Entecavir for Patients with Hepatitis B Decompensated Cirrhosis in China: a meta-analysis

Evidence about the clinical effects of entecavir (ETV) for patients with hepatitis B decompensated cirrhosis remain controversial. Therefore, we perform this meta-analysis to assess the treatment outcomes of ETV in participants with hepatitis B decompensated cirrhosis. Relevant studies were identified by searching databases until the March 2016. A random-effects model was used to estimate summary relative risks (RRs) and 95% confidence intervals (CIs). GRADEprofiler3.6 was used to evaluate the quality of the evidence. A total of 26 studies (involving 2040 patients) were included. The quality of the evidence was classified from very low to high by the GRADED approach for all included RCTs. Meta-analysis showed that patients were more likely to experience HBV-DNA loss (RR:1.85, 95%CIs: 1.41 to 2.43, P < 0.0001 at 48 weeks), have normalized alanine aminotransferase levels (ALT) (P = 0.003 at 24 weeks, P = 0.02 at 48 weeks), and have a low mortality rate at 24 weeks (P = 0.003) when treated with ETV. There was no significant different between ETV and the control groups at the total mortality (P = 0.06) and HBeAg seroconversion (P = 0.14). In conclusion, ETV could be the first line therapy for patients with HBV related decompensated cirrhosis, because ETV could reduce the early mortality and move HBV DNA load down.

Scientific RepoRts | 6:32722 | DOI: 10.1038/srep32722 of ETV group reduced more than ADV group after a treatment of 48 weeks. Keating GM 3 showed that patients had a significant liver function improvements from baseline after 12 months treatment of ETV in patients with decompensated cirrhosis.
There are several meta-analysis about ETV for patients with HBV related decompensated cirrhosis. For example, Peng H et al. 10 only compared LAM combined with ADV with ETV. Ye X et al. 11 just showed the effects of LAM and ETV for hepatitis B decompensated cirrhosis. Singal A. K. et al. 12 did not study the difference of patients without anti-viral agents with patients who used ETV for HBV related decompensated liver cirrhosis. All records included in Singal A. K. et al.'s meta-analysis are published before 2010. In our review, 26 studies and 2040 patients are involved. We used the HBV DNA loss, ALT normalization, mortality and HBeAg seroconversion to evaluate the effect of ETV for patients with HBV related decompensated cirrhosis.

Results
Description of the included studies. A total of 26 RCTs, with 2040 patients fulfilled the included criteria ( Fig. 1 and Table 1). 930 patients were treated with ETV, 1110 patients were treated by other ways. The number of patients who were treated with other NAs drugs except ETV were 561, 549 patients were not treated by NAs drugs. The dominative outcomes are the HBV DNA loss, the recovery of ALT, the mortality and the HBeAg seroconversion.

Risk of bias in included studies.
The summary results of the risk of bias were showed in Fig. 2. All trials were free from baseline imbalance bias and incomplete outcome bias. All trials were random control trials. None of the trials had adequate allocation concealment. One trial 13 had adequate blinding. All trials might have academic bias and funding bias.   Meta-analysis results. We used risk ratio (RR) as summary measures. We also stated the 95% confidence intervals.
HBV DNA loss. HBV DNA loss at 12 weeks. 7 RCTs were included. There were 568 patients in total. The analysis of heterogeneity showed that I 2 = 71%. These trials were considered statistically significant heterogeneity. We got RR = 3.52, 95%CI [1.77, 6.99], P = 0.0003. The experiment groups were higher than the control groups. This revealed a statistically significant. The result was showed in Fig. 3.
Subgroups of HBV DNA loss at 12w. There are four subgroups of HBV DNA loss in our research. Firstly, the control groups were patients with other NAs on the basis of CT.4 RCTs were included. There were 302 patients in total. The analysis of heterogeneity showed that I 2 = 57%. These trials were considered statistically significant heterogeneity. We got RR = 4.37, 95%CI [1.58, 12.09], P = 0.004. The experiment groups were higher than the control groups. This revealed a statistically significant. Secondly, the control groups were patients with CT without any NAs. 3 RCTs were included. There were 217 patients in total. The analysis of heterogeneity showed that I 2 = 31%. We got RR = 39.44, 95%CI [8.54, 182.03], P < 0.00001. The experiment groups were higher than the control groups. This revealed a statistically significant. Thirdly, the control groups were patients with LAM. 3 RCTs were included. There were 171 patients in total. The analysis of heterogeneity showed that I 2 = 81%. These trials were considered statistically significant heterogeneity. We got RR = 1.23, 95%CI [0.23, 6.76], P = 0.81. There was no statistically significant between the experiment groups and the control groups. Fourthly, the control groups were patients with ADV only. 3 RCTs were included. There were 181 patients in total. The analysis of heterogeneity showed that I 2 = 0%. These trials were no statistically significant heterogeneity. We got RR = 8.01, 95%CI [3.22, 19.94], P < 0.00001. The experiment groups were higher than the control groups. This revealed a statistically significant.
HBV DNA loss at 24 weeks. 17 RCTs were included. Total patients were 1324. The analysis of heterogeneity showed that I 2 = 93%. These trials were considered statistically significant heterogeneity. We got RR = 4.51, 95%CI [2.51, 8.12], P < 0.00001. The experiment groups were higher than the control groups. This revealed a statistically significant. The result was showed in supplementary information. And the funnel pool was showed in Fig. 4.

