Two trained researchers systematically searched the following databases from their inception to July 2017: PubMed, Cochrane Central Register of Controlled Trials, Chinese National Knowledge Infrastructure, Chinese VIP information and WanFang database. The search strategy of PubMed was as follows, and it was modified to suit other English or Chinese databases.

PubMed search strategy:

#1. Wilson's disease [mh]

The RoB of included articles was assessed using the 7-item criteria from the Cochrane' s collaboration tool (Higgins et al., 2011). Two authors independently evaluated the study quality, and the final result was identified by discussion when countering the disagreement.

The frequency of use of the particular herb was calculated and those used at a high frequency were described in detail.

In order to assess the reporting completeness with a rating system quality of the clinical studies, we used a rating system according to our previous articles (Wang et al., 2019) as follows: (1) high quality: full information about the botanical material is provided, including a voucher specimen; (2) moderate quality: only partial information about the botanical material is provided and a voucher specimen is missing; there are taxonomic inaccuracies; (3) low quality: inadequate information and overall taxonomically is inadequate.

The pooled analyses were carried out with RevMan 5.3 software. Heterogeneity was assessed using the Cochrane Q-statistic test (p < 0.05 was considered statistically significant) and the I²-statistic test. A fixed effects model (I² < 50%) or a random effects model (I² > 50%) was used depending on the value of I². Funnel plots were used to visually estimate publication bias. We calculated the standard mean difference (SMD) with 95% Confidence Intervals (CIs). Sensitivity analyses omitting everyone, which study at a time from the original analysis were conducted to demonstrate our main results to be robust.

We identified 1,049 hints, of which 364 articles remained after removal of duplicates. Through screening titles and abstracts, 320 studies were excluded because they were case reports, they lacked a comparison group, they were not CHM studies or reports of clinical trials. After full-text evaluation of the remaining 44 articles, 26 studies were excluded for the following reasons: (1) 7 articles were not RCTs; (2) 3 articles included CHM treatment in control groups; (3) 2 articles included oxiracetam or tiopronin in control group; (4) 6 articles have inappropriate outcome measures; (5) 8 articles were suspected of being published more than once. Eventually, 18 eligible studies were identified (Figure 1).

Eighteen studies with 1,220 participants were included from 1997 to 2016. The sample size ranged from 33 to 146 with an age of 3 to 59 years. The duration of diseases lasted from 1 month to 31 years. The course of treatment ranged from 28 to 90 days. Five studies (Han et al., 2014; Hu, 2014; Zhang et al., 2014a; Fang, 2015; Jiang, 2016) were diagnosed according to Chinese Medical Association of Neurology (2008), 9 studies (Han et al., 1999; Hong et al., 2000; Xue et al., 2007; Zhang, 2007; Chen and Wang, 2010; Wang et al., 2010; Xu et al., 2012a,b; Zhang et al., 2014b) were diagnosed according to Yang criteria (1995) and 4 studies used comparable definitions (Ren et al., 1997; Cui and Zhao, 2001; Xiao, 2003; Chen and Wang, 2008). Three studies (Han et al., 1999; Chen and Wang, 2008, 2010) used CHM monotherapy, and the others used CHM paratherapy. The control group used penicillamine (Xiao, 2003; Zhang, 2007; Chen and Wang, 2008, 2010), DMSA (Ren et al., 1997; Xiao, 2003; Zhang et al., 2014a; Fang, 2015), DMPS (Hong et al., 2000; Xue et al., 2007; Wang et al., 2010; Xu et al., 2012a,b; Han et al., 2014; Hu, 2014; Zhang et al., 2014b; Jiang, 2016), Zinc sulfate (Cui and Zhao, 2001; Xiao, 2003). The characteristics of the 18 trials are summarized in Table 1. In four studies (Han et al., 1999; Hong et al., 2000; Xiao, 2003; Chen and Wang, 2010), the preparations were made in hospitals including the associated pharmaceutical quality control. Six studies (Wang et al., 2010; Xu et al., 2012b; Han et al., 2014; Zhang et al., 2014a; Fang, 2015; Jiang, 2016) used a commercial preparation and in 8 studies (Ren et al., 1997; Cui and Zhao, 2001; Xue et al., 2007; Zhang, 2007; Chen and Wang, 2008; Xu et al., 2012a; Hu, 2014; Zhang et al., 2014b) no data on quality control were reported. The constituent of CHM and pharmaceutical quality control in each included study was listed in detail in Table 2.

