Effectiveness of Chinese herbal medicine in treating liver fibrosis: a systematic review and meta-analysis of randomized controlled trials

Background The studies on the effectiveness of Chinese herbal medicines (CHM) in treating liver fibrosis (LF) were not consistent. This study aims to systematically review the effectiveness of CHM on treating LF patients. Methods Databases including MEDLINE, AMED, EMBASE, The Cochrane Central Register of Controlled Trials, China National Knowledge Infrastructure, TCMOnline, Chinese Biomedical Literature Database, and Chinese Medical Current Contents were searched up to March 2011. Randomized controlled trials (RCTs) involving LF patients receiving CHM, Western medicine, combined CHM and Western medicine compared with placebo, Western medicine or no intervention were included. LF markers including serum hyaluronic acid (HA), laminin (LN), procollagen type III (PC-III), type IV collagen (IV-C), matrix metalloproteinase (MMP), and tissue inhibitors of metalloproteinase (TIMP) were measured as primary outcomes. Liver biochemistry, including alanine aminotransferase (ALT) and aspartarte aminotransferase (AST), and improvement of related clinical symptoms were measured as secondary outcomes. Risk of bias of allocation sequence, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and other biases were assessed. Results Twenty-three RCTs with 2123 participants were analyzed in subgroups of types of comparison and study quality. Fifteen studies were graded as good quality. CHM alone and combined with Western medicine showed significant improvements in HA, LN, PC-III and IV-C compared with Western medicine alone. However, there were no significant differences observed between CHM and placebo treatments. Conclusion The current inconclusive results in determining the effectiveness of CHM treatment on LF, due to the poor methodological quality and high heterogeneity of the studies, suggests that large RCTs using standardized Chinese medicine syndrome diagnosis and CHM formulae with longer follow-up are required for further evaluation.


Introduction
Liver fibrosis (LF), as a result of wound-healing response to recurrent liver injury, is thought to be an early reversible stage of liver cirrhosis [1]. It is characterized by the formation of fibrotic scar tissue with abnormal accumulation of fibroblasts and myofibroblasts, and excessive synthesis and deposition of extracellular matrix (ECM) proteins. The development of anti-fibrotic therapy is important for patients with chronic liver diseases, especially for chronic hepatitis B (HBV) and C virus (HCV) infections [2], which are the most prevalent blood-borne viral infection and the major causes of LF worldwide, especially in mainland China [3][4][5]. Few LF treatments are effective and inexpensive without adverse side effect [6][7][8].
Categories of current research into Chinese herbal medicine (CHM) treatment of LF include (1) the prevention of anti-fibrosis effects, (2) mechanisms, and (3) clinical efficacy, safety and quality control [3]. Clinically, several studies reported the efficacy of CHM on LF [9][10][11]. A previous systematic review of 11 studies on LF suggested that "Fuzheng Huayu Capsule" had beneficial effects on LF [12]. However, the review included only one CHM compound, and the results might not be representative of all CHM. Moreover, with advanced progress on LF in recent years, the review should be updated to include recent studies.
This article aims to systematically review the published randomized controlled trials (RCTs) for evaluating the effectiveness of CHM on LF treatment.

Methods
This study was conducted according to the Cochrane practice [13,14], including pre-specified objectives, search strategy, inclusion criteria, quality assessment, data collection and meta-analysis.

Search strategy
Published RCTs on CHM treating LF patients were searched via the following electronic databases from their inception to March 2011: MEDLINE since 1948, AMED since 1985, EMBASE since 1974, and The Cochrane Central Register of Controlled Trials since 1996. In addition, four Chinese electronic databases including China National Knowledge Infrastructure (CNKI), TCMOnline, Chinese Biomedical Literature Database (CBM), and Chinese Medical Current Contents (CCMC) were searched since January 2000. The bibliographies of studies identified in the systematic search were checked for potentially relevant publications. Unpublished data were not included.

Types of studies
This review included only RCTs on the effectiveness of CHM. The studies with quasi-randomized and non-randomized study design were excluded.

Participants
The studies recruited patients suffering from chronic hepatitis diseases, fatty liver or schistosomiasis japonica, and having histologically significant LF were included. The studies involved patients having co-infection of two or more types of hepatitis or fatty liver with other chronic liver diseases, or having decompensated liver diseases were excluded.

