The template of hepatitis B virus transcription, the covalently closed circular DNA (cccDNA), plays a key role in the life cycle of the virus and permits the persistence of infection. It has been suggested that hepatitis B surface antigen (HBsAg) quantification reflects the concentration of cccDNA in the liver. In hepatitis B e antigen (HBeAg) positive chronic hepatitis B, HBsAg levels are higherduring the immune tolerance phase than during the immune clearance phase. During the natural history of chronic hepatitis B, serum HBsAg declines progressively from the immune-tolerant to the low replicative phase. In HBeAg negative patients, the combination of low hepatitis B virus (HBV) DNA (<2000 IU/ml) and low HBsAg levels (<1000 IU/ml) can predict inactive carrier status, low risk of hepatocellular carcinoma, and the probability of HBsAg loss. HBsAg in combination with HBV DNA predicts the outcome of Peg-Interferon therapy: An absence of decline at week 12 is a good predictor of non-response and to stop therapy. Any decline at week 24, suggests that therapy should be continued to 48 weeks. Although the decrease in HBsAg decline slow with nucleos(t)ide analogue therapy, a rapid decline can predict future HBsAg seroclearance. A level <100 IU/ml during six consecutive months could be a marker of a sustained response after treatment cessation.
The hepatitis B surface antigen (HBsAg) was identified more than 40 years ago . Detection of HBsAg in serum is still the hallmark of HBV infection and remains the cornerstone for the diagnosis of chronic hepatitis B (CHB). Over the years, HBsAg has proven to be a steady marker that can be used to predict clinical outcomes.
HBsAg synthesis during the HBV viral life cycle is complex and usually occurs in the endoplasmic reticulum [2, 3]. Upon entry into the hepatocyte the virion sheds its protein coat and its genome is transported into the nucleus where it resides as stable fully double stranded covalently closed circular HBV DNA (cccDNA) and acts as a template for the transcription of the viral gene . HBsAg is translated from mRNA with the transcriptional template-active cccDNA, which is the reflection of the number of infected hepatocytes. The clinical relevance of HBsAg levels is inferred from the relationship of this marker to the intrahepatic amount of cccDNA. There is a correlation between serum HBsAg concentrations and the intrahepatic levels of cccDNA, with the highest levels occurring in HBeAg positive hepatitis B and the lowest in patients with resolved hepatitis [5-7]. Through this association, the amount of circulating HBsAg is thought to indirectly measure the control of infection by the immunological response independent from the antiviral response, which can be assessed by measuring HBV DNA levels in serum. Serum HBsAg can be considered to be a surrogate marker of the number of infected cells.
In the early 1990s HBsAg quantification was considered to be a promising, simple and inexpensive method to monitor viral replication in CHB patients who were receiving IFN treatment . More recently, studies have reported that the management of CHB can be optimized in daily practice by the HBsAg quantification [9, 10]. Indeed, the availability of standardized commercial assays has renewed interest in quantitative serum HBsAg as a biomarker and to stratify the risk of disease progression and predict treatment response mainly in patients receiving pegylated interferon (PEG-IFN) therapy [11, 12]. In clinical practice, HBsAg quantification cannot replace viral load measurement. The combination of both has been shown to be very useful for monitoring the natural history of the disease and treatment outcome . The aim of this article is to describe the use of HBsAg quantification in clinical practice especially for the evaluation of disease severity and response to treatment.
Several automated assays have been developed for the quantitative measurement of HBsAg. The most widely used assays are the Architect HBsAg QT assay (Abbott Diagnostics, Abbott Park, IL, USA) and the Elecsys HBsAg II assay (Roche Diagnostics, Indianapolis, IN, USA). . The range of the Architects Assay is 0.05–250 IU/ml and a manual or automatic 1/500 dilution is needed for higher levels. The Elecsys Assay has an automatic on board dilution with range of quantification from 0.05 to 52 000 IU/ml. The assays express results in international units per millilitre (IU/ml), based on the WHO reference standard; 1 IU/ml is equivalent to 1–10 ng/ml of HBsAg or 5 × 107 virions . Results with both assays are highly correlated and in close agreement for each HBV genotype. In this review, we will discuss the use of this new biomarker to optimize the management of CHB patients.
