Graphical Abstract
INTRODUCTION
Chronic hepatitis B (CHB) infection affects 254 million people globally, and was responsible for 1.1 million deaths in year 2022 from cirrhosis and hepatocellular carcinoma (HCC) [1]. Antiviral treatment can reduce the risk of these complications in at-risk individuals living with CHB. Viral biomarkers reflective of the viral burden are indispensable to identify at-risk individuals to start antiviral treatment, assess treatment response, and provide prognostic information in the long run.
Several blood-based viral biomarkers including hepatitis B virus (HBV) RNA and hepatitis B core-related antigen (HBcrAg) correlate well with intrahepatic covalently-closed circular DNA (cccDNA) [2,3]. Circulating HBV RNA is encapsidated pre-genomic RNA (pgRNA) in virus-like particles or naked capsids, as full-length, spliced or 3’ terminally truncated forms [4]. The predominant species of circulating HBV RNA is 3.5-kilobase HBV RNA when measured by reverse transcription polymerase chain reaction (RT-PCR) [5,6]. HBV RNA levels in the blood are typically 1–2 log lower than the HBV DNA levels in untreated patients. This relationship is reversed after a period of nucleoside analogue (NUC) treatment, when the blood HBV RNA levels were decreased to a lesser extent than HBV DNA, leading to a reverse in RNA:DNA ratio [7]. This phenomenon is in line with primary action of NUC which only inhibits the reverse transcription of pgRNA to HBV DNA. The reduction of HBV RNA would likely be the result of dilutional effect of hepatocyte turnover with NUC reducing de-novo infection of the newly formed hepatocytes.
Most studies characterize RNA kinetics in the early phase of antiviral treatment [8], with the longest FU up to 2 years [7,9]. These studies revealed a large proportion (up to 77.5%) of NUC-treated patients having persistently detectable HBV RNA in the blood despite DNA undetectability [7]. The effect of long-term NUC on pgRNA remains unknown. Characterizing the pgRNA profile upon long-term NUC treatment is particularly important when considering new approaches to achieve functional cure or partial cure, such as the use of novel compounds to assess target engagement or the ‘stop-to-cure’ approach where on-treatment RNA levels can predict success of off-therapy virological control after cessation of long-term NUC [10,11].
Even with long-term NUC treatment, the risk of HCC is not eliminated in CHB infection. As such, predictive scores like the PAGE-B score (derived from age, gender and platelet) were developed to estimate the risk of HCC in NUC-treated patients. Although the discrimination power is good (c-index 0.82), the score was mainly developed in Caucasian patients [12]. In this context, measuring blood pgRNA has also been shown to play a role in HCC risk prognostication in NUC-treated Chinese patients, as on-treatment detectable serum RNA is associated with 3.5-fold higher risk of HCC [13].
The aim of this study is to determine the longitudinal pgRNA levels in patients receiving up to 5 years of NUC.
MATERIALS AND METHODS
Patient source
This is a retrospective study that involved patients with CHB aged ≥18 years from the Department of Medicine, Queen Mary Hospital, The University of Hong Kong, Hong Kong which was an academic-affiliated tertiary hospital. Patients were persistently seropositive for hepatitis B surface antigen (HBsAg) for at least 6 months and were first started on first-line NUC (entecavir [ETV], tenofovir disoproxil fumarate [TDF], or tenofovir alafenamide [TAF]) between year 2006 and year 2018 for guideline-approved indications at the material time. In principle, treatment was indicated for patients with high HBV DNA coupled with persistently raised alanine aminotransferase (ALT) or presence of cirrhosis. The day of starting NUC was regarded as baseline for this study. Patients were excluded in the presence of known HCC or diagnosis of HCC within 6 months after NUC initiation, concomitant etiologies for chronic liver disease (such as hepatitis C virus co-infection, Wilson’s disease, alcohol-related liver disease, etc.), pregnancy, received liver transplantation, participated in drug trials for CHB, in the absence of retrievable plasma samples, or did not have ≥5 years of follow-up since NUC initiation. Patients were followed up every 4–6 months for clinical assessment, blood tests, and review of ultrasonography for HCC surveillance which was suggested by attending physicians during follow-up.
