To investigate potential predictors for treatment response to nucleos(t)ide analogues (NAs) in hepatitis B e antigen (HBeAg)-positive chronic hepatitis B (CHB) patients.

Seventy-six HBeAg-positive CHB patients received 96-wk NAs optimized therapy (lamivudine and adefovir dipivoxil) were studied retrospectively. Serum hepatitis B surface antigen, HBeAg, hepatitis B core antibody, hepatitis B virus (HBV) DNA and alanine aminotransferase levels were quantitatively measured before and during the treatment at 12 and 24 wk. Stepwise logistic regression analyses were performed to identify predictors for treatment response, and areas under the receiver operating characteristic curves (AUROC) of the independent predictors were calculated.

Forty-three CHB patients (56.6%) achieved virological response (VR: HBV DNA ≤ 300 copies/mL) and 15 patients (19.7%) developed HBeAg seroconversion (SC) after the 96-wk NAs treatment. The HBeAg level (OR = 0.45, P = 0.003) as well as its declined value (OR = 2.03, P = 0.024) at 24-wk independently predicted VR, with the AUROC of 0.788 and 0.736, respectively. The combination of HBeAg titer < 1.3 lg PEIU/mL and its decreased value > 1.6 lg PEIU/mL at 24-wk predicted VR with a sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) of 85%, 100%, 100% and 83%, respectively, and the AUROC increased to 0.923. The HBeAg level (OR = 0.37, P = 0.013) as well as its declined value (OR = 2.02, P = 0.012) at 24-wk also independently predicted HBeAg SC, with the AUROC of 0.828 and 0.814, respectively. The HBeAg titer < -0.5 lg PEIU/mL combined with its declined value > 2.2 lg PEIU/mL at 24-wk predicted HBeAg SC with a sensitivity, specificity, PPV, NPV of 88%, 98%, 88% and 98%, respectively, and the AUROC reached 0.928.

The combination of HBeAg level and its declined value at 24-wk may be used as a reference parameter to optimize NAs therapy.

Hepatitis B virus (HBV) infection is a global public health problem and an estimated 240 million persons are chronically infected worldwide, among which 20%-30% will develop cirrhosis and hepatocellular carcinoma (HCC)-the major complications of chronic hepatitis B (CHB)[1]. Antiviral treatment has been proved to be an effective and potent way to reverse the process of liver fibrosis or cirrhosis and decrease the incidence rate of liver complications[2]. However, due to the persistence of HBV covalently closed circular DNA (cccDNA) in the nucleus of infected hepatocytes, HBV cannot be completely eradicated by current antiviral drugs, and a long-term treatment are necessary for most patients. It is now clear that sustained viral suppression and hepatitis B e antigen (HBeAg) seroconversion (SC) are two important markers of treatment response for CHB patients receiving antiviral therapy, which is usually associated with a good long-term outcome[3]. Generally, after a 1-year course of the current available nucleos(t)ide analogues (NAs) or peginterferon (Peg-IFN) therapy, 7%-76% of patients achieved undetectable serum HBV DNA and 16%-32% developed HBeAg SC for patients with HBeAg-positive CHB[3]. Therefore, it is crucial to identify pre-treatment and early on-treatment biomarkers that can effectively predict long-term treatment response and use these biomarkers to choose appropriate antiviral drugs and treatment regimens to optimize therapy and improve efficacy.

Serum HBV DNA is the most widely used virological marker in the management of CHB patients[2,3]. A study from Zeuzem et al[4] reported that both baseline ALT ≥ 2 × upper limit of normal (ULN) (OR = 2.47, P = 0.0012) and non-detectable serum HBV DNA at treatment week 24 (OR = 2.61, P < 0.001) were associated with HBeAg SC after 2-year telbivudine (LdT) treatment, and among patients with non-detectable serum HBV DNA at 24-wk as well as favorable pretreatment characteristics [alanine aminotransferase (ALT) ≥ 2 × ULN and HBV DNA < 9 lg copies/mL], 52% obtained HBeAg SC at 2-year of therapy[4]. However, the detection of serum HBV DNA is costly and may not always objectively serve as a reliable indicator of sustained response to antiviral therapy[5]. Unlike HBV DNA, serum hepatitis B surface antigen (HBsAg), HBeAg and hepatitis B core antibody (anti-HBc) are classical serological markers for HBV infection and are used in clinical diagnosis routinely. The level of HBsAg was identified as an outcome predictor for Peg-IFN therapy among HBeAg-positive CHB patients[6]; however, its predictive value in NAs treatment was inconsistent based on the reported data[7,8]. Serum HBeAg level was proposed to be a better outcome predictor for NAs treatment according to recent studies[8-12]. However, most of studies applied the semi-quantitative measurement of HBeAg, and some had a limited sample size or a short period of follow-up. Thus the predictive value of the quantitative HBeAg level needs to be further evaluated. In addition, benefiting from a newly developed double-sandwich anti-HBc immunoassay, anti-HBc quantification was identified as a novel biomarker for predicting treatment response[13]. Nevertheless, very few studies have systematically evaluated the predictive power of these biomarkers for NAs treatment response.

