Drug-induced liver injury (DILI) is a worldwide health problem and is the fifth leading cause of death associated with liver disease (1). It is essentially an adverse reaction to conventional drugs, herbs, dietary supplements, etc. during treatment, especially with anti-tuberculosis and antipsychotic drugs (2, 3). About 1.7 million cases of death in 10.4 million cases of tuberculosis were estimated by the World Health Organization (WHO) (4). The incidence of DILI will continue to increase in the future due to the use of anti-tuberculosis drugs (5, 6).

Silymarin is a remedy for hepatoprotection with anti-oxidant, anti-inflammatory, anti-fibrotic, and liver protection effects (7-9). Animal and in vitro experiments have demonstrated significant hepatoprotective effects of silymarin on DILI. Ramanathan and Sivanesan reported that 100 mg/kg silymarin significantly improved zidovudine and isoniazid-induced hepatotoxicity in rats (10). Silymarin reduced hepatotoxicity caused by resorcinol and rifampicin by inhibiting glutathione reductase and glutathione peroxidase activity in mice (11). Tasduq et al. (12) confirmed that silymarin could reverse the abnormal increase of liver enzyme indices (including AST and ALT) in anti-tuberculous DILI rats. However, clinically, the therapeutic effect of silymarin on antituberculosis DILI remains controversial. Asgarshirazi et al. (13) reported that silymarin could significantly improve anti-epileptic DILI. Chen et al. (14) reported that silymarin had a preventive effect on DILI due to thioacetamide. A previous study (15) reported that silymarin significantly improved anti-TB DILI. Conversely, randomized clinical trials designed by Marjani et al. (16), Heo et al., Zhang et al. (17) and Gu et al. (18) showed no significant effect of silymarin on anti-TB DILI. Therefore, it is necessary to systematically evaluate the hepatoprotective effect of silymarin on anti-TB DILI.

Recently, Tao et al. (19) in a systematic review, reported that silymarin effectively reduced the occurrence of anti-TB DILI, and the best effect was obtained after four weeks of treatment. However, it is well known that the clinical efficacy of a drug is related not only to the duration of treatment, but also to the administration dose. According to reports, silymarin at 720 mg/d significantly reduced the serum levels of liver enzymes in patients with hepatitis C, while 420 mg/d silymarin had no significant effect (20), suggesting that the dose is a key factor in the hepatoprotective effect of silymarin. Therefore, it is necessary to clarify the role of dose and treatment time in the anti-TB DILI effect of silymarin. Besides, the uncontrolled dose and treatment time of antipsychotic drugs can often cause adverse reactions (21). To date, there is no study evaluating the clinical efficacy of silymarin in preventing antipsychotic DILI. It seems it is time to systematically evaluate the role of dose and treatment time in the anti-TB/antipsychotic DILI effects of silymarin.

We conducted the first systematic review and meta-analysis study to evaluate the clinical efficacy of silymarin in the treatment of anti-TB/antipsychotic DILI in several subgroups based on follow-up time and dose.

2.1. Date Source

Two authors independently searched the literature in PubMed, Scopus, Web of Science, Cochrane, EBSCOhost, clinicaltrials.gov, China Journal Full-text Database (CNKI), Wanfang Database, and China Medical Bio-Documentary Database (CBM). Keywords and free words were as follows: “silymarin” or “silibinin” or “silybum” or “silybin” or “silydianin” or “silychristin” or “milk thistle”, “anti-tuberculosis drug-induced liver injury, antipsychotic drug-induced liver injury” and all results intersected with “randomized controlled trial OR randomized”. The search time was from the establishment of the library to September 2019. The detail of the literature search process is summarized in Figure 1.

Figure 1. Flow diagram of the study selection process in the meta-analysis

2.2. Inclusion and Exclusion Criteria

The inclusion criteria were as follows: (1) randomized clinical trial, (2) patients receiving antituberculosis treatment, (3) test group treated with silymarin and control group with accepted placebo, (4) Chinese language literature with observations or control groups of more than 100 patients, and (5) incidence, side effects, time of occurrence, AST, ALT, and TBIL with at least one outcome indicator. Meta-analyses, reviews on outcome indicators, meeting abstracts, duplicated studies, and editorial comments were excluded. Two reviewers independently screened relevant articles, and inconsistency was discussed with the third author. The Kappa coefficient was 0.8134.

2.3. Data Extraction

The first author, sample size, intervention methods, treatment time, dosage, and outcome indicators were extracted by two authors (SK and HY). Differences and inconsistencies between the two reviewers were coordinated by the third author (ZL). The Jadad quality scale was used to evaluate the quality of the literature in terms of random sequence generation, allocation concealment, double-blinding, and description of withdrawals and drop-outs. Scores 0 - 3 indicated a low quality and 4 - 7 indicated a high quality.

