Abstract
Background
Methods
Results
Conclusions
Introduction
Chronic infection with hepatitis B (HBV) affects more than 248 million people globally and may lead to liver cirrhosis, liver failure, hepatocellular carcinoma and death [1]. In spite of preventative vaccination, HBV is one of the few infectious diseases where there has been an increase in related morbidity and mortality over the last 20 years, and viral hepatitis is now the 7th leading cause of death worldwide [2]. Current treatment options for chronic hepatitis B include the nucleoside/ nucleotide analogues and pegylated interferon-α (PEG-IFN). In contrast to nucleoside/nucleotide analogues, which directly inhibit viral polymerase, interferon-α (IFN-α) has both direct antiviral and immunomodulatory effects. For instance, IFN-α enhances the antiviral responses via induction of IFN-stimulated genes, increased cell-mediated antiviral activity and degradation of cccDNA via activation of APOBEC3 family members [3,4]. Although PEG-IFN therapy is associated with off-treatment sustained responses and hepatitis B surface antigen (HBsAg) loss, its use in current clinical practice is limited because of low overall response rates and significant side effects [5]. PEG-IFN therapy is, however, expected to experience a revival in combination with novel antiviral agents, like capsid assembly modulators and nucleic acid polymers, that are currently in early phase clinical trials [6,7]. In addition, new immune modulators that exploit the IFN pathways but have fewer side effects are being developed [8]. Prediction of response to immune modulating therapy therefore remains important.
Hepatitis B core-related antigen (HBcrAg) is a new serum biomarker comprising hepatitis B e antigen (HBeAg), HBcAg and precore protein p22, all of which are strongly correlated with covalently closed circular DNA (cccDNA) levels and intrahepatic replicational activity and may in turn be associated with response to PEG-IFN therapy [9–14]. Serum HBcrAg levels may therefore be useful for monitoring treatment efficacy, but whether it offers advantages over HBsAg monitoring in Caucasian patients is unknown.
Methods
Patient samples
In this post hoc analysis of an investigator-initiated multicentre randomized trial [15], HBcrAg and HBsAg levels [16] were measured at baseline, on-treatment (weeks 4, 8, 12, 24), end-of-treatment (week 52) and 6 months post-treatment (week 78) in HBeAg-positive chronic HBV patients treated for 52 weeks with PEG-IFN-α2b with or without lamivudine. Patient inclusion criteria for this trial have been described elsewhere [15]. In the current study, patients were eligible for post hoc analysis if they had a known outcome at week 78 and an available serum sample in our biobank for HBcrAg measurement. The study was conducted in accordance with the guidelines of the Declaration of Helsinki and the principles of Good Clinical Practice.
Laboratory testing
Serum levels of HBcrAg (log U/ml) were measured using the Lumipulse G HBcrAg assay (Research Use Only, distributed by Fujirebio Europe, Ghent, Belgium) on a LUMIPULSE G1200 analyser (Fujirebio Inc., Tokyo, Japan). The upper limit of quantitation of the Lumipulse HBcrAg assay was 7 log U/ml and samples with higher levels were diluted according manufacturer's instructions and retested. The lower limit of detection was 2 log U/ml, with a validated lower limit of quantification of 3 log U/ml [6,16,17]. Serum alanine aminotransferase (ALT) levels were measured using automated techniques. Serum quantitative HBeAg using the Roche ELECSYS assay (Roche Diagnostics, Indianapolis, IN, USA) and quantitative HBsAg with the Abbott Architect (Abbott, Abbott Park, IL, USA). HBV DNA was measured using an in-house developed TaqMan polymerase chain reaction assay based on the EuroHep standard [18]. The presence of PC (at nucleotide position G1896) and BCP (at nucleotide positions A1762 and G1764) mutants was assessed using the INNO-LiPA HBV PreCore assay (Innogenetics, Ghent, Belgium) [19]. HBV genotype was assessed using the INNO0LiPA assay (Innogenetics).
