A Systematic Review of Herbal Medicine in the Treatment of Cirrhotic Ascites

1) Database selection and literature search

The Korean literature databases of National Digital Science Library (http://www.ndsl.kr), Korean Medical Database http://kmbase.medric.or.kr), Korean studies Information Service System (http://kiss.kstudy.com), Society KISTI (http://society.kisti.re.kr), KoreaMed (http://www.koreamed.org), Korean Traditional Knowledge Portal (http://www.koreantk.com), and Korea Institute of Oriental Medicine (http://oasis.kiom.re.kr) were searched to identify Korean articles. The keywords used in the Korean search engines were ‘liver cirrhosis’, ‘ascites’, or ‘cirrhotic ascites’, ‘herbal medicine’, and ‘randomized controlled trial’.

The databases of PubMed (http://www.ncbi.nlm.nih.gov/pubmed/), EMBASE (http://www.embase.com), Cochrane CENTRAL (http://www.thecochranelibrary.com), and CNKI (http://www.cnki.net were searched to identify international articles, and the keywords used in the search engines were ‘liver cirrhosis’ or ‘ascites’, ‘herbal medicine’ or ‘traditional Chinese medicine’, and ‘randomized controlled trial’

2) Study selection

Two investigators (LYR and CNK) independently searched and selected the studies. The identified studies were combined, and duplicates were removed. The titles and abstracts of studies were initially screened based on the inclusion and exclusion criteria, and only RCTs that used herbal medicine treatment for the purpose of treating cirrhotic ascites were selected through full-text review. Any disagreements between the two investigators were resolved by majority vote after review by a third investigator (KKS).

The search was conducted between March 14 and 16, 2019 for studies published up to February 2019 without language restriction. Of the identified articles, RCTs that provided oral herbal medicine prescription in patients with cirrhotic ascites were selected, and studies were excluded if they were non-clinical in vivo or in vitro studies, not an original publication, reported as abstracts only, or provided herbal medicine treatment as an enema or a patch were excluded (Fig. 1).

Fig. 1

Flow diagram of the selection process

3) Data analysis

Following a full-text review of the selected studies, data on publication year, number of participants, treatment method and length of intervention and control groups, assessment tools, therapeutic effect, and incidence of adverse events of each study were extracted. Any disagreements on data extraction were resolved by consultation with a third investigator.

4) Risk of bias assessment

The Cochrane’s Risk of Bias (RoB) tool was used as this systematic review only included RCTs. This tool assesses bias across seven domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessors, handling of incomplete outcome data, selective outcome reporting, and other sources of bias that compromise the validity. For detailed criteria of each domain, NECA’s systematic review manual12) was used.

5) Publication bias

Publication bias is a type of reporting bias that occurs when the likelihood of a study being published is associated with statistical significance of study’s outcome. It is commonly assessed using a funnel plot, and Egger’s regression test was performed as a statistical test.

6) Meta-analysis

Statistical analysis of the studies included in the systematic review was performed using the Cochrane Collaboration’s RevMan (Review Manager) version 5.3. To measure effect size, risk ratio (RR) with two-sided 95% confidence interval (CI) was calculated for dichotomous outcome variables mean difference (MD) with 95% CI was calculated for continuous outcome variables. After assessing heterogeneity using Higgins I2) test, the summary statistics from each study were combined for meta-analysis using both fixed-effects and random-effects models.

1) Study design

The thirteen studies ultimately selected were RCTs conducted in China and Iran between 2005 and 2018. These articles investigated the effects of oral herbal medicine on cirrhotic ascites when used as a stand-alone therapy or in combination with acupuncture and western medicine (Table 1).

Table 1

Summary of randomized controlled trials

2) Intervention and control groups

(2) Intervention group design

Of the selected studies, 10 studies^{13–16,18–20,22,24–25)} provided herbal medicine treatment in combination with supportive treatment of western medicine and compared its effect with that of supportive treatment alone. One study21) compared herbal medicine stand-alone therapy with supportive treatment. Two studies17,23) provided additional herbal medicine; one study provided herbal medicine, acupuncture, and supportive treatment and compared it with supportive treatment17), and one study provided herbal medicine in combination with moxibustiom and supportive treatment and compared it with supportive treatment23). The herbal preparations used for intervention were as follows (Table 2).

Table 2

Composition of the herbal preparations used in treatment

3) Outcome measures

Outcome measures used in the selected studies were effective rate, ascites regression time, liver function test, 24 hour urine volume, and Child-Pugh score. Of these, effective rate was used in 10 studies^{15–19,21–25)}, liver function test was used in six studies^{13–15,22–24)}, and ascites regression time was used in five studies^{14,19–20,24–25)} as outcome measures. Of the outcome measures, effective rate, liver function test, and ascites regression time were included in the meta-analysis.

