Accurate Identification of Mucinous Pancreatic Cystic Lesions Using Small-Volume Analytes

Introduction: The accurate identification of mucinous pancreatic cystic lesions (PCLs) is paramount for cancer risk stratification. Cyst fluid carcinoembryonic antigen (CEA), the only routinely used test, requires high volumes and has low sensitivity. We aimed to compare the performance of two investigational small-volume biomarkers, glucose and the protease gastricsin, to CEA for PCL classification.

Methods: We obtained cyst fluid samples from 81 patients with pathologically confirmed PCLs from four institutions between 2003 and 2016. Gastricsin activity was measured using an internally quenched fluorescent substrate. Glucose levels were measured with a standard glucometer. CEA levels were obtained from the medical record. Models using Classification and Regression Trees were created to predict mucinous status. Model performance was evaluated using nested cross-validation.

Results: Gastricsin activity, CEA, and glucose levels from patients with mucinous (n = 50) and nonmucinous (n = 31) PCLs were analyzed. Area under the curve (AUC) was similar for individual classifiers (gastricsin volume normalized [GVN] 0.88; gastricsin protein concentration normalized [GPN] 0.95; glucose 0.83; CEA 0.84). The combination of two classifiers did not significantly improve AUC, with CEA + GVN (0.88) performing similarly to CEA + GPN (0.95), GVN + glucose (0.87), GPN + glucose (0.95), and CEA + glucose (0.84). The three-analyte combination performed similarly to single and dual classifiers (GPN + glucose + CEA AUC 0.95; GVN + glucose + CEA AUC 0.87). After multiple comparison corrections, there were no significant differences between the individual, dual, and triple classifiers.

Conclusions: Gastricsin and glucose performed similarly to CEA and required <5% of the volume required for CEA; these classifiers may be useful in patients with limited cyst fluid. Future multicenter prospective studies are needed to validate and compare these novel small-volume biomarkers.