The consent for the current study was waived, as it only involved the use of de-identified archived blood samples. The sample usage for testing in the present study was approved by the Institution Review Board, The University of Hong Kong/ Hong Kong West Cluster Institutional Review Board (UW 24-470).
Clinical and biochemical parameters
Clinical variables including age, sex and type of NUC were recorded. Biochemical parameters including liver function test, alpha fetoprotein and complete blood count were recorded.
Plasma pgRNA, DNA and HBcrAg measurement
Plasma pgRNA was measured by Research Use Only RealTime HBV RNA v2.0 (0.2 mL) (Abbott Diagnostics, Abbott Park, IL, USA) which has a lower limit of detection (LLOD) and quantification (LLOQ) of 0.8 log U/mL (~20 copies/mL). Levels were log transformed and expressed in log10 U/mL. For the sake of data analysis, to differentiate detectable but unquantifiable RNA from undetectable RNA, a value of 0.0 log U/mL was arbitrarily assigned to samples with undetectable plasma pgRNA, while a value of 0.8 log U/mL was assigned to samples with detectable but unquantifiable plasma pgRNA. The assay was designed for duplexed detection of the X and core target in the 5’ and 3’ ends of the full-length pgRNA. HBV DNA measurements were performed by Cobas Taqman assay (Roche Diagnostics) with the LLOD of 10 IU/mL. Levels were log transformed and expressed in log10 U/mL. A value of 0.9 log IU/mL was arbitrarily assigned to samples with undetectable plasma HBV DNA. pgRNA and HBV DNA were measured at baseline, 1 year, 3 years and 5 years of NUC treatment. HBcrAg was measured in a subgroup of baseline samples by the Lumipulse G HBcrAg chemiluminescence Enzyme Immunoassay (Fujirebio, Tokyo, Japan), with LLOD of 100 IU/mL and linear range of 1,000 to 10,000,000 U/mL.
Definition
The primary outcome of this study is the kinetics of plasma pgRNA under the effects of NUC for up to 5 years. Secondary outcomes are the factors associated with lower plasma HBV RNA levels in these patients. The PAGE-B score at baseline is used to define the risk of HCC in this NUC-treated cohort. A score of ≤9, 10–17 and ≥18 denotes low risk, intermediate risk and high risk, respectively [12]. Cirrhosis was diagnosed by Fibrosis-4 Index (FIB-4) ≥2.67, which is a score calculated based on 4 clinical parameters: age, platelet count, ALT and aspartate aminotransferase (AST) [14]. HCC was diagnosed by classical radiological appearance on contrast-enhanced cross-sectional imaging (computed tomography or magnetic resonance imaging) with or without histological confirmation [15].
Statistical analysis
Continuous variables were expressed in median and interquartile range. Comparison of continuous variables between groups was performed using Mann–Whitney U-test or Kruskal–Wallis test. Categorical variables were expressed in percentages. Comparison of categorical variables between groups was performed using Pearson’s χ2 test or Fisher’s exact test, as appropriate. Pearson correlation coefficients were generated to appreciate the linear relationship between viral biomarkers. Differences in pgRNA levels between different timepoints were calculated through paired sample analysis by subtraction, and patients with samples at baseline with any subsequent timepoints (year 1, year 3, year 5) were included for analysis. We performed analysis on modelling the kinetics of pgRNA after NUC treatment in this study. The linear mixed regression was applied to estimate the individual reductions of pgRNA based on the decline curves from baseline at year 1, 3 and 5. The estimated individual pgRNA reductions were also used to assess the HBV RNA levels at different time points during treatment and the duration of NUC treatment required to achieve HBV RNA undetectability. Plasma samples from patients who developed HCC during the 5-year follow-up period would be censored. Cox proportional hazard regression was performed to estimate the hazard ratios for factors associated with HCC. Subgroup analyses were performed according to different clinical parameters. All statistical analysis was performed using Statistical package for Social Sciences (SPSS) version 27.0 (IBM Co., Armonk, NY, USA) and STATA version 17 (StataCorp LLC, College Station, TX, USA). A two-tailed P-value of <0.05 was considered statistically significant.
