Potential conflict of interest: Nothing to report.
The assays used to determine serum hepatitis B virus DNA levels (Cobas Taqman assay) and hepatitis B surface antigen levels (Elecsys HBsAg II assay) performed in our laboratory were supported by an unrestricted grant from Roche Diagnostics.
The kinetics of hepatitis B surface antigen (HBsAg) levels preceding spontaneous HBsAg seroclearance has not been fully investigated. The kinetics of HBsAg and hepatitis B virus (HBV) DNA of 203 treatment-naïve, hepatitis B e antigen (HBeAg)-negative patients with spontaneous HBsAg seroclearance were compared with 203 age- and sex-matched HBeAg-negative controls. Serum samples at 3 years, 2 years, 1 year, and 6 months before HBsAg seroclearance and at the time of HBsAg loss were tested. Median HBsAg levels at these respective time points before HBsAg seroclearance were 23.5, 3.51, 0.524, and 0.146 IU/mL. For all time points, patients with HBsAg seroclearance had significantly lower median HBsAg and HBV DNA levels, compared to those of the controls (all P < 0.001). Median HBsAg and HBV DNA levels declined significantly until HBsAg seroclearance (P < 0.001). Although median HBsAg levels also decreased significantly with time (P = 0.006) in controls, median HBV DNA levels remained similar (P = 0.414). Serum HBsAg levels, followed by HBsAg log reduction, were the best predictors of HBsAg seroclearance, with an area under the receiving operator characteristic (AUROC) of 0.833 (95% confidence interval [CI]: 0.792-0.873) and 0.803 (95% CI: 0.755–0.849), respectively. The optimal cut-off HBsAg level and HBsAg reduction to predict HBsAg seroclearance were <200 IU/mL (sensitivity, 84.2%; specificity, 73.4%) and 0.5 log IU/mL/year (sensitivity, 62.8%; specificity, 88.7%), respectively. For patients with HBsAg levels ≥200 IU/mL, an annual 0.5-log reduction was highly predictive of subsequent HBsAg seroclearance (AUROC, 0.867; 95% CI: 0.778-0.956). Conclusion: To conclude, serum HBsAg <200 IU/mL and 0.5-log reduction in HBsAg were predictive of HBsAg seroclearance within 3 years of follow-up. These parameters may serve as good indicators for the consideration of treatment duration and cessation for chronic hepatitis B. (HEPATOLOGY 2012;56:812–819)
Seroclearance of the hepatitis B surface antigen (HBsAg) during the natural history of chronic hepatitis B (CHB) is associated with favorable long-term outcomes,1, 2 although the development of hepatocellular carcinoma (HCC) remains possible.3-5 The incidence of HBsAg seroclearance ranges between 0.5% and 2.26% per year.3, 5, 6 HBsAg seroclearance is the ultimate treatment endpoint for CHB, but occurs only infrequently after pegylated interferon (Peg-IFN)7 or nucleoside analog therapy.8-10
The recent development of serum HBsAg quantification has provided an additional tool in monitoring both treated and untreated CHB patients.11 Serum HBsAg titers were initially proposed as a surrogate marker for hepatitis B virus (HBV) covalently closed circular DNA. But, a recent study found such a correlation to exist only in hepatitis B e antigen (HBeAg)-positive, and not in HBeAg-negative, disease.12
Several recent studies have highlighted the differences in HBsAg titers throughout the natural history of CHB, but are limited by their cross-sectional nature.13, 14 A recent longitudinal study demonstrated the variations in HBsAg levels in different disease phases of CHB. A serum HBsAg reduction of more than 1 log reflects improved immune control and increases the probability of HBsAg seroclearance.15 Two recent studies from Asia followed up 390 and 103 HBeAg-negative patients, respectively, and found a HBsAg level of <100 IU/mL predictive of eventual HBsAg seroclearance.16, 17 These longitudinal studies, however, were all limited by the very small number of patients with HBsAg seroclearance (n < 20). Another recent study consisting of 46 patients with HBsAg seroclearance suggested the optimal level to predict HBsAg seroclearance to be HBsAg <200 IU/mL.18 However, the relationship between HBsAg and HBV DNA preceding HBsAg seroclearance and the possible combined use of both markers in predicting HBsAg seroclearance have not been studied.
