Biobanks containing formalin-fixed paraffin-embedded tissue, as well as frozen serum or plasma, are important resources for molecular epidemiologic studies. However, few studies have compared the reliability of formalin-fixed tissue samples and archival plasma samples for genotyping. We determined the genotype of four proposed genetic risk factors for hepatocellular carcinoma [hereditary hemochromatosis (HFE 63 and 282), alpha(1)-antitrypsin deficiency (AAT 342) and cystic fibrosis (CFTR 508)] on formalin-fixed tissue samples, stored for up to 25 years, from 318 patients diagnosed with hepatocellular carcinoma and on plasma or serum samples from 31 of these patients. The genotypes were analyzed by RFLP or allele-specific amplification as well as by TaqMan assays. In addition, genotyping was attempted after whole genome amplification by multiple displacement amplification (MDA). Genotyping was successful in 94% of the tissue samples and successful and identical to the tissue samples from the same subjects in 98% of the plasma/serum samples. DNA from plasma samples could be amplified >5,000-fold by MDA and genotyping after MDA gave identical results to the genotyping of the same subjects before whole genome amplification. MDA amplification of the tissue samples was not successful. In summary, archival plasma was found to be an adequate source of efficiently amplifiable DNA. MDA on plasma samples allows analysis of multiple genotypes in epidemiologic studies.