Adverse Events Associated With Treatment of Tripterygium wilfordii Hook F: A Quantitative Evidence Synthesis

Background: Tripterygium wilfordii Hook F can cause adverse effects (AEs) in clinical application and may be harmful to human health. This study aim to summarize the AEs caused by T. wilfordii tgpolyglycoside (TWP), the most common preparation of T. wilfordii Hook F for clinical use. Methods: The Cochrane Library, EMBASE, PubMed, and Web of Science were searched to identify potential articles on this topic. All single-arm trials, controlled clinical trials, and randomized controlled trials were selected and summarized. Meta-regression was used to determine the sources of heterogeneity, and subgroups were used to identify factors leading to AEs. Results: Forty-six studies, comprising 25 randomized controlled trials, 13 controlled clinical trials, and 8 single-arm trials, were included in this meta-analysis, representing 2437 enrolled TWP-treated participants. Combined intervention, drug dosage, medication treatment, pharmaceutical manufacturers, and specific organ toxicity were identified as potential factors leading to TWP-induced AEs in this meta-analysis. In patients treated with TWP, the global incidence of AEs was 30.75% (95% confidence interval [21.18–40.33], I 2 = 97%), and that of severe grade AEs was 4.68% (95% confidence interval [0.00–12.72], I 2 = 53%). Organ-specific analyses indicated that TWP treatment elicited intestinal toxicity, reproductive toxicity, hepatotoxicity, nephrotoxicity, hematotoxicity, cutaneous toxicity, and other damages. The AEs analyzed in the subgroups of combined intervention, drug dosage, medication treatment, and pharmaceutical manufacturers were considered as primary outcomes, and organic-specific AEs were considered as secondary outcomes. Conclusions: The occurrence of TWP-induced AEs was systemic, organ-specific, and related to medication course, combined intervention, and drug dosage.


INTRODUCTION
Tripterygium wilfordii Hook F (TwHF), a vine plant, is used as a Chinese herbal medicine (CHM). Currently, the root of TwHF is the main officinal part used in medicinal applications, although its stem and leaf also have the same chemical composition and function as those of the root. TwHF has anti-inflammatory, anti-fertility, anti-tumor, antibacterial, (Reno et al., 2015) and other biological activities. It has determinate and curative effects on immune disorders and is an efficient substitute drug for autoimmune diseases (Liu et al., 2019). Owing to the adverse effects (AEs) of TwHF in clinical preparations, methods for reducing the toxicity and increasing the efficacy of TwHF are being extensively investigated. In order to unify TwHF preparations in the statistics, we analyzed studies that included only the use of T. wilfordii tgpolyglycoside (TWP) treatment. TWP is the most widely used TwHF preparation in clinical application for kidney disease, rheumatism, Crohn's disease, skin diseases, thyroid problems, and Sjogren's syndrome . Several active ingredients such as triptolide, triptonide, celastrol, wilforlide A, and wilforine are contained in TWP, which extracted from TwHF with elimination of the main toxic components (Tian et al., 2012). Despite the sophistication of the drug purification process, related AEs still occur intermittently.
Different types of complex and challenging AEs have evolved owing to the wide clinical application of TwHF. Previous studies have shown that the AEs caused by TWP mainly include leukopenia, gastrointestinal reactions, menstrual disorders, and liver dysfunction . Incidences of leukopenia and liver dysfunction have significantly increased in patients treated for immune system disease, and independent incidences of AEs caused by TwHF have been associated with various organ toxicities, such as intestinal toxicity, reproductive toxicity, hepatotoxicity, nephrotoxicity, hematotoxicity, and cutaneous toxicity Zhang et al., 2018). This indicates that the therapeutic dose of TwHF is close to its toxic dose, resulting in unavoidable occurrence of AEs. However, its safety can be controlled within acceptable limits . Hence, the China State Food and Drug Administration issued a drug warning on TwHF preparations in April 2012 to restrict its clinical application (The State Food and Drug Administration, 2012).
Although some patients experience AEs due to TwHF, reasonable treatment and management strategies can help improve their tolerance to the plant. The dosage and treatment courses of TwHF should be strictly controlled, and patients treated with TwHF should be closely observed for AEs. TwHFinduced AEs attracted public attention for the first time in 2013 (Chen et al., 2013). However, researchers have focused on the incidence of AEs in a single systemic disease (primary nephrotic syndrome), although TwHF is a multi-systemic drug used for treating multi-systemic diseases. Therefore, high-quality metaanalyses of the AEs associated with TwHF are needed.
TWP, the most common preparation of TwHF in clinical practice, was the object of analysis of this study. We collected and analyzed articles on AEs caused by TWP and investigated the main AEs and risk factors associated with TwHF to provide references for future clinical safety.