RESULTS
Baseline characteristics
A total of 1,354 patients were recruited. At baseline, the median age was 49.8 (interquartile range [IQR] 40.2–57.3) years old, with 65.2% male and majority (83.9%) being hepatitis B e antigen (HBeAg)-negative. Cirrhosis was present in 28.6% of patients. The median HBV DNA and pgRNA were 5.76 (IQR 3.54–7.17) log IU/mL and 3.68 (IQR 2.42–5.19) log U/mL, respectively. The majority (80.9%) was started on ETV, with the remaining 16.3% on TDF and 2.7% on TAF. Up to 45.2% patients had high PAGE-B score at baseline (Table 1).
Longitudinal profile of plasma HBV RNA and DNA
Retrievable plasma samples for HBV DNA and pgRNA analysis were present in 1,035, 1,136 and 1,186 patients, and 932, 1,098 and 839 patients at year 1, year 3 and year 5, respectively. For HBV DNA, the median levels were 1.00 (IQR 1.00–2.23), 1.00 (IQR 1.00–1.00) and 1.00 (IQR 1.00–1.00) log IU/mL, for year 1, year 3 and year 5, respectively. For pgRNA, the median levels were 2.45 (IQR 1.82–3.62), 2.23 (IQR 1.67–3.05) and 2.14 (IQR 1.48–2.86) log U/mL, for year 1, year 3 and year 5, respectively (Fig. 1). To address the missing data of HBV RNA levels at year 1, 3 and 5 during the NUC therapy, Multiple Imputation by Chained Equations (MICE) was also performed in this study. Missing value of HBV pgRNA data was imputed 20 times using other observed variables, including baseline age and gender. After MICE, the data completion rate of HBV pgRNA levels at year 1, 3 and 5 during NUC therapy was increased to 100%, 100% and 99.9%, respectively. The median HBV pgRNA levels after MICE were 2.60 (IQR 1.86–3.82), 2.23 (1.61–3.10) and 2.08 (1.25–2.84) log U/mL, for year 1, year 3 and year 5, respectively, confirming the negligible impact of missing data.
Baseline plasma pgRNA demonstrated strong linear correlation with baseline HBcrAg (r=0.858, P<0.001), moderate linear correlation with baseline HBV DNA (r=0.553, P<0.001) and weakly with age (r=–0.277, P<0.001) and PAGE-B score (r=–0.162, P=0.004). Upon NUC treatment, plasma pgRNA levels were well correlated with pgRNA levels at all other timepoints (r=0.584–0.609, P<0.001), but the correlation coefficient with HBV DNA weakens with NUC treatment (1 year: r=0.257, P<0.001; 3 year: r=0.276, P<0.001; 5 year: r=0.224, P<0.001) (Supplementary Table 1 and Supplementary Figs. 1, 2, 6).
The median decline in pgRNA from baseline at year 1, year 3 and year 5 was 0.82 (IQR 0.09–1.96) log U/mL, 1.20 (IQR 0.30–2.30) log U/mL and 1.54 (IQR 0.53–2.67) log U/mL, respectively (Fig. 2). Younger patients (<50 years) had a bigger decline of pgRNA at 3 years (1.35 vs. 1.13 log U/mL decline, P=0.004) (Supplementary Table 2). When compared to ETV, tenofovir (TDF or TAF) led to smaller decline in plasma pgRNA at 1 year (–1.02 vs. –0.49 log U/mL, P=0.007), and 3 year (–1.37 vs. –0.95 log U/mL, P=0.029) but this difference was no longer observed at 5 year (Supplementary Table 2, Supplementary Fig. 3). Sex, HBeAg status, cirrhosis and PAGE-B score were not associated with significant differences in pgRNA decline at all timepoints (Supplementary Table 2).