We accessed the reporting completeness of the material in each study with a rating system, which is related to the information about the botanical material and voucher specimens. Only two studies (Han et al., 2014; Zhang et al., 2014a) are of high quality, which provided the full information about the botanical material and included voucher specimens. Twelve studies (Ren et al., 1997; Han et al., 1999; Hong et al., 2000; Cui and Zhao, 2001; Xiao, 2003; Xue et al., 2007; Zhang, 2007; Chen and Wang, 2008, 2010; Xu et al., 2012a; Hu, 2014; Zhang et al., 2014b) are of moderate quality, which provided partial information about the botanical material and did not provide voucher specimens. Four studies (Wang et al., 2010; Xu et al., 2012b; Fang, 2015; Jiang, 2016) are of low quality with inadequate information and were overall taxonomically inadequate. The quality of each included clinical study is summarized in Table 3.

The score of RoB ranged from 2/7 to 4/7. Of which, 10 studies got two points (Ren et al., 1997; Cui and Zhao, 2001; Xiao, 2003; Xue et al., 2007; Zhang, 2007; Chen and Wang, 2008, 2010; Xu et al., 2012b; Fang, 2015; Jiang, 2016); 7 studies got three points (Han et al., 1999; Hong et al., 2000; Wang et al., 2010; Xu et al., 2012a; Hu, 2014; Zhang et al., 2014a,b); and 1 study got four points (Han et al., 2014). Two studies (Xu et al., 2012a; Han et al., 2014) described the detailed methods for random sequence generation and no studies described allocation concealment. No blinding on patients or personnel was applied. All studies reported drop-out data. Ten studies (Cui and Zhao, 2001; Xiao, 2003; Xue et al., 2007; Zhang, 2007; Chen and Wang, 2008, 2010; Xu et al., 2012a,b; Fang, 2015; Jiang, 2016) were judged as unclear risk of bias for selective reporting. There were baseline comparisons and patients' consent were well reported, and other biases were not found in all included studies. The RoB in each included study is concluded in Table 4.

Twenty-five herbs were included in the 18 studies. The top 13 most frequently used herbs were Radix et Rhizoma Rhei, Rhizoma Coptidis, Rhizoma Curcumae Longae, Rhizoma Curcumae, Radix Salviae Miltiorrhizae, Herba Andrographis, Herba Lysimachiae, Herba Scutellariae Barbatae, Caulis Spatholobi, Rhizoma Alismatis, Radix Curcumae, Radix Scutellariae and Rhizoma Acori Tatarinowii, and all of them were used more than 5 times. The full and validated botanical names of herbs were generalized in Table 5.

The possible mechanisms of the most frequently used herbs and the main active ingredients are as follows: (1) Curcumin: an active ingredient from commonly used herbs like Rhizoma Curcumae Longae, Rhizoma Curcumae, Radix Curcumae and Radix Curcumae can partially restore protein expression of most ATP7B mutants to restore functional copper export (van den Berghe et al., 2000; Zhang et al., 2011; European Association for Study of Liver, 2012). Furthermore, curcumin is an ideal antioxidant, an effective scavenger of reactive oxygen species (Samarghandian et al., 2017), and it exerts anti-fibrotic effect through regulating hepatic stellate cells (HSCs) function (Jin et al., 2016; Liu et al., 2016; Mustafa, 2016). (2) Radix et Rhizoma Rhei: Rhubarb root and its active components have anti-oxidation (Shia et al., 2009), anti-fibrotic (Jin et al., 2005), and anti-inflammation effects (Hwang et al., 2013). (3) Rhizoma Coptidis: Berberine from Rhizoma Coptidis exerted anti-fibrotic and anti-oxidation effects (Zhang et al., 2008). (4) Herba Scutellariae Barbatae: P-coumaric acid from Herba Scutellariae Barbatae possess anti-oxidative activities (Ibrahim et al., 2007) and reverse the ATP7B function defect via regulating pre-mRNA splicing (Lin et al., 2015). (5) Herba Andrographis: Andrographolide from Herba Andrographis displayed anti-inflammatory activity through reducing the expression of pro-inflammatory mediators (Panossian et al., 2002) and exhibited hepatoprotective effects through anti-oxidative effect (Vetriselvan et al., 2011).