Natural history of the hepatitis B virus infection
HBeAg positive chronic hepatitis B
In the first 20–30 years of life, HBV infection is characterized by an immune tolerance phase in which patients usually have positive HBeAg, very high HBV DNA, normal alanine aminotransferase (ALT) levels and minimal histological damage. This is followed by an immune clearance phase, which may lead to HBeAg seroconversion . Two studies, one European and one Asian, reported that HBsAg levels varied during the natural history of the disease [7, 16]. They showed that the HBsAg titre was higher during the immune tolerance phase than during the immune clearance phase as well as being higher in HBeAg(+) than in HBeAg(−) patients. The European study  reported an HBsAg level of 4.96 vs 4.37 log10 IU/ml in patients during the immune tolerance phase vs the immune clearance phase, respectively, while the Asian study  reported an HBsAg level of 4.53 vs 4.03 log10 IU/ml in patients during the immune tolerance phase vs immune clearance phase respectively (Fig. 1) . The observation of lower titres in Asian patients may be related to different HBsAg expression by different genotypes .
HBeAg negative chronic hepatitis B
Identification of inactive carriers. After HBeAg seroconversion, the clinical spectrum of HBeAg(−) CHB ranges from ‘inactive carrier’ status to aggressive HBeAg(−) CHB. The latter is generally differentiated from ‘inactive carriers’ by serial serum ALT and HBV DNA level determinations. However, in 45–65% of cases ALT activity can fluctuate with long periods of normal ALT levels resulting in misclassification. Inactive carriers have no or mild histological lesions in the liver with an excellent prognosis for survival and a low incidence of cirrhosis and HCC, while patients with HBeAg(−) CHB with fluctuating activity have a more severe disease progression with frequent cirrhosis [19, 20]. It is important to differentiate true inactive carriers from patients in remission as the latter are at risk of progression to cirrhosis and therefore candidates for the treatment. According to National Institute of Health and European Association for the Study of the Liver guidelines [21, 22] the differentiation between inactive and active phases of HBeAg(−) CHB is based on an HBV DNA cut-off of 2000 IU/ml. This cut-off has led to several controversial studies [9, 23-26].
However, the major concern is the identification of patients with high risk of reactivation who are good candidates for therapy. In a recent study, we showed that HBsAg ≥1000 IU/ml and HBV DNA ≥200 IU/ml could be used to identify patients with high risk of reactivation with a negative predictive value of 96 and 92% sensitivity .
Prediction of HBsAg seroconversion. Loss of HBsAg is the most reliable indicator to measure a cure of hepatitis B. In HBeAg(+) patients, it has been reported that after HBeAg seroconversion, a rapid decrease in HBV DNA is the best predictor of HBsAg loss . More recently this same group has shown that the predictive value for HBsAg loss at 5 and 10 years was more accurate when HBsAg and HBV viral load were combined . The study by Chen et al.  reported that HBsAg quantification every 2 years in HBeAg(−) patients with persistently normal ALT measures the magnitude of HBsAg decline. A decline ≥1 log10 IU/ml during this period, or a single measurement below 200 IU/ml are the best predictors of HBsAg loss [positive predictive value (PPV) 100%]. Finally, a threshold of HBsAg decline ≥0.3 log10 IU/ml/year identifies patients with high probability of HBsAg loss with a negative predictive value (NPV) of 95% and a PPV of 85% .
HBsAg and liver histology. Two studies have reported a negative correlation between the stage of fibrosis and HBsAg level in patients with HBeAg(+) chronic hepatitis (Fig. 2) [30, 31]. This correlation was not observed in patients with HBeAg(−) chronic hepatitis (Fig. 2) or with HBV DNA. In genotypes B and C patients (frequently encountered in Asia) an HBsAg 3.85 log10 IU/ml threshold differentiates patients with minimal fibrosis (≤F1) from those with moderate or severe fibrosis (≥F2), with a 91% NPV .
Prediction of CHC. The REVEAL study  indicated that HBV DNA is the major driver of disease progression in Asian populations. Patients with a viral load ≥2000 IU/ml have an increased risk of HCC, while patients with HBV DNA <2000 IU/ml are at low-risk.
