Article Text
Abstract
Introduction Pancreatic cancer is a devastating disease with a poor outcome. Current biomarkers fall short in accessibility, sensitivity, specificity and their ability to distinguish malignant from benign conditions. Metabolomics aims to decipher molecular signatures that will distinguish disease from controls, ultimately leading to novel targets for diagnosis and treatment. The aim of this novel study was to characterise potentially new pancreatic cancer biomarkers.
Method Urine and serum samples from 32 patients with pancreatic cancer (PC), were compared with healthy controls and patients with chronic pancreatitis and benign jaundice. Two analytical platforms, UHPLC-MS and GC-TOF-MS, were used in this study in order to minimise false negatives and identify a quantitative compliment of all metabolites across a wide molecular weight range. GC-TOF-MS can identify very low molecular weight polar metabolites <350 Da, where-as reverse phase UHPLC-MS is useful in resolving and detecting non-polar metabolites. Subsequent metabolite identification was subject to univariate and multivariate analysis (p < 0.05).
Results After univariate analysis, a total of 747 serum metabolites and 638 urinary metabolites were identified as significantly different in comparison to healthy controls. When chronic pancreatitis and benign jaundice controls were used for comparison to pancreatic cancer then 185 and 176 serum and 333 and 51 urinary metabolites were identified as statistically significant at differentiating these groups respectively. Multivariate analysis showed good discrimination between resectable and non-resectable cases, based on metabolite profile. Urine profiles, in particular, showed consistent separation between resectable and non-resectable disease. When participants without definitive pathology were removed from analysis, the separation improved further. It was so clear that only an unsupervised PCA analysis was sufficient to visualise group separation. Importantly there did not seem to be any effect from diabetes or jaundice on the differentiation of metabolite profiles.
Conclusion In this novel study, metabolite profiling can distinguish patients with pancreatic cancer from controls. Furthermore, multivariate analysis of metabolite profiles is able to distinguish resectable from irresectable disease with 95% confidence intervals. Multiple putative biomarkers have been identified and will be presented.
Disclosure of interest None Declared.