Proportion achieving undetectable plasma HBV DNA and undetectable or unquantifiable pgRNA under NUC treatment
After 1, 3 and 5 years of NUC treatment, 69.2%, 88.9% and 93.0% patients achieved undetectable plasma HBV DNA. The corresponding percentages for achieving undetectable/ unquantifiable plasma pgRNA are 13.5%, 15.9% and 20.1% (Fig. 3). Older age at baseline is associated with higher proportion of undetectable/ unquantifiable pgRNA at baseline and year 3, but not at year 1 and year 5 (Supplementary Fig. 4). Compared to ETV treatment, tenofovir treatment was more likely to result in undetectable/ unquantifiable plasma pgRNA at 1 year (11.8% vs. 20.7%, P=0.003), 3 year (14.5% vs. 20.8%, P=0.023) and 5 year (18.9% vs. 23.3%, P=0.092) (Supplementary Fig. 5).
Linear mixed regression analysis was performed to estimate the trajectory of plasma pgRNA and the time needed to reach undetectable/ unquantifiable pgRNA. The model adjusted for the biphasic nature of pgRNA decline and the simulated results are shown in Figure 4. The median duration required to reach unquantifiable and undetectable pgRNA for patients with CHB was 22.3 (IQR 10.6–40.1) years and 30.3 (IQR 17.0–50.1) years after the start of NUC treatment, respectively.
Factors associated with lower pgRNA levels
Baseline clinical parameters were examined to identify subgroups with lower pgRNA levels. Age ≥50, HBeAg-negativity and the subgroup with high baseline PAGE-B score were consistently associated with lower pgRNA at all timepoints, whereas male sex was associated with lower pgRNA at NUC treatment of 1 year onwards. Cirrhosis was found to be associated with lower pgRNA at 1 year of NUC. Type of NUC was not associated with significantly lower pgRNA levels (Table 2, Figs. 5, 6).
Among 218 patients with baseline HBeAg-positivity, 63.8% developed HBeAg seroclearance during initial 5 years of follow-up. The baseline pgRNA levels were not significantly different comparing those who developed HBeAg seroclearance compared to those who remained HBeAg positive (5.59 vs. 5.10 log U/mL, P=0.152), but was associated with smaller magnitude of pgRNA decline at 1 year of NUC (–0.36 vs. –1.07 log U/mL, respectively, P=0.005).
Upon extended period of long-term follow up (median duration of 11.4 [IQR 9.5–12.9]) years, 54 patients developed HBsAg seroclearance at a median interval of 11.0 years (IQR 8.5–12.9), with annual HBsAg seroclearance rate of 0.35%. None of the HBsAg seroclearance events occurred during the initial 5 years of NUC, thus all patients were maintained on NUC as of year 5 timepoint. In those who developed HBsAg seroclearance later on, pgRNA level was lower at baseline (2.42 vs. 3.71 log U/mL, P=0.084), and were significantly lower at year 1 (1.78 vs. 2.48 log U/mL), year 3 (0.98 vs. 2.24 log U/mL) and year 5 (0.80 vs. 2.16 log U/mL) of NUC therapy (P<0.001), compared to those who did not lose HBsAg upon long-term follow-up.
HCC
Of the 1,354 patients, 84 subsequently developed HCC at a median interval of 3.6 (IQR 1.9–5.2) years. Compared to patients who remained free from HCC, those who developed HCC were older (57.3 vs. 49.1 years old, P<0.001), more likely to be male (88.1% vs. 63.7%, P<0.001), more likely to be HBeAg-positive at baseline (25% vs. 15.5%, P=0.031), had higher FIB-4 score (3.70 vs. 1.67, P<0,001), had lower baseline HBV DNA (5.2 vs. 5.8 log IU/mL, P=0.014), and had higher AFP (5.1 vs. 3.0 ng/mL, P<0.001). At multivariate Cox regression, age (HR 1.062, 95% CI 1.039–1.086), male sex (HR 3.478, 95% CI 1.757–6.884), HBeAg positive (HR 3.146, 95% CI 1.786–5.542) and baseline HBV DNA (HR 0.836, 95% CI 0.755–0.925) were independently associated with HCC (all P<0.001; Supplementary Table 3). pgRNA levels at all timepoints and the dynamic changes were not associated with HCC development (Supplementary Table 4). Results were similar after various ratios of propensity score matching (1:1, 1:2, 1:5 and 1:10) for age and FIB-4 score (Supplementary Tables 5 and 6).