The value of serum HBV DNA levels in predicting HBsAg seroclearance remains controversial. In spite of a poor correlation between serum HBsAg levels and HBV DNA levels in HBeAg-negative disease observed in some studies,12, 19 another study found a persistently low serum HBV DNA level and a reduction in HBV DNA both predictive of eventual HBsAg seroclearance.6 The combination of single-point serum HBsAg and HBV DNA quantification also enabled the identification of inactive carriers among CHB genotype D patients.20 It remains to be seen whether the combination of serum HBsAg and HBV DNA could predict HBsAg seroclearance.
In our current study, we proposed to study serum HBsAg kinetics from 3 years preceding HBsAg seroclearance until the time of HBsAg seroclearance and compare them with the kinetics of HBsAg-positive, HBeAg-negative age- and sex-matched controls. By enrolling a large population of CHB patients spontaneously clearing HBsAg without treatment (n = 203), the predictive values of different serum HBsAg and HBV DNA levels could be unequivocally determined.
ALT, alanine aminotransferase; anti-HBe, antibody to hepatitis B e antigen; anti-HBs, antibody to hepatitis B surface antigen; AUCs, areas under the curves; AUROC, area under the receiving operator characteristic; CHB, chronic hepatitis B; CI, confidence interval; HBeAg, hepatitis B e antigen; HBsAg, hepatitis B surface antigen; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; Peg-IFN, pegylated interferon; ROC, receiver operating characteristic.
Patients and Methods
In our two previous studies, we investigated the clinical characteristics of 92 and 298 treatment-naïve Chinese CHB patients with HBsAg seroclearance.1, 4 One hundred and twelve of these patients, with HBsAg seroclearance documented between June 2001 and December 2006, had stored serum available for 3 years before the occurrence of HBsAg seroclearance and were recruited for the present study. In addition, between January 2007 and February 2011, an additional 91 treatment-naïve CHB patients with documented HBsAg seroclearance were also enrolled, bringing the total number of the study population with spontaneous HBsAg seroclearance to 203. All patients had HBsAg positivity documented for more than 6 months and were all HBeAg negative on presentation to our clinic.
Patients were followed up every 6 months for clinical assessment and measurement of liver biochemistry, alpha-fetoprotein, and HBV serology. Serum HBsAg seroclearance was defined as loss of serum HBsAg with or without the appearance of antibody to hepatitis B surface antigen (anti-HBs) for two samples taken at least 6 months apart. All patients followed up at our clinic had been informed and agreed to the collection of serum during every follow-up. Serum samples collected at every visit were stored at −20°C until tested. Serum HBV DNA and HBsAg levels were performed 3 years, 2 years, 1 year, and 6 months before HBsAg seroclearance and at time of HBsAg seroclearance (i.e., baseline). The number of stored serum available for these tests were 203, 190, 185, 136, and 203 at the time points of 3 years, 2 years, 1 year, 6 months, and baseline, respectively.
In addition, among the approximately 5,000 treatment-naïve CHB patients followed up at our clinic between May 2010 and May 2011, 203 treatment-naïve age- and sex-matched CHB patients were selected as controls. These patients had 3 years of stored sera preceding recruitment (taken with informed consent) and were HBsAg-positive and HBeAg-negative during these 3 years. Serum HBV DNA and HBsAg levels were measured at five time points: 3 years, 2 years, 1 year, and 6 months before recruitment and at date of recruitment (i.e., baseline). These control patients were matched with patients with HBsAg seroclearance at a ratio of 1:1 for age and sex at all time points. The number of stored serum available for these tests were 203, 189, 187, 190, and 197 at the time points of 3 years, 2 years, 1 year, 6 months, and baseline, respectively.
None of the patients from the two groups received any antiviral therapy during the entire follow-up period. This study was approved by the Institutional Review Board, the University of Hong Kong and West Cluster of Hospital Authority, Hong Kong.
Serologic markers, including serum HBsAg, HBeAg, anti-HBs, and antibody to hepatitis B e antigen (anti-HBe), were measured by Abbott Laboratories (Chicago, IL). Serum HBV DNA levels were measured using the Cobas Taqman assay (Roche Diagnostics, Branchburg, NJ), with a lower limit of detection of 20 IU/mL. Serum HBsAg levels were measured using the Elecsys HBsAg II assay (Roche Diagnostics, Gmbh, Mannheim, Germany),21 with a linear range of 0.05-52,000 IU/mL. Samples with HBsAg levels higher than 52,000 IU/mL were retested at a dilution of 1:100, according to the manufacturer's instructions.
One hundred randomly chosen patients with HBsAg seroclearance, followed by 100 age- and sex-matched controls, were chosen for the determination of HBV genotype using the INNO-LIPA HBV genotyping assay, which was performed according to the manufacturer's instructions (Innogenetics, Gent, Belgium).
All continuous values were expressed in median (range). For patients with undetectable serum HBV DNA or HBsAg, the results were taken as the lower limit of detection (20 and 0.05 IU/mL, respectively). The HBsAg (log IU/mL)/HBV DNA (log IU/mL) ratio, which reflects the percentage of subviral particles over virions, was measured. To compare the characteristics between the two patient groups, Mann-Whitney's U test or Kruskal-Wallis' test, when appropriate, was used for continuous variables with a skewed distribution; the chi-squared test was used for categorical variables. Correlation between serum HBsAg levels and other variables, because of the repeated observations noted per patient, was performed using Pearson's weighted correlation coefficient.22 The predictions of HBsAg seroclearance were first examined by the construction of corresponding receiver operating characteristic (ROC) curves, followed by the assessment of overall accuracy by areas under the curves (AUCs). Then, the optimal level of prediction was attained by Youden's index,23, 24 which is defined as the sensitivity plus the specificity minus 1. Statistical analyses were performed using SPSS software (version 18.0; SPSS, Inc., Chicago, IL) and R Project for Statistical Computing software (version 2.14.1; R Foundation, Vienna, Austria). A two-sided P value of <0.05 was considered statistically significant.
Baseline demographics of both groups at the time of recruitment are shown in Table 1. There were no significant differences noted in the distribution of age, gender, and liver biochemistry and genotype. For patients with HBsAg seroclearance, the median age of HBsAg seroclearance was 51.9 years (range, 16.6-82.4). At the time of this writing, 63 patients (31.0%) had developed anti-HBs. Patients with HBsAg seroclearance had significantly lower serum HBsAg, HBV DNA levels, and HBsAg/HBV DNA ratios at baseline (all P < 0.001), compared to controls.
Table 1. Baseline Clinical Demographics of the Two Patient Groups (Time Point 0 Years)
For patients with detectable HBsAg and HBV DNA levels, there was no correlation noted between these two markers for both patients achieving HBsAg seroclearance (r = 0.005; P = 0.941) and controls (r = −0.003; P = 0.973).
Median HBsAg levels over the 3-year study period are depicted in Fig. 1. Patients with HBsAg seroclearance had a significant decline in HBsAg levels (P < 0.001). HBsAg levels in patients with HBsAg seroclearance were significantly lower at all time points, compared to controls. In total, 74.4% of patients with HBsAg seroclearance had serum HBsAg <100 IU/mL 3 years before seroclearance, with the percentage of patients achieving HBsAg <100 IU/mL significantly increasing with time (P < 0.001). In the control group, serum HBsAg levels also decreased significantly, but more gradually (P = 0.006). Using the time point of 3 years as baseline, 135 (66.5%) controls showed variations in HBsAg levels of more than 50% during the entire study period.
Median rates of annual HBsAg level decline for the two patient groups are depicted in Table 2. When combining all time points, the median annual rates of HBsAg decline in patients with HBsAg seroclearance and controls were 0.751 log IU/mL/year (range, −2.678-3.356) and 0.083 log IU/mL/year (range, −3.936-2.896), respectively (P < 0.001). When compared with controls, a significantly larger proportion of patients with HBsAg seroclearance achieved more than 1 log reduction in HBsAg levels per year (all P < 0.001).
Table 2. Median Annual HBsAg Log Reductions Over 3-Year Study Period in Patients
Median HBsAg Reduction (Log IU/mL Per Year) (Range)
Percentage of Patients With >1 Log Reduction in HBsAg
Patients With HBsAg Seroclearance
Patients With HBsAg Seroclearance (%)
Time point is defined as the period before HBsAg seroclearance: 0 years indicates date of seroclearance.