The same authors retrospectively studied the relationship between HBsAg and the risk of HCC in the same population. They reported that like HBV DNA, the cumulative risk for HCC is also highly correlated with HBsAg levels. The 20 year cumulative risk of HCC was 1.4, 4.5 and 9.2% in patients with serum HBsAg levels <100, 100–999 and ≥1000 IU/ml respectively . Similarly, in another Asian group the ERADICATE study  confirmed that the incidence of HCC in HBe(−) patients was strongly associated with HBsAg levels but not with viral load. They reported that in patients with HBV DNA <2000 IU/ml, an HBsAg level below 1000 IU/ml was associated with a 2% incidence of HCC at 20 years which increased to 8% with an HBsAg level above 1000 IU/ml. This association between HBsAg and the development of HCC is not observed if HBV DNA is above 2000 IU/ml. This association was not observed in HBe(+) patients. The authors propose an algorithm to categorize the risk of disease progression and corresponding management in Asian HBeAg(−) patients (Fig. 3) .
However, all these results must be generalized with caution since most of the studies are Asian studies in patients mainly infected with genotypes Band C [32-35].
HBsAg level and therapy
The goal of antiviral therapy in chronic hepatitis B is to clear cccDNA from the liver. The therapeutic endpoints include sustained suppression of viral replication, normalization of serum ALT and histological improvement in particular the regression of fibrosis, with the goal of HBsAg seroconversion. Patients and clinicians have the choice of several therapies. Two treatment strategies can be used; either immune-mediated control of HBV infection obtained with 48 weeks of PEG-IFN therapy, or viral control obtained with long-term nucleoti(si)des (NAs) that are potent HBV inhibitors.
One of the major advantages of PEG-IFN is a finite course of therapy that can potentially lead to sustained disease remission in the subsequent decades. Oral antiviral NAs are very well-tolerated, require minimal clinical and laboratory monitoring and in most cases must be continued for years.
The mechanisms underlying HBsAg clearance during antiviral therapy are unknown. There are two strategies, either to obtained sustained off-therapy virological control with a finite course of PEG-IFN or definitive suppression of HBV replication with oral NAs.
PEG-IFN has limited direct antiviral efficiency but stimulates the induction of host immune response against HBV. Selection of patients with the highest probability of achieving a response to PEG-IFN is essential for optimal use and to help the clinician decide whether to start treatment with PEG-IFN. In HBe(+) patients a finite course of PEG-IFN results in a 24 week post treatment response (sustained HBeAg seroconversion) in approximately 25–30% of patients. The predictors of response are older age, female sex, higher serum ALT, lower HBV DNA levels and genotypes A or B. In HBe(−) patients a 24 week post treatment sustained response (HBV DNA <2000 IU/ml and normal ALT) occurs in 25% of patients. The predictors of response are younger age, female gender, higher serum ALT and lower HBV DNA levels. The efficacy of PEG-IFN has been shown in two large pivotal studies [36, 37]. HBeAg seroconversion was observed in 32% of HBeAg(+) patients  and HBV DNA <400 IU/ml in 19% of HBeAg(−) patients . The percentage of decline in HBsAg levels during PEG-IFN therapy may reflect its efficacy and provide useful information on the prediction of treatment outcome.
HBeAg positive patients
Current data [36-43] indicate that on treatment HBsAg quantification could help identify patients with a high probability of sustained virological response (SVR) and non-responders. In a study by Chan et al. , SVR rates were 62 and 11% in patients with HBsAg ≤300 IU/ml and in those with HBsAg >300 IU/ml at 24 weeks of therapy. The authors also showed that patients with a combined response of HBsAg ≤300 IU/ml and a decrease ≥1 log10 IU/ml at 24 weeks had higher SVR rates than those without the combined response (75% vs 15%), with a PPV and NPV of 75 and 85% respectively. Tangkijvanich et al.  report that baseline HBsAg levels were lower in patients with HBeAg seroconversion at the end of therapy than in non-responders. In the Neptune study [41-43], the highest HBeAg sero-conversion rates were observed in patients with HBsAg levels ≤1500 IU/ml at 12 or 24 weeks of therapy with a PPV of 57 or 54% and a NPV of 72 or 76% respectively (Fig. 4). These results confirm those of the phase III PEG-IFN alfa 2a trial . Sonneveld et al.  showed that HBsAg patients who received a combination of PEG-IFN plus lamivudine (LAM) had a more pronounced on-treatment decrease than those who received PEG-IFN alone, although the former relapsed. The absence of decline at week 12 had a NPV of 97% for SVR and no chance of HBsAg loss. The study by Lau reported a lower NPV (82%) than the Sonneveld study. The discrepancy between the two studies might be related to the different populations. Indeed, in the study by Sonneveld most of the patients were genotypes A and D while the Lau study included patients with genotypes B and C. It is interesting to note that certain studies have reported a greater decrease in HBsAg levels in patients with HBV genotypes A and B during PEG-IFN therapy [44-48]. These observations were confirmed in the study by Sonneveld et al.  suggesting that response-guided therapy could be used (Fig. 5).
HBeAg negative patients
The response rate to PEG-IFN therapy is low (<20%) in HBeAg(−) patients . These patients are difficult to monitor. Indeed, most of them achieve undetectable serum HBV DNA at the end of therapy, but relapse when treatment is stopped [50-52]. Therefore, on-treatment predictive factors of post-treatment outcomes may help define more appropriate treatment strategies. The most recent evidence suggests that HBsAg quantification is a worthwhile marker for monitoring these patients during PEG-IFN therapy [52-54].
Rijckborst et al.  reported that a combined decrease in HBsAg and HBV-DNA is the best predictor of SVR in patients receiving PEG-IFN. A lack of decrease in HBsAg and a serum HBV DNA decline of less than 2 log10 IU/ml have a NPV of 100% for SVR. Another study has shown that a serum HBsAg decrease ≥0.5 log IU/ml at week 12 had a high PPV of SVR (PPV 89%) and a lack of decrease of ≥1 log10 IU/ml at week 24 had a high NPV of non-SVR (NPV 92%) . More recently, Marcellin et al.  reported that an HBsAg decline ≥10% at week 12 therapy was significantly associated with a long-term post-treatment response (>5 years).
Studies have investigated end- and post-treatment HBsAg titres to predict SVR and loss of HBsAg during post-treatment follow-up. In a cohort of 127 HBeAg(−) genotype D patients who received 48 weeks of PEG-IFN treatment, Brunetto et al.  showed that an on-treatment ≥1 log10 IU/ml HBsAg decrease and an HBsAg level <10 IU/ml at the end of therapy were highly predictive of a SVR at 24 weeks post treatment and for HBsAg loss at 3 years post treatment. In this study, at the 3 year post-treatment follow-up an HBsAg loss was observed in 52% of the patients with an HBsAg <10 IU/ml at week 48 of treatment, compared with only 2% of the patients with an HBsAg >10 IU/ml at week 48. The authors did not find any association between serum HBV DNA response (HBV DNA<400 IU/ml) at the end of therapy and HBsAg loss. A French retrospective study also reported that most of the patients who lost HBsAg during post-treatment follow-up had a ≥1 log10 IU/ml decrease in HBsAg levels at the end of 48 weeks of IFN treatment followed by a steady decrease until seroclearance .
Week 12 stopping rule
Most studies report that the absence of a decrease in HBsAg or a HBV DNA decline of less than 2 log10 after 12–24 weeks of a 48 weeks course of PEG-IFN is associated with a NPV of 84–100% for SVR [38-43, 48, 50-56] (Table 1). Rijckborst et al.  confirmed these results in patients in the Parc study  and performed external validation in the phase III registration study  and the PegBLiver study , both of which included patients with HBV genotypes A to D. There was a NPV ≥95% for SVR in both validation studies in the absence of a decrease in HBsAg or a decline in HBV DNA of less than 2 log10 IU/ml at week 12 of treatment. The authors propose a response-guided treatment algorithm based on HBsAg kinetics at week 12 of treatment. Treatment could be discontinued or changed if non-responders were identified early (Fig. 6). Finally, in a systematic review of the articles and abstracts published in English between January 2004 and June 2012 an Italian study reported that the cost effectiveness of the management of chronic HBe(−) hepatitis B could be significantly improved by shifting to response-guided first line therapy with PEG-IFN followed by a switch to NAs in patients who meet the week-12 HBV DNA/HBsAg stopping rule (no decrease in serum HBs/HBV DNA decrease <2 log10) or in week-48 non-responders/relapsers . The main reports are summarized on Fig. 7.
