Cyclooxygenase-2 (COX-2) influences carcinogenesis through immune response suppression, apoptosis inhibition, regulation of angiogenesis and tumor cell invasion, and metastasis. It is now well established that COX-2 is overexpressed in many premalignant, malignant, and metastatic cancers, including hepatocellular carcinoma (HCC). DNA sequence variations in the COX-2 gene may lead to altered COX-2 production and/or activity, and so they cause inter-individual differences in the susceptibility to HCC. Functional coding region polymorphisms -1195A>G (rs689466), -765G>C (rs20417), and +8473T>C (rs5275) in the COX-2 gene have recently been shown to be associated with several human cancers but their association with HCC has yet to be investigated. We used hospital-based case-control study to assess the hypothesis that the functional COX-2 variation may affect individual susceptibility to the HCC. COX-2 polymorphisms were investigated in 129 confirmed subjects with HCC and 129 cancer-free control subjects matched on age, gender, smoking, and alcohol consumption using a polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) assay. The distribution of the COX-2 -1195A>G and +8473T>C genotypes were not significantly different between HCC cases and control. However, proportion of the COX-2 -765CC genotype which leads to a 30% reduction of the COX-2 promoter activity was significantly lower in patients with HCC (3.1%) when compared to control subjects (11.6%) (P < 0.05). Logistic regression analyses revealed that the COX-2 -765G>C variant genotype (-765CC) was associated with a significantly decreased risk of HCC compared with the -765GG wild-type homozygotes [P < 0.05, odds ratio (OR) = 0.25, 95% confidence interval (CI) = 0.08-0.79]. Our results suggest for the first time that the -765CC genotype of COX-2 -765G>C polymorphism, causing lower COX-2 gen expression, is a genetic protective factor for HCC. However, because this is the first report concerning the COX-2 -1195A>G, -765G>C, and +8473T>C polymorphisms and the risk of HCC, independent studies are needed to validate our findings.