A Systematic Review and Network Meta-Analysis about the Efficacy and Safety of Tripterygium wilfordii Hook F in Rheumatoid Arthritis

Objective This study aims to evaluate the efficacy of various conventional synthetic DMARDs, including Tripterygium wilfordii Hook F (TwHF) for treating rheumatoid arthritis (RA) by network meta-analysis. Methods We retrieved the related literature from online databases and supplemented it by using a manual retrieval method. Data was extracted from the literature and analyzed with STATA software. Results A total of 21 trials (5,039 participants) were identified. Assessment of ACR20 response found that TwHF combined with methotrexate (MTX) had the greatest probability for being the best treatment option among the treatments involved, while TwHF used singly was second only to TwHF combined with MTX. Assessment of ACR50 response found that TwHF combined with MTX ranked second in all treatment options after cyclosporine A (CsA) combined with leflunomide (LEF) and TwHF alone, followed by TwHF combined with MTX. Assessment of ACR70 response found that CsA combined with LEF ranked first, TwHF combined with LEF ranked second, TwHF combined with MTX ranked third, and TwHF used singly ranked fourth. In the safety analysis, TwHF had the least probability of adverse event occurrence, followed by TwHF combined with MTX, which ranked first and second, respectively. Conclusion Compared with the current csDMARDs for treating RA, the efficacy of TwHF was clear, and TwHF combined with MTX performed well under various endpoints. In the future, large, rigorous, and high-quality RCTs are still needed to confirm the benefits of TwHF therapy on RA.


Introduction
Rheumatoid arthritis (RA) is a common systemic immune disease which is characterized by joint inflammation, destruction, and deformity associated with chronicity and a high rate of disability. Improvement in treatment, stopping progression, and optimizing quality of life are priorities in the field of rheumatology in China. Tripterygium wilfordii Hook F (TwHF) refers to the dry root or root xylem of the celastraceae plant Tripterygium wilfordii, a widely used herb in traditional Chinese medicine (TCM). In accordance with TCM theory, TwHF is considered a key herb for treating persistent rheumatoid arthritis, due to its strong efficacy in eliminating wind-damp and promoting blood circulation to dredge collaterals. In recent years, TwHF prepared by extracting the essence of the active components of Tripterygium wilfordii has been used in clinical practice to treat a variety of rheumatic immune diseases, including RA [1][2][3][4]. It is noted that TwHF has been found to possess toxicity and is associated with having adverse events, such as hepatorenal toxicity, reproductive toxicity, and hematologic toxicity. Network meta-analysis (NMA) is a technique for comparing three or more interventions simultaneously in a single analysis by combining direct and indirect evidence and ranking the efficacy. Compared with traditional metaanalysis, NMA may assist in comparing the efficacy of multiple interventions for a disease more comprehensively, to provide more rigorous evidence through greater synthesis of information. ere are many meta-analyses on TwHF in treating RA [5,6], but most are pairwise comparisons, which have limited ability to illustrate the individual differences among multiple disease-modifying antirheumