The purpose of this study was to evaluate the predictive ability of PTGIS for liver metastasis. Protein expression of PTGIS was analyzed on tissue microarray consisting of 117 CRC cases with liver metastasis (M1) and 104 cases of CRC without liver metastasis at least 5 years after resection of primary CRC (M0) by immunohistochemistry. Expression of PTGIS in 147 of 221 of primary lesions exhibited positive staining. Moreover, the PTGIS expression was significantly higher in CRC-M1 than CRC-M0 group. More importantly, the 87% (20/23) heterochronous metastatic cases showed positive staining for PTGIS. Collecting the primary and liver metastatic tumor samples from the same colon cancer patients, we tested the expression of PTGIS and revealed that the expression level of PTGIS in the hepatic metastases was noticeably higher than in the matched primary colon cancer tissues from the same patient in 9 out of 16 cases examined. Logistic regression analysis indicated that the expression of PTGIS and lymph node involvement were risk factors in colon cancer liver metastasis independent of the other variables. In leave-one-out validation model, the combination of PTGIS and lymph node involvement yielded the 89.7% satisfactory sensitivity and 83% specificity for detection of hepatic metastasis. Kaplan-Meier survival analysis revealed a correlation between higher PTGIS expression levels and shorter overall survival times. In conclusion, our results suggest that PTGIS combined with lymph node involvement may be used as accurate predictors of liver metastasis in colorectal cancer.