Efficacy and safety of pregnant women with chronic hepatitis B receiving nucleos(t)ide analogs therapy: a network meta-analysis

Objective Reducing mother-to-child transmission(MTCT)of hepatitis B virus(HBV) is one of the key ways to eliminate hepatitis B. Although studies have demonstrated the effectiveness of oral nucleos(t)ide analogs(NAs),drug’s comparisons are lacking. The network meta-analysis aims to comprehensively compare and summarize the efficacy and safety of the three drugs (Lamivudine(LAM), Telbivudine(TBV) and Tenofovir(TDF)), providing a basis for drug selection. A We performed pair-wise meta-analysis and Bayesian of

mothers with high viral loads have received NAs therapy to avoid MTCT (17). We believe that the previous excellent meta-analysis about antiviral drugs for MTCT of HBV provide a good reference for clinical practices.In fact, due to the late time to market of TDF and the lack of head-to-head studies between two drugs (especially with TDF), it is not good for the choice of drugs in our clinical practices. In this study, we included newly published studies (especially about TDF), as well as more studies on head-to-head comparisons between LAM, TBV and TDF. More notably, for the problem that traditional meta-analysis cannot compare two or more drugs (unless they are compared in the same study), we conducted a network meta-analysis to compare drugs. Through comprehensive analysis of the included studies, indirect data on drug pairwise comparison are obtained. These data can provide the basis for our choice of drugs and provide evidence for pregnancy management of pregnant women with CHB.

Search Strategy
These relevant published studies of the analysis were comprehensively and independently searched by two authors. They searched PubMed(Medline), Web of Science, the Cochrane Library, EMBASE, CNKI and SinoMed up to December 2019, based on the combination of medical subject headings (MeSH) and key words including "lamivudine", "telbivubine", "tenofovir", "hepatitis B", "HBV" and "mother-to-child transmission" (Supplementary Table 1). We also manually retrieved other potentially eligible studies such as reference lists of included studies, systematic reviews and relevant guidelines. Our search is not restricted by language, publisher and publication status.

Inclusion and Exclusion Criteria
Randomized controlled trials(RCTs) and non-randomized controlled trials(NRCTs) are included. In these studies, pregnant women who receive one of the three NAs during pregnancy are both HBsAg and hepatitis e surface antigen (HBeAg) positive and their HBV DNA ≥ 2 × 10 5 IU/mL. We excluded patients who were co-infected with HCV, HDV, and HIV, and those who had received α-INF or other antiviral treatment before six months of pregnancy were also excluded. Two independent reviewers use uniform inclusion and exclusion criteria, reconciling by a third independent reviewer if they have disagreements. A total of two rounds of screening were carried out, all the studies being screened based on the title and abstract,then the full text. For studies excluded if they didn't meet the criteria, we record the reasons for exclusion.

Data Extraction and Outcomes
Data extraction was tested using a pre-designed data extraction format, the two researchers used the same standard to extract the data of interest in the included studies after adjustment. Similarly, any differences were reconciled by consensus or by a third researcher. We were mainly interested in the following outcomes: name of the first author, year of publication, type of study, country, patients' baseline characteristics, maternal serum HBV DNA and HBsAg level before delivery, HBV DNA and HBsAg levels in newborns and 6-12 months after birth, and adverse events.

Quality Evaluations of Included Study
The evaluation of research quality was conducted in two parts: RCTs and NRCTs.The Cochrane Risk of Bias assessment tool was used to assess the risk of bias in RCTs. NOS scale was used to assess the quality of observational studies.The former mainly was used to evaluate from the aspects of randomization, allocation concealment, blindness and unknown risk. The latter was used to evaluate the choice and comparability of case and control, exposed and non-exposed or other.

Statistical analysis
Firstly, we performed a traditional meta-analysis on the results to compare the efficacy of antiviral drugs to patients who used the same drug at the same dose. These outcome indicators were expressed as binary variables, then risk ratio (RR) and 95% confidence(95%CI) were calculated. Heterogeneity between studies was evaluated using a chi-squared Q test and I 2 statistics (p < 0.1 or I 2 > 50% considered heterogeneity between studies). When there was no heterogeneity between the combined studies, a fixed-effects model is used, otherwise a random-effects model was used.
Bayesian Markov chain-Monte Carlo model was used for network meta analysis, We incorporate both direct and indirect evidence into our model to compare the efficacy of antiviral drugs if treatments that have not, or cannot be compared directly. P value and 95% CI were evaluated the significance of the outcome. We setted appropriate Model adjustment iterations and Model simulation iterations, they were setted of 20000 and 50000 at the beginning.Model convergence was evaluated by potential scale reduction factor ( PSRF ). When PSRF was equal to one, it indicates that the model has good convergence and high reliability. The most likely ranking of the effects of each drug was achieved by calculating the Markov chain iteration ratio of each drug, which was represented by a rank probability plot. We also use the node split model to analyze the consistency of direct and indirect comparison results.
During the entire analysis process, if there is heterogeneity, we look for the cause of heterogeneity through subgroup analysis and sensitivity analysis.Statistical indicators I 2 and 95% CI were showed by forest plots. The statistical analysis process was completed using GeMTC 14.3 and Stata 23.0 software.