Statistical analysis
We compared serum HBcrAg levels across treatment arms and studied the relationship between HBcrAg kinetics during therapy and off-treatment response. Response to treatment was defined as HBeAg loss with HBV DNA <2,000 IU/ml at 6 months post-treatment (week 78) [5]. For association testing between variables we used the Student's t-test, one-way ANOVA, χ
2
, Pearson correlation or their non-parametric equivalents when appropriate. A grid-search was used to identify an HBcrAg cutoff that could be used to identify non-responders to PEG-IFN therapy; performance of this cutoff was compared and combined with the currently recommended stopping-rule based on an HBsAg level >20,000 IU/ml at week 24 of therapy [20,21]. Presented error bars in figures represent the standard error of the mean (
Results
Baseline patient characteristics
Out of 266 patients, 222 individuals (83%) were eligible for post hoc analysis (114 treated with PEG-IFN monotherapy). The 44 ineligible patients were excluded due to missing samples or unknown treatment outcomes. The characteristics of the excluded patients were similar to those of the overall cohort. HBV genotypes A/B/C/D/ other were detected in 75/21/36/90/0. Ninety-one of the patients were male (80%). Most patients were Caucasian (160 [72%]) or Asian (46 [21%]). Response was observed in 42 (19%), HBsAg loss in 19 (9%); rates were similar in the two treatment arms. The cohort characteristics are shown in Table 1.
Patient characteristics
Multiples of upper limit of the normal range.
Hepatitis B e antigen (HBeAg) loss and HBV DNA <2,000 IU/ml. ALT, alanine aminotransferase; HBcrAg, hepatitis B core-related antigen; HBsAg, hepatitis B surface antigen; PEG-IFN, pegylated interferon.
Correlation between baseline variables
Baseline serum HBcrAg levels strongly correlated with serum HBV DNA (r=0.635; P<0.0001), HBsAg (r=0.456; P<0.0001) and HBeAg (r=0.607; P<0.0001) levels. HBcrAg levels were higher in patients with precore mutations compared with patients who harboured wild-type virus, a sole BCP mutation or combined precore and BCP mutations (HBcrAg 8.8 versus 8.4 versus 8.3 versus 8.0 log U/ml; P<0.0001). Baseline HBcrAg levels for genotypes A/B/C/D were comparable (HBcrAg levels 8.2 versus 8.5 versus 8.3 versus 8.5 log U/ml; P=0.23). There was no association between baseline HBcrAg level and age (r=-0.05; P=0.48), body mass index (r=0.037; P=0.60), ethnicity (Caucasian 8.4 versus Asian 8.5 versus other 8.2 log U/ml; P=0.38) or sex (male 8.4 log U/ml versus female 8.3 log U/ml; P=0.79).
Relationship between baseline HBcrAg levels and response to treatment
Baseline serum HBcrAg levels were 8.4 log U/ml in responders and 8.4 log U/ml in non-responders (P=0.91). Similar results were obtained after adjustment for other established predictors (HBV genotype, presence of precore and/or basal core promotor mutations, ALT, HBV DNA and HBsAg levels, and age): the adjusted OR was 2.095 (95% CI: 0.871, 5.042; P=0.10).
On-treatment kinetics of serum HBcrAg levels and HBsAg levels
One year of PEG-IFN monotherapy resulted in a significant decline in both serum HBcrAg levels and HBsAg levels, which was sustained after stopping therapy (mean decline from baseline to week 78: HBcrAg 1.5 log U/ml; Figure 1A and HBsAg 0.93 log IU/ml; Figure 1B). Patients treated with combination therapy experienced more pronounced HBcrAg and HBsAg declines during therapy (mean declines at week 52: HBcrAg decline 2.44 versus 1.37 log U/ml; P<0.0001; Figure 1A and HBsAg decline 1.47 versus 0.87; P=0.016; Figure 1B), but these differences were not sustained off-treatment (mean HBcrAg declines 1.47 versus 1.43 log U/ml at week 78; P=0.85; Figure 1A).