DISCUSSION
Our study described the long-term kinetics of plasma pgRNA in patients on first-line NUCs. We confirmed the slow drop in plasma pgRNA levels upon potent NUC therapy, with a median of 1.54 log decline at 5-year of NUC. In addition, only 16.3% could achieve undetectability at 5 years of treatment. This reflects limited action of NUC on the upstream pathway of the viral cycle. pgRNA is directly transcribed from cccDNA, and is regarded as its surrogate marker. Once NUCs are started, reverse transcription is halted and DNA synthesis from RNA is ceased, resulting in rapid decline in DNA but relatively stagnant amount of circulating HBV RNA formed from direct transcription of cccDNA. Another point to note is the lack of data to confirm infectivity of HBV RNA virions. Even if HBV RNA still circulates and populates the liver micro-environment, these virions or pgRNA-containing capsids cannot infect new hepatocytes [16]. As such, the observed decline in RNA pool likely results from hepatocyte turnover [17]. Nevertheless, the fact that many patients are still having measurable pgRNA after being treated with NUCs up to 5 years can provide a clinical utility to assess the virus replication in terms of cccDNA transcriptional activity inside the liver. For those with undetectable pgRNA, we can have a higher confidence in grading these patients as having silenced viral activity which is the expected outcome from functional cure, hopefully achieved at a higher rate by all the novel agents being investigated in this decade. Furthermore, the long duration of NUC therapy needed to achieve plasma pgRNA undetectability underscores the potential role of this viral biomarker as an additional tool in selecting patients to consider the ‘stop-to-cure approach’. As plasma HBV RNA has been shown to predict off-treatment virological response [18], even if all criteria are met including on-treatment sustained HBV DNA undetectability [19]—which was achieved in >93% patients after 5 years of NUC in our cohort, additional measurement of plasma pgRNA will help to identify patients with high risk of post-NUC cessation relapse.
We deciphered the profile of pgRNA according to real world clinical characteristics. Older age, male sex, HBeAg-negativity and high PAGE-B score were consistently associated with lower plasma pgRNA compared to the respective counterparts. In contrast, type of NUC or cirrhosis was not associated with significant differences in the absolute plasma pgRNA level upon NUC treatment. These findings suggest the need to define specific reference ranges of plasma pgRNA measurement for patients with different clinical characteristics. When looking at the relative change in pgRNA upon NUC, ETV was associated with bigger magnitude of decline initially compared to tenofovir, but the difference was no longer significant at 5-year of treatment. The observed smaller pgRNA reduction seen in tenofovir-treated patients than in ETV-treated patients could be due to two reasons. Firstly, there was a higher proportion of patients started on TDF with undetectable pgRNA at baseline than those started on ETV (9.7% vs. 4.0%, P=0.01; Supplementary Fig. 5). Secondly, tenofovir treatment was more likely than ETV to achieve pgRNA undetectability at subsequent timepoints (Supplementary Fig. 5). Therefore, tenofovir rendered pgRNA undetectable in more patients, resulting in an apparently smaller magnitude of pgRNA reduction compared to ETV treatment.