Table 1. Negative predictive value for sustained virological response according to HBsAg level/decrease at week 12 and week 24 peginterferon therapy
The goal of treatment with NAs is to maintain HBV suppression. Although the mechanism of HBsAg decline is unclear during NA therapy it may reflect a better degree of host immune control against the virus or a decrease in the amount of cccDNA in the hepatocyte. All NAs are competitive inhibitors of the natural endogenous intracellular nucleotide. Their incorporation in nascent viral DNA results in premature chain termination by preventing the incorporation of the next nucleotide in the virals DNA strand, which is therefore very effective in suppressing HBV replication. Second generation NAs, tenofovir (TDF) and entecavir (ETV), are more effective in suppressing HBV and less likely to be associated with drug resistance than original NAs. The patterns of response observed with both molecules are generally similar although these agents have a different structure and inhibit different phases of HBV replication. Once it is internalized in the cell TDF (Nucleotide) inhibits reverse transcription of RNA to DNA. ETV (Nucleoside) inhibits DNA synthesis in the infected cell, thus reducing viral load. In addition they are orally administered and have a good safety profile and tolerance, wide applicability and proven histological improvement . NAs can be given to all patients while PEG-IFN has several absolute or relative contraindications . However, they must be taken indefinitely because withdrawal is generally associated with viral reactivation and HBsAg seroconversion is rarely reported .
Why quantify HBsAg during NAs therapy
Although HBV-DNA becomes rapidly undetectable, studies have clearly shown that the decline in HBsAg during NAs therapy is less pronounced than that during PEG-IFN treatment [60, 61]. The decline appears more marked in HBe(+) than in HBe(−) patients .
The study by Boglione et al.  reported a significant difference in the decline in HBsAg levels at 2 years of therapy depending on the NAs administered (Fig. 8). Several studies have reported that baseline HBsAg levels and on treatment HBsAg quantification are good predictors of end of treatment response and SVR [60-67]. Zoutendijk et al.  investigated HBsAg kinetics in patients who were successfully treated with long-term ETV or TDF. The authors used linear mixed regression analysis of individual HBsAg declines to estimate the duration of therapy required to achieve an HBsAg decline of 1 log10 IU/ml from baseline and HBsAg clearance. They showed that the median durations of therapy to achieve a 1 log10 IU/ml decrease were: 6.6 [1.7–18] years and 8 [0.5–15] years in HBeAg(+) and HBeAg(−) patients respectively. Median durations for HBsAg clearance were: 36 [10–93] years and 39 [1.3–90] years in HBeAg(+) and HBeAg(−) patients respectively.
In the two pivotal TDF studies in HBeAg(−)  and in HBeAg(+)  patients the authors confirm that HBsAg responses are greater in HBeAg(+) patients than in HBeAg(−) patients receiving TDF monotherapy. HBsAg loss was only observed in HBeAg(+) patients with a 24 week HBsAg decrease from baseline >2 log10 IU/ml, high baseline HBsAg levels and genotypes A or D. In a cohort of 162 HBeAg(+) patients with HBV DNA <60 IU/ml after 3 years of telbuvidine (LdT) therapy, Wursthorn et al.  showed that 25% of patients with a ≥1 log10 IU/ml HBsAg decrease after 1 year of therapy had an HBsAg loss compared with 1.4% of patients with a <1 log10 IU/ml HBsAg decrease. In the study by Reijnders et al.  in HBeAg(+) and (−) patients who received 48 weeks of either PEG-IFN or ETV, an HBsAg decline was only observed in HBeAg+ patients after 48 weeks of either PEG-IFN or ETV therapy. No decline in HBsAg levels was observed in HBeAg(−) patients receiving ETV. Studies including treatment-naive patients receiving NAs report that low baseline HBsAg levels and an early decline (24 weeks therapy) are good predictors of SVR [60-67, 71-73]. Furthermore, studies suggest that an HBsAg cut-off below 2–3 log10 IU/ml could be used for treatment discontinuation [74, 75]. In the study by Liang et al.  121 consecutive patients were prospectively recruited after treatment cessation (LAM, ADV, ETV). The relapse rate was higher (31%) in patients with treatment HBsAg levels ≤3 log10 IU/ml than in those with HBsAg levels ≤2 log10 IU/ml (9%). Similar results were reported by Cai et al.  in patients receiving 2 years of LdT therapy in which HBsAg levels ≤2 IU/ml at treatment cessation had a PPV and NPV for SVR of 100 and 93% respectively.