Search results and study characteristics
Through the developed search strategy, a total of 1,521 articles were retrieved from the listed databases. The search results was listed in Fig. 1. After screening, a total of 35 articles were included in the study.Among these studies there were 9 RCTs (18, 19, 23-25, 29, 36, 37, 47) Figure 2 is a network of these studies. The characteristics of all the included studies were summarized in Table 1 and the baseline information of the patients, as well as the immunization of the newborns after birth. We evaluated the quality of all the studies. Since most of the studies are NRCTs, they are difficult to control the allocation and blindness, and there is a high risk of bias. However, because the researchers better maked the patients in each group take medicine in compliance, so the risk of bias is low. Our assessment of the risk of bias was showed in Fig.S1. All analyses were based on previous published studies, no ethical approval and patient consent are required.

HBsAg positive rate during delivery
Of the 11 studies that included lamivudine and the control group, HBsAg positive rates in newborns is reported in 8 (18-20, 22-24, 26, 27) Fig. 4]. The three drugs could reduce the HBsAg positive rates by 6.98%, 9.99%, and 7.74%, respectively.

Hbv Dna Positive Rate Of 6-12 Month Infants
A network meta-analysis was performed on the HBV DNA positive rate of infants aged 6-12 months reported in studies. The results were consistent with the results of the direct meta-analysis. Treatment with NAs is effective, but there is no reliable comparison between drugs. (Table.3)

Net-work Meta Results
HBsAg positive rate of 6-12 month infants The Bayesian Markov Chain-Monte Carlo model parameter values were set to: number of chains was 4, tuning iterations was 2000, simulation iterations was 5000, initial values scaling was 2.5, and thinning interval was 10, at this time the model has good convergence with PSRF being 1, and when the number of occurrences is zero, we also add a value to the number of non-occurrences at the same time to correct the model accuracy .
Random-effects standard deviation of the inconsistency model and the consistency-model did not show a significant deviation and the inconsistency test performed by the node split method showed that there were no local inconsistencies (P > 0.05). In summary, a net meta-analysis was performed on the HBsAg-positive rate of infants aged 6-12 months reported in 31 studies by using a consistent effect model. The results of the summary analysis were showed in Table 2. The use of any antiviral drug can significantly reduce the HBsAg positive rate than without the drug, but we have no sufficient reason to believe that the results of pair-wise comparison of the three drugs are reliable(95% confidence interval does not include 1). Table 2 The results of network meta analysis of HBsAg positive rates in infants at 6-12 months  Table 3 The results of network meta analysis of HBV DNApositive rates in infants at 6-12 months We sortted the effects of three antiviral drugs on blocking MTCT by drawing the figure of rank probability. LAM was most likely to be the drug with the worst effect that blocks MTCT of HBV. TBV and TDF have the same probability of medium effect.The possibility of TBV and TDF being the most effective drugs were 45% and 50% (Fig. 5)