METHODS
This systematic review was performed following the Cochrane Handbook for Systematic Reviews of Interventions, (Higgins and Green, 2018) presented under the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines shown in Supplementary Table 3.

Selection Criteria
In this analysis, we included studies of randomized controlled trials (RCTs), controlled clinical trials (CCTs), and single-arm trials reporting AEs after the intervention of TWP, irrespective of sex, age, and ethnicity.
The exclusion criteria were as follows: (i) studies on non-TWP extract or compound; (ii) studies not reporting AEs; (iii) articles of meeting abstracts, cell or animal studies, reviews, systematic reviews, and meta-analysis; and (iv) articles with full text not available.

Outcomes
The primary outcomes considered in this study were the incidence of AEs and their grades, which were recorded according to the Common Terminology Criteria for Adverse Events 5.0 (CTCAE 5.0) of the National Cancer Institute. AE grades ≥3 were considered severe. AE incidence was evaluated in terms of drug dosage, treatment duration, combined interventions, and pharmaceutical manufacturers as the primary outcome. Organspecific AEs were measured as the secondary outcomes.

Selection of Studies and Data Extraction
The Cochrane Library, Excerpta Medica Database (EMBASE), PubMed, and Web of Science (WOS) were searched in this study. Studies dating from the earliest citation in the databases till November 2018 were included irrespective of language. The search terms used were "safety, " "side effects, " and "adverse events, " combined with "Tripterygium wilfordii Hook F., " "Tripterygium wilfordii tgpolyglycoside, " "Lei Gong Teng, " and "clinical. " Three investigators independently screened the studies according to the inclusion criteria using self-designed data-extraction templates, and extracted information that included the first author, study characteristics (article type, year, and design), drug characteristics (medication dosage, pharmaceutical manufacturers, combined interventions, administration route, and drug duration), participant characteristics (region, sex, mean age, sample size, and diagnosis), and AE characteristics (incidence rate, occurrence time, specific AE manifestations, and solutions), and then measured the outcomes. Two authors assessed the risk of bias, three authors performed data analysis and interpretation, four authors performed statistical analyses, three authors drafted the manuscript, and all authors read and approved the final manuscript.

Quality Assessments
The Cochrane Handbook (Higgins and Green, 2018) was used to evaluate the methodological quality of the included RCTs regarding the following characteristics: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of the outcome assessment, incomplete outcome data, selective reporting, and other biases. The terms "low, " "unclear, " and "high" referred to low, uncertain, and high risks of bias, respectively. The quality of non-RCTs was evaluated using the Newcastle-Ottawa Scale (NOS) (Stang, 2010) The results were cross-checked by two investigators, and disagreements were settled via discussion.

Statistical Analysis
The incidence of AEs was estimated for the studies included in this meta-analysis. We pooled the incidence of AEs in TWP treatment. Heterogeneity between studies was assessed using the Q test and I 2 statistics. When the heterogeneity was small, that is when I 2 < 50%, the fixed-effect model was used for analysis, and when the heterogeneity was large, that is when I 2 < 50% and P > 0.1, the random-effect model was used for analysis. The overall heterogeneity of the results in this meta-analysis was large; hence, the random-effect model was used to calculate the average statistics of the weighted combination of multiple research statistics. Potential publication bias was examined using funnel plots. Incidence and meta-regression analyses were performed using R software [version 3.8.6 (2018-3-15); Nordic Cochrane Centre, Copenhagen, Denmark] with the package Meta and Metafor function.