Interventions
The studies comparing CHM (such as pills, tablets, capsules, decoctions, and injections) with placebo, Western medicine, or no intervention were included. CHM intervention could be a sole anti-fibrotic therapy or an adjunct treatment. The studies assessing combined effects of CHM with other intervention (e.g. CHM plus acupuncture, injection of CHM into acupoint, and acupoint application) were excluded. The studies used non-conventional herbal medicines or complementary medicines as control groups were also excluded. Co-intervention, including those supplements such as vitamins, was allowed if both arms of the randomized allocation received the same co-intervention.

Outcome measures
Primary outcome measures were mean differences (MD) of LF biomarkers, which indicate ECM metabolism, including serum hyaluronic acid (HA), laminin (LN), procollagen type III (PC-III), type IV collagen (IV-C), matrix metalloproteinase (MMP), and tissue inhibitors of metalloproteinase (TIMP) [16]. Secondary outcomes included liver biochemistry which including alanine aminotransferase (ALT) and aspartarte aminotransferase (AST), and the improvement of related clinical symptoms was defined as the alleviation of subjective symptoms after the interventions. Both non-serious and serious adverse events were evaluated. A serious adverse event included event of death, life-threatening incidents, or inpatient or prolonged of hospitalization which resulted in a persistent or significant disability [17].

Data extraction and assessment of methodological quality
Two authors (FC and NW) independently assessed studies for eligibility, extracted data in duplicate using a structured data extraction form, and cross-checked for transcription errors. The data extraction form comprised the items of primary author, study citation, study design, participants, interventions, outcome measures, and adverse events according to pre-specific selection criteria. In case duplicate publications were found, only the most informative and updated version was included. The quality of included studies was evaluated independently by the two authors (FC and NW) using a tool for evaluating 'risk of bias' which tool was adapted from the Cochrane Handbook for Systematic Review of Interventions [13] with slight modifications for transforming the bias codes of "yes", "unclear" and "no" into 2, 1 and 0, respectively, for presenting clearer results. The following six questions were asked: (1) Was the allocation sequence adequately generated?
(3) Was knowledge of the allocated interventions adequately prevented during the study? (4) Were incomplete outcome data adequately addressed? (5) Are reports of the study free of suggestion of selective outcome reporting? (6) Was the study free of other problems that could put it at a risk of bias?
Every item would be given 2 points for answering 'yes', 1 for answering 'unclear' and 0 for answering 'no'. Prevention of knowledge of the allocated interventions, i.e. blinding (patient, personnel, and outcome assessor blinding), was assessed separately. There were eight items in total as three items from blinding (including blinding of patients, personnel, and outcome assessor), and five from the rest five questions. The scale ranged from 0 to 16 points with 0 to 7 regarded as poor quality, and 8 or above as good quality. Any disagreement was resolved by consensus. If necessary, the third author (YF) was consulted for resolution.

Data synthesis of outcome measures
Review Manager, Version 5.1 for Windows (The Nordic Cochrane Centre, Copenhagen, Denmark) and STATA 10.1 (StataCorp, College Station, TX, USA) were used for data analysis. Meta-analysis was carried out on the intention-to-treat (ITT) basis regardless whether the subjects were lost to follow-up. Subgroup analysis was conducted among different comparisons (including CHM versus Western medicine, CHM versus placebo, and combined treatment versus Western medicine) and study quality (all studies versus the studies with good quality). Inversevariance random effects model was used MD and 95% confidence intervals (CI) were calculated. Heterogeneity was assessed by examining the forest plots and I 2 statistics, where I 2 values of 25%, 50%, and 75% were regarded as low, moderate, and high heterogeneity, respectively [18]. Sensitivity analysis for primary outcomes was performed to assess the impact of excluding outlier studies when there was high heterogeneity (I 2 > 75%) between studies. In a three-arm study that had two control groups of conventional medicine and no intervention, the treatment group was split into two groups to create two comparisons in the meta-analysis. Funnel plot and Egger's regression asymmetry test were applied to detect for the potential publication bias [19]. P value less than 0.05 were considered statistically significant. Figure 1 shows the process of the study selection. Twenty-three RCTs with a total of 2123 participants met the eligibility criteria were included in this review (Table 1). Sample sizes ranged from 44 to 164. Among these included studies, a study recruited schistosomiasis japonica patients only [20], the remaining recruited patients with chronic HBV infection. All of the identified studies were conducted in China and published in Chinese language, and only one study was published in English language [21]. Three studies used three-arm study design (2 intervention groups compared with 1 control group, or 1 intervention group compared with 2 control groups) [21][22][23] and others used two-arm study design (1 intervention group versus 1 control group) [20,[24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42]. No study reported mortality, liver cirrhosis or cancer, quality of life or cost as outcomes. The information about the study design, participants, intervention, outcome assessment, and quality was presented in Additional file 1.

