Feasibility of identifying pancreatic cancer based on serum metabolomics.

BACKGROUND We postulated that the abundance of various metabolites in blood would facilitate the diagnosis of pancreatic and biliary lesions, which could potentially prevent unnecessary surgery. METHODS Serum samples from patients with benign hepatobiliary disease (n = 43) and from patients with pancreatic cancer (n = 56) were examined by ¹H NMR spectroscopy to quantify 58 unique metabolites. Data were analyzed by "targeted profiling" followed by supervised pattern recognition and orthogonal partial least-squares discriminant analysis (O-PLS-DA) of the most significant metabolites, which enables comparison of the whole sample spectrum between groups. RESULTS The metabolomic profile of patients with pancreatic cancer was significantly different from that of patients with benign disease (AUROC, area under the ROC curve, = 0.8372). Overt diabetes mellitus (DM) was identified as a possible confounding factor in the pancreatic cancer group. Thus, diabetics were excluded from further analysis. In this more homogeneous pancreatic cancer group, compared with benign cases, serum concentrations of glutamate and glucose were most elevated on multivariate analysis. In benign cases, creatine and glutamine were most abundant. To examine the usefulness of this test, a comparison was made to age- and gender-matched controls with benign lesions that mimic cancer, controlling also for presence of jaundice and diabetes (n = 14 per group). The metabolic profile in patients with pancreatic cancer remained distinguishable from patients with benign pancreatic lesions (AUROC = 0.8308). CONCLUSIONS The serum metabolomic profile may be useful for distinguishing benign from malignant pancreatic lesions. IMPACT Further studies will be required to study the effects of jaundice and diabetes. A more comprehensive metabolomic profile will be evaluated using mass spectrometry.