Study Selection
Our search strategy identified 2138 potential studies from electronic databases, and the detailed steps are shown in Figure 1. In total, 818 articles were excluded because of duplication and 1321 studies were reserved temporarily. The remaining articles were screened for titles and abstracts, and 172 studies were saved based on our inclusion and exclusion criteria. Furthermore, 126 studies were excluded as they did not meet our expectations upon perusing the entire text. Finally, we included 46 studies (Ao et al., 1994;Jiang, 1994;Zhang et al., 1994;Tao et al., 1995;Ji et al., 1998;Zhou et al., 1999) in the meta-analysis.

Study Characteristics
Specific details of the clinical trials are presented in Table 1 and Figure 2. Overall, 2437 enrolled TWP-treated participants were included in the meta-analysis. There were 25 RCTs, 13 CCTs, and 8 single-arm trials. Seven of these were retrospective studies and 39 were prospective studies (Figure 3). Among the 46 studies, at least one study used TWP as a treatment approach, which was defined as an experimental group or a control group. The sample size of the 46 trials ranged from 1 to 210. Twenty-one studies (with 1075 patients) used TWP as the primary and only treatment approach. Twenty-nine studies (with 1441 patients) used TWP as the active group, and 15 studies (with 658 patients) were used as the control group.
The overall incidence of AEs presented heterogeneity was as high as 30.75% [95% confidence interval (CI) (21.18-40.33), I 2 = 97%] (Supplementary Figure 1). The results of meta-regression of current AEs among patients treated with TWP revealed that high-quality studies, combined medications, medication course, drug dosage, pharmaceutical manufacturers, and organ-specific AEs (P < 0.05) significantly affected heterogeneity (Table 2). However, studies of lower quality were associated with higher P value, suggesting that studies with multivariate adjustment retained some residual confounding observations.
In 36 trials, the TWP-induced AEs usually disappeared after discontinuing the medicine. Only one (with one patient) trial (Li, 2000) showed that the AEs improved at the end of the treatment and disappeared in 3 weeks after withdrawal. However, there was no improvement in nephrotoxicity after 3 months of drug withdrawal in one trial (with one patient) (Tao et al., 1995).

Course of Medication and Combined Medication
In 46 studies, 39.1% (18/46) of the diagnoses were related to urinary system diseases, 23.9% (11/46) to RA, 10.9% (5/46) to digestive system diseases, and 8.7% (4/46) to skin diseases (including 2 cases of psoriasis, 1 case of chronic urticaria, and 1 case of gangrenous pyoderma). There were two cases of Graves' eye disease, one case of uterine fibroids, one case of endometriosis, one case of oral lichen planus, one case of ankylosing spondylitis, one case of lupus erythematosus, and one case of periostitis. The details are shown in Figure 2.
The incidence of AEs of all grades in patients without any combined medication was 40.16%

Dosage-Dependent Analysis of AEs
Twenty-nine of the 46 trials showed the number and grade of AEs after treatment with specific doses of TWP.     Table 3 and Supplementary Figure 5. Forty-three of the 46 studies showed the dose frequency in the meta-analysis. We selected 15 of those trials that involved therapeutic doses ranging from 1 to 1.5 mg/kg/d to avoid heterogeneity, and counted the frequency of medicine consumption. However, in most studies (13/15), patients consumed the medicine thrice a day; hence, we could not determine the correlation between medicine consumption frequency and AEs.

Pharmaceutical Manufacturer
Thirty-five of the 46 studies clearly mentioned the name of the pharmaceutical manufacturer with specific events.  Table 3 and Supplementary Figure 7.

Organ-Specific AEs
We analyzed the incidence of AEs associated with intestinal toxicity, reproductive toxicity, hepatotoxicity, nephrotoxicity, hematotoxicity, and cutaneous toxicity during drug use. The most common AE of all grades caused by TWP was intestinal toxicity, followed by reproductive toxicity. The detailed information regarding organ-specific AEs is shown in Figure 4 and Supplementary Figure 8.