Control groups
Comparison groups included Western medicines (including IFN, Entecavir, Adefovir Dipivoxil, and LVD), placebo, and no intervention. Three studies used placebos, with one using physiological saline and food coloring [30], and two using similar shape capsules [20,41].

Duration of follow-up
The range of intervention duration in the studies ranged from 60 days to two years with mostly were six months (N = 7) and one year (N = 7). The duration of followup was only reported in three studies with a range from three to six months [27,30,34].

Methodological quality
Fifteen studies were graded as good quality and others as poor quality (as shown in Additional file 1). Out of 23 included studies, only two studies reported adequate generation of allocation sequence using random number tables or drawing of lots for assigning groups [29,41]. None of the studies described the method of allocation concealment. Only one study reported using blinding design (single blind without description of the blinding method) [21] and three studies used placebos [20,30,41]. Six studies did not provide the information on missing   data [28,29,[33][34][35]41]. None reported the use of ITT in their analysis.

Outcomes
Primary outcomes

Adverse events
Although adverse events were reported in nine studies, none was serious. Among these, three reported no adverse event in both groups [31,32,35] while one reported no adverse event for CHM [30]. Five reported gastrointestinal discomfort and drug-allergic symptoms for CHM, placebo, IFN and LVD [20,21,27,30,41]. One reported two cases with dizziness using the combined treatment [21]. Other adverse symptoms of flu-like symptoms or mild leucopenia and thrombocytopenia were related to the use of IFN or LVD [21,22,27]. No study described the method of data collection for adverse events.

Evaluation of publication bias
Publication bias was found for HA (P = 0.003), PC-III (P = 0.001) and LN (P = 0.047) although non-significant for IV-C (P = 0.814) according to Egger's test. However, visual inspection of the funnel plots (please see funnel plots in Additional file 2) found no obvious basis.

Overall findings
The levels of LF markers were significantly reduced in patients receiving CHM or combined treatment compared with Western medicine. The levels were significantly decreased in CHM group compared with no intervention although the effect was not significantly different when in comparison with Western medicine or placebo. Moreover, CHM was found to be effective in symptom improvements. It should be noted that CHM was not consistently better than placebo.

Methodological quality of studies
Eight studies [21,25,28,31,33,34,36,37] were assessed to be poor quality according to modified Cochrane 'risk of bias' scale. Only two studies reported the method of randomization [34,41]. The method of allocation concealment was not reported by all studies, which should be alert to the possibility of selection bias and overestimation of intervention effects [43]. Blinding was not reported or inappropriately reported by most of the studies. Although most studies used objective outcome measures, it did not rule out the possibility of performance bias and detection bias [44].

Potential biases
Strict eligibility criteria were used to reduce heterogeneity. The included studies, however, had various participants' characteristics and different CHM or combined treatment against different control interventions. Small sample sizes, methodological differences between studies and variations in study objectives might contribute to heterogeneity. To investigate the high levels of heterogeneity in this metaanalysis we performed subgroup analysis and sensitivity analysis. These analyses did not find inconsistency. Some high quality studies might be missed due to the strict eligibility criteria such as a multicentre, double blinded RCT comparing "Fuzhenghuayu capsule" with "Heluoshugan capsule" [45]. Other reviews including CHM as control may be needed in the future. As most studies were of small scale and poor methodological quality, large RCTs of high quality would be required for determining the effectiveness of CHM on LF treatment.
In order to minimize bias in the review, we did not restrict the publication type and language, and searched many commonly accessed databases. However, all identified studies were conducted in China, and studies more likely reported positive results, which may be influenced by publication and location bias [46][47][48]. Our analysis of publication bias using Egger's test did show publication bias in the outcomes of HA, LN, and PC-III, although the funnel plots were symmetric in distribution (Additional file 2).

Limitations and further research
Most of the current studies only focused on the effectiveness of intervention [49] but neglected monitoring the harmful effect from CHM. Further studies should assess both the safety and effectiveness. Well-designed, multi-centre and large sample size RCTs in compliance with the CONSORT guideline [50] should be implemented. Studies with CHM should be registered before their conduct.

Conclusion
The current inconclusive studies are of poor methodological quality and high heterogeneity do not adequately support the effectiveness of CHM treatment on LF. Large RCTs using standardized Chinese medicine syndrome diagnosis and CHM formulae with longer followup are required for further evaluation.