Time point 3 to 2 years
Time point 2 to 1 year
Time point 1 to 0 years
Among patients with HBsAg seroclearance with genotype performed, there were no differences in median HBsAg levels at 3 years (genotype B, 26.8 IU/mL; genotype C, 48.1 IU/mL; P = 0.623) or in median annual log reduction of HBsAg (genotype B, 0.553 log IU/mL/year; genotype C, 0.686 log IU/mL/year; P = 0.310).
Patients with HBsAg seroclearance who subsequently developed anti-HBs (n = 63) had a higher median HBsAg level at 3 years, compared to those who were negative for anti-HBs (n = 140) (52.5 and 12.1 IU/mL, respectively; P = 0.002). However, it should be noted that the HBsAg levels at 3 years for both groups of patients were very low levels. Patients with the development of anti-HBs were significantly younger than those without anti-HBs (median age of HBsAg seroclearance: 47.5 and 53.4 years, respectively; P = 0.013), although there were no significant differences in gender (P = 0.837), genotype (P = 0.855), alanine aminotransferase (ALT) (P = 0.680), or HBV DNA at 3 years (P = 0.112) between the two groups. There was also no significant difference in annual HBsAg decline between the two groups (0.667 and 0.565 log IU/mL/year, respectively; P = 0.174).
HBV DNA Levels.
Median HBV DNA levels over the 3-year study period are depicted in Fig. 2. For patients with HBsAg seroclearance, there was a gradual, but significant, decline in median serum HBV DNA levels (P < 0.001). Serum HBV DNA levels in the control group remained similar (P = 0.414). Comparing the two groups, there was a significant difference in median HBV DNA levels over all time points (P < 0.001). Among patients with HBsAg seroclearance with genotype performed, patients with genotype C had a significantly lower HBV DNA at 3 years, when compared to patients with genotype B (29 and 252 IU/mL, respectively; P = 0.003).
Median rates of annual HBV DNA level decline for patients with detectable viremia (>20 IU/mL) are listed in Supporting Table 1. When combining all time points, the median annual rates of HBV DNA decline in patients with HBsAg seroclearance and controls were 0.543 (range, −2.078-3.646) and −0.023 log IU/mL/year (range, −5.618-4.771), respectively (P < 0.001).
In the control group, using time point 3Yr as baseline, 175 (86.2%) patients had variations in HBV DNA levels of more than 50% during the entire study period, significantly more than that of HBsAg levels (p < 0.001).
HBsAg/HBV DNA Ratio.
Median HBsAg/HBV DNA ratios for both patient groups from 3 years to baseline are depicted in Fig. 3. During the 3-year study, in patients with HBsAg seroclearance, median HBsAg/HBV DNA levels decreased from 0.527 (range, −0.733-3.661) at 3 years to −1 (range, −1 to −0.464) at HBsAg seroclearance (P < 0.001). Median HBsAg/HBV DNA ratios in the control group did not show any significant change over time (P = 0.125). The difference between the two patient groups was significant at all time points (P < 0.001).
Predictors of HBsAg Seroclearance.
ROC curves and AUC values of different parameters used to predict HBsAg seroclearance are depicted in Fig. 4. Among the five parameters compared, serum HBsAg levels achieved the best AUC (0.833), followed by log reduction of HBsAg (0.802), both better than HBV DNA and log reduction of HBV DNA (0.743 and 0.648, respectively). Youden's indices, sensitivities, and specificities of different levels of serum HBsAg and HBsAg log reductions are depicted in Table 3. The optimal HBsAg level to predict HBsAg seroclearance was HBsAg <200 IU/mL (Youden's index, 5.76; sensitivity, 84.2%; specificity, 73.4%), followed by HBsAg <100 IU/mL (Youden's index, 5.42; sensitivity, 74.9%; specificity, 79.3%) One hundred and seventy (83.7%) patients with HBsAg seroclearance had serum HBsAg <200 IU/mL, compared to 53 (26.1%) in the control group (P < 0.001). Median HBsAg log reductions among the two groups with HBsAg <200 IU/mL were 0.567 (range, −0.984-2.233) and 0.106 log IU/mL/year (range, −0.375-1.189), respectively (P < 0.001).