HBcrAg and HBsAg kinetics during the study
The kinetics of HBcrAg levels among the patients with genotypes A/B/C/D were similar for the first 12 weeks. However, at week 24, patients with genotype A and B experienced a more pronounced HBcrAg decline than patients with genotype C and D, which was sustained during and off-treatment (P=0.0005, week 78; Additional file 1).
Among the patients treated with PEG-IFN mono-therapy, those with a response experienced a more pronounced on-treatment decline in both HBcrAg and HBsAg (mean declines at week 52: HBcrAg 3.4 versus 1.0 log U/ml [P<0.0001; Figure 2A] and HBsAg 3.03 versus 0.44 log IU/ml [P<0.0001; Figure 2B]), which were sustained off-treatment. HBcrAg remained detectable in all patients who achieved a response, even in 19 patients with HBsAg loss (mean HBcrAg 6.8 log U/ml [range: 4.0–9.5] at week 78).
HBcrAg and HBsAg decline from baseline in responders and non-responders in patients treated with pegylated interferon-α monotherapy
Prediction of response using on-treatment HBcrAg levels and HBsAg levels
Given the more pronounced HBcrAg decline in patients treated with combination therapy, these analyses were limited to patients treated with PEG-IFN alone. At week 12, HBcrAg levels did not discriminate between responders and non-responders (HBcrAg area under the receiver operating characteristic curve [AUROC] 0.57; 95% CI: 0.41, 0.72; P=0.42). HBcrAg at week 24 did discriminate between responders and non-responders with an AUROC of 0.76 (95% CI: 0.56, 0.97; P=0.008). Our grid search identified week 24 HBcrAg >8.35 log U/ml as the optimal cutoff for identification of non-responders. The cutoff identified 19 patients (19%) of whom 1 achieved a response, resulting in a sensitivity of 23% and an NPV of 95%. However, the HBcrAg based stopping rule was not superior to the one based on HBsAg of >20,000 IU/ml at week 24, which identified 45 patients (42%) with an NPV of 100% and 31% sensitivity, when applied to this cohort.
Similarly, a grid-search identified a cutoff of <5.73 log U/ml with optimal positive predictive value (PPV). This cutoff identified 10 patients, of whom 8 achieved a response (PPV of 80%). This cutoff had a sensitivity and specificity of 47% and 97%, respectively.
Prediction of response using HBcrAg levels in patients not meeting the HBsAg based stopping-rule
Among the 56 patients with HBsAg <20,000 IU/ml at week 24 (that is, patients who did not meet the HBsAg-based stopping-rule), HBcrAg was significantly lower in responders than in non-responders (6.3 log U/ml versus 7.2 log U/ml; P=0.023), with an AUROC of 0.67 (95% CI: 0.50, 0.84; P=0.050). An HBcrAg cutoff of >8.35 log U/ml identified nine additional patients (16%) with a 21% sensitivity and an NPV of 89%.
Discussion
Currently available predictors of response to PEG-IFN identify only small subgroups and therefore have limited clinical utility. As a result, PEG-IFN may currently be underutilized.
Serum HBcrAg is an emerging biomarker that appears to have clinical utility in the management of CHB. In HBeAg-positive HBV patients, both HBsAg and HBcrAg levels refect viral replication, as they correlate with serum HBV DNA levels, as well as intrahepatic HBV transcriptional activity. In HBeAg-negative patients, HBsAg levels correlate less well with intrahepatic replication, possibly due to production from integrated HBV DNA. In these patients, the association between HBcrAg and viral replication is preserved [22]. Given the association with intrahepatic viral replication, on-treatment kinetics could refect therapy response and correlate well with intrahepatic cccDNA levels and replicative activity [12,14], which in turn is associated with response to PEG-IFN. In our study, patients with a sustained off-treatment response achieved significantly more pronounced HBcrAg declines than did non-responders. These differences were sustained off-treatment and were observed in patients treated with PEG-IFN alone or in combination with lamivudine. Using a grid-search of cutoff points we identified an HBcrAg level of >8.35 log U/l as the optimal cutoff for identification of non-responders. Performance of this stopping-rule was adequate but was not superior to the established stopping-rule based on HBsAg >20,000 IU/ml. Application of our stopping rule in patients who did not meet the HBsAg-based stopping-rule identified an additional nine patients (16%) with an NPV of 89%, thus allowing therapy discontinuation for futility in more patients. However, these findings will need to be confirmed in further validation studies.