Our paper did not show a positive correlation between plasma pgRNA and HCC development, in contrast to another recently published study which indirectly estimated the HBV RNA levels by subtracting HBV DNA from total nucleic acids determined by RT-PCR [13]. This would allow the detection of more forms of RNA than the assay used in the current study that was designed to measure or detect full-length pgRNAs. It would be worthwhile to look into the role of other circulating forms of RNA (e.g., spliced variants, truncated variants) in HCC prediction. In addition, the number of events of HCC is relatively low within a 5-year follow-up frame. Longer duration of follow-up would allow more events to be observed for comparison. Moreover, this cohort presented with high PAGE-B score at baseline (45.2%) and relatively high proportion of them with cirrhosis (>25%), probably representing patients at higher risk of HCC. This may outweigh the other risk factors (age, low platelet, sex) and hence mask association of plasma pgRNA, if any, on risk of HCC. While it was generally believed that lower viral load is associated with better prognosis, the level of plasma pgRNA were lower in those with risk factors for HCC (old age, male sex) as shown in our cohort. This again calls for age- and sex-specific reference ranges for plasma pgRNA measurement for a fair comparison between patients with different clinical characteristics, in addition to adjusting for the duration of NUC therapy. Indeed, plasma HBV RNA measurement per se should be standardized. A number of different methods have been described, including RT-PCR, droplet digital PCR, or rapid amplification of complimentary DNA ends (RACE)-based real-time PCR method [5,20,21]. The amplification targets used in different systems non-negligibly differ from each other, resulting in a heterogenous profile of HBV RNA species detected with varying lengths observed. Standardized assays are needed to ensure optimized accuracy and repeatability of findings to inform clinical decisions. We also showed that baseline HBV DNA was negatively associated with HCC (HR 0.836) even after adjusting for age, sex and HBeAg, in contrast to the general concept that viral load positively correlates with HCC risk. It is worth noting that in the REVEAL-HBV cohort where the REACH-B score was derived from, only 2% of patients were classified as cirrhotic [22], compared to our cohort with >25% of patients being cirrhotic. In fact, the association between viral load and HCC among patients with cirrhosis is not as clear as in non-cirrhotic patients [23,24]. Increasing data also suggests that the relationship between viral load and risk of advanced liver disease is not linear [25], and might even be an inverse association in HBeAg positive population [26]. Evidence also suggested a negative correlation between intrahepatic viral nucleic acid detection and tumor aggressiveness [27], tumor size [28], and a positive association with better survival [27]. More data is needed to examine the role of viral load and HCC risk especially in cirrhotic population. Practically, the priority for viral nucleic acid quantification might not be as high in the setting of cirrhosis—a strong risk factor for HCC per se—when NUC should be initiated and HCC surveillance strategy should be implemented regardless of viral load.
The strengths of our study include the large sample size, the longest follow-up to date determining pgRNA dynamics upon NUC treatment, and well characterized clinical characteristics. Our study has several limitations. Firstly, cirrhosis was diagnosed by FIB-4, a well-established non-invasive index for liver fibrosis assessment. We did not have histological or imaging confirmation for most patients. Secondly, HBV genotype data is not known, but it is expected to be predominantly B and C for this population of Chinese ethnicity [29]. Thirdly, the study involved a single-center retrospective cohort. We were not able to evaluate patients who were lost to follow-up or without retrievable plasma samples. Fourthly, the impact of data imputation for unquantifiable pgRNA values could not be neglected in this study, especially the dynamic changes in pgRNA levels were small. Fortunately, unquantifiable pgRNA only accounted for 0.7–3.8% in the various timepoints. Lastly, the correlation analyses were exploratory in nature and not explicitly designed to test a specific hypothesis. The observations should be interpreted with considerations of multiple comparisons, unadjusted confounding factors, and small sample sizes for certain subgroups. Future prospective studies with larger sample size and external validation would be required to confirm the findings from this study.
In conclusion, plasma pregenomic HBV RNA declines much slower than DNA upon NUC therapy, with a median of 1.54 log decline and only 16% achieving RNA undetectability after 5 years of first-line NUC indicating cccDNA silencing has not been achieved in the majority of patients. Age, sex, HBeAg status, PAGE-B score and duration of NUC should be taken into account when interpreting the plasma pgRNA level.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
