Three assays, one based on monoclonal antibodies and the others on polyclonal antibodies, were employed to detect hepatitis B surface antigen (HBsAg)-reactive samples in both vaccinated and unvaccinated populations in areas of the world where hepatitis B virus (HBV) is endemic. Any discordant sera were tested by polymerase chain reaction (PCR) to confirm current infection, and sequence data were obtained from the DNA coding for the major hydrophilic region (MHR) of HBsAg of those samples positive for PCR. In all countries studied, samples that reacted in one HBsAg assay but not another were found. In the most extreme case, about 5% of viremic sera in Papua New Guinea were nonreactive in the monoclonal HBsAg assay; 9 of the 13 PCR-positive samples had novel or once-described variants, or a variant out of its usual genotype context. In South Africa, samples with sequences of subtype ayw2 reacted poorly, particularly in the polyclonal assay. Two had novel variants. In Sardinia, antibody to hepatitis B core antigen (anti-HBc) was analyzed as a marker of infection. A significant proportion of anti-HBc-positive, but monoclonal HBsAg-negative, vaccinees and unvaccinated persons were found to be PCR positive, as were some individuals without any markers of hepatitis B virus infection. Five more novel variants were found in these groups. There are implications for the design of HBsAg assays, which may have to be modified according to local sequence variability. Not all discordant samples were explained by variants, indicating that assay sensitivity is fundamental to diagnostic efficacy. Overall, this study defined 16 novel variants and 2 new potential epitope clusters.