Eighteen RCTs involving 1,220 patients suffering from WD were identified. The main findings of this study were that CHM adjuvant therapy could increase 24 h urinary copper excretion, and improve liver function and the total clinical efficacy rate for WD. Two trails (Chen and Wang, 2008, 2010) indicated that CHM monotherapy was not superior to the WCM. Eight out of eighteen studies reported no serious adverse events relevant to CHM formulas, indicating that CHM formulas were generally safe and well tolerated for patients with WD. The possible mechanisms are associated with reversing the ATP7B mutants, and exerting anti-oxidation, anti-inflammation and anti-fibrotic effects. Thus, the findings of the present study suggested, to a limited extent, that CHM paratherapy can be used for WD according to the methodological flaws, whereas the beneficial use of CHM monotherapy for WD still lacks evidence.

There are several limitations in the primary studies. Firstly, although we included RCTs, some inherent and methodological weaknesses still existed in the primary studies: only 2 trials provided sufficient information on how the random allocation was generated, while none of the other trials included reported the allocation concealment. No study employed the blinding procedure, making it difficult to bias results intentionally or unintentionally and to help ensure the credibility of study conclusions. A placebo effect is conceptually defined as the beneficial effect associated with an intervention that does not include the presumed active ingredients; however, CHM placebo are hard to mimic identical interventional herbal prescription due to the fact that CHM is special in color, smell and taste. Thus, placebo-controlled randomized trials are well-recognized method when evaluating the efficacy of CHM treatment. In addition, most trials are without calculating the formal pretrial sample size. The trials with inadequate sample sizes seem to be one risk in exaggerating intervention benefits. Secondly, WD is a chronic disease, which needs life-long treatments. Long-term efficacy and safety are important assessments to determine the clinical effectiveness of an agent in treatment. However, in the present study, treatment duration ranged from 28 to 90 days. Long-term safety of CHM for WD could not be determined because duration of treatment is short and dropouts were only reported in one study. According to other clinical trials for WD (Brewer et al., 2009; Weiss et al., 2015; Nicholl et al., 2017), it is recommended that the treatment duration of further trials must not be >60 days, and must last more than 1 year. Thirdly, clinical heterogeneity would be very significant due to the variations in study quality, intervention of CHM prescriptions, comparators, and outcome measures. Owing to being highly variable in composition and dosage of CHMs, it is difficult to assess the efficacy of a specific CHM by performing pooling analysis. Fourthly, all trials were conducted in China, which may limit the generalizability. Further international multicenter RCTs of CHM for WD are needed, in order to generalize the results worldwide.

Use of CHM for WD patients has increased in the past decades. However, the choice of CHM is mainly empirical and lacking consensus among clinical doctors. The available evidence from the present study supported, to a limited extent, that CHM paratherapy can be used for patients with WD but should not be recommended as monotherapy in WD. In addition, the most frequently used herbs selected by the present study should be considered as herbal prescription for WD and as a candidate for further clinical trials.

In the present study, we identified an area that is worthy of further study. Firstly, the potential benefit of CHM as an adjunct treatment for WD still needs to be further confirmed by high-quality RCTs. Thus, we recommend that CONSORT 2010 statement (Schulz et al., 2010), CONSORT for CHM Formulas (Cheng et al., 2017), and RCTs investigating CHM (Flower et al., 2011) should be used as the guidelines when the designing, registering and reporting of further RCTs. Secondly, WD was thought of as a “rare” autosomal disorder by neurologists, and it proved difficult to conduct large sample RCT. However, this review identified 1,220 subjects with WD from 1997 to 2016. If the primary clinical data of all RCTs were recorded in standard, the evidence would be more reliable. Thus, it is necessary to promote clinical data sharing, as has been suggested by the International Committee of Medical Journal Editors (ICMJE) (Taichman et al., 2017).

