Abstract
BACKGROUND: Timely diagnosis and classification of colorectal cancer (CRC) are hindered by unsatisfactory clinical assays. Our aim was to construct a blood-based biomarker series using a single assay, suitable for CRC detection, prognostication and staging. METHODS: Serum metabolomic profiles of adenoma (N=31), various stages of CRC (N=320) and healthy matched controls (N=254) were analysed by gas chromatography-mass spectrometry (GC-MS). A diagnostic model for CRC was derived by orthogonal partial least squares-discriminant analysis (OPLS-DA) on a training set, and then validated on an independent data set. Metabolomic models suitable for identifying adenoma, poor prognosis stage II CRC and discriminating various stages were generated. RESULTS: A diagnostic signature for CRC with remarkable multivariate performance (R(2)Y=0.46, Q(2)Y=0.39) was constructed, and then validated (sensitivity 85%; specificity 86%). Area under the receiver-operating characteristic curve was 0.91 (95% CI, 0.87-0.96). Adenomas were also detectable (R(2)Y=0.35, Q(2)Y=0.26, internal AUROC=0.81, 95% CI, 0.70-0.92). Also of particular interest, we identified models that stratified stage II by prognosis, and classified cases by stage. CONCLUSIONS: Using a single assay system, a suite of CRC biomarkers based on circulating metabolites enables early detection, prognostication and preliminary staging information. External population-based studies are required to evaluate the repeatability of our findings and to assess the clinical benefits of these biomarkers.