Gastric cancer has one of the highest morbidities and mortalities worldwide. Early detection is key measure to improve the outcome of gastric cancer patients. In our efforts to identify potential markers for gastric cancer detection, we coupled xenotransplantation mouse model with a plasma proteomic approach. MKN45 gastric cancer cells were subcutaneously injected into nude mice and plasma samples from mice bearing different sizes of tumors were collected and subjected to iTRAQ and mass spectrometry analysis. ITIH3 protein was found to be more highly expressed in plasma of tumor bearing mice compared to control. Subsequent screening of ITIH3 expression in 167 clinical plasma samples, including 83 cancer-free subjects and 84 gastric cancer patients, revealed higher ITIH3 level in the plasma of gastric cancer patients. A receiver operating characteristics (ROC) curve estimated a maximal sensitivity of 96% at 66% specificity for ITIH3 in gastric cancer detection. In addition, plasma from early stage gastric cancer patient has significantly (p < 0.001) higher level of ITIH3 compared to that from noncancer subject. Our data suggest that ITIH3 may be a useful biomarker for early detection of gastric cancer.