Intestinal Toxicity
The most common digestive AE during treatment with TWP was gastrointestinal discomfort, with an incidence of 13.88% [95% CI (9.12-20.56), I 2 = 85%). In addition, gastric hemorrhage was a serious symptom in the digestive system, although the probability of occurrence was low (Supplementary Figure 9).

Reproductive Toxicity
Amenorrhea was the most frequent AE in patients treated with TWP, and its incidence was 10.16% [95% CI (5.54-18.16), I 2 = 78%). No serious AE occurred in the reproductive system (Supplementary Figure 10).

Cutaneous Toxicity
The incidence of all grades of skin system AEs was less than 5%. The general AEs of the skin included scall, ulcers, and herpes zoster, which were not of severe grade (Supplementary Figure 14).

Other Damages
The incidences of other damages caused by TWP, such as palpitation, headache, elevated blood sugar, and hair loss,    were all <5%. The incidence of other damages was 4.43% [95% CI (2.92-6.67), I 2 = 72%] at a low rate of AEs (Supplementary Figure 15).

Quality Assessment, Heterogeneity, and Publication Bias
The Cochrane Collaboration's tool was used to evaluate the RCTs' methodological qualities in the meta-analysis, and the NOS was used for CCTs and single-arm trials. The overall risk of bias was assessed as low risk. Hence, the quality of the studies included was satisfactory ( Supplementary Tables 1 and 2). The meta-regression analysis revealed that high-quality studies, combined medications, medication course, drug dosage, pharmaceutical manufacturer, and organ-specific AEs (P < 0.05) had a significant effect on heterogeneity in the meta-analysis ( Table 2).
The funnel plot and Egger's test were used for detecting publication bias, which showed an asymmetric distribution of trials along the side of the funnel (Supplementary Figure 16). TWP was used in the different groups of the included studies. Owing to the different variable properties in each article, the results of the meta-analysis may have been affected by interference from unrelated variables. Egger's test showed that P = 0.02364 (< 0.05) indicated the presence of publication bias (Supplementary Figure 17).