WD is caused by ATP7B mutations, resulting in copper accumulation and toxicity. The possible mechanisms of CHM for WD not only involve the targets of the ATP7B gene, but also the multiple targets of copper accumulation in various tissues and organs. Curcumin and P-coumaric acid were reported to reverse the ATP7B function defect. Curcumin could partially restore protein expression by directly enhancing the protein expression of mutant ATP7B with residual copper export activity (van den Berghe et al., 2000; Zhang et al., 2011; European Association for Study of Liver, 2012). The EASL Guidelines recommended that treatment with curcumin might be a novel therapeutic strategy in WD (European Association for Study of Liver, 2012). P-coumaric acid, another ingredient of herbs, can also reverse the ATP7B function defect via a different mechanism by regulating pre-mRNA splicing (Lin et al., 2015).

Copper accumulates in hepatocytes where it induces damage through oxidative stress due to its highly reactive redox capacity (Rosencrantz and Schilsky, 2011). In addition, necrosis and apoptosis triggered immune reaction and inflammation to activate the quiescent HSCs, causing hepatic fibrosis (Jin et al., 2016). The possible pharmacological mechanisms of CHM for copper accumulations of WD are as follows: (1) Antioxidant effects: Curcumin, Anthraquinone (from Radix et Rhizoma Rhei), Danshensu and Salvianolic acid B (from Radix Salviae Miltiorrhizae), were shown to ameliorate the oxidative stress by reducing oxidative stress parameters malondialdehyde, thereby improving the hepatic glutathione content and hepatic superoxide dismutase (SOD) (Liu et al., 2016; Samarghandian et al., 2017), inhibiting the formation of superoxide anions (Shia et al., 2009), and exerting a low level of lipid peroxidase (Mishra et al., 2014; Lee et al., 2016, 2017), leading to maintenance of mitochondrial activity (Zhou et al., 2015). Radix Scutellariae improved the antioxidant capacity by induction of the antioxidative enzymes and removal of reactive oxygen species (ROS) (Pan et al., 2015). P-coumaric acid (Ibrahim et al., 2007), Andrographolideand (Vetriselvan et al., 2011), Caulis Spatholobi (Jeon et al., 2008) and Tanshinone IIA, (Shu et al., 2016) have also been shown to exhibit antioxidant effects; (2) Anti-inflammatory effects: Emodin (from Radix et Rhizoma Rhei), Andrographolide (from Herba Andrographis), Radix Salviae Miltiorrhizae and Curcumin analog demonstrated anti-inflammatory properties by reducing the expression of pro-inflammatory mediators via the NF-kB activation pathway (Lee et al., 2003; Hwang et al., 2013; Yue et al., 2014) and MAPK/AP-1 pathway (Choi et al., 2013), and by inhibiting iNOS and COX-2 expression (Paulino et al., 2016). The bioactive components from Radix Scutellariae (Liu et al., 2016) and Quercetin from Herba Lysimachiae (Wang et al., 2015) have been reported to exhibit anti-inflammatory activity; (3) Anti-fibrotic effects: activation of quiescent HSCs is the major event in hepatic fibrosis (Jin et al., 2016). Skullcapflavone I (from Radix Scutellariae) (Park et al., 2005) and Curcumin (Jin et al., 2016) exerted anti-fibrotic effects by inducing apoptosis or senescence in activated HSCs. Furthermore, Curcumin was found to be an anti-fibrotic mediator that inhibits HSCs activation and the transition to myofibroblast-like cells (Mustafa, 2016). In contrast, Radix et Rhizoma Rhei exerted anti-fibrotic effects by the direct inhibition of stellate cell activation without reducing hepatocyte cell death (Jin et al., 2005). Salvianolic acid A and B from Danshen (Tsai et al., 2011), Berberine (Zhang et al., 2008), Radix Scutellariae (Chen et al., 2013) and Radix et Rhizoma Rhei (Pan et al., 2015) have been reported to prevent hepatic fibrosis in different aspects, including inhibition of proliferation and fibrogenesis of HSCs, and regulation of the antioxidant system and lipid peroxidation. Thus, CHM is likely to be useful as a multi-targeting therapy for WD pathogenesis (Figure 6).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.