Metabolomic-derived novel cyst fluid biomarkers for pancreatic cysts: glucose and kynurenine

Background: Better pancreatic cyst fluid biomarkers are needed. Objective: To determine whether metabolomic profiling of pancreatic cyst fluid would yield clinically useful cyst fluid biomarkers. Design: Retrospective study. Setting: Tertiary-care referral center. Patients: Two independent cohorts of patients (n = 26 and n = 19) with histologically defined pancreatic cysts. Intervention: Exploratory analysis for differentially expressed metabolites between (1) nonmucinous and mucinous cysts and (2) malignant and premalignant cysts was performed in the first cohort. With the second cohort, a validation analysis of promising identified metabolites was performed. Main Outcome Measurements: Identification of differentially expressed metabolites between clinically relevant cyst categories and their diagnostic performance (receiver operating characteristic [ROC] curve). Results: Two metabolites had diagnostic significance-glucose and kynurenine. Metabolomic abundances for both were significantly lower in mucinous cysts compared with nonmucinous cysts in both cohorts (glucose first cohort P=.002, validation P=.006; and kynurenine first cohort P=.002, validation P=.002). The ROC curve for glucose was 0.92 (95% confidence interval [CI], 0.81-1.00) and 0.88 (95% CI, 0.72-1.00) in the first and validation cohorts, respectively. The ROC for kynurenine was 0.94 (95% CI, 0.81-1.00) and 0.92 (95% CI, 0.76-1.00) in the first and validation cohorts, respectively. Neither could differentiate premalignant from malignant cysts. Glucose and kynurenine levels were significantly elevated for serous cystadenomas in both cohorts. Limitations: Small sample sizes. Conclusion: Metabolomic profiling identified glucose and kynurenine to have potential clinical utility for differentiating mucinous from nonmucinous pancreatic cysts. These markers also may diagnose serous cystadenomas.