DISCUSSION
The purpose of this study was to review the safety of TWP and provide a reference for TwHF preparations. It must be noted that the conclusions of this study can only serve as a reference for AEs caused by preparation types besides TWP, as the main toxicity component triptolide in TwHF is removed in TWP (Tian et al., 2012). In addition, owing to the different forms of articles included in this meta-analysis, the amount of information contained in each article differs. As much as possible, we performed subgroup analysis for studies with the same characteristics, which may have led to differences in total study counts and subgroup study counts. However, incorporating different types of articles is also one of the major strengths of this meta-analysis. A study (which included 428 trials) published in 2016 reported that the incidence of all grades of AEs caused by TwHF preparations was 26.7% ]. The results of that systematic review were similar to our findings, but it included all types of TwHF preparations and 94% (402/428) of the included trials were from Chinese databases . The use of all types of TwHF preparations might increase statistical heterogeneity, and literatures of low quality might lead to low quality of methodology. In this analysis, we excluded other types of TwHF preparations, and the articles were all from English databases to ensure the quality of the outcomes.
We included 46 trials with 2437 patients in this study. The incidence of AEs of all grades in patients treated with TWP was 30.75% (Supplementary Figure 1). In comparison, the incidence of severe-grade AEs was only 4.68% (Supplementary Figure 2). This indicated that the incidence of severe grade AEs caused by TWP was fairly low and that the drug was safe to use. We also observed that the incidence of AEs caused by treatments with TWP without any combination medication was higher than that of AEs caused by TWP treatment combined with other therapeutic methods (Supplementary Figure 3). This indicated that TWP had relatively high toxicity, and hence, AEs could be triggered without combining TWP with other drugs. Next, we conducted a subgroup analysis of the trials according to the combinations of different interventions with TWP and observed the highest incidence in the TWP plus MTX group, followed by the TWP plus NSAIDs and TWP plus antibiotics groups. The group with the lowest incidence was TWP plus Chinese medicine (Table 3). Interestingly, the incidence of FIGURE 5 | Effective constituents of TwHF (such as triptolide, triptonide, celastrol, and wilforine) are both the active ingredients and toxicity source. TwHF induces toxicity in various target organs. In the digestive system, TwHF causes gastric mucosa irritation and increases the infiltration of inflammatory cells in the gastrointestinal tract. TwHF also disrupts the signaling of sex hormones and myeloperoxidase in reproductive organs, causes oxidative stress, increases the ratio of Th17/ Treg, and suppresses hepatic gluconeogenesis. In the kidneys, TwHF decreases the activities of renal SOD and GSH and increases the level of MDA and BUN, resulting in serious oxidative stress and structural damage. TwHF can affect mitochondrial function and reduce the number of platelets, WBC, and lymphocytes. TwHF, Tripterygium wilfordii Hook F; ROS, reactive oxygen species; Plt, blood platelet; WBC, white blood cell; ALT, alanine aminotransferase; AST, aspartate aminotransferase; SCr, serum creatinine; BUN, blood urea nitrogen; SOD, superoxide dismutase; GSH, glutathione; MDA, malondialdehyde; WBC, white blood cell.
AEs caused by TWP with all types of combined medications was lower than that caused by TWP without any combination. This indicated that combination with other drugs may weaken the toxicity caused by TWP and reduce the occurrence of AEs (Supplementary Figure 4). A previous study (Cao et al., 2018) performed a similar subgroup analysis between TwHF monotherapy and combination therapy. Many of the symptoms, including leukopenia, gastrointestinal reactions, irregular menstruation, and abnormal liver function, were observed in patients on TwHF without any combinations at a higher frequency than in patients on combined therapy. However, this previous study included 36 RCTs that were reported in 2015. With the emergence of more evidence, systematic reviews should be updated in a timely manner. In our meta-analysis, high-quality RCTs, CCTs, and single-arm trials were added, and a set of rigid inclusion and exclusion criteria was used. The results proved that AEs observed after treatment with combined medications were related to the original toxicity of the drugs.
Furthermore, we observed that almost all AEs occurred during the duration of the medication. Meanwhile, AEs could be alleviated by reducing the dosage of the drugs or by changing the medicine consumption habit, such as consuming the medicine after meals. The AE grade was reduced after 3 months of drug withdrawal. This showed that the incidence of AEs decreased in patients treated with TWP when the drug was completely metabolized because of the recovery of bodily functions. However, the results might be related to case loss.
The recommended dosage of TWP was 1−1.5 mg/kg/d, and we observed that the incidence of AEs with this dosage was 29.01% ), I 2 = 82%] (Supplementary Figure 18), which was lower than that in any other dosage groups. This indicated that either higher or lower dosage may lead to more serious AEs. In addition, the official recommended usage time is 3 months. We also observed an increase in AE incidence after more than 3 months of use (Supplementary Figure 19). Thus, TwHF should be used in moderation and for a short term.