Table 3. Youden's Indices, Sensitivities, and Specificies of Different HBsAg Levels and Log Reductions in Predicting HBsAg Seroclearance
Number of Patients (%)
Patients With HBsAg Seroclearance
Time point 3 years selected.
Time points 3 to 2 years selected.
Time points 3 to 2 years selected; n = 33 and 150 for patients with HBsAg seroclearance and controls, respectively.
The optimal HBsAg annual log reduction to predict HBsAg seroclearance was 0.5 log (Youden's index, 5.15; sensitivity, 62.8%; specificity, 88.7%). One hundred and seven patients with HBsAg seroclearance (52.7%) achieved ≥0.5 log reduction from 3 to 2 years, significantly more than 17 (8.4%) patients in the control group (P < 0.001).
We further examined patients with serum HBsAg ≥200 IU/mL at 3 years (n = 33 and 150 for patients with HBsAg seroclearance and controls, respectively). In this subgroup of patients, the AUC for HBsAg log reduction was 0.867 (P < 0.001; 95% confidence interval [CI]: 0.778-0.956), with a 0.5-log reduction most optimal in predicting HBsAg seroclearance (Youden's index, 6.35; sensitivity, 74.1%; specificity, 89.4%). For patients with serum HBsAg <200 IU/mL (n = 170 and 53, respectively), the AUC for HBsAg log reduction was comparably lower at 0.796 (P < 0.001; 95% CI: 0.724-0.868).
We also examined whether the addition of HBV DNA into HBsAg levels could improve the AUC for predicting HBsAg seroclearance. We found that there was no increase in AUCs using different combinations of HBsAg and HBV DNA in terms of their absolute levels and reductions (data not shown). Analyzing HBsAg among patients with undetectable HBV DNA levels produced an AUC of only 0.648 (P = 0.013; 95% CI: 0.538-0.823). Among the subgroups of patients with HBsAg ≥200 IU/mL, HBV DNA log reduction also produced an AUC of only 0.735 (P < 0.001; 95% CI: 0.623-0.848).
Our current study demonstrated the kinetics of serum HBsAg and HBV DNA levels preceding HBsAg seroclearance in a large population of CHB patients with HBsAg seroclearance. To our knowledge, this is a study with the largest number of patients with HBsAg seroclearance to date (n = 203).
Our present study outlines the changes in HBsAg kinetics before spontaneous HBsAg seroclearance. The enrollment of age- and sex-matched controls would allow us to optimally delineate the differences in serologic and virologic kinetics between the two patient groups. With 3 years of serial data, we were able to show a marked difference in HBsAg levels between patients with HBsAg seroclearance and controls. In our study, the median HBsAg levels of controls were between 366 and 846 IU/mL at different time points, levels which were similar to those reported in other studies on serum HBsAg levels in HBeAg-negative CHB.13-15 The results of our control group also provide additional insight into the natural history of HBsAg levels in HBeAg-negative CHB. Serum HBsAg levels decreased gradually over time and appears to be a much more stable marker than HBV DNA levels, which are known for their fluctuating nature.25
Our study confirms that serum HBsAg measurements can be an important tool for physicians in weighing the chances of HBsAg seroclearance in the long term. Because serum HBsAg decreases gradually with time, lower levels of HBsAg would eventually lead to HBsAg seroclearance. By achieving the highest AUC in our study (0.833), the absolute HBsAg level offers the best predictive value of eventual HBsAg seroclearance. From our study, HBsAg <200 IU/mL is already optimal in predicting eventual HBsAg seroclearance (Youden's index, 5.76), although HBsAg <100 IU/mL also had good predictive value (Youden's index, 5.42). Whether the slightly inferior predictive value of HBsAg <100 IU/mL (n = 151 in patients with HBsAg seroclearance) to that of <200 IU/mL (n = 170) is a result of the statistical underpower for detection needs further clarification.