2.4. Statistical Analysis

The meta-analysis was performed using RevMan 5.0 software. Dichotomous variables of count data were calculated by the Mantel-Haenszed (M-H) method, and the effect size was expressed by the odds ratio (OR) and the 95% confidence interval (CI). Continuous variables were analyzed by the Inverse Variance (IV) method, and the effect value was expressed by the Mean Difference (MD) or Standard Mean Difference (SMD) and 95% CI. The Q test and chi-square test were used for heterogeneity analysis. P ≥ 0.1, I² ≤ 50%, combined analysis using the fixed-effect model; P ≤ 0.1, I² ≥ 50%, using the random-effects model for combined analysis. Sensitivity analysis was performed to test the influence of a single study on the overall effect size by the leave-one-out method.

3.1. Outcome Studies

Our search retrieved 496 articles related to silymarin or anti-TB/antipsychotic DILI. However, 370 duplicate articles were excluded. By reading the title and the abstract of articles, 109 articles were excluded because of being a review study, animal study, or doing an inappropriate intervention. After reading the full texts, eight articles did not meet the inclusion criteria (no available data, non-RCT, and observations or control groups of less than 100 patients in Chinese language literature), as shown in Table 1. In total, nine studies were included. RCTs were screened independently by two authors (SY and HY) through all databases, as shown in Figure 1.

3.2. Basic Characteristics of the Study

Nine RCTs with 2,712 participants (1,351 cases in the silymarin group and 1,361 cases in the control group) were evaluated. The first author, study design, sample size, intervention measures, follow-up, dosage, and related outcome indicators are shown in Table 2.

3.3. Result of Data Analysis

Nine studies contributed to the occurrence of anti-TB/antipsychotic DILI analysis and participants were separated into five subgroups with different dosages and follow-up periods. Sensitivity analysis showed that silymarin at less than 300 mg/d significantly reduced the occurrence of anti-TB/antipsychotic DILI, whether it was applied for two weeks, four weeks, or eight weeks [subtotal OR: 0.78, 95% CI (0.41, 1.32), P = 0.56, I² = 0% for two weeks; subtotal OR: 0.28, 95% CI (0.13, 0.60), P = 0.28, I² = 22% for four weeks; subtotal OR: 0.65, 95% CI (0.31, 1.35), P = 0.60, I² = 0% for eight weeks; pooled OR: 0.53, 95% CI (0.35, 0.78), P = 0.42, I² = 3%] (Figure 2). Silymarin at 315 mg/d significantly reduced the occurrence of anti-TB/antipsychotic DILI [subtotal OR: 0.17, 95% CI (0.08, 0.39), P = 0.04, I² = 76%] for eight weeks but no significant difference was found between the over 400 mg/d silymarin group and the control group [subtotal OR: 0.93, 95% CI (0.20, 4.39), P = 0.02, I² = 76%] (Figure 2). No significant difference was found in the occurrence of adverse events between the control and silymarin groups [pooled OR: 0.94, 95% CI (0.71, 1.25), P = 0.46, I² = 0%] (Figure 3).

Figure 2. Effect of different dosages and treatment times of silymarin on the occurrence of anti-TB/antipsychotic DILI

Figure 3. The occurrence of adverse events by silymarin

Compare to the control group, 315 mg/d silymarin significantly reduced serum liver enzymes AST, ALT, and TBIL in eight weeks [subtotal MD: -29.00, 95% CI (-31.57, -26.43) for AST; subtotal MD: -124.00, 95% CI (-126.76, -121.24) for ALT; subtotal MD: -12, 95% CI (-12.72, -11.28) for TBIL], all with low heterogeneity (P = 1.00, I² = 0%). Silymarin at less than 300 mg/d significantly reduced serum liver enzymes AST and ALT in two weeks, four weeks, and eight weeks [AST subtotal MD: -2.89, 95% CI (-6.87, 1.09) for two weeks; AST subtotal MD: -5.74, 95% CI (-10.69, -0.80) for four weeks; AST subtotal MD: -5.22, 95% CI (-9.92, -0.51) for eight weeks, with low heterogeneity (P > 0.29, I² < 10%); ALT subtotal MD: -3.30, 95% CI (-7.15, 0.56) for two weeks; ALT subtotal MD: -3.55, 95% CI (-10.91, 3.80), P = 0.07, I² = 70% for four weeks; ALT subtotal MD: -3.48, 95% CI (-10.58, 3.62) for eight weeks, with low heterogeneity (P > 0.61, I² = 0%)]. No significant difference was found between the silymarin and placebo groups in terms of TBIL change [TBIL subtotal MD: 0.02, 95% CI (-0.07, 0.11) for two weeks; TBIL subtotal MD: -0.12, 95% CI (-0.50, 0.25) for four weeks; TBIL subtotal MD: -0.02, 95% CI (-0.12, 0.08) for eight weeks, with low heterogeneity (P > 0.61, I² = 0%)], as shown in Figure 4.

Figure 4. Effect of silymarin on serum liver enzymes AST, ALT, and TBIL

The time of the occurrence of anti-TB/antipsychotic DILI in the silymarin group and control group was analyzed. The results showed that silymarin prolonged the occurrence time of DILI, which was 1.78 times that of the control group (Figure 5).