Subgroups of HBV DNA loss at 24 weeks.
There are four subgroups of HBV DNA loss at 24 weeks. Firstly, the control groups were patients with other NAs. 8 RCTs were included. Total patients were 689. The analysis of heterogeneity showed that I 2 = 80%. These trials were considered statistically significant heterogeneity. We got RR = 1.64, 95%CI [1.16, 2.32], P = 0.005. There was statistically significant between the experiment groups and   ALT normalization. ALT normalization at 24 weeks. 6 RCTs were included. There were 501 patients. The analysis of heterogeneity showed that I 2 = 62%. These trials were considered statistically significant heterogeneity. We got RR = 1.62, 95%CI [1.17, 2.23], P = 0.003. The experiment groups were higher than the control groups. This revealed a statistically significant. The result was showed in Fig. 5.
ALT normalization at 48 weeks. 7 RCTs were included. Total patients were 622. The analysis of heterogeneity showed that I 2 = 77%. These trials were considered statistically significant heterogeneity. We got RR = 1.38, 95%CI [1.06, 1.80], P = 0.02. There was statistically significant between the experiment groups and the control groups.
Mortality. 9 RCTs were included. There were 727 patients. The analysis of heterogeneity showed that I 2 = 0%. These trials were not considered statistically significant heterogeneity. We got RR = 0.55, 95%CI [0.30, 1.03], P = 0.06. This did not reveal a statistically significant. The result was showed in Fig. 6. There are two subgroups of mortality. One subgroup was the mortality till 24 weeks.12 RCTs were included. There were 765 patients. The analysis of heterogeneity showed that I 2 = 0%. These trials were not considered statistically significant heterogeneity. We got RR = 0.38, 95%CI [0.20, 0.71], P = 0.003. This revealed a statistically significant. The other was mortality till 48 weeks. 9 RCTs were included. There were 627 patients. The analysis of heterogeneity showed that I 2 = 0%. These trials were not considered statistically significant heterogeneity. We got RR = 0.58, 95%CI [0.33, 1.03], P = 0.06. This did not reveal a statistically significant.
HBeAg seroconversion. 7 RCTs were included. There were 555 patients. The analysis of heterogeneity showed that I 2 = 35%. These trials were not considered statistically significant heterogeneity. We got RR = 1.46, 95%CI [0.89, 2.40], P = 0.14. This did not reveal a statistically significant. The result was showed in Fig. 7. Evidence quality. The results of the evidence quality were showed in supporting information. Discussion ETV an oral deoxyguanosine nucleoside analogue, inhibits serum HBV DNA efficiently, improves the biochemical and histological characters of HBV related diseases 4,14 . S. Amini-Bavil-Olyaee et al. 15 proved ETV was in short term a safe option for HBeAg negative patients. There were several meta-analysis 10-12 about the oral anti-viral agents for patients with decompensated HBV related liver cirrhosis. However, there were not enough evidence to prove that ETV could be the first line drug for HBV related decompensated liver cirrhosis.
Chen FZ et al. 16 , Feng J et al. 17 , Zhang DH et al. 18 , Hu XM 19 and Bi YL 20 proved that the patients with ETV could undergo more HBV DNA loss than the patients with LAM at 24 weeks. In the studies of Shao JB et al. 21 , ETV made more HBV DNA loss than LAM at 12 weeks, but less at 24 weeks. Our data showed that ETV could significantly move viral load down to undetectable levels compared to patients without NAs treatment. The patients with ETV also experienced more HBV DNA loss than patients with ADV therapy. Although, at 12 th and 24 th weeks there were no significant differences in undetectable viral load between ETV and LAM in patients with Hepatitis B virus-related decompensated liver cirrhosis. ETV efficiently improve the outcome of HBV DNA loss than LAM at 48 th week. ETV's long-term efficacy is superior to LAM at the part of HBV DNA loss.
ETV causes statistically significant sharp decline in ALT level at 24 th and 48 th week. Although, there were no significant differences between ETV and control groups at 48 th week.
Xu Y 22 considered the mortality of ETV group was higher than the group with LDT and ADV. In our research, ETV reduces the mortality of patients at 24 th week. ETV could reduce early mortality.
Lin XS et al. 23 , Han ZQ et al. 24 , Yang J et al. 9 and Feng J et al. 17 thought ETV could improve the rate of HBeAg seroconversion. Liaw et al. 8 thought HBeAg seroconversion was higher with ADV at 24 weeks. In our research, there were no significant different between the ETV group and the control group.
Liaw yf et al. 13 thought there were no significant different between ETV and TDF among three treatment regimens(HBV DNA loss, ALT normalization and mortality), TDF was superior to ETV in terms of HBeAg seroconversion. Xu Y 22 thought LDT combined with ADV early acting was better than ETV. However, there was only one paper to support their conclusion, more trials were needed.
We used Funnel pool to evaluate the publication bias and found that almost all the related meta-analysis had the publication bias. The results would be affected.
The degree of evidence quality about patients' mortality (ETV versus other treatments) and HBV DNA loss (ETV versus patients only take CT) at 24 weeks was high. Night outcomes of the degree of evidence quality is moderate. Other results' degrees were from very low to low. The low and very low quality of the evidence would affect the reliability of the results.
There are still some limits of our research. (i) We only evaluated four outcomes (the HBV DNA loss, the rates of ALT normalization, the mortality, the HBeAg seroconversion). Other results (such as pathological changes  of liver tissue, cost-effectiveness issues are not mentioned in our study. (ii) There were no specific descriptions of the lower magnitude to the decline of detectable HBV DNA and the decrease of ALT. (iii) The number of RCTs included in this study is limited and the included samples' number is insufficient. (iv) The quality of RCTs included in our study is not high. We still need RCTs of multi-center, high qualities and a large of samples to obtain a comprehensive Meta-analysis.
Despite the shortcomings of the studies included in our review, these studies constitute the best level of evidence that is currently available. Overall, the evidence from systematic review and meta-analysis is more trustworthy than observational studies and expert opinions. In our research, ETV could be the first line therapy for patients with HBV related decompensated cirrhosis, because ETV could reduce the early mortality and move HBV DNA load down. independently. We searched the terms of ETV, decompensated cirrhosis, hepatitis B, and randomized controlled trial. The results were limited by the MeSH terms of these words. Finally, we expanded the search results by the free word retrieval for the newest reports. All the citations of the identified trials were checked. We also checked the citations of published reviews meta-analysis or guidelines. Manual search was made to augment the search strategy.