Our results now confirm – in a cohort of predominantly Caucasian patients – the predictive properties of HBcrAg for PEG-IFN-based therapies. This contradicts the conclusions of a previously reported small Thai study comprising HBeAg-positive patients with genotypes B and C, that reported superior performance of an HBcrAg stopping rule of >8.0 log U/ml [13]. Our results are consistent with studies from our group concerning HBeAg-negative patients treated with PEG-IFN and HBeAg-positive patients who received PEG-IFN add-on to ETV [10,13,23].
In the current study, we observed strong HBcrAg declines in patients who achieved a sustained off-treatment response. Remarkably the kinetics of HBcrAg and HBsAg during PEG-IFN therapy were quite similar in magnitude, however, HBsAg kinetics identified twice the number of patients who could be stopped. HBcrAg alone are therefore not superior to HBsAg levels for identification of non-responders to PEG-IFN-based therapy. Still, our findings suggest that a combination of HBsAg levels and HBcrAg may help further optimize individualized clinical decision making by identifying additional patients that can discontinue treatment for futility.
Our study also shows that combination therapy yields more pronounced on-treatment HBcrAg decline, which is not sustained off-treatment. The mechanism underlying this observation is currently unclear, and any more insight should be generated using new in vitro or animal models that elucidates the effect of lamivudine and PEG-IFN combination on the intrahepatic immune system and HBV-infected hepatocytes. What we do know is that the degree of on-treatment viral suppression is the greatest in those patients that receive combination treatment, which is also refected in more pronounced declines of other viral components like HBeAg and HBsAg, which show similar kinetics [21,24]. The apparent synergistic effects of PEG-IFN with nucleoside/nucleotide analogues is the basis for ongoing studies that combine new compounds with a PEG-IFN backbone [25].
Another important observation in our cohort is that HBcrAg levels remained detectable early after loss of HBsAg, this in concordance with Seto et al. [26,27] who detected HBcrAg levels up to 42 months after HBsAg seroclearance. Since another recent study in patients treated with nucleoside/nucleotide analogues suggested that high HBcrAg levels predict HBeAg relapse after HBeAg seroconversion [11], future studies with long-term follow-up after HBsAg loss should explore whether HBcrAg after HBsAg loss may predict HBsAg relapse.
Finally, the persistent detectability of HBcrAg after IFN-α induced HBsAg loss may also be relevant for the monitoring of the efficacy of novel antiviral agents that directly interfere with HBsAg production and or release [25].
In summary, HBeAg-positive CHB patients that respond to PEG-IFN therapy achieve a more pronounced HBcrAg decline during treatment, and this is sustained during off-treatment follow-up. HBcrAg levels at week 24 predict response to treatment, and assessment of HBcrAg levels in combination with existing HBsAg-based stopping-rules may further optimize individualized antiviral therapy for CHB.
Footnotes
Acknowledgements
The authors would like to thank Laura Vernoux (Fujirebio, Ghent, Belgium) for the helpful and interesting discussions. This study was sponsored by the Foundation for Liver and Gastrointestinal Research (SLO) in Rotterdam, the Netherlands. Financial support was provided by Health∼Holland (Stichting LSH-TK, funded by the Dutch government project number LSHM15032) and Fujirebio Europe. The funding sources did not have any influence on study design, data collection, analysis and interpretation of the data, writing of the report or the decision to submit for publication
This study was performed in accordance with the guidelines of the 1964 Declaration of Helsinki, the ethical standards of our institution and the principles of Good Clinical Practice. All subjects gave written informed consent.
MJS received research support and consultancy fees from Fujirebio, Roche, Gilead. HLAJ received grants from and is a consultant for: Bristol-Myers Squibb, Gilead Sciences, Novartis, Roche, Merck, Innogenetics. AB received research support from Fujirebio and Gilead. The other authors declare no competing interests.
