Stored sera from 181 Greek patients with hepatocellular carcinoma (HCC), 35 patients with metastatic liver cancer, and 416 hospital controls with diagnoses other than malignant neoplasm or liver disease were examined with first and second generation hepatitis C virus (HCV) enzyme immunoassays as well as with five HCV supplemental assays based on structural and nonstructural HCV peptides. Second generation HCV enzyme immunoassays were more sensitive than first generation assays. However, both assays had suboptimal specificity using the standard reactivity criterion (absorbance of sample to cutoff greater than or equal to 1.0). Specificity was improved by centrifugation and by using a sample's optical density to cutoff ratio greater than or equal to 3.0 or supplemental assays; in this instance the prevalence of antibodies to HCV was 13.3% (24 of 181), 0 (0 of 35), and 1.4% (6 of 416) in HCC, metastatic liver cancer, and hospital controls, respectively. A similar estimation of prevalence of antibody to HCV in HCC (12.5% or 4 of 32) was obtained when the recombinant immunoblot assay, second generation, was used to screen a random sample of HCC patients. The relative risk linking HCV to HCC was estimated as 10.4 (95% confidence interval, 4.2-26.0; P less than 0.0001). These data suggest that the prevalence of antibodies to HCV in HCC using stored sera has been previously overestimated even though the evidence of a causal association of HCV with HCC persists.