Safety Of Nas Therapy: Infants
Apgar scores, length, weight at birth, congenital malformations and preterm birth data were used to assess the safety of the NAs therapy for the infant. Interestingly, despite studies reporting low birth weight infants, we did not find that the babies who is in the NAs group had significantly different weights at birth[WMD=-0.03,95%CI=-0.06-0.00,p = 0.151, I 2 = 24.0%. Figure 6]. In addition, the Apgar score and height of the infants in the treatment group did not increase or decrease. [Apgar:WMD=-0.00,95%CI=-0.02-0.02,p = 0.991,I 2 = 0.0%. Fig.S2;Height:WMD = 0.03,p = 0.170,95%CI=-0.07-0.12,I 2 = 22.3%. Fig.S3]. Three studies reported the NAs therapy does not cause premature babies. Congenital malformations were reported in five studies, and it is worth noting that researchers evaluated that the occurrence of these adverse events was independent of medication. A multicenter prospective cohort study by Xu(19) et al. states that for mothers with elevated ALT at baseline, the median ALT levels decreased to 1xULN at 4 and 6 weeks after administration in both groups, respectively. But, it increased again at 4 weeks after delivery in no NAs therapy. ALT normalization rate in TBV group was higher than that in control group[RR = 1.55,95%CI = 1.24-1.95, p = 0.00,I 2 = 0.00%]. Only one study (50) reported rates of ALT normalization in the use of TDF. The study noted that mothers receiving TDF had higher rates of ALT normalization (82% vs 61%). Pan et al. (47) pointed out that those patients with ALT levels that were 10 times above the normal range had alanine aminotransferase flares resolved after the initiation of antiviral therapy. Research by Chen et al. (51) found that during the study period, the incidence of ALT elevation > 2 xULN in the TDF group over 3 months was significantly lower than that in the control group (3.23% vs 14.29%, P = 0.0455).

Publication Bias
We used funnel plots to evaluate publication bias in included studies. Due to limited data resources available for the study, we only performed publication bias detection for the main outcome, MTCT rate.The results of funnel plot showed good symmetry and dispersion among the studies. There was no significant publication bias (Fig. 7).