The recent study by Seto et al.  including patients from Hong Kong with a favourable virological response to more than 10 years of LAM treatment, reported that the best indicator of HBsAg seroclearance was a baseline HBsAg level <1000 IU/ml (NPV 98%; PPV 38%) and an HBsAg decrease ≥0.166 log10/IU/ml/year (NPV 98%; PPV 26%). Interestingly, the same authors reported that a decrease in HBsAg levels after 3 years of TDF was highly correlated with baseline levels. A greater decrease was observed in patients with baseline ≥3 log10 IU/ml than in those with baseline <3 log10 IU/ml. HBsAg levels remained stable in the latter  (Fig. 9).
The study by Gish et al.  performed in HBe positive genotype A to D patients receiving 48 weeks of TDF reported a greater end of treatment decrease in HBsAg (−2.04 log10 IU/ml) in patients with HBeAg seroconversion than in patients who remained HBeAg(+). Furthermore, they observed a greater decrease in genotypes A and D than in genotype B while genotype C patients remained unchanged.
Add on therapy
The addition of ETV or TDF to PEG-IFN therapy has been shown to increase on treatment HBsAg decline, HBeAg seroconvertion and SVR rate. [80-82]. In the study by Marcellin et al.  SVR and HBsAg loss in HBe(+)/(−) patients treated with 48 weeks PEG-IFN plus TDF was associated with low baseline HBsAg levels. The NPV for a SVR and HBsAg loss was 79 and 88%, respectively, in patients with baseline HBsAg levels above 1500 IU/ml. Similarly in HBe(+)/(−) patients who received a 48 week course of PEG-IFN plus ADV, HBsAg loss at 2 years post treatment was 11 and 17% respectively. Low baseline HBsAg levels were an independent predictor of HBsAg loss in HBe(−) but not in HBe(+)patients. An HBsAg cut-off of 20 IU/ml at week 24 post treatment was highly predictive of HBsAg loss at week 144 (PPV 89%; NPV 100%) . Ouzan et al.  reported that in HBeAg(−) patients with HBV DNA that was fully suppressed by long-term NAs treatment, the addition of PEG-INF for a maximum of 96 weeks based on HBsAg-titre monitoring led to a loss of HBsAg and cessation of NAs therapy in 6/10 patients.
HBsAg titres are now a subject of research to monitor the natural history of HBV infection and predict treatment outcome. Several recent studies on serum HBsAg monitoring have shown that HBsAg levels change during the natural course of chronic hepatitis B and during ongoing therapy. These results can be used to determine the best patient management strategy.
During the natural history of HBV, HBsAg quantification can be used to differentiate between inactive carriers (no need for treatment) and HBeAg(−) chronic hepatitis B patients, who are likely to reactivate (closer monitoring) and who can benefit from therapy. Early HBsAg monitoring could be used during PEG-IFN therapy to develop a response-guided algorithm to stop or switch therapy at week 12 in poor responders, to continue standard 48-week treatment in most patients with a favourable response, and to extend therapy for intermediate on-treatment responders to improve the chances of response (Table 1). The role of HBsAg monitoring during NA therapy must be clarified. The development of stopping rules should be determined in these lifelong therapies. Several studies suggest that baseline treatment and on-treatment HBsAg levels might help identify patients who can stop therapy with a reduced risk of reactivation.