HBsAg seroclearance with HBV DNA <100 IU/mL and long-term risks of adverse liver events, all-cause mortality, and cause-specific mortality in chronic hepatitis B: a prospective cohort study

Summary

Background

Sustained HBsAg seroclearance with HBV DNA suppression is associated with favorable outcomes in chronic hepatitis B (CHB), but its associations with overall survival and cause-specific mortality remain insufficiently quantified.

Methods

In a prospective, multicenter, community-based cohort of 6018 Chinese CHB patients, sustained HBsAg seroclearance with HBV DNA <100 IU/mL was modeled as a time-dependent exposure using propensity score-matched Cox regression. We estimated population attributable fractions (PAFs) for burden associated with persistent infection and restricted mean survival time (RMST) for survival difference.

Findings

During a median 11.1-year follow-up, 1029 participants with sustained HBsAg seroclearance and HBV DNA <100 IU/mL (97.0% spontaneously) were compared with 2931 matched participants with persistent infection. This serovirological state was associated with lower risks of cirrhosis (adjusted hazard ratio [aHR], 0.45; 95% CI, 0.27–0.74) and hepatocellular carcinoma (aHR, 0.52; 95% CI, 0.33–0.81). It was also associated with lower all-cause mortality (aHR, 0.47; 95% CI, 0.36–0.60), as well as liver-related, extrahepatic cancer-related, and cardiovascular mortality. The RMST difference was 5.17 years within the observed age range.

Interpretation

Sustained HBsAg seroclearance accompanied by very low HBV DNA levels was associated with lower risks of adverse liver outcomes and selected extrahepatic mortality, as well as longer RMST, highlighting the prognostic relevance of this serovirological state for long-term risk stratification in CHB.

Funding

National Natural Science Foundation of China (82373654, 82674792); Jiangsu Provincial Key Discipline of Epidemiology (ZDXK202250); Scientific Research Project of Jiangsu Provincial Health Commission (H2023041); Science and Technology Plan Project of Jiangsu Provincial Disease Control and Prevention Administration (JSJK2026M00015).

Graphical abstract

Research in context.

Evidence before this study

We searched PubMed for articles published up to December 31, 2025, using combined search terms related to HBsAg seroclearance (including “HBsAg seroclearance”, “functional cure”, “HBsAg loss”, and “HBsAg seroconversion”) and clinical outcomes (including “cirrhosis”, “hepatocellular carcinoma”, “mortality”, “cause-specific mortality”, and “extrahepatic mortality”). Existing evidence has consistently shown that HBsAg seroclearance or functional cure is associated with lower risks of HCC and other adverse liver outcomes. However, evidence remains limited on whether HBsAg seroclearance is associated with subsequent extrahepatic mortality, particularly extrahepatic cancer-related and cardiovascular mortality. Given the aging population of individuals with chronic hepatitis B (CHB) and the increasing burden of comorbidities, clarifying these associations is important for defining the broader prognostic relevance of HBsAg seroclearance and informing long-term risk stratification.

Added value of this study

In this prospective and community-based cohort of 6018 Chinese adults with CHB followed from 2012 to 2023, sustained HBsAg seroclearance accompanied by HBV DNA <100 IU/mL was associated not only with lower risks of cirrhosis and HCC but also with lower all-cause, extrahepatic cancer-related, and cardiovascular mortality. These associations remained broadly consistent after accounting for competing risks and across multiple sensitivity analyses. Model-based analyses estimated a 42.3% population attributable fraction associated with persistent infection and a 5.17-year difference in restricted mean survival time between the two states within the observed age range.

Implications of all the available evidence

These findings extend the prognostic relevance of HBsAg seroclearance beyond conventional hepatic outcomes to overall survival and selected extrahepatic mortality outcomes. They support consideration of HBsAg seroclearance as an informative long-term prognostic marker in CHB and highlight the potential population health relevance of broader attainment of this serovirological state.

Introduction

Hepatitis B virus (HBV) infection affected approximately 254 million people worldwide in 2022, resulting in an estimated 1.1 million deaths.¹ Notably, China bears the world's heaviest share of the burden, with approximately 75 million people infected.² HBV infection is a major risk factor for the development of liver cirrhosis and hepatocellular carcinoma (HCC) and has also been associated with an increased risk of digestive system malignancies and kidney diseases.^3,4 Although the relationship between HBV infection and the development of cardiovascular disease remains debated, accumulating evidence indicates that patients with chronic HBV infection experience a significantly higher risk of mortality from cardiovascular and cerebrovascular diseases (as well as extrahepatic malignancies and endocrine disorders) compared with the general population.^5,6 To address this substantial public health challenge, the World Health Organization has endorsed a global strategy to eliminate viral hepatitis by 2030, emphasizing not only the prevention of new infections (e.g., mother-to-child transmission) but also, crucially, the scaling up of diagnosis and treatment to reduce mortality among those already infected.⁷

Loss of HBsAg is a key therapeutic goal in chronic hepatitis B (CHB) and forms the basis of the treatment endpoint commonly termed functional cure.⁸ Spontaneous HBsAg seroclearance can also occur during the natural course of CHB, albeit infrequently.⁹ Previous studies have shown that HBsAg seroclearance is associated with substantially lower risks of adverse liver outcomes, including cirrhosis and HCC.^10,11 However, its associations with overall survival and cause-specific mortality, particularly extrahepatic mortality, remain less well characterized. Clarifying these associations may help define the broader prognostic relevance of HBsAg seroclearance for long-term risk stratification in CHB.

Furthermore, given the rarity of HBsAg seroclearance (occurring in approximately 1% per year), accurate estimation of its long-term prognostic associations may be affected by potential methodological biases.⁹ Specifically, HBsAg seroclearance is a time-dependent event occurring during follow-up: individuals must remain event-free until the point of seroclearance is observed. Consequently, conventional analyses that treat subsequent HBsAg seroclearance as a fixed baseline characteristic are prone to immortal time bias, which can bias the estimated association with survival. Therefore, prospective cohorts with regular interval surveillance are essential, as they allow for more accurate ascertainment of the timing of seroclearance—a prerequisite for correctly allocating person-time and mitigating this bias.

To address these gaps, we used data from the Chronic Hepatitis B Infection and Liver Disease (HEPCARE) study, a large multicenter, prospective community-based cohort in China.¹² Using time-dependent Cox regression to account for changes in serovirological status during follow-up, we aimed to evaluate the associations between sustained HBsAg seroclearance with HBV DNA <100 IU/mL and three prespecified domains of long-term clinical outcomes: adverse liver events, all-cause mortality, and cause-specific mortality. We further estimated model-based population attributable fractions (PAFs) to characterize the population-level burden associated with persistent infection and restricted mean survival time (RMST) to quantify between-state survival differences. Together, these analyses were intended to provide a more comprehensive assessment of the long-term prognostic and public health relevance of sustained HBsAg seroclearance with very low HBV DNA levels in CHB.

