-mucinous cysts.

Design 1169 pancreatic cysts (1999–2021) with CEA results were identified. 394 cases had histological confirmation as the diagnostic standard. Additionally, 237 cysts without histological confirmation demonstrated KRAS, GNAS, or RNF43 mutations by molecular testing and were combined with the histologically confirmed cysts for separate analysis on a total cohort of 631 cysts.

Results Median CEA was significantly higher in mucinous cysts (323.9 ng/mL, n=314) versus non-mucinous cysts (204.6 ng/mL, n=80) (p<0.001). Receiver operating characteristic curve analysis demonstrated an optimal CEA cut-off of 20 ng/mL (area under the curve: 80%), though the specificity was lower than desired (sensitivity 89%, specificity 64%). At the previously established threshold of 192 ng/mL, sensitivity and specificity were 56% and 78%, respectively. To achieve a specificity of 85% as originally reported, a CEA threshold of 250 ng/mL was needed; the 13 false positive cases at this threshold included 4 benign simple cysts, 2 squamoid cysts, 1 serous cystadenoma, 1 lymphoepithelial cyst and 5 more uncommon entities. All results remained similar within the total cohort after including additional cases with KRAS/GNAS/RNF43 mutations only.

Conclusion Cyst fluid CEA continues to be a useful test in the diagnosis of mucinous pancreatic cysts but does not appear as specific as previously reported. Raising the CEA threshold to 250 ng/mL to maintain specificity for differentiating mucinous from non-mucinous cysts may be considered.