Cyclooxygenase-2 (COX-2) is up-regulated in 85-90% of primary human colorectal cancers and is a putative target for the chemopreventative activity of non-steroidal anti-inflammatory drugs. However, COX-2 expression by human colorectal cancer liver metastases has been poorly characterized. We studied a consecutive series of 38 patients who underwent liver resection for metastatic disease, for whom long-term (up to 57 months), prospective follow-up data were available. Semi-quantitative immunohistochemistry for COX-2 was performed on 54 metastases from 35 patients, for whom adequate histological material was available. Diffuse cytoplasmic staining for COX-2 protein was detected in cancer cells in 100% of metastases (COX-2 score 1, n = 25; score 2, n = 29). There was no relationship between metastasis size or differentiation grade and the level of COX-2 protein expression. There was no difference in colorectal cancer-free or overall survival between patients with high (score 2) and low (score 1) COX-2 scores (Kaplan-Meier survival analysis and log rank test, both P = 0.97). Multivariate Cox regression analysis identified age, incomplete resection and presence of extra-hepatic disease as independent predictors of disease-free and overall survival following surgery. COX-2 protein was also localized to a subset of stromal fibroblasts and mononuclear cells within metastases as well as hepatocytes from resection specimens. COX-2 protein was expressed by cancer cells in all human colorectal cancer liver metastases which were studied. Investigation of the effect of selective COX-2 inhibition on metastasis growth and metastasis cancer cell proliferation/apoptosis in vivo are warranted.