Methods

Study population

Details of the HEPCARE cohort can be found elsewhere.¹² In brief, this cohort includes 7250 individuals with chronic hepatitis B infection, who were enrolled through two rounds of serological surveys conducted between September 2009 and November 2010 in Jiangsu Province. These eligible individuals were required to be HBsAg-positive, hepatitis C virus antibody negative, and free of cirrhosis and liver cancer. Follow-up assessments were conducted in 2012, 2013, 2014, 2016, 2018, 2020, and 2023. At each assessment, participants underwent face-to-face questionnaires, in-person physical examinations, laboratory testing—including HBsAg and HBV DNA measurements—and imaging examinations. The exact follow-up dates varied across regions. Participant retention was high across follow-up waves (Table S1). All participants provided signed informed consent before participation.

Because HBV DNA data collection began at the 2012 follow-up, this study defined the 2012 follow-up visit as the baseline and followed participants through December 2023. In this study, participants were excluded if they met any of the following criteria: pre-existing cirrhosis or HCC; erroneous data records; fewer than two follow-up measurements of HBsAg and HBV DNA; or more than two missing covariates. After the above exclusions, there were 6551 participants in the subsequent analyses (Fig. 1).

Fig. 1.

Flow diagram of patient selection and propensity score matching. Abbreviations: HBsAg, hepatitis B surface antigen; HCC, hepatocellular carcinoma; HEPCARE, Chronic Hepatitis B Infection and Liver Disease; OBI, occult hepatitis B infection.

Covariates and laboratory assessment

Detailed descriptions of the investigation and laboratory methods have been reported.¹² For each participant, demographic characteristics, lifestyle factors, and antiviral therapy details (i.e., interferon therapy, nucleoside/nucleotide analogs) were collected using a structured questionnaire conducted by trained interviewers at study entry and at each follow-up visit.

During follow-up visits conducted between 2012 and 2014, HBsAg was measured using an enzyme-linked immunosorbent assay kit (KeHua Bioengineering Co., Ltd., Shanghai, China), with a lower detection limit of 0.2 IU/mL. Beginning in 2016, HBsAg was measured using chemiluminescent microparticle immunoassay (Abbott Architect-i1000), with a lower detection limit of 0.05 IU/mL. HBV DNA levels were quantified using a one-step, extraction-free Hepatitis B Virus Nucleic Acid Diagnostic Kit (PCR-Fluorescence Probing) (Sansure Biotech Inc., China). According to the manufacturer's instructions (Registration No. 20153400083), the lower limit of quantification (LLOQ) is 100 IU/mL, with a linear quantification range spanning from 1.0 × 10² to 5.0 × 10⁹ IU/mL.

Definition of HBV serovirological states

Because most participants with negative HBsAg results at baseline (2012) did not undergo further HBV DNA testing, HBsAg positivity was used as the sole marker to define infection status at baseline. During all subsequent follow-up visits, both HBsAg and HBV DNA were used for group classification.

In this community-based observational study, the primary serovirological exposure was defined as sustained HBsAg seroclearance with HBV DNA <100 IU/mL, confirmed at two or more available consecutive follow-up measurements. Intermittent missed follow-up visits between the two qualifying measurements were permitted. After excluding two participants whose serovirological profiles were compatible with occult hepatitis B infection (OBI; HBsAg-negative with HBV DNA ≥100 IU/mL), the persistent infection group comprised participants who remained HBsAg-positive throughout follow-up, irrespective of HBV DNA level.

The transition time to the HBsAg-seroclearance state was determined using the actual follow-up dates. Because the exact timing of the serovirological transition was known only to fall between the last assessment not meeting the exposure definition and the first assessment meeting it, the midpoint of this interval was used as the estimated transition time, consistent with previous methodological practice.¹³ For participants meeting the exposure definition at baseline, the baseline date was used.

In an additional analysis, participants were classified at each follow-up into three time-varying serovirological states based on concurrent HBsAg and HBV DNA status: (1) HBsAg-positive/HBV DNA ≥100 IU/mL; (2) HBsAg-positive/HBV DNA <100 IU/mL; and (3) HBsAg-negative/HBV DNA <100 IU/mL. This analysis assessed whether HBsAg seroclearance provided prognostic information beyond very low HBV DNA levels alone.

Ascertainment of outcomes

The prespecified study outcomes were organized into three domains: (1) adverse liver events, comprising incident cirrhosis, HCC, decompensated cirrhosis, and liver failure; (2) all-cause mortality; and (3) cause-specific mortality, comprising liver-related, extrahepatic cancer-related, cardiovascular and respiratory mortality. Each outcome was analyzed separately.

The adverse liver events were ascertained by integrating data from the medical record system, tumor registration system, death registration system, as well as imaging examinations and self-reporting. Information on the date and causes of death was obtained from the death registration system, with causes of death classified according to the 10th version of the International Classification of Diseases (ICD-10). In detail, liver-related mortality was defined as death caused by HCC, cirrhosis, or other liver diseases (C22, K70–K76); extrahepatic cancer-related mortality was defined as death due to extrahepatic cancers with codes C00–C97 (except C22); cardiovascular mortality and respiratory mortality were defined using ICD-10 codes I00–I87 and J00-J99, respectively.

Notably, discrepancies across different data sources were resolved with priority given to the death registration system, followed by the tumor registration system, the medical record system, abdominal imaging, and participant self-reports.

Propensity score matching and inverse probability of treatment weighting

To improve comparability in selected key baseline characteristics between the groups, we conducted propensity score matching (PSM) using the “MatchIt” R package. Propensity scores were calculated using the following covariates in logistic regression models: age, sex, area, and baseline antiviral treatment status. Nearest-neighbor matching without replacement was performed at a target ratio of up to 1:3, with a caliper of 0.02 SDs on the logit propensity-score scale, and MatchIt-derived weights based on the achieved matching ratio were retained in subsequent analyses. Covariate balance was assessed using the standardized mean differences (SMDs) before and after PSM, with values less than 0.1 indicating adequate balance.

As a secondary analysis, baseline HBV DNA was additionally incorporated into the propensity score model to further account for differences in baseline viral load. Because its inclusion substantially reduced the availability of suitable matches and the resulting sample size, inverse probability of treatment weighting (IPTW) was performed separately within each of the 10 imputed datasets using stabilized and untruncated inverse probability weights.