In the current study, HBeAg-positive CHB patients received 96-wk NAs optimized therapy [lamivudine (LAM) and adefovir dipivoxil (ADV)] were retrospectively investigated. Serum HBsAg, HBeAg, anti-HBc, HBV DNA and ALT were quantitatively tested, and the baseline as well as early on-treatment levels of these parameters were analyzed using logistic regression model to assess their functions in predicting 96-wk virological response (VR) and HBeAg SC.

Achieving a long-term suppression of serum HBV DNA through antiviral therapy is one of important targets of treatment for CHB patients. Several studies tried to investigate the value of an early on-treatment change of HBV markers such as HBeAg in predicting VR to NAs therapy. A study presented that the HBeAg titer decreased by 1 lg PEIU/mL at 12-wk predicted VR (HBV DNA < 20 IU/mL) after 48-wk entecavir (ETV) therapy with a sensitivity, specificity, PPV, NPV of 67.6%, 87.9%, 86.2% and 70.7%, respectively[10]. However, the predictive value of HBeAg titer and its declined value at 12-wk for a long-term treatment response (≥ 2 years) was not reported. In this study, we found that HBeAg titer < 0.8 lg PEIU/mL combined with its declined value > 0.84 lg PEIU/mL at 12-wk predicted VR after 96-wk LAM and ADV optimized therapy with a sensitivity, specificity, PPV, NPV of 71%, 91%, 91% and 71%, respectively, and accompanied with an AUROC of 0.812. In addition, although some studies indicated that the on-treatment HBeAg level or declined value at 24-wk was a predictor for NAs treatment response[9,12], the sample size was small or HBeAg was detected using a semi-quantitative method which cannot accurately quantify HBeAg level, thus limited the validity of the prediction. For example, Zhang et al[12] detected serum HBeAg semi-quantitatively and found that the on-treatment declined value of HBeAg (> 65%) at 24-wk was the best predictor for treatment response (HBeAg seroconversion and accompanied by undetectable serum HBV DNA) after 96-wk ETV therapy, and the PPV, NPV, AUROC was 83.3%, 93.6% and 0.885, respectively. However, the finding may have difficulty in applying clinical practice since the absent of accurate quantitative HBeAg levels. In our study, the HBeAg titer (OR = 0.45, P = 0.003) and its declined value (OR = 2.03, P = 0.024) at 24-wk were found to predict 96-wk VR with the AUROC of 0.788 and 0.736, respectively. Moreover, the combination of HBeAg titer < 1.3 lg PEIU/mL and its decrease > 1.6 lg PEIU/mL at 24-wk predicted 96-wk VR with a sensitivity, specificity, PPV, NPV of 85%, 100%, 100% and 83%, respectively, and the AUROC increased to 0.923, which had a better predictive value than the semi-quantitative method reported by Zhang et al[12]. With respect to the cost of semi-quantitative and quantitative tests of HBeAg, it is comparable between them. Twenty-two patients in our cohort matched the combination standard and they all achieved 96-wk VR. The results suggested that if patients reached HBeAg titer < 1.3 lg PEIU/mL and declined > 1.6 lg PEIU/mL at 24-wk, there will be a better viral suppression during the continued NAs therapy.

