Virological Changes of Chronic Hepatitis B Patients with Minimally Elevated Levels of Alanine Aminotransferase: A Meta-Analysis and Systematic Review

Background Chronic hepatitis B (CHB) patients with normal or minimally increased levels of alanine aminotransferase (ALT) are still at the risk of hepatocellular carcinoma, cirrhotic events, and mortality. However, there is a debate over the initiation of antiviral treatment for these patients. This systematic review and mate-analysis aimed to explore this problem. Methods MEDLINE (PubMed), EMBASE, Cochrane Central Register of Controlled Trials, and Web of Science databases were systematically searched for retrieving relevant studies with risk ratios (RRs) or risk differences (RDs) for virological changes between antivirus-treated and no antivirus-treated CHB patients with ALT levels less than two-fold of the upper limit of normal. Retrieved data ranged from January 1990 to October 2020. Results Of 6783 abstracts screened, 9 studies met the criteria for inclusion in the systematic review and had a low risk of bias. Among studies that were involved in the meta-analyses, it was found that the rates of HBsAg loss (RR = 12.22, 95% confidence interval (CI): 4.28–34.95, P < 0.001), HBsAg seroconversion (RR = 19.90, 95% CI: 2.75–144.09, P=0.003), and undetectable HBV DNA (RR = 11.89, 95% CI: 2.44–57.89, P=0.002) were both higher in the antiviral treatment group compared with placebo or no treatment group. Subgroup analysis suggested that patients who received interferon (IFN)-based therapy were more inclined to achieve HBsAg loss (P=0.010), HBsAg seroconversion (P=0.020), and HBeAg loss (P=0.002). Conclusion From a sizable population, it was revealed that CHB patients with normal or minimally increased levels of ALT could benefit from the antiviral therapy, especially those who received IFN-based treatment.


Introduction
Chronic hepatitis B (CHB) virus infection remains a worldwide health burden, aficting approximately 257 million people [1]. Up to 40% of untreated CHB patients progress to cirrhosis, and these patients are at the risk of developing decompensated cirrhosis and hepatocellular carcinoma (HCC) [2,3]. Studies have shown that a signifcant proportion (40%) of CHB patients have normal or minimally increased levels of alanine aminotransferase (ALT), while they may still be at the risk of HCC, cirrhotic events, and mortality in patients with CHB-related cirrhosis [4][5][6][7]. Hence, timely treatment can delay disease progression and improve the prognosis.
At present, it is still an area of ongoing controversy on whether CHB patients with mildly raised ALT levels (more than one-fold, while less than two-fold of the upper limit of normal (ULN)) should be treated. Te American Association for the Study of Liver Diseases (AASLD, updated in 2018) [8] and the Asian-Pacifc guideline (updated in 2015) [9] concluded that the threshold of the ALT level for initiating antiviral therapy is no less than two-fold of the ULN. Conversely, the European Association for the Study of the Liver (EASL, updated in 2017) recommended treatment for patients with a serum ALT concentration of more than the ULN and hepatitis B virus (HBV) DNA >2000 IU/L [1]. In the current Chinese guideline, patients with the ALT level more than the ULN and detectable HBV DNA should receive treatment [10]. Terefore, the agreement on the initiation of antiviral therapy was not reached worldwide. In a retrospective cohort study with involvement of 3624 untreated CHB patients that was conducted in South Korea, it was revealed that patients whose conditions did not meet therapeutic indications for EASL, AASLD, and other authoritative guidelines still had a cumulative 5-year HCC incidence of 2.1-3.2% [11]. At present, the notion that whether this group of CHB patients should be treated remains controversial and whether they can beneft from the antivirus therapy is still undiscovered.
Hepatitis B surface antigen (HBsAg) loss with or without HBsAg seroconversion is regarded as a functional cure and the ultimate endpoint for CHB therapy [1,8,9]. For patients who achieved partial cure, HBV DNA is undetectable in their serum [12]. Te present study aimed to identify the proportion of HBsAg loss with or without the seroconversion to hepatitis B surface antibody (HBsAb) in CHB patients with only normal or minimally elevated ALT levels. Te secondary objective was to identify the proportions of undetectable HBV DNA, hepatitis B e antigen (HBeAg) loss, and HBeAg seroconversion.

Search Results.
Totally, 6783 articles were retrieved through searching in the four databases mentioned below, of which 512 articles were eliminated because of duplicate publication. After screening of titles and abstracts, the full texts of 22 studies were downloaded to assess their eligibility. Finally, 7 articles [13][14][15][16][17][18][19][20] and 2 abstracts [21] which met the eligibility criteria are included in this meta-analysis (Figure 1). Te details of the study selection are shown in Table 1. During the analysis, one of the abstracts was published and our collected data were renewed subsequently [18].

Study and Patients' Characteristics.
In the primary analysis, there were a total of 898 patients in the 9 included studies. Most of the included studies were randomized controlled trials (RCTs), except for 2 studies [13,15], which did not assign patients randomly. In addition to one study published in 2002 [20], the year of publication of the other studies was between 2014 and 2021. Of note, patients were treated exclusively with nucleotide analogs (tenofovir) in Hsu et al. [18], nucleoside analogs (entecavir) in Tseng et al. [17], nucleotide analogs plus nucleoside analogs (tenofovir and emtricitabine) in Chan et al. [16], and IFN combined with NAs in 5 studies. In 2 studies [19,20], two or more diferent regimens were used in the experimental group, whereas only one study [20] reported the outcome of each regimen. In 3 studies [13,14,21], patients in the experimental group were given IFN at the beginning, and in 2 studies [13,14], they additionally received NAs based on their conditions. It was found that 2 studies [17,18] enrolled both HBeAgpositive and HBeAg-negative patients, regardless of their HBeAg status, in which enrolled subjects were grouped in one research [17]. However, the remaining 7 studies involved either in HBeAg-negative or HBeAg-positive patients. As for 4 studies that enrolled HBeAg-positive patients, one study [20] reported HBV DNA level in pg/mL, and the others [14][15][16] reported in IU/mL and enrolled patients with high viral load (>20000 IU/mL). Furthermore, regarding 3 studies [13,19,21] that enrolled only HBeAgnegative individuals, the level of HBV DNA ≤20000 IU/ml (all reported in IU/mL unit) was covered in the inclusion criterion.
Most of the studies mainly enrolled male adult patients. Nevertheless, 2 HBV-infected children with immunetolerant characteristic and HBV postpartum women were enrolled, respectively. Te majority of patients were Asian except for those in 3 studies [16,19,20]. Besides, 2 trials measured the outcomes at 192 weeks [16] and 144 weeks [18], and 7 studies performed measurement within 96 weeks.

Results of Meta-Analysis.
Of the 9 studies, 6, 5, 4, 5, and 6 studies reported the outcomes of HBsAg loss, HBsAg seroconversion, HBeAg loss, HBeAg seroconversion, and undetectable HBV DNA, respectively. Stratifed by the treatment strategy, treatment duration, or baseline parameters, subgroup analysis was performed to assess the association between antiviral therapy and each endpoint in the meta-analysis using a random-efects model.

HBsAg Loss.
Compared with the control group, antiviral therapy was associated with a signifcantly higher incidence of HBsAg loss (RR � 12.22, 95% CI: 4.28-34.95, P < 0.001) (Figure 2(a)). Ten, it was attempted to explore how the characteristics of viral replication could afect HBsAg loss, and patients were stratifed by the status of HBeAg and HBV DNA. Both subjects with HBeAg-positive and HBV DNA >20000 IU/mL (RR � 15.11, 95% CI: 2.08-109.69, P � 0.007) and subjects with HBeAg-negative and HBV DNA ≤20000 IU/mL (RR � 9.93, 95% CI: 2.84-34.67, P < 0.001) exhibited a superior efect of HBsAg loss than that of the control one. Te diference between the two subgroups (subjects with HBeAg-positive and HBV DNA>20000 IU/ mL subgroup and subjects with HBeAg-negative and HBV DNA ≤20000 IU/mL subgroup) was also not signifcant (P � 0.730) (Figure 2(b)). Besides, in studies with IFN treatment, antiviral therapy was associated with a signifcantly higher incidence of HBsAg loss (RR � 11.19, 95% CI: 3.89-32.22, P < 0.001). In addition, there was no event of HBsAg loss in the interferon-free studies, regardless of antiviral or viral event (Figure 2(c)). When the outcomes were calculated as RD, there was a remarkable diference in HBsAg loss between IFN-based subgroup and IFN-free subgroup (P � 0.010) (Supplementary Figure 1). Tese results indicated that IFN therapy had a higher incidence of HBsAg loss in CHB patients. Furthermore, stratifed by treatment strategy, both combination therapy (RR � 9.83, 95% CI: 1.31-74.03, P � 0.030) and monotherapy (RR � 8.00, 95% CI: 1.11-57.49, P � 0.040) exhibited a superior efect of HBsAg loss than that of the control one. However, the diference between the subgroups (monotherapy subgroup and combination therapy subgroup) was not signifcant (P � 0.890) (Figure 2(d)). In conclusion, patients with antiviral treatment were inclined to have the clearance of HBsAg.

HBsAg Seroconversion.
Compared with nonantiviral therapy, antiviral therapy was associated with a signifcantly higher incidence of HBsAg seroconversion (RR � 19.90, 95% CI: 2.75-144.09, P � 0.003) (Figure 3(a)). Among 5 studies that reported the events of HBsAg seroconversion, all events happened in the combination group. In studies with IFNbased treatment, antiviral therapy was associated with a signifcantly higher incidence of HBsAg seroconversion (RR � 15.75, 95% CI: 2.19-113.47, P � 0.006). In contrast, in studies that adopted IFN-free regimen, no event of HBsAg seroconversion occurred (Figure 3(b)). When the outcomes were calculated as RD, there was a signifcant diference between IFN-based subgroup and IFN-free subgroup (P � 0.020) (Supplementary Figure 2). Collectively, patients with antiviral therapy, especially those undergoing IFN treatment, were more inclined to have HBsAg seroconversion.
2.6. HBeAg Loss. Compared with nonantiviral treatment, antiviral therapy was associated with no signifcant diference in the rate of HBeAg loss (RR � 1.68, 95% CI: 0.14-19.67, P � 0.680; with a noticeable heterogeneity I 2 � 75%) (Figure 4(a)). Terefore, sensitivity analysis was performed to fnd the source of heterogeneity. Te results showed that with the removal of Lu's study (2014), the heterogeneity could reduce to a degree where it was not signifcant (Supplementary Figure 3). In studies that involved IFN-based treatment, antiviral therapy was associated with a signifcantly higher incidence of HBeAg loss (RR � 23.86, 95% CI: 3.03-187.80, P � 0.003). In contrast, in studies that adopted IFN-free regimen, there were no signifcant diference between the two groups in HBeAg loss (RR � 0.47, 95% CI: 0.12-1.81, P � 0.280). Besides, the difference between the subgroups (IFN-based subgroup and IFN-free subgroup) was statistically signifcant (P � 0.002) (Figure 4(b)). Both combination therapy (RR � 8.66, 95% CI: 0.34-222.33, P � 0.190; with a noticeable heterogeneity I 2 � 79%) and monotherapy (RR � 0.80, 95% CI: 0.17-3.86, P � 0.780; with a noticeable heterogeneity I 2 � 60%) exhibited the same efect of HBsAg loss than that of the control one. In addition, the diference between the two subgroups (monotherapy subgroup and combination therapy subgroup) was not signifcant (P � 0.200) (Figure 4(c)). To further investigate the association between therapy duration and HBeAg loss, subgroup analysis was conducted in short-term (≤96 weeks) and long-term (>96 weeks). Although a signifcant diference was found between the two subgroups (P � 0.030), neither the long duration subgroup nor the other subgroup showed a signifcant diference between the antiviral therapy group and control group (Figure 4(d)). Collectively, only patients who underwent IFN-based antiviral therapy were more inclined to have HBeAg loss, and patients may not beneft from a long-term treatment.

HBeAg Seroconversion.
Compared with nonantiviral treatment, antiviral therapy was associated with no significant diference in the rate of HBeAg seroconversion (RR � 1.66, 95% CI: 0.32-8.60, P � 0.540; with a noticeable heterogeneity I 2 � 53%) ( Figure 5(a)). In studies that adopted IFN-based treatment, antiviral therapy was associated with        Subgroup analysis showed no signifcant diference between IFN-based subgroup and IFN-free subgroup (P � 0.070) ( Figure 5(b)). Both combination therapy (RR � 2.75, 95% CI: 0.68-11.00, P � 0.150) and monotherapy (RR � 0.71, 95% CI: 0.17-2.95, P � 0.630) exhibited the same efect of HBsAg loss compared with the control one. Moreover, the diference between the two subgroups (monotherapy subgroup and combination treatment subgroup) was not signifcant (P � 0.180) ( Figure 5(c)). In addition, stratifed by treatment duration, subgroup analysis showed that there was a signifcant diference between the two subgroups (long-term subgroup and short-term subgroup) (P � 0.006). Only in the short-term subgroup, antiviral therapy exhibited a superior efect of HBeAg seroconversion than that of the control one (RR � 6.34, 95% CI: 1.51-26.52, P � 0.010) ( Figure 5(d)). Generally, only patients who underwent IFN-based antiviral therapy were more inclined to have HBeAg seroconversion. Besides, the extension of the treatment duration did not promote the occurrence of HBeAg seroconversion.

Discussion
In this systematic review, we evaluated the proportion of HBsAg loss with or without the seroconversion to HBsAb in CHB patients with only normal or minimally elevated ALT levels. It was revealed that antiviral therapy increased the rate of achieving HBsAg loss, HBsAg seroconversion, and undetectable HBV DNA compared with no or placebo treatment. IFN exhibited a more important role in HBsAg loss with or without HBsAg seroconversion than NAs. Besides, no synergism was found between IFN and NAs in virological response.
Tere was a consensus that antiviral therapy should be actively administered in patients with elevated ALT levels (>2 ULN), cirrhosis, and liver cancer, while there was a controversy over antiviral therapy in CHB patients with normal or mildly elevated ALT levels. Studies have suggested that CHB patients with normal or low ALT levels have a certain histological damage, and they may eventually develop liver failure, cirrhosis, or HCC [22,23]. A retrospective cohort study found that long-term antiviral therapy reduced the incidence of liver cancer in patients with CHB, with no signifcant association with ALT levels [24]. In addition, a meta-analysis indicated that the rate of fbrosis is more than 40% in patients with CHB and minimal increased ALT levels [25]. Combined with the results of our study, it is likely to conclude that given the possibility of liver disease progression and the available benefts of antiviral therapy, the threshold for the initiation of treatment must be individualized.
For CHB patients with normal or minimally increased ALT levels, the overall treatment goal is to inhibit and even eliminate hepatitis B virus infection, alleviate necrosis and infammation, and suppress the disease progression. IFN and NAs are two efective antiviral drugs for CHB patients to delay the disease progression and to improve the longstanding prognosis. A meta-analysis included 24 studies and found that combination therapy took an advantage on promoting HBsAg loss [26]. Another research demonstrated that NAs combined with IFN strategy could improve efcacy on HBeAg seroconversion compared to monotherapy with NAs [27]. However, our study found the equivalence of combination therapy and monotherapy in achieving virological response. Considering the high heterogeneity and small sample size, additional RCTs are required to verify the synergy between IFN and NAs on virological response.
NAs can suppress the replication of hepatitis B virus effectively, while IFN has dual functions on viral inhibition and immunomodulation. In clinical practice, for the vast majority of CHB patients, long-term NA therapy is the top choice for them, apart from the few achieved HBsAg clearance or the conversion to HBsAb. IFN, compared with NAs, is inferior in preventing the virus from replicating itself, while it is superior in achieving HBeAg seroconversion and HBsAg loss [28,29]. Te present study revealed that CHB patients with ALT levels <2 ULN can beneft from IFN-based treatment in achieving primary outcomes. Besides, there was a signifcant diference in HBeAg clearance, and it could be due to immunomodulatory efects of IFN. Previous studies have demonstrated that IFNbased treatment was associated with greater sustained virological and serological responses and a higher chance of HBsAg loss, and the induced functional cure was durable [30,31]. However, our research indicated that IFN-based therapy was not superior to IFN-free therapy in achieving HBeAg seroconversion and undetectable HBV DNA, which might be attributed to the same reason as mentioned above.
A previous prospective cohort study revealed that the rate of HBeAg seroconversion increased along with the prolonged treatment [32]. In the subgroup analysis stratifed by the treatment duration, short-term treatment, compared with long-term treatment, could improve the efcacy of HBeAg loss and HBeAg seroconversion. Te possible reason for the diference was that all the studies included in the long-term treatment subgroup adopted IFN-free strategy and tenofovir treatment, while most of the studies in the short-term treatment subgroup adopted IFN-based strategy, and the diference might stem from IFN treatment.
Te present study exhibited the following defciencies. Firstly, several factors might infuence patients' response to the therapy, including demographic characteristic, HBV genotype, HBeAg status, HBV DNA levels, and with or without history of undergoing antiviral treatment. However, the limitation of the available studies impeded further subgroup analysis to explore the infuences of the abovementioned factors. Secondly, because the unknown approach for measuring ALTand ULN was not mentioned in some articles, and it was infeasible to evaluate the efcacy of antiviral therapy for CHB patients with normal ALT level. Tirdly, high heterogeneity and limited research studies in each subcohort restricted us from exploring optimal antiviral strategies for CHB patients with ALT level <2 ULN. Fourthly, the small quantity of the available studies and certain methodological limitations, which were related to randomization and blinding processes, diminished the quality of our study.
In general, for CHB patients with ALT level <2 ULN, antiviral therapy could signifcantly improve the HBsAg loss rate. Trough the immune modulatory function of IFN and our results, IFN is highly recommended to CHB patients with ALT level <2 ULN for achieving the objective of functional treatment.

Methods
Tis study was conducted and reported according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. Te study protocol was registered in the PROSPERO database (Registration No. CRD42020209639).

Data Sources and Searches.
On October 3, 2020, the relevant studies were retrieved by searching four English language databases, such as MEDLINE (PubMed), EMBASE, Cochrane Central Register of Controlled Trials, and Web of Science, from January 1990 to October 2020. Te following strategy was applied: ((("Hepatitis B"(Mesh)) OR ("Hepatitis B virus"(Mesh))) OR ("Hepatitis B, Chronic"(Mesh))) AND ((alanine aminotransferase (Title/Abstract)) OR ("Alanine Transaminase"[Mesh))). To avoid missing potentially relevant articles, the additional citations of all retrieved articles were also searched manually.

Study Selection.
Studies fulflling the following specifc criteria were considered for inclusion in the primary analysis: (i) patients: both infected with hepatitis B virus and with ALT level ≤2 ULN (ULN shall be subjected to each article reported); (ii) treatment strategy: patients were divided into antiviral therapy group (treated with nucleos(t)ide analogs (NAs) or interferon (IFN)) and control group (no treatment or placebo); and (iii) outcomes: including rates of undetectable HBV DNA, HBeAg loss, HBeAg seroconversion, HBsAg loss, or HBsAg seroconversion. We excluded studies where study cohorts included patients with (i) coinfection with hepatitis C virus, hepatitis D, or human immunodefciency virus; (ii) remarkable alcohol abuse; or (iii) other liver diseases. Ten, four of the authors (C. X. Y, Z. X. R, W. H. W, and Z. B. X) reviewed the articles independently. Information about the source or author of the report was not blinded to the investigators. However, when a decision was made on which studies should be included, two observers (C. X. Y and Z. B. X) analyzed in parallel and without interference. In case of the occurrence of any disagreement, they conferred with other two authors (W. H. W and Z. X. R). Te fnal decision was confrmed by four observers and the senior author (X. C). Te decision to include these data was not infuenced by the results of the recruited study.

Data Extraction and Study
Quality. Te modifed Newcastle-Ottawa Scale was applied to assess the risk of bias in cohort studies by two reviewers (C. X. Y and Z. B. X) separately. Te included studies in the meta-analysis were rated on a scale of zero to 9 points. Only studies that scored six or higher points were regarded as good quality. For each study considered for being included, the following data were fetched from them: (i) study characteristics (frst author's name, year of publication, geographic locale, study design, sample size, number of patients included in analysis, and quality score); (ii) patients' demographics (age, gender, HBeAg-positive rate, baseline HBV DNA level, and baseline ALT level); (iii) inclusion and exclusion criteria; (iv) treatment details (i.e., antiviral agent, treatment duration, lower limit of detection of HBV DNA, and data collected at timepoints).

Statistical Analysis.
Using the Mantel-Haenszel method, the pooled outcome was calculated as risk ratio (RR) or risk diference (RD) with 95% confdence interval (CI) for various indicators reported as dichotomous variables, involving undetectable HBV DNA, HBsAg loss, HBsAg seroconversion, HBeAg loss, and HBeAg seroconversion. Te χ 2 test was applied for the exploration of heterogeneity, and statistical signifcance was set to P < 0.05. To quantify heterogeneity, I 2 statistic was taken, with a maximum of 50% that was defned as low heterogeneity, while >50% indicated signifcant heterogeneity. If there was a high degree of heterogeneity, the random-efects model was used, otherwise, the fx-efects model was utilized. Te Review Manager 5.2 software (Cochrane Collaboration, Oxford, UK) was used to calculate RRs and draw the forest plots to display the results of the meta-analysis. Te squares around the estimates were proportional to the weights used in the meta-analysis, with horizontal lines that represented 95% CI. Sensitivity analysis was performed by leave-one-out analysis to test the infuences of individual studies on aggregate estimates. Te following subgroup analyses were carried out, wherever possible: therapeutic strategy (combination therapy (NAs and IFNs), NAs-monotherapy, and IFN-monotherapy; patients who were treated with two or more types of NAs were divided into the monotherapy group), treatment duration, baseline HBeAg status, and baseline HBV DNA level. Once a signifcant diference was found between the subgroups (test for interaction, P < 0.05), the results were reported separately. A formal test was also performed for subgroup interactions.

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon request.

Disclosure
Xiyao Chen and Xingrong Zheng are the co-frst authors of the article.

Supplementary Materials
Supplementary Figure 1