4) Therapeutic effect

Of the selected studies, 10 studies that compared herbal medicine-supportive treatment combination therapy with supportive treatment stand-alone therapy^{13–16,18–20,22,24–25)} were included in the meta-analysis.

(1) Effective rate

Effective rate of herbal medicine-supportive treatment combination therapy was 1.27 times higher than that of the supportive treatment stand-alone therapy, indicating herbal medicine-supportive treatment combination therapy is significantly more effective in treating cirrhotic ascites. A low degree of heterogeneity was observed across the studies (Risk Ratio=1.27, 95% CI: 1.17 to 1.38, p<0.001, I²=0%, Fig. 2)

Fig. 2

Results of Meta-analysis for the Effective Rate(HM+ST vs ST)

Abbreviations: HM, herbal medicine; ST, supportive treatment

(2) Liver function test

A subgroup analysis was performed to compare liver function test of herbal medicine-supportive treatment combination therapy and supportive treatment stand-alone therapy. The intervention group showed significant effect than the control group and a high degree of heterogeneity of 98% was observed across the studies (Mean Difference =−11.39, 95% CI:−15.08 to −7.69, p<0.001, I²=98%, Fig. 3).

Fig. 3

Results of Meta-analysis for the Liver Function Test(HM+ST vs ST)

Abbreviations: HM, herbal medicine; ST, supportive treatment; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TB, total biilirubin; ALB, albumin

(3) Ascites regression time

Although not statistically significant, ascites regression time of the herbal medicine-supportive treatment combination therapy was 30.76 days less than that of the supportive treatment stand-alone therapy. A high degree of heterogeneity of 100% was observed across the studies (Mean Difference=−30.76, 95% CI:−62.37 to 0.84, p=0.06, I²=100%, Fig. 4).

Fig. 4

Results of Meta-analysis for the Ascites Regression Time(HM+ST vs ST)

Abbreviations: HM, herbal medicine; ST, supportive treatment;

5) Bias assessment

Assessment of risk of bias was performed for the 13 articles included in this review using Cochrane risk of bias tool. With respect to random sequence generation domain, three articles13,21,23) that used a table of random numbers were assessed as having low risk, while the rest of the articles provided no details and were assessed as having unclear risk. All articles provided no details on allocation concealment, and therefore were assessed as having unclear risk. With respect to blinding of participants and personnel domain, one article13) that used placebo was assessed as having low risk, while the rest of the articles were assessed as having high risk because blinding was deemed not possible due to the nature of intervention method that included herbal decoction. With respect to blinding of outcome assessor domain, one article13) that reported using blinding was assessed as having low risk, while the rest of the articles provided no details and were assessed as having unclear risk. With respect to incomplete outcome data domain, 10 articles^{13,16–17,19–25)} that reported missing values and confirmed no errors were assessed as having low risk, while the rest of the articles did not provide details and were assessed as having unclear risk. With respect to selective reporting domain, 10 articles^{13–17,19–22,25)} that did not have study protocol but reported all pre-specified outcomes and all expected outcomes were assessed as having low risk, while the rest of the articles provided no details and were assessed as having unclear risk. With respect to other sources of bias domain, most articles did not provide sufficient information or evidence, and therefore all articles were assessed as having unclear risk (Fig. 5, Fig. 6).

Fig 5

Risk of bias graph

Fig. 6

Risk of bias summary

6) Publication bias

Inspection of the funnel plot showed low risk of publication bias (Fig. 7)

Fig. 7

Funnel plot for the outcome of effective rate (HM+ST vs ST)

Abbreviations: HM, herbal medicine; ST, supportive treatment;

7) Safety

Of the 13 studies included in this review, a total of three studies20,22,25) reported adverse events. One study 20) one case of hyponatremia and one case of renal function occurred in the intervention group, whereas three cases of hyponatremia, three cases of renal dysfunction, and two cases of hepatic encephalopathy occurred in the control group, indicating that the safety of the intervention group was significantly higher than that of the control group (p<0.05). In another study22), symptoms of headache and nausea occurred, but there was no significant difference between the intervention and control groups. In the third study25), four cases of hyponatremia and one case of hepatic encephalopathy occurred in the intervention group, whereas 14 cases of hyponatremia, eight cases of renal dysfunction, and seven cases of hepatic encephalopathy occurred in the control group, indicating the safety of the intervention groups was significantly higher than that of the control group (p<0.05)