Supporting Method Quantitative HBeAg assay The development of the quantitative HBeAg assay was based on the method of Perrillo et al (14). Serum HBeAg was measured by EIA (ARCHITECT platform; Abbott Laboratories). This commercial HBeAg kit is not marketed as a quantitative assay but produces a signal which is linear within a restricted range. The HBeAg reference preparation was obtained from the Paul Ehrlich (PE) Institute (Paul-Ehrlich Institute, Langen, Germany). The preparation has a defined HBeAg activity of 100 PE IU/mL. An in-house working HBeAg standard was then prepared from a pool of high-titre HBeAg positive specimens received for routine diagnostic testing, and calibrated against the PE reference preparation. The standard was subsequently determined to have an HBeAg titre of 2900 PE IU/mL, and was then used to generate a calibration curve within the linear range of the assay. Dilutions of the in-house working standard (diluted twofold, from 1 in 2 to 1 in 4,096, or 1450 to 0.7 PE IU/mL) were prepared using the ARCHITECT HBsAg Manual Diluent (Abbott Diagnostics, Sligo, Ireland), which contains recalcified human plasma that is non-reactive for HBsAg, HIV-1 RNA or HIV-1 Ag , anti-HIV-1/HIV-2, anti-HCV, and anti-HBs. The standards were aliquoted and stored at -20C. The performance characteristics of the assay for use as a quantitative tool were evaluated. The in-house HBeAg working standard was used to generate a calibration curve within the linear range of the assay. The linear range was approximately 0.5 – 100 PE IU/mL. For determination of the inter-assay variability, 8 dilutions of the standard (1 in 2 – 1 in 4096) were tested in four separate runs. No significant deviation from linearity was present for any of the following dilutions: 1 in 16, 1 in 32, 1 in 64, 1 in 128, 1 in 256, 1 in 512, 1 in 1024 and 1 in 2048 (data not shown). Intra-assay variation was tested using duplicate samples of the reference preparation at 1 in 64 and 1 in 2048 dilutions, in four separate assay runs. The percentage coefficient of variation ranged from 0.68 – 7.6%, with the higher coefficient of variation at lower HBeAg levels. The intra-assay variation was less than 5% (range 0.54 – 2.7%) for all samples (results not shown). Each day that clinical samples were tested, aliquots of the diluted reference sample (1 in 16 – 1 in 2048) were used to generate a standard curve. Linear regression was used to convert the assay result into PE IU/mL for each sample. Where samples fell outside the linear range of the assay (approximately 0.5 – 90 PE IU/mL, but defined by the standard curve for each assay run), serial dilution (1 in 10, 1 in 100, 1 in 1000, ± 1 in 10,000) was performed using the ARCHITECT HBsAg Manual Diluent. Non-reactive human serum was used as the negative control.