Figure 5. Time of occurrence of anti-TB/antipsychotic DILI in silymarin group and control group

3.4. Evaluation of Publication Bias

Considering all nine RCTs, the funnel plot of the occurrence of liver injury in each subgroup was basically symmetrical (Figure 6) with no significant publication bias. However, heterogeneity was found in the subgroups of 315 mg/d for eight weeks and over 400 mg/d.

Figure 6. Publication bias of RCTs

Recently, the hepatoprotective effect of silymarin as a liver-protecting drug has been controversial. Heo et al. (15) reported that silymarin significantly reduced serum biochemical parameters in patients with anti-TB DILI. Fathalah et al. (20) reported that 720 mg/d silymarin significantly reduced the serum liver enzyme index in patients with hepatitis C. However, clinical trial results by Fried et al. (27) showed that high doses of silymarin did not significantly reduce the serum ALT levels in patients with chronic hepatitis. A meta-analysis showed that silymarin did not significantly improve serum biochemical parameters in patients with liver disease (28). Therefore, it was important to evaluate whether silymarin has a liver-protecting effect and to determine the dosage of silymarin. Besides, Zhang et al. (17) reported that silymarin had no significant effect on anti-TB DILI, but instead it caused liver damage, suggesting that silymarin may be ineffective in anti-TB DILI. However, Tao et al. (19) reported that silymarin significantly reduced the incidence of anti-TB DILI and for four-week treatment was the best duration to protect the liver. It indicated that the treatment time is one of the effective factors in the treatment of anti-TB DILI with silymarin. Our meta-analysis aimed to systematically evaluate the clinical efficacy of silymarin in the treatment of anti-TB/antipsychotic DILI divided into several subgroups based on follow-up time and dose.

Overall, silymarin significantly reduced the occurrence of anti-TB/antipsychotic DILI compared to the control group and no significant adverse reaction was found. Under the premise of consistent doses, the results were similar to those reported by Tao et al. (19). The optimal treatment time for silymarin for preventing anti-TB/antipsychotic DILI was four weeks. The occurrence of anti-TB/antipsychotic DILI significantly reduced by treatment with 315 mg/d silymarin for eight weeks. It was worth to mention that no significant change was found by treatment with a dose of greater than 400 mg/d (including 400 mg/d and 420 mg/d). There are two possible explanations for the difference. The small sample size may be one of the reasons. Additionally, over 400 mg/d subgroup underwent treatment for two weeks, four weeks, and eight weeks. Different treatment times may have led to inconsistencies between the results obtained by Marjani et al. (16), Zhang et al. (17), and Luangchosiri et al. (22) reported that silymarin effectively inhibited the occurrence of anti-TB DILI, while Marjani et al. (16) and Zhang et al. reported no effect.

Besides, we performed a subgroup analysis of serum liver enzyme indicators in anti-TB/antipsychotic DILI patients. The results showed that the patients’ serum liver enzymes significantly reduced by treatment with 315 mg/d silymarin for eight weeks. Serum AST and ALT levels significantly reduced after two weeks, four weeks, and eight weeks by treatment with less than 300 mg/d and TBIL levels significantly reduced after four weeks by treatment with less than 300 mg/d. However, serum TBIL did not change significantly after two weeks and eight weeks by treatment with less than 300 mg/d. It was suggested that silymarin could improve the serum liver enzyme index of anti-TB/antipsychotic DILI patients to a certain extent, which was related to the treatment time. The duration of four weeks was the best time for silymarin to treat anti-TB/antipsychotic DILI without considering the dosage. It may be related to the presence of autoimmunity in anti-TB/antipsychotic DILI patients. According to reports, DILI patients have autoimmune properties, mostly in the first month of treatment for anti-TB DILI (29, 30). That is, the autoimmunity of anti-TB/antipsychotic DILI patients prolonged the time of DILI and silymarin had a little effect on treatment in two weeks.

Indeed, this meta-analysis had some limitations. The RCTs Jadad rating scale scores for the subgroup of 315 mg/d silymarin treatment for eight weeks showed low literature quality (Table 1). In addition, the > 400 mg/d silymarin subgroup comprised of treatment times of two weeks, four weeks, and eight weeks, reducing the reliability of the results. Heterogeneity was found in the subgroup of silymarin treated with 315 mg/d silymarin for eight weeks and a dose of greater than 400 mg/d (Figure 5). Last but not least, five of nine RCTs did not report serum liver enzymes, and only reported the incidence of anti-TB/antipsychotic DILI. Therefore, the efficacy of 315 mg/d and > 400 mg/d silymarin in preventing anti-TB/antipsychotic DILI remains to be further studied.

Above all, silymarin significantly reduced the occurrence of anti-TB/antipsychotic DILI compared to the control group and no significant adverse reaction was found. The optimal treatment time for < 300 mg/d silymarin to prevent anti-TB/antipsychotic DILI was four weeks. Silymarin treatment with 315 mg/d for eight weeks was the best choice and the quality of RCTs was not considered. The optimal treatment time of silymarin to prevent anti-TB/antipsychotic DILI was related to its dose.