Background: Gastric cancer (GC) is one of the most common malignant tumors. Metabolomics has shown promise to be an important novel tool in cancer detection. This research was conducted to identify whether GC had a divergent urinary metabolic phenotype compared with healthy controls (HCs). Methods: Urines from 77 GC patients and 67 matched HCs were analyzed using gas chromatography-mass spectroscopy (GC-MS). Univariate and multivariate statistical analysis were employed. Differential metabolites were identified using orthogonal partial leastsquares- discriminant analysis (OPLS-DA). Potential GC vs. HCs biomarker model was identified using logistic regression analysis. Diagnostic performance was evaluated using receiver-operating characteristic (ROC) curve analysis. Results: The 68 metabolites were identified using GC-MS. GC patients had a distinct urinary metabolic phenotype. There were 26 differential metabolites identified by OPLS-DA model, and a potential biomarker model consisting of five metabolites-lactic acid, 1-methylnicotinamide, glutamine, myo-inositol and 3-indoxylsulfate-were generated by logistic regression analysis. ROC curve analysis showed the good diagnostic performance with an area under the receiver operating characteristic curve (AUC) of 0.952 (95% CI=0.913-0.995) and 0.961 (95% CI=0.924-1.000) in the training and testing set, respectively. Conclusion: These results demonstrated that the clinical applicability of metabolic profiling for early GC diagnosis showed great promise and should be explored further.