Serum hepatitis B virus (HBV) markers are the most important data for epidemiological screening and clinical diagnosis of HBV infection, especially in endemic areas. We compared the results of the Roche Modular Analytics E170 assay, the Abbott Architect i2000 assay, and an immunoradiometric assay (IRMA) for HBV surface antigen (HBsAg), anti-HBV surface antigen (anti-HBs), HBV e antigen (HBeAg), and anti-HBV e antigen (anti-HBe). A number of serum samples (264, 263, 224, and 202 for HBsAg, anti-HBs, HBeAg, and anti-HBe, respectively) were studied. For samples giving discrepant results for HBeAg between methods, real-time PCR assays were performed. The concordance rates among the three methods were high for HBsAg (100%) and HBeAg (94.6), but low for anti-HBs (91.6%) and anti-HBe (82.2%). For anti-HBs, which could be measured quantitatively by the Modular E170 and Architect i2000 procedures, discrepant results were observed at low levels of anti-HBs. For anti-HBe, the positive rate was highest with Modular E170 (60.9%) followed by the IRMA kit (54.1%) and Architect i2000 (51.0%). This study shows substantial differences between the assay results by the three methods, which should be taken into account in determinations of serum HBV markers.