Search method.
Criteria for considering studies for this review. All the included studies satisfied the following selection cri- (iii) incomplete data; (iv) co-infection with other viruses (eg. Hepatitis A virus); (v) other decompensated liver disease (eg. autoimmune liver disease).

Assessment of risk of bias.
We assessed the risk of bias in the trials by following the instructions given in Cochrane Handbook for Systematic Reviews of Interventions. We assessed the following procedures of each trials because the methodological quality of the trials could have an influence on intervention effects. We assessed the following parts: (i) random sequence generation (ii) allocation concealment (iii) blinding (iv) incomplete outcome data (v) selective outcome reporting (vi) baseline imbalance (vii) academic bias (viii) funding bias. Every domain was evaluated by three degrees which are low risk of bias, unclear risk of bias and high risk of bias.
Subgroup analysis. We planned to perform the following subgroup analyses: (i) ETV versus other NAs. (ii) Trials with ETV versus trials only take CT without any antiviral drugs. (iii) ETV versus LAM. (iv) ETV versus ADV. Statistical methods. We use the software package Review Manager 5.3.5 to perform the meta-analysis according to the recommendation of the Cochrane Collaboration 25 . We use risk ratio (RR) to calculate the 95% confidence interval for our research. In our research, all indices we included were dichotomous variables. We used a random-effects model for all studies. The random-effect model is DerSimonian-Laird. The heterogeneity was explored by chi-squared test with significance set at P value 0.10, and the quantity of heterogeneity 26 was measured by I 2 . I 2 < 50% was considered there was heterogeneity of the trials included. Generally, if I 2 > 50% was considered statistically significant heterogeneity 27 .
We performed intention-to-treat analysis for the participants who could not finish the treatment. The patients who did not finish the treatment included patients who died, patients who gave up the treatment, and patients we could not connect with them. We considered these participants as negative results.
Quality of the evidence. We used GRADEprofiler3.6 to evaluate the quality of the evidence according to the guideline of GRADES of Recommendations Assessment Development and Evaluation (GRADE). There are four degrees in GRADE: high, moderate, low and very low. The results of GRADE were showed by evidence profile (EP).
Funnel plot. We intended to use funnel plot to measure the publication bias. Lau J et al. 28 thought that at least