We intended to statistically determine the relationship of age and gender with TWP application, although it could not be completed due to the lack of information. A study (Liu et al., 2010) showed that AEs caused by TwHF were more frequent in females than in males. Furthermore, toxicokinetic data indicated that TWP has gender-related differences in drug accumulation and toxicities, and that females are more sensitive to TWP toxicity than males . These results suggest that women should consider a protective agent for target organs when using TwHF. Owing to the effects of TWP on female menstruation and male sperm production, the most vulnerable age is the childbearing age.
Previous data show that TwHF has several effective constituents, such as triptolide, triptonide, celastrol, and wilforine, with the most important being triptolide (Fan et al., 2018). These constituents are both the active ingredients and the source of toxicity (Gao et al., 2019). The mechanism of TwHF-induced AEs is complicated, and AEs occur in multiple FIGURE 6 | TwHF inhibits the activity of immune cells and prevents T-cell proliferation with the reduction in proinflammatory cytokines and chemokines; TwHF alters intracellular GSH levels, resulting in the overproduction of ROS, leading to mitochondrial damage; TwHF increases the generation of anion superoxide (O 2-), inhibits the activity of antioxidant enzymes such as SOD and catalase, and induces oxidative stress and dysfunction in organs. Meanwhile, oxidative stress is an important factor in the induction of apoptosis. TwHF treatment activated caspase-3, caspase-8, and caspase-9, and p53 and downregulated Bcl-2 levels in mitochondriainitiated apoptosis. TwHF can increase the expression of GRP78 to inhibit endoplasmic reticulum stress and induce protective autophagy; it also inhibits apoptosis by eliminating dysfunctional mitochondria. TwHF, Tripterygium wilfordii Hook F; GSH, glutathione; ROS, reactive oxygen species; SOD, superoxide dismutase. systems and targets. The effect of TwHF on different systems is shown in Figure 5. In addition, the immunoregulatory effects of TwHF are mainly directed to immune cells, immune molecules, and cellular signaling. TwHF exerts multidimensional regulatory activities on T cells and dendritic cells in vivo (Xi et al., 2017) (Figure 6).
In addition, the quality of a CHM is closely related to the quality of its active ingredients, which is important for both quality evaluation and control. The standard for the extraction of TWP in the market often considers wilforine as the main evaluation index, which is proposed as a biologically active and toxic component of TwHF (Xi et al., 2017). However, TwHF preparations produced by different manufacturers vary in chemical composition, which is consistent with clinical observations . To overcome this problem, we analyzed the drug produced by different pharmaceutical manufacturers (Table 3, Supplementary Figure 7) and observed that the heterogeneity between groups was large. This might be due to the lack of information and the fact that all the pharmaceutical companies were from China.
TWP is used in the treatment of various diseases, although it induces a wide spectrum of AEs. In the included trials, the most common AE was intestinal toxicity (Supplementary Figure 9), followed by reproductive toxicity (Supplementary Figure 10). To summarize, the main AEs that appeared in the included trials were as follows: (i) gastrointestinal tract stimulation, (ii) damage to the reproductive system, (iii) liver or kidney damage, (iv) AEs of skin mucosa, (v) leukopenia, (vi) induced arrhythmia, and (vii) infection or acute poisoning.
In summary, the toxicity of TwHF depends on the dose and time of administration. Notably, both high/short-term doses and low/long-term doses cause AEs (Li et al., 2014). As each individual reacts differently to TWP, which may affect the accuracy of the meta-analysis, we used meta-regression to analyze and divide the subgroups to identify the sources of heterogeneity in this meta-analysis. As TwHF preparations have a prominent anti-inflammatory effect and are clinically irreplaceable, (Yao et al., 2010) we suggest that the clinical dosage of TwHF should be strictly controlled and its usage should be stringently monitored. Furthermore, toxicological studies on TwHF should be conducted to ensure the safety of TwHF preparations.
Our study has some limitations. First, some of the included studies were not RCTs, which was the main source of publication bias. Second, some key information was missing from some of the included studies; for instance, the number of certain specific AEs was not mentioned, and most studies showed the total number of AEs. Furthermore, the clinical environment was highly variable, and hence, the use of the random-effects models was important for controlling variability. Therefore, control variables were used to ensure the quality of the meta-analysis in the subgroup analysis of the drug characteristics and system. Simultaneously, the sample size of different subgroups was small, and some subgroups could not achieve the test effect.
In conclusion, our meta-analysis provided a statistical outline of AEs in patients treated with TWP. The included evidence was entirely from English databases, and the pattern of AE occurrence was organ-specific and related to TwHF. GLOSSARY TWP, Tripterygium wilfordii tgpolyglycoside; TwHF, Tripterygium wilfordii Hook F; AE, adverse event; CHM, Chinese herbal medicine; RCT, randomized controlled trial; CCT, controlled clinical trial; EMBASE, Excerpta Medica Database; WOS, Web of Science; CTCAE 5.0, the Common Terminology Criteria for Adverse Events 5.0; NOS, Newcastle-Ottawa Scale; CI, confidence interval; SCr, serum creatinine; BUN, blood urea nitrogen; AST, aspartate aminotransferase; ALT, alanine aminotransferase; NSAID, non-steroidal anti-inflammatory drug; MTX, methotrexate; RA, rheumatoid arthritis; ER, endoplasmic reticulum; HPLC, high-performance liquid chromatography.

DATA AVAILABILITY STATEMENT
The data sets analysed during the current study are available from the corresponding author upon reasonable request.