Statistical analysis

Missingness was 5.7% for HBV DNA and less than 1% for other covariates (Table S2). Missing values were addressed using multiple imputation with 10 imputed datasets generated in the pooled cohort. Continuous variables were summarized as means (SDs) when normally distributed and as medians (IQRs) otherwise, whereas categorical variables were presented as counts and percentages.

In the propensity score-matched population (n = 3960), HBsAg-seroclearance state was treated as a time-dependent exposure, and weighted Cox proportional hazards models were used to evaluate its associations with adverse liver events and mortality. Matching weights according to the actual matching ratio within each matched set were incorporated into all outcome models. Robust sandwich variance estimators clustered by matched set were used to account for within-set correlation.

Simon–Makuch curves, an extension of the Kaplan–Meier approach for time-varying exposure status, were used to visualize survival according to dynamic serovirological status during follow-up. Differences between the curves were assessed using the score test of the univariate Cox model, which is a generalized log-rank test for start-stop data for a time-dependent exposure.

Competing risks were considered in analyses of hepatic events and cause-specific mortality. Non-liver death was treated as the competing event for cirrhosis, HCC, decompensated cirrhosis, and liver failure, whereas deaths from all other causes were treated as competing events for each cause-specific mortality outcome. Associations with HBsAg seroclearance were estimated using Fine–Gray subdistribution hazard models with the same covariate adjustment sets. For implementation of the time-varying exposure, the last observed serovirological status before a competing event was carried forward thereafter. Model-based g-computation was then used to estimate adjusted cumulative incidence functions (CIFs) and risk differences at 5 and 10 years under two hypothetical sustained exposure scenarios—HBsAg seroclearance throughout follow-up and persistent infection throughout follow-up—standardized to the covariate distribution of the matched population. For all-cause mortality, standardized risks and their differences were estimated using the same framework without competing risks.

To further evaluate the population-level burden associated with persistent infection, PAFs were calculated using model-based standardization from the time-dependent Cox regression models. For each outcome, the observed risk in the matched cohort was compared with the risk under a hypothetical scenario in which all individuals were assigned to HBsAg seroclearance throughout follow-up, with PAF calculated as $\text{PAF}=\frac{R_{\text{obs}}-R_{\text{sc}}}{R_{\text{obs}}}$, where $R_{obs}$ denotes the observed risk and $R_{sc}$ the model-standardized risk under the HBsAg seroclearance state. Estimates were pooled across the imputed datasets. For RMST analysis, age was used as the time scale. RMSTs were obtained by trapezoidal integration of the state-specific survival curves, and the difference was calculated between the two states. For both PAF and RMST, 95% CIs were estimated using cluster bootstrap of matched subclasses with 1000 replicates.

Several sensitivity analyses were conducted to assess the robustness of the findings (Table S3). First, the analyses were repeated in the full unmatched cohort to assess whether the findings were sensitive to matching. Second, the models were additionally adjusted for four liver function markers or metabolic factors to assess potential residual confounding. Third, landmark and 1-year lag analyses were performed to reduce potential immortal-time bias and reverse causation, respectively. Finally, analyses were repeated after rematching the cohort, following exclusion of 16 participants with subsequent HBsAg seroreversion and/or HBV DNA reactivation (n = 3861), and in participants who attended at least five follow-up visits (n = 3225), to assess robustness to exposure misclassification and follow-up intensity. Considering the potential interaction, we analyzed the associations stratified by sex, age, baseline HBV DNA level and antiviral treatment history.

All data analyses were performed using R software, version 4.4.2 (R Project for Statistical Computing). The significance level was set at P < 0.05, and all P values were 2-sided. To control the false discovery rate, the Benjamini-Hochberg procedure was applied separately within each model across the nine prespecified outcomes.

Ethics approval

This study involves human participants and was approved by the Ethical Review Committee of the Jiangsu Provincial Center for Disease Control and Prevention (JSJK2018-B027-02). All participants provided signed informed consent for blood collection and interviews.

Role of the funding source

The funders had no role in the study design, data collection, data analysis, data interpretation, writing of the report, or the decision to submit the manuscript for publication.

Results

Baseline characteristics and propensity score matching

During the entire follow-up period (median [IQR]: 133.5 [132.5–133.5] months), 1030 participants met the definition of sustained HBsAg seroclearance with HBV DNA <100 IU/mL after excluding four individuals who experienced outcomes before the transition. To minimize potential misclassification bias, 527 participants with HBsAg loss and HBV DNA <100 IU/mL documented at only a single follow-up visit were excluded from the final analysis, including 241 participants who were first observed with such serological transition at the 2023 follow-up (Fig. 1). After excluding two participants whose serovirological profiles were compatible with OBI, a total of 4988 participants remained in a state of persistent infection. Before matching, participants in the HBsAg seroclearance group were significantly older, more often male, had lower levels of antiviral treatment exposure and education, and were more likely to be smokers and to have a history of hypertension compared with those in the persistent infection group. In addition, the HBsAg seroclearance group had a higher proportion of unquantifiable HBV DNA (<100 IU/mL) and lower baseline ALT and AST levels (Table 1).

Table 1.

Baseline characteristics of study groups before and after propensity score matching.

Variables	Before matching				After matching
	HBsAg seroclearance n = 1030	Persistent infection n = 4988	P value	SMD	HBsAg seroclearance n = 1029	Persistent infection n = 2931	P value	SMD
Female sex	472 (45.8%)	2667 (53.5%)	<0.001	0.153	472.0 (45.9%)	1405.7 (48.0%)	<0.001	0.042
Age, years, mean (SD)	52.10 (12.27)	47.79 (13.21)	<0.001	0.338	52.07 (12.24)	52.57 (12.18)	<0.001	0.040
Age group, years			<0.001	0.321			<0.001	0.081
<30	32 (3.1%)	341 (6.8%)			32.0 (3.1%)	77.9 (2.7%)
≥30 and <50	398 (38.6%)	2484 (49.8%)			398.0 (38.7%)	1128.9 (38.5%)
≥50 and <70	531 (51.6%)	1887 (37.8%)			531.0 (51.6%)	1470.7 (50.2%)
≥70	69 (6.7%)	276 (5.5%)			68.0 (6.6%)	253.5 (8.6%)
Area			<0.001	0.135			0.737	0.011
Zhangjiagang	318 (30.9%)	1285 (25.8%)			317.0 (30.8%)	888.7 (30.3%)
Danyang	246 (23.9%)	1432 (28.7%)			246.0 (23.9%)	706.4 (24.1%)
Taixing	466 (45.2%)	2271 (45.5%)			466.0 (45.3%)	1335.9 (45.6%)
Baseline antiviral treatment			0.002	0.154			0.366	0.052
No treatment	1018 (98.8%)	4829 (96.8%)			1017.0 (98.8%)	2893.0 (98.7%)
NAs	7 (0.7%)	131 (2.6%)			7.0 (0.7%)	29.4 (1.0%)
IFN/peg-IFN	3 (0.3%)	15 (0.3%)			3.0 (0.3%)	6.6 (0.2%)
NAs plus IFN/peg-IFN	2 (0.2%)	13 (0.3%)			2.0 (0.2%)	1.9 (0.1%)
BMI, kg/m2, mean (SD)	24.66 (7.35)	24.04 (5.28)	0.010	0.097	24.66 (7.35)	24.30 (5.81)	0.149	0.055
BMI categorya			<0.001	0.137			0.056	0.098
Underweight	34.2 (3.3%)	172.7 (3.5%)			34.2 (3.3%)	79.7 (2.7%)
Normal	455.5 (44.2%)	2493.8 (50.0%)			454.5 (44.2%)	1402.8 (47.9%)
Overweight	389.4 (37.8%)	1771.6 (35.5%)			389.4 (37.8%)	1097.1 (37.4%)
Obesity	150.9 (14.7%)	549.9 (11.0%)			150.9 (14.7%)	351.5 (12.0%)
Married	949.0 (92.1%)	4595.7 (92.1%)	1	<0.001	948.0 (92.1%)	2710.0 (92.5%)	0.709	0.012
Education			<0.001	0.196			0.938	0.011
Primary school and below	456.8 (44.3%)	1737.1 (34.8%)			455.8 (44.3%)	1289.4 (44.0%)
Middle school	409.9 (39.8%)	2291.9 (45.9%)			409.9 (39.8%)	1182.8 (40.4%)
High school and above	163.3 (15.9%)	959.0 (19.2%)			163.3 (15.9%)	458.8 (15.7%)
Cigarette Smoking			<0.001	0.148			0.285	0.052
Non-smoker	672.0 (65.2%)	3595.2 (72.1%)			672.0 (65.3%)	1979.7 (67.5%)
Former smoker	25.2 (2.4%)	93.1 (1.9%)			25.2 (2.4%)	58.5 (2.0%)
Current smoker	332.8 (32.3%)	1299.7 (26.1%)			331.8 (32.2%)	892.7 (30.5%)
Alcohol consumption			<0.001	0.172			0.494	0.052
Never	758.5 (73.6%)	4023.5 (80.7%)			758.5 (73.7%)	2223.3 (75.9%)
Occasional	176.9 (17.2%)	596.1 (12.0%)			175.9 (17.1%)	448.7 (15.3%)
Frequent	75.6 (7.3%)	286.2 (5.7%)			75.6 (7.3%)	205.7 (7.0%)
Daily	19.0 (1.8%)	82.2 (1.6%)			19.0 (1.8%)	53.3 (1.8%)
History of chronic diseases	226 (21.9%)	813 (16.3%)	<0.001	0.144	225.0 (21.9%)	587.7 (20.1%)	0.203	0.045
Tumor	12 (1.2%)	41 (0.8%)	0.374	0.035	12.0 (1.2%)	23.3 (0.8%)	0.286	0.038
Hypertension	192.1 (18.7%)	697.8 (14.0%)	<0.001	0.126	191.1 (18.6%)	507.4 (17.3%)	0.350	0.033
Diabetes	26 (2.5%)	98 (2.0%)	0.303	0.038	26.0 (2.5%)	76.4 (2.6%)	0.890	0.005
Coronary heart disease	15 (1.5%)	60 (1.2%)	0.608	0.022	15.0 (1.5%)	38.0 (1.3%)	0.695	0.014
Stroke	7 (0.7%)	15 (0.3%)	0.121	0.054	7.0 (0.7%)	13.3 (0.5%)	0.417	0.030
Family history of hepatitis B	148.1 (14.4%)	771.9 (15.5%)	0.401	0.031	148.1 (14.4%)	435.0 (14.8%)	0.727	0.013
HBV DNA level, IU/ml			<0.001	0.890			<0.001	0.853
<100	784.8 (76.2%)	1881.0 (37.7%)			783.8 (76.2%)	1145.1 (39.1%)
≥100 and <2000	163.4 (15.9%)	1415.1 (28.4%)			163.4 (15.9%)	823.5 (28.1%)
≥2000 and <200,000	65.9 (6.4%)	1116.7 (22.4%)			65.9 (6.4%)	669.5 (22.8%)
≥200,000	15.9 (1.5%)	575.2 (11.5%)			15.9 (1.5%)	292.9 (10.0%)
ALT, U/L, median (IQR)	20.00 (14.00, 27.00)	21.00 (16.00, 32.00)	<0.001	0.129	20.00 (14.00, 27.00)	22.00 (16.00, 32.00)	<0.001	0.121
Abnormal ALTb	68 (6.6%)	557 (11.2%)	<0.001	0.161	68.0 (6.6%)	309.1 (10.5%)	<0.001	0.141
AST, U/L, median (IQR)	26.00 (21.00, 33.00)	28.00 (22.00, 37.00)	<0.001	0.036	26.00 (21.00, 33.00)	29.00 (23.00, 38.00)	<0.001	0.042
Abnormal ASTc	175 (17.0%)	1184 (23.7%)	<0.001	0.168	175.0 (17.0%)	740.1 (25.3%)	<0.001	0.203
Abnormal TBILd	65 (6.3%)	337 (6.8%)	0.651	0.018	65.0 (6.3%)	206.0 (7.0%)	0.443	0.029
Abnormal ALBe	18 (1.7%)	107 (2.1%)	0.487	0.029	18.0 (1.7%)	76.4 (2.6%)	0.122	0.059
Abnormal GLBf	33 (3.2%)	122 (2.4%)	0.197	0.046	33.0 (3.2%)	80.2 (2.7%)	0.443	0.028

ALB, albumin; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; GLB, globulin; HBsAg, hepatitis B surface antigen; IFN, interferon; NAs, nucleos(t)ide analogs; peg-IFN, pegylated interferon; SMD, standardized mean difference; TBIL, total bilirubin.

Values are presented as pooled counts (%), means (SD), or medians (IQR), as appropriate. HBsAg seroclearance group was defined as sustained HBsAg seroclearance with HBV DNA <100 IU/mL confirmed at two or more consecutive assessments. Non-integer counts arise from pooling across 10 imputed datasets and, after matching, from the application of matching weights. An SMD <0.1 was considered indicative of adequate balance.

^aBMI categories were defined as underweight, <18.5 kg/m²; normal weight, ≥18.5 to <24 kg/m²; overweight, ≥24 to <28 kg/m²; and obesity, ≥28 kg/m².

^bAbnormal ALT was defined as >50 U/L in males and >40 U/L in females.

^cAbnormal AST was defined as >40 U/L in males and >35 U/L in females.

^dAbnormal TBIL was defined as >20 μmol/L.

^eAbnormal ALB was defined as outside the normal range of 40–55 g/L.

^fAbnormal GLB was defined as outside the normal range of 20–40 g/L.

Given the substantial baseline differences between the two groups, PSM was performed to enable a more balanced comparison of long-term outcomes. Matching was conducted on age, sex, area, and baseline antiviral treatment status. The final matched cohort included 1029 participants with HBsAg seroclearance and 2931 with persistent infection. Excellent balance was achieved for four matching covariates (Fig. S1), and improved balance in most other measured baseline characteristics (Table 1). However, substantial residual imbalance remained in baseline HBV DNA (SMD = 0.853), with participants who achieved HBsAg seroclearance having markedly lower viral replication at baseline. A smaller residual imbalance was observed for ALT (SMD = 0.121) (Table 1).

Association of HBsAg seroclearance with adverse liver events

Among the matched populations, there were 147 incident cases of cirrhosis, 189 of HCC, and 18 of liver failure. Among the patients with cirrhosis, 36 subsequently progressed to decompensated cirrhosis (Fig. 2).

Fig. 2.

Associations between HBsAg seroclearance and adverse liver events: time-dependent Cox regression analysis after propensity score matching. Abbreviations: HBsAg, hepatitis B surface antigen; HCC, hepatocellular carcinoma; HR, hazard ratio; PY, person-years. Model 1 was adjusted for area, sex, baseline antiviral treatment status and baseline age, and Model 2 was additionally adjusted for education level, marital status, drinking and smoking status, hypertension, family history of HBV and BMI. P values shown in the plots were obtained from the score test of a univariate Cox model. ^†P_FDR < 0.1, ∗P_FDR < 0.05, ∗∗P_FDR < 0.01 and ∗∗∗P_FDR < 0.001.

Participants who achieved HBsAg seroclearance experienced markedly lower incidence rates of cirrhosis (2.02 vs. 3.97 per 1000 person-years [PY]) and HCC (2.97 vs. 4.97 per 1000 PY) than those with persistent infection (Fig. 2). The differences in incidence of cirrhosis and HCC were statistically significant between the two groups (cirrhosis: P < 0.001; HCC: P = 0.001). In the multivariable model (Model 2), HBsAg seroclearance was associated with a lower risk of cirrhosis (adjusted hazard ratio [aHR] = 0.45; 95% CI: 0.27–0.74; P_FDR = 0.007) and HCC (aHR = 0.52; 95% CI: 0.33–0.81; P_FDR = 0.011), with these associations remaining consistent across sequential adjustment models.

Although the incidence rates of decompensated cirrhosis (4 incident cases; 0.47 per 1000 PY) and liver failure (1 incident case; 0.12 per 1000 PY) were also lower in the HBsAg seroclearance group, the limited number of events resulted in lower statistical power and imprecise estimates. No statistically significant association was observed for decompensated cirrhosis (P_FDR = 0.104), whereas the significant association with liver failure was only observed in Model 2 (aHR = 0.14; 95% CI: 0.03–0.61; P_FDR = 0.013).

Association of HBsAg seroclearance with all-cause and cause-specific mortality

During follow-up, 469 deaths occurred in the matched population. Of these, 64 occurred in the HBsAg seroclearance group (incidence rate: 7.55 per 1000 PY), which was significantly lower than the 405 deaths observed in the persistent infection group (12.17 per 1000 PY; P < 0.001). Hepatic and extrahepatic risks were numerically higher among those reporting antiviral treatment before seroclearance (Table S4). After full multivariable adjustment (Model 2), HBsAg seroclearance remained associated with a lower risk of all-cause mortality (aHR = 0.47; 95% CI: 0.36–0.60; P_FDR < 0.001) (Fig. 3).

Fig. 3.

Associations between HBsAg seroclearance and all-cause and cause-specific mortality: composition of death, hazard ratios, and Simon-Makuch curves. Abbreviations: HBsAg, hepatitis B surface antigen; HR, hazard ratio; PY, person-years. Hazard ratios and 95% confidence intervals were estimated using time-dependent Cox proportional hazards model with the persistent infection group as the reference. Model 1 was adjusted for area, sex, baseline antiviral treatment status and baseline age, and Model 2 was additionally adjusted for education level, marital status, drinking and smoking status, hypertension, family history of HBV and BMI. P values shown in the Simon–Makuch plots were obtained from the score test of a univariate Cox model. ^†P_FDR < 0.1, ∗P_FDR < 0.05, ∗∗P_FDR < 0.01 and ∗∗∗P_FDR < 0.001.

When deaths were further classified by causes, 146 were attributed to extrahepatic cancers, 117 to cardiovascular diseases, 112 to liver-related diseases, and 22 to respiratory diseases (Fig. 3). Consistent with the hepatic findings, HBsAg seroclearance was associated with a lower risk of liver-related mortality (1.53 vs. 2.97 per 1000 PY; aHR = 0.46; 95% CI: 0.26–0.82; P_FDR = 0.013). HBsAg seroclearance was also associated with lower risks of extrahepatic cancer-related mortality (2.48 vs. 3.76 per 1000 PY; aHR = 0.51; 95% CI: 0.32–0.81; P_FDR = 0.011) and cardiovascular mortality (2.36 vs. 2.91 per 1000 PY; aHR = 0.56; 95% CI: 0.35–0.89; P_FDR = 0.018). When examined by specific cancer type, the strongest association was observed for digestive system cancer-related mortality (Table S5). No statistically significant association was observed for respiratory mortality (P_FDR = 0.237).

Associations and cumulative incidence accounting for competing risks

In Fine–Gray models accounting for competing events, HBsAg seroclearance remained associated with lower subdistribution hazards of cirrhosis (subdistribution HR = 0.47; 95% CI: 0.29–0.78), HCC (0.55; 0.35–0.85), liver-related mortality (0.49; 0.27–0.86), and extrahepatic cancer-related mortality (0.56; 0.35–0.89). Estimates for cardiovascular mortality and respiratory mortality were also below 1 but were less precise (Table S6).

For all-cause mortality, which has no competing event, the model-standardized 5-year cumulative incidences were 2.16% in the HBsAg seroclearance state and 4.45% in the persistent infection state (risk difference = −2.29%; 95% CI: −2.98 to −1.60), and the corresponding 10-year estimates were 6.42% and 12.46% (risk difference = −6.04%; 95% CI: −7.80 to −4.28). After accounting for competing risks, 10-year cumulative incidence was also lower in the HBsAg seroclearance state for cirrhosis, HCC, cardiovascular mortality, and extrahepatic cancer-related mortality, with corresponding risk differences of −2.09 (95% CI: −3.17 to −1.01), −2.21 (95% CI: −3.61 to −0.81), −1.01 (95% CI: −1.96 to −0.06), and −1.72 (95% CI: −2.89 to −0.55) percentage points, respectively (Table S7). Corresponding 5-year estimates showed a similar pattern.

Accounting for baseline HBV DNA and low HBV DNA levels without HBsAg seroclearance

To further address potential confounding by baseline viral burden, we performed an IPTW analysis additionally incorporating baseline HBV DNA, which improved but did not fully eliminate the between-group imbalance in HBV DNA (Fig. S2). Point estimates for most outcomes remained similar, although precision was reduced for several endpoints (Table S8). HCC showed the most notable attenuation and was no longer statistically significant (HR = 0.62; 95% CI: 0.32–1.18), suggesting that this association may partly depend on differences in baseline HBV DNA levels. By contrast, the associations with all-cause and extrahepatic cancer-related mortality remained statistically significant.

We further examined whether the observed associations could be explained by HBV DNA suppression alone by classifying follow-up into three time-varying states: HBsAg-positive/HBV DNA <100 IU/mL, HBsAg-negative/HBV DNA <100 IU/mL, and HBsAg-positive/HBV DNA ≥100 IU/mL (Table 2). Compared with the HBsAg-positive/HBV DNA <100 IU/mL state, the HBsAg-negative/HBV DNA <100 IU/mL state was associated with lower hazards of cirrhosis (HR = 0.42; 95% CI: 0.24–0.73), all-cause mortality (0.45; 0.34–0.60), extrahepatic cancer-related mortality (0.41; 0.25–0.67), HCC (0.58; 0.35–0.95), and cardiovascular mortality (0.57; 0.34–0.96). By contrast, the HBsAg-positive/HBV DNA ≥100 IU/mL state was associated with higher hazards of HCC (HR = 1.43; 95% CI: 1.03–1.98) and liver-related mortality (2.23; 1.46–3.43). These findings suggest that HBsAg seroclearance may provide additional prognostic information beyond low HBV DNA levels alone.

Table 2.

Associations between three time-varying serovirological states and clinical outcomes.

Outcome	Serovirological state	Events	Person-years	Incidence rate per 1000 PY	Model 1 HR (95% CI)	P value	Model 2 HR (95% CI)	P value
Cirrhosis	HBsAg+, HBV DNA <100 IU/mL	62.2	13,941.0	4.46	1.00 (Reference)	–	1.00 (Reference)	–
	HBsAg−, HBV DNA <100 IU/mL	17.0	8547.5	1.99	0.41 (0.24–0.72)	0.002	0.42 (0.24–0.73)	0.002
	HBsAg+, HBV DNA ≥100 IU/mL	66.0	18,080.3	3.65	0.89 (0.62–1.29)	0.552	0.90 (0.62–1.29)	0.560
HCC	HBsAg+, HBV DNA <100 IU/mL	62.7	14,061.2	4.46	1.00 (Reference)	–	1.00 (Reference)	–
	HBsAg−, HBV DNA <100 IU/mL	23.0	8547.4	2.69	0.57 (0.35–0.93)	0.025	0.58 (0.35–0.95)	0.032
	HBsAg+, HBV DNA ≥100 IU/mL	104.0	18,217.7	5.71	1.38 (1.00–1.91)	0.051	1.43 (1.03–1.98)	0.032
Decompensated cirrhosis	HBsAg+, HBV DNA <100 IU/mL	16.1	14,166.7	1.14	1.00 (Reference)	–	1.00 (Reference)	–
	HBsAg−, HBV DNA <100 IU/mL	4.0	8585.7	0.47	0.37 (0.12–1.08)	0.070	0.37 (0.12–1.11)	0.076
	HBsAg+, HBV DNA ≥100 IU/mL	16.1	18,279.2	0.88	0.83 (0.41–1.67)	0.596	0.82 (0.40–1.66)	0.582
Liver failure	HBsAg+, HBV DNA <100 IU/mL	12.3	14,237.6	0.87	1.00 (Reference)	–	1.00 (Reference)	–
	HBsAg−, HBV DNA <100 IU/mL	1.0	8595.7	0.12	0.09 (0.02–0.47)	0.005	0.08 (0.02–0.30)	<0.001
	HBsAg+, HBV DNA ≥100 IU/mL	3.8	18,308.5	0.21	0.23 (0.06–0.85)	0.027	0.24 (0.07–0.83)	0.024
All-cause mortality	HBsAg+, HBV DNA <100 IU/mL	201.0	14,249.2	14.10	1.00 (Reference)	–	1.00 (Reference)	–
	HBsAg−, HBV DNA <100 IU/mL	58.0	8597.3	6.75	0.45 (0.34–0.60)	<0.001	0.45 (0.34–0.60)	<0.001
	HBsAg+, HBV DNA ≥100 IU/mL	232.1	18,316.0	12.67	1.08 (0.88–1.33)	0.440	1.11 (0.90–1.36)	0.330
Liver-related mortality	HBsAg+, HBV DNA <100 IU/mL	30.0	14,249.2	2.10	1.00 (Reference)	–	1.00 (Reference)	–
	HBsAg−, HBV DNA <100 IU/mL	11.0	8597.3	1.28	0.59 (0.31–1.14)	0.118	0.61 (0.31–1.20)	0.155
	HBsAg+, HBV DNA ≥100 IU/mL	71.2	18,316.0	3.89	2.10 (1.37–3.23)	<0.001	2.23 (1.46–3.43)	<0.001
Extrahepatic cancer-related mortality	HBsAg+, HBV DNA <100 IU/mL	70.3	14,249.2	4.93	1.00 (Reference)	–	1.00 (Reference)	–
	HBsAg−, HBV DNA <100 IU/mL	20.0	8597.3	2.33	0.44 (0.27–0.72)	<0.001	0.41 (0.25–0.67)	<0.001
	HBsAg+, HBV DNA ≥100 IU/mL	61.2	18,316.0	3.34	0.78 (0.53–1.14)	0.203	0.74 (0.51–1.08)	0.123
Cardiovascular mortality	HBsAg+, HBV DNA <100 IU/mL	53.3	14,249.2	3.74	1.00 (Reference)	–	1.00 (Reference)	–
	HBsAg−, HBV DNA <100 IU/mL	18.0	8597.3	2.09	0.54 (0.32–0.89)	0.016	0.57 (0.34–0.96)	0.035
	HBsAg+, HBV DNA ≥100 IU/mL	57.4	18,316.0	3.14	1.11 (0.73–1.68)	0.629	1.18 (0.78–1.80)	0.430
Respiratory mortality	HBsAg+, HBV DNA <100 IU/mL	10.0	14,249.2	0.70	1.00 (Reference)	–	1.00 (Reference)	–
	HBsAg−, HBV DNA <100 IU/mL	4.0	8597.3	0.47	0.56 (0.17–1.82)	0.333	0.55 (0.16–1.89)	0.343
	HBsAg+, HBV DNA ≥100 IU/mL	9.0	18,316.0	0.49	0.78 (0.29–2.08)	0.622	0.88 (0.32–2.46)	0.810

HBsAg, hepatitis B surface antigen; HBsAg+, HBsAg-positive; HBsAg−, HBsAg-negative; HCC, hepatocellular carcinoma; HR, hazard ratio; PY, person-years.

Events and person-years were weighted using the original propensity score-matching weights. Therefore, event counts may be non-integer. Model 1 was adjusted for area, sex, baseline antiviral treatment status, and baseline age. Model 2 was additionally adjusted for education level, marital status, drinking, and smoking status, hypertension, family history of HBV and BMI. P values were not adjusted for multiple comparisons.

Population attributable fractions associated with persistent infection

Based on the above association models, Fig. 4 shows the model-based PAFs for major adverse liver events and mortality after PSM and IPTW. After full adjustment, the PAFs associated with persistent infection were 48.1% for cirrhosis, 46.9% for liver-related mortality, 42.3% for all-cause mortality, and 40.6% for HCC after PSM. The PAFs for extrahepatic cancer-related mortality and cardiovascular mortality were 40.0% and 35.3%, respectively. The corresponding IPTW estimates were generally comparable. Sex-stratified analyses after PSM showed consistently higher PAF estimates in men across all evaluated outcomes, whereas wider confidence intervals were observed in women (Fig. S3).

Fig. 4.

Population attributable fractions of adverse outcomes after propensity score matching and inverse probability of treatment weighting. Abbreviations: HCC, hepatocellular carcinoma; IPTW, inverse probability of treatment weighting; PAFs, Population attributable fractions; PSM, propensity score matching. The HBsAg seroclearance group was used as the reference. Error bars indicate 95% confidence intervals. Outcomes not shown in the figure had insufficient event numbers to allow reliable estimation of model-based PAFs.

Restricted mean survival time associated with HBsAg seroclearance

Fig. 5 presents the Simon–Makuch survival curves for all-cause mortality using age as the time scale. Across nearly the entire age range, the survival probability remained consistently higher in the HBsAg seroclearance group than in the persistent infection group (P < 0.001). The divergence between the two curves became more pronounced after approximately 60 years of age and widened substantially in older age.

Fig. 5.

Simon-Makuch curves for all-cause mortality using age as the time scale. Abbreviations: HBsAg, hepatitis B surface antigen. P values shown in the Simon–Makuch plots were obtained from the score test of a univariate Cox model.

The estimated RMST was 83.21 years (95% CI: 81.16–85.47) in the HBsAg seroclearance state compared with 78.04 years (95% CI: 76.47–79.39) in the persistent infection state, corresponding to a 5.17-year difference in RMST within the observed age range (Table S9). Sex-stratified analyses showed that the RMST difference was larger in men (Fig. S4).

Sensitivity analysis and subgroup analysis

Sensitivity analyses were conducted to assess the robustness of the primary findings to several potential sources of bias and alternative analytical assumptions (Table S3). Overall, no directional changes were observed, and the effect estimates were generally consistent with the primary findings (Fig. S5).

Repeating the time-dependent Cox analyses in the full cohort without PSM yielded similar estimates for cirrhosis, HCC, decompensated cirrhosis, all-cause mortality, liver-related mortality, and extrahepatic cancer-related mortality, although estimates for less frequent outcomes were less precise (Table S10). Additional adjustment for liver-function markers had little influence on the results. After further adjustment for metabolic factors, point estimates remained below 1 across all estimable outcomes; however, precision was reduced because fewer events were available after shifting the analytical baseline to 2018 (Table S11).

Landmark analyses addressing uncertainty in exposure timing also showed broadly similar patterns. At the 2012 landmark, HBsAg seroclearance remained associated with lower risks of cirrhosis, HCC, all-cause mortality, and liver-related mortality, whereas at the 2016 landmark, significant associations were observed for cirrhosis, HCC, all-cause mortality, and extrahepatic cancer-related mortality (Table S12). A 1-year lag analysis produced comparable estimates, with no directional changes across outcomes (Table S13). Excluding participants with subsequent HBsAg seroreversion and/or HBV DNA reactivation or restricting the analysis to those attending at least five follow-up visits also had little influence on the estimates (Tables S14 and S15).

Stratified analyses were broadly consistent across age and baseline HBV DNA strata, with no significant interactions (all P for interaction >0.05). Significant sex interactions were observed for all-cause mortality (P = 0.005) and liver-related mortality (P = 0.011), with stronger inverse associations in men. For cirrhosis, HCC, and cardiovascular mortality, point estimates also suggested stronger inverse associations in men, although the corresponding interaction tests were not statistically significant (Figs. S6–S8). Due to the limited numbers of events, we also found no statistical evidence of interaction between antiviral treatment history and HBsAg seroclearance (Fig. S9).

Discussion

Based on a multicenter CHB cohort with long-term follow-up, this study systematically evaluated the associations of sustained HBsAg seroclearance with HBV DNA <100 IU/mL with liver-related outcomes, all-cause mortality, and cause-specific mortality. This serovirological state was associated with lower risks of HCC, cirrhosis, extrahepatic cancer-related mortality, and cardiovascular mortality. Notably, after accounting for competing risks, the associations remained evident for multiple hepatic and extrahepatic adverse outcomes. Together, these findings provide a broader view of the long-term prognostic relevance of HBsAg seroclearance by encompassing not only hepatic outcomes, but also major extrahepatic risks and overall survival.

The associations with lower risks of HCC and cirrhosis are consistent with extensive existing evidence. Several cohorts and meta-analyses have reported substantially lower risks of HCC and cirrhosis following HBsAg clearance.^10,14,15 Moreover, clinical studies have demonstrated that sustained virological suppression is associated with reduced progression of liver fibrosis and lower risks of HCC recurrence and HCC-related death.^16,17 Beyond these established hepatic associations, a notable finding of the present study was the association between sustained HBsAg seroclearance and lower risks of extrahepatic cancer-related and cardiovascular mortality. Previous epidemiological studies have linked chronic HBV infection to several extrahepatic malignancies, including gastric, colorectal, pancreatic, and hematologic cancers.^18,19 By contrast, associations between chronic HBV infection and cardiovascular disease incidence have been less consistent, although some studies have suggested a possible association with cardiovascular mortality.^20,21 Notably, these associations were consistently observed across the Fine–Gray, three-state, and multiple sensitivity analyses, supporting the robustness of the overall findings. These findings may be particularly relevant to the long-term risk stratification in individuals with CHB, in whom comorbid conditions are becoming increasingly common.^22,23

Several biological hypotheses may provide context for these extrahepatic associations. For extrahepatic cancer-related mortality, persistent HBV antigen exposure and associated immune dysregulation may impair tumor immune surveillance, whereas HBsAg seroclearance may reflect a more favorable long-term viral-host trajectory relevant to cancer development and progression.24, 25, 26 For cardiovascular mortality, the higher prevalence of carotid plaques and subclinical atherosclerosis observed in patients with CHB suggests a possible link between active HBV disease and atherosclerotic burden.²⁷ More broadly, HBsAg seroclearance is accompanied by a marked reduction in viral antigen burden and partial restoration of HBV-specific immune responses.28, 29, 30 The resulting changes in chronic immune activation and systemic inflammatory status may represent shared, nonspecific pathways potentially relevant to extrahepatic cancer-related, cardiovascular, and more broadly all-cause mortality.^31,32 However, these extrahepatic associations should be regarded as hypothesis-generating rather than as evidence of a direct systemic effect of HBsAg loss, and require confirmation in independent cohorts and further mechanistic investigation.

We further explored whether HBsAg seroclearance provides prognostic information beyond HBV DNA suppression alone. In both the IPTW analysis additionally incorporating baseline HBV DNA and the dynamic serological-state analyses, HBsAg seroclearance remained associated with lower risks of several outcomes beyond those observed with isolated HBV DNA suppression, consistent with previous cohort evidence.³³ Conversely, compared with persistent viral replication, isolated HBV DNA suppression was not associated with lower risks of extrahepatic mortality. These findings suggest that even among individuals with effective viral suppression, a conventional clinical treatment goal, progression toward HBsAg seroclearance or functional cure may provide additional prognostic value.

Our study also provided model-based estimates of attributable burden and survival differences to quantify the potential population health relevance of HBsAg seroclearance. The estimated PAF associated with persistent infection was approximately 42% for all-cause mortality, while the RMST analysis showed a 5.17-year between-state difference within the observed age range. Together, these estimates highlight the potential population-health relevance of the observed prognostic differences between the two serovirological states.

Sex-stratified analyses suggested potential heterogeneity in the prognostic associations of HBsAg seroclearance. Inverse associations were generally stronger in men, with significant interactions for all-cause and liver-related mortality and similar directional patterns for cirrhosis, HCC, and cardiovascular mortality. Sex-specific PAF and RMST estimates also suggested numerically larger attributable fractions and survival differences in men. These findings raise the possibility that the long-term prognostic relevance of HBsAg seroclearance may differ by sex, consistent with previous evidence that men with CHB generally experience less favorable long-term clinical outcomes than women.³⁴

It should be noted that most HBsAg seroclearance events in this cohort occurred spontaneously. Therefore, our findings primarily reflect the long-term prognosis associated with such events. Spontaneous and treatment-induced HBsAg seroclearance (or functional cure) may occur in patients with different characteristics, disease stage, and underlying biological processes.^35,36 Accordingly, the associations observed in our cohort may partly reflect a favorable long-term viral-host profile and overall health status among individuals who undergo spontaneous seroclearance, as well as residual confounding from unmeasured factors such as socioeconomic status and healthcare utilization. Although previous studies have reported broadly similar associations with HCC and composite clinical outcomes following spontaneous and nucleos(t)ide analogue-induced HBsAg seroclearance, these findings require further validation in independent cohorts, particularly those enriched for treatment-induced HBsAg seroclearance or functional cure.³⁶

This study has several strengths, including a stringent longitudinal definition of the serovirological exposure, requiring sustained HBsAg seroclearance with HBV DNA <100 IU/mL on at least two consecutive assessments, long-term prospective follow-up with low loss to follow-up and centralized laboratory testing, and time-dependent modeling with extensive sensitivity analyses. By extending evaluation beyond hepatic outcomes to cause-specific mortality, PAF, and RMST, the study also broadens the clinical relevance of HBsAg seroclearance.

Several limitations should be considered. Serovirological assessments were conducted at relatively long intervals, introducing uncertainty in the timing of HBsAg seroclearance, and the HBV DNA assay LLOQ of 100 IU/mL may have classified some low-level viremia as below the quantification threshold. In addition, residual confounding remains possible because detailed alcohol-consumption data were unavailable during early follow-up and metabolic measurements were insufficient to consistently ascertain metabolic dysfunction-associated steatotic liver disease. Fine–Gray analyses should be interpreted cautiously because, with a time-varying serovirological exposure, the assumed exposure status after a competing event cannot be verified. Moreover, sparse events for liver failure and respiratory mortality resulted in imprecise estimates that should be interpreted cautiously. Finally, as the study was based on a single provincial, community-based cohort, the generalizability of the findings to other regions, ethnic groups, and clinical populations requires further validation.

In conclusion, this study shows that sustained HBsAg seroclearance with very low HBV DNA was associated with lower risks of hepatic outcomes, all-cause mortality and several cause-specific mortality outcomes, together with favorable model-based estimates of attributable burden and RMST. These findings highlight sustained HBsAg seroclearance as an informative marker for long-term hepatic and extrahepatic risk stratification in CHB.

Contributors

Study concept and design: Ci Song. Cohort follow-up conduct and acquisition of data: Xiangjun Zhai, Jie Jiang, and Liguo Zhu. Data curation: Jiao Qian, Chu He, Xiaoxiao Chen, Siyu Wang, Yuxiang Yan, Qingyu Jiang, and Huanjie Bai. Statistical analysis: Ying Zhang, Wentao Yao. Writing—original draft: Ying Zhang, Ci Song. Funding acquisition: Zhibin Hu, Hongbing Shen, Xiangjun Zhai, and Ci Song. Writing—review & editing: Xiangjun Zhai, Ci Song. Data validation: Tao Jiang, Ci Song, and Xiangjun Zhai. Ci Song was responsible for the decision to submit the manuscript.

Data sharing statement

Individual-level data from the HEPCARE cohort are not publicly available because of participant confidentiality and ethical approval requirements. However, de-identified data may be made available to researchers for non-commercial academic purposes upon reasonable request. Requests should be directed to the corresponding authors, Dr. Xiangjun Zhai (jszxj@jscdc.cn) or Dr. Ci Song (songci@njmu.edu.cn).

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used GPT-5.2 (OpenAI, USA) in March 2026 for language polishing and refinement. After using this service, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Declaration of interests

The authors do not have any conflict of interest.