The second-best method in predicting HBsAg seroclearance would be using the annual log reduction in HBsAg (AUC, 0.803). Serum HBsAg reduction is especially useful in patients with serum HBsAg ≥200 IU/mL (AUC, 0.867), when compared to HBsAg <200 IU/mL (AUC, 0.796). Therefore, adapting an annual 0.5-log reduction of HBsAg levels to predict subsequent HBsAg seroclearance is recommended in patients with baseline HBsAg ≥200 IU/mL. In the control group, annual 1-log reductions in HBsAg levels were uncommon, accounting for less than 5% for all time points, in contrast to 20.7%-48.7% of 1-log reductions noted in patients eventually clearing HBsAg. Thus, our study provides evidence that serial serum HBsAg measurements can be useful in identifying CHB patients with good immune control and eventual HBsAg seroclearance. From our study, an annual HBsAg reduction of 0.5 log already offers the best predictive value of HBsAg seroclearance, for all patients and also for patients with serum HBsAg ≥200 IU/mL.
Serum HBV DNA levels and their reductions were less useful in predicting HBsAg seroclearance. In addition, there was poor correlation between HBV DNA and HBsAg in both patient groups. It has been previously suggested that the relationship between viral replication and HBsAg production breaks down in HBeAg-negative CHB, probably because viral integration, a nonessential event in the life cycle of HBV, produces HBsAg in the absence of viral replication.12, 26 Also, HBsAg is produced in excess by replicating viruses. The significant decrease in HBsAg/HBV DNA ratios over time among patients with HBsAg seroclearance in our study implies a decrease in subviral particle production occurring in the absence of marked changes in viral replication before HBsAg seroclearance. Unlike the identification of inactive carriers,20 the combined analysis of HBV DNA and HBsAg levels in our study did not yield favorable AUCs and is less useful in predicting HBsAg seroclearance.
Among patients achieving HBsAg seroclearance, patients developing anti-HBs (n = 63) had a significantly younger age of HBsAg seroclearance (P = 0.013). Studies of vaccinated subjects show a higher rate of anti-HBs development among younger patients,27 although the implication of anti-HBs development in CHB patients with HBsAg seroclearance, and its association with HBsAg kinetics, needs further investigation. One study has shown the development of anti-HBs to have no influence over the subsequent occurrence of HCC.4
Besides providing important clinical data on serologic and virologic parameters before spontaneous HBsAg seroclearance, our present study also offers a reference for future studies investigating the usefulness of serum HBsAg measurements of CHB patients undergoing antiviral therapy. Serum HBsAg levels have already been shown to be useful in predicting favorable outcomes in Peg-IFN therapy.28, 29 In contrast, patients commenced on nucleoside analog therapy do not show significant decline in serum HBsAg up to 2 years,30 although a 0.5-log reduction in HBsAg is also predictive of subsequent HBsAg seroclearance.31 The achievement of low HBsAg levels or a strong reduction in HBsAg should thus be investigated in the future for suitability as treatment endpoints. Future studies should also consider matching baseline HBsAg and HBV DNA levels for a more detailed comparison of HBsAg kinetics.
A limitation of our study is that our patient population might not be totally representative of all treatment-naïve CHB populations, with no HBeAg-positive patients at initial presentation included. Although HBsAg loss is possible shortly after HBeAg seroconversion,16 the average age of HBeAg seroconversion in our population is 35 years32 and the average age of HBsAg seroclearance is 50 years4; hence, the proportion of patients with HBsAg seroclearance within 3 years of HBeAg seroconversion is likely to be small. Therefore, the validity of our study results, when applied to spontaneous HBsAg seroclearance, should not be affected by the absence of HBeAg-positive patients. In addition, HBV genotyping was not performed in all patients. Nevertheless, the lack of significant difference in genotype distribution among the two patient groups is in line with findings suggesting HBV genotypes as not being a key factor in determining HBsAg seroclearance.16 Further studies on this aspect are needed.
In conclusion, in CHB patients with spontaneous HBsAg seroclearance, low levels of serum HBsAg could be detected up to 3 years before HBsAg seroclearance and were more predictive of HBsAg seroclearance than low levels of serum HBV DNA. Serum HBsAg levels <200 IU/mL already offered a good prediction of eventual HBsAg seroclearance in 3 years. In patients with serum HBsAg ≥200 IU/mL, an annual 0.5-log reduction in serum HBsAg increases the prediction of HBsAg seroclearance. Both absolute and serial measurements of serum HBsAg would offer valuable clinical data in determining the probability of long-term seroclearance. These may also serve as good indicators for the consideration of treatment duration and cessation for CHB.