In addition to achieve long-term suppression in serum HBV DNA, HBeAg SC is another important indicator to evaluate the efficacy of antiviral therapy in HBeAg-positive CHB patients. In our study, the declined value of HBeAg at 12-wk (OR = 2.47, P = 0.001) independently predicted 96-wk HBeAg SC with an AUROC of 0.767, and for the predictive value of HBeAg declined value at 24-wk, an AUROC increased to 0.814. Lee et al[8] found that the decline of HBeAg at month 6 was a strongest predictor for HBeAg SC after 2 years of ETV treatment with an AUROC of 0.820 (P = 0.004), which was similar to our result (AUROC = 0.814). However, Lee’s study used the declined value of HBeAg alone to predict treatment response and did not consider combining other indicators. Our data showed that both HBeAg titer as well as its declined value at 24-wk were independent predictors for 96-wk HBeAg SC, and it strongly predicted HBeAg SC with an AUROC of 0.928 if combining HBeAg titer < -0.5 lg PEIU/mL and declined value > 2.2 lg PEIU/mL at 24-wk. A study from Shin et al[11] revealed that HBeAg titer < 0.62 lg PEIU/mL after 48 wk of ETV therapy was a strongest predictor for HBeAg SC at year 3 with an AUROC of 0.86 (P < 0.001), which was inferior to the combined prediction validity of HBeAg level and declined value at 24-wk in our study (AUROC = 0.928). Our study presented that 88% (7/8) of patients with HBeAg titer < -0.5 lg PEIU/mL and declined value > 2.2 lg PEIU/mL at 24-wk achieved 96-wk HBeAg SC, whereas only 2% (1/51) of patients without meting above standards obtained HBeAg SC, thus got a NPV of 98%. Among the 51 patients with unfavorable 24-wk HBeAg titer, 46 patients were added on ADV at 24-wk due to serum HBV DNA > 300 copies/mL. The results suggested that patients with HBeAg titer > -0.5 lg PEIU/mL as well as declined value < 2.2 lg PEIU/mL after 24 wk of LAM therapy will rarely achieve HBeAg SC during the following NAs treatment, even adding on ADV still cannot improve the efficacy. So the patients with unfavorable 24-wk HBeAg titer should be considered to switch to other drugs or use other regimens for a better treatment outcome.

Results of logistic regression analysis in the current study showed that no baseline parameters were associated with 96-wk HBeAg SC, which was consistent with Lee’s report[8], who did not find correlations between baseline level of HBeAg (or HBsAg, HBV DNA) and HBeAg SC after 2 years of ETV therapy either. However, other studies showed that pre-treatment serum HBeAg was associated with HBeAg SC during ETV treatment[10,12]. The inconsistent results may due to that treatment period was relatively short or HBeAg was detected semi-quantitatively in the latter’s studies[10,12]. Several recent studies revealed that B lymphocytes played a key role in the regulation of host immune responses to HBV, while anti-HBc was produced and secreted by hepatitis B core antigen-specific B lymphocytes, therefore the serum anti-HBc level may be a surrogate marker for the host immune response to HBV[16,17]. Fan et al[13] demonstrated that the baseline level of anti-HBc independently predicted HBeAg SC after 2 years of LdT and ADV optimized therapy (OR = 1.99, P = 0.001). However, our results presented that baseline anti-HBc was not associated with HBeAg SC or VR after 96-wk LAM and ADV optimized therapy (HBeAg SC, OR = 1.60, P = 0.459; VR, OR = 1.68, P = 0.308). The discrepancy may due to the little difference of baseline anti-HBc level between patients with and without HBeAg SC (0.10 lg IU/mL) in our study, while the difference in baseline anti-HBc was 0.24 lg IU/mL in Fan’s study. Besides, previous studies identified that the elevation of ALT was contributed to T lymphocyte mediated hepatolysis occurred in CHB patients, therefore baseline ALT level may reflect T lymphocyte immune response to HBV which is related to the outcome after antiviral treatment[18]. Zeuzem et al[4] reported that the pre-treatment ALT level was associated with VR and baseline ALT ≥ 2 × ULN could independently predict non-detectable serum HBV DNA after 2 years of LdT treatment (OR = 2.00, P = 0.0071). In agreement with these findings, we also found that baseline ALT ≥ 5 × ULN independently predicted 96-wk VR (OR = 7.09, P < 0.001) with the AUROC of 0.700 (P = 0.003, data not shown), which suggested that patients with a higher pre-treatment ALT level maybe situated in a better immune status and will have an active virological response to NAs treatment.

All the parameters (HBsAg, HBeAg, anti-HBc and HBV DNA) presented continuous descent during the 96-wk treatment. Serum HBV DNA dropped significantly from baseline 8.16 lg copies/mL to 96-wk 2.85 lg copies/mL (P < 0.001) and 56.6% (43/76) patients obtained HBV DNA ≤ 300 copies/mL in our cohort, showing a similar data to the GLOBE study[19], in which 55.6% HBeAg-positive patients achieved HBV DNA < 300 copies/mL with serum HBV DNA declined by 6.1 lg copies/mL after 2-year LdT treatment. The antiviral effect of NAs agents was reducing viral replication by the inhibition of HBV DNA polymerase, but having limited impacts on the level of HBsAg[20]. In our study, HBsAg decreased slowly by 0.59 lg IU/mL after 96-wk therapy. Heathcote et al[21] also reported that the mean HBsAg level in HBeAg-positive CHB patients decreased by 0.66 lg IU/mL after 2 years of tenofovir disoproxil (TDF) treatment, which was similar to our data. In addition, anti-HBc level decreased from baseline 4.77 lg IU/mL to 96-wk 3.45 lg IU/mL in our cohort, with an average decrease of 1.32 lg IU/mL. In Fan’s report[13], 1.26 lg IU/mL of an average decrease of anti-HBc level was observed in HBeAg-positive CHB patients after 104-wk LdT and ADV optimized therapy (baseline 4.20 lg IU/mL to 104-wk 2.94 lg IU/mL). The average declines in anti-HBc levels were comparable between the two studies, although the baseline anti-HBc level was relatively higher in our study. Concerning the dynamic change of HBeAg, Shin et al[11] reported that HBeAg level reduced from baseline 2.23 lg PEIU/mL to 0.96 lg PEIU/mL after 96-wk ETV therapy and 17.1% (14/82) achieved HBeAg SC. In the present study, HBeAg level decreased from baseline 2.24 lg PEIU/mL to 96-wk 0.20 lg PEIU/mL and 19.7% (15/76) obtained HBeAg SC, which presented comparable HBeAg reduction and HBeAg SC rate as compared to Shin’s study.

Comparing dynamic changes in HBeAg level between patients with and those without 96-wk VR (or HBeAg SC), our results showed that baseline HBeAg levels were comparable between two groups. Further, HBeAg levels in patients obtained 96-wk VR (or HBeAg SC) were significantly lower than those without VR (or HBeAg SC) from 12-wk to the end of follow-up (P < 0.05, as shown in Figures 1B and 2B). With respect to dynamic changes of HBV DNA levels, the same trend of decline was observed. The patients achieved 96-wk VR had a significant lower HBV DNA level at every follow-up time-point except baseline when compared with those without VR (P < 0.05), while the significant differences in HBV DNA levels between patients with and without HBeAg SC were only observed at week 24, 72 and 96 (Figures 1D and 2D). These results indicated that HBeAg may have a better predictive value than HBV DNA for treatment response in HBeAg-positive CHB. At the same time, our finding showed that both the HBV DNA level as well as its declined value at 24-wk (or 12-wk) were significantly associated with 96-wk VR and HBeAg SC when performed univariate analysis (P < 0.05). However, HBV DNA levels and its declines would not associate with 96-wk VR and HBeAg SC anymore if all parameters including HBeAg levels and its declines were enrolled in multivariate analysis. Other studies also pointed out that HBeAg levels as well as its declines may maintain a better predictive value than HBV DNA to predict NAs treatment response in HBeAg-positive CHB patients[10,11]. HBeAg is generated by transcription and translation of HBV cccDNA, and previous studies had reported that serum HBeAg level was significantly correlated with intrahepatic HBV cccDNA (r = 0.507, P = 0.010)[22]. Thus, the decline in serum HBeAg level may reflect the reduction of HBV cccDNA and represented a good treatment outcome. In addition, viral persistence and the development of CHB was associated with viral manipulation and evasion of the host’s immune system, while HBeAg has been reported to attenuate the host immune response to the nucleocapsid protein and down-regulate the innate and adaptive immune responses[23,24]. Therefore, the decline in HBeAg level might weaken this effect and thus the patients may present a better immune control for HBV infection.

There were some limitations in our study. Firstly, LAM and ADV used in the cohort are no longer the first-line antiviral drugs. However, NAs drugs have a similar antiviral mechanism, and HBeAg SC rates in CHB patients are comparable between ETV (or TDF) therapy and an optimized therapy with LAM and ADV. Hence, we propose that the combination parameter of HBeAg level and its declined value at 24-wk might be used as a reference parameter to predict the efficacy of ETV or TDF treatment. Secondly, treatment endpoint evaluated in our study was on-treatment response. For virological response, off-treatment response may be more important than on-treatment response.

To our knowledge, this is the first report that identified the combination of on-treatment quantitative HBeAg level and its decline as the predictor of positive response to long term NAs therapy among HBeAg-positive CHB patients. In particular, the combination of HBeAg titer and its decline at 24-wk strongly predicted 96-wk VR and HBeAg SC with the AUROC of 0.923 and 0.928, respectively. This combination predictor was identified through a retrospective investigation based on the cohort of HBeAg-positive CHB patients received LAM and ADV optimized therapy. We will conduct a prospective study to evaluate and confirm the predictive validity of the combination parameter in the cohort of ETV or TDF therapy in the future, and anticipate that the combination of on-treatment HBeAg level and its declined value could serve as a reference parameter to guide NAs therapy.

The authors are grateful to Dr. Shumei Yun, from Missouri Department of Health and Senior Services, Jefferson City, MO, United States for proofreading and editing the manuscript.