The branched DNA (bDNA) assay for hepatitis B virus (Chiron Corporation Emerville, USA) was investigated by application to HBV-infected patients in Taiwan, where the B and C genotypes of hepatitis B virus are most prevalent. The study group included sera with hepatitis B surface antigen (HBsAg) and e antigen (HBeAg); Group 1 (n=70) without treatment; Group 2 (n=28) lamivudine treatment less than 3 months; Group 3 (n=73) lamivudine treatment 3-12 months; Group 4 (n=45) HBeAg-negative sera after 1 year treatment with lamivudine; control group (n=36) HBsAg-negative sera. Comparison of identical-sample results showed a significantly higher coefficient of variation for low-level HBV DNA (<100 MEq/ml) than for high-level (> or =100 MEq/ml), indicating increasing assay inaccuracy uncertainty as the sample HBV DNA concentration decreased. It is thus concluded that low-titered sera should receive special careful pipetting and processing. It was also found that using the relative luminescence of the negative control plus two standard deviations (S.D.) as a new cutoff could promote sensitivity (97.1-->97.1%, 89.3-->100%, 76.7-->84.9%, and 17.8-->22.2% in Groups 1-4, respectively) and specificity (94.4-->97.2%). In summary, the bDNA HBV assay showed only moderate assay performance for samples with low HBV DNA levels. This problem can be improved partially by choosing a new cutoff value based on the relative luminescence of the negative controls in the kit.