Epidemiological studies have implicated the inducible form of cyclooxygenase (COX-2) in the pathogenesis of colorectal cancer; however, its role is not fully understood. To examine the relationship between the expression of COX-2 in human colorectal cancer and patient survival. Patients diagnosed as having colorectal cancer were evaluated and followed up for up to 9.4 years (median follow-up, 2.7 years). Tumor sections were stained for COX-2 using a rabbit polyclonal antibody raised against human COX-2. The extent of COX-2 staining was graded by 2 observers blinded to outcome. Preabsorption of the anti-COX-2 antibody with a COX-2 peptide abolished the staining, demonstrating the specificity of the assay. Gastrointestinal unit of a large general teaching hospital in Dublin, Ireland. Seventy-six patients (median age, 66.5 years) with colorectal cancer (Dukes tumor stage A, n = 9; Dukes B, n = 30; Dukes C, n = 25; Dukes D, n = 12) whose diagnosis was made between 1988 and 1991. Fourteen normal colon biopsies were stained for COX-2 as controls. Survival in years following diagnosis compared by extent of COX-2 epithelial staining (grade 1, <1%; grade 2, 1%-19%; grade 3, 20%-49%; grade 4, > or = 50%), Dukes stage, tumor size, and lymph mode metastasis. COX-2 was found in tumor epithelial cells, inflammatory cells, vascular endothelium, and/or fibroblasts. The extent of epithelial staining was heterogeneous, varying markedly among different tumors. Normal tissue adjacent to the tumors also stained weakly for COX-2. No COX-2 was detected in control tissue samples. The Kaplan-Meier survival estimate was 68% in patients who had grade 1 tumor epithelial staining compared with 35% in those with higher grades combined (log-rank chi2 = 5.7; P = .02). Greater expression of COX-2 correlated with more advanced Dukes stage (Kendall tau-b, 0.22; P = .03) and larger tumor size (Kendall tau-b, 0.21; P = .02) and was particularly evident in tumors with lymph node involvement (Kendall tau-b, 0.26; P = .02). Our data indicate that COX-2 expression in colorectal cancer may be related to survival. These data add to the growing epidemiological and experimental evidence that COX-2 may play a role in colorectal tumorigenesis.