Discussion
MTCT mainly refer to intrauterine transmission, transmission during delivery and transmission through nursing or breast milk (53). In areas with high hepatitis B prevalence, at least half of them develope from HBV infection during perinatal or early childhood (54). Hepatitis B vaccine is believed to be effective in preventing HBV infection, especially for babies born to mothers who carry hepatitis B virus. In addition, HBIG is recommended for babies born to HBeAg-positive mothers on the basis of vaccination (55). Use of HBV vaccine and HBIG within 12 hours of birth, followed by two additional HBV vaccination inoculations, has been demonstrated to reduce the MTCT rate from 90% to approximately 5%-10% (56,57). However, these 5% -10% of people who have failed immunity will enter a chronic hepatitis state once their condition is not effectively controlled. As the condition worsens, decompensated cirrhosis and liver failure may occur, and they may develop liver cancer, eventually dying (58). High HBV DNA load in pregnant women is a key risk factor for immune failure (11,59,60). It is estimated that the risk of HBV vertical transmission is more than 90% when the maternal viral load is greater than 2 × 10 5 IU /ml. Amniocentesis performed on HBV infected mothers with HBV DNA level higher than 2 × 10 7 IU/ml significantly increased the frequency of vertical HBV transmission (60)(61)(62). Guidances from the European Association for the Study of Liver, the American Association for the Study of Liver Diseases, and the Asian Pacific Association for the Study of the Liver recommend that antiviral treatment should be used for pregnant women with hepatitis B virus to avoid MTCT (63)(64)(65).
LAM was the first approved nucleoside reverse transcriptase inhibitor that has therapeutic effect on HBV and the first drug used to block MTCT of HBV. Long-term clinical practice has proved its good effect.However, in the course of treatment, it was prone to drug resistance mutations (66,67).TBV is a class of deoxynucleoside drugs approved by FDA in 2006 and China in 2007.Drug resistance and virological breakthrough were lower than lamivudine in numerous clinical trials (68,69). TDF is a nucleotide analogue that inhibits reverse transcriptase and ribonucleic acid, and has been shown to be effective and safe in HIV or HBV single infection and co-infection in a number of studies with virtually no drug resistance (70)(71)(72).
The U.S. food and drug administration classified TBV and TDF as category B pregnancy (animal experiments found no has any harm to the embryos, but it is lack of clinical studies in pregnant women), LAM, ADV and ETV as pregnancy category C drugs (animal experiments prove that the fetus has teratogenic effect, but did not study in pregnant women).In this analysis, we included a total of 35 studies involving 6,109 pregnant women with HBV DNA ≥ 2 × 105 IU/mL. The results of direct meta-analysis showed that the HBsAg serum positive rate of the three antiviral treatment groups was significantly lower than that of the control group at birth. More importantly, the combined results of 6-12 months showed that antiviral treatment could effectively reduce the incidence of MTCT,. In addition, antiviral drugs have also shown a good effect in reducing HBV DNA load in pregnant women, and NAs therapy can better inhibit HBV DNA level in babies born. We performed a reticulated meta-analysis to compare three antiviral drugs. The results also showed the effectiveness of antiviral drugs for treatment. However, because there are fewer RCTs experiments available, there are fewer studies comparing pairs, Comparisons between drugs are not reliable.According to the comparison of grade probabilities, we found that TBV and TDF showed greater advantage than LAM in blocking MTCT of HBV, but the relative advantage of TDF and TBV was difficult to determine, and the probability of TBV being the best drug was 4% higher than TDF. TDF, which was first reported in 2012 for the prevention of MTCT has had few high-quality clinical studies, which may account for its lack of superiority in this meta-analysis (73). In recent years, some scholars have done clinical studies or other meta-analyses comparing the two, and the conclusions have been somewhat controversial. In the study by Zeng et al., There were no cases of MTCT in 51 and 58 infants delivered in the TDF group and TBV group, respectively. Among the 32 infants in the untreated group, 4 had hepatitis B Antigen positive, the difference was statistically significant (p < 0.01). The study by Xiao et al. also obtained the same result ( 0/60 vs 0/62, p < 0.05). However, in the study of Chen et al., the serological positive rate of HBsAg in 12month-old infants showed a difference between the TBV group and the TDF group (3.79% vs. 3.33%,p < 0.05), indicating that tenofovir was slightly more effective.However, these comparison samples are small and not RCT, so the conclusions may be biased. Deng (74). conducted a meta-analysis of 306 mothers receiving tibivudine and 270 mothers not receiving antiviral drugs. Serological positive rates of HBsAg and HBV DNA in infants in the tibivudine group were significantly lower at birth and 6-12 months later.
In the systematic evaluation conducted by Brown et al. (75) the efficacy and safety of LAM, TBV and TDF for pregnancy were evaluated. The results showed that antiviral drugs not only significantly reduced the incidence of MTCT, but also that the former two drugs were safe for pregnant women and did not increase the incidence of adverse events in infants. However, the safety of TDF was not described in this study, perhaps because of the limited number of studies involving the safety of tenofovir that could be included and the lack of controlled outcome data.
In our analysis, although there is less study on drug safety data, according to the existing evidence to merge, no matter what kind of medication was not increased pregnant women cesarean section rate and the rate of postpartum hemorrhage, and did not cause higher CK. These limited safety data prove that our antiviral treatment is feasible. However, these studies show that pregnant women are receiving treatment for a short period of tim and also lack of data on maternal and fetal late effects. This requires researchers to obtain through long-term follow-up.
The strength of this analysis is that we comprehensively and thoroughly evaluate the information contained in the included studies, which includes both newly published studies and later TDF what is used for clinical lately.
We have not only conducted traditional meta-analysis, but also used Bayesian network model analysis to solve the shortcomings of the lack of comparative study between drugs in reality, obtaining comparisons between drugs indirectly. However, there are still many shortcomings. First, due to the limited published RCTs, more NRCTs was included, which also expanded the low quality of evidence. However, these NRCTs are not the cause of greater heterogeneity in subgroup analysis. Second, more of the included studies were from Chinese studies, which made the results limited to the global scale. Third, there are fewer head-to-head studies between the two drugs included, which results in low reliability of the combined results, which is also reflected in our network analysis. Fourth, we did not integrate more factors, such as drug resistance, hepatitis B virus genotype, costeffectiveness and so on.
We also need more multicenter, rct studies comparing head-to-head comparisons between drugs, especially more studies comparing the effectiveness and safety of TBV and TDF. This allows us to have more high-quality evidence in the choice between the two drugs, so that we can obtain both direct evidence and more reliable indirect evidence through meta-analysis.
In conclusion, we found that it is effective and safe to block MTCT of HBV through Nas therapy for pregnant women with CHB and high viral load in the body through analysis. TBV and TDF are more effective than LAM. TBV and TDF have similar blocking effects.However, because of the higher resistance of TBV than TDF, we seem to recommend the use of TDF.

Availability of data and material
The datasets used or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
Not applicable.

Consent for publication
All authors are informed and agree to publish

Conflict of interest
There are no financial or other relations that could lead to a conflict of interest.

Funding
Provincial Science and Technology Plan Project of Hebei Province (17397765D).

Authors' contributions
Erhei Dai conceived and designed the study. Literature search: Bo Zhu,Xiuli Chen. Data extraction: Bo Zhu, Xiuli Chen, Fumin Feng. Data analysis: Bo Zhu. Drafting manuscript: Bo Zhu. ErHei Dai, Fumin Feng reviewed and edited the manuscript. All authors read and approved the manuscript.

Figure 1
The PRISMA flow of this study    Forest plots of the weight of newborns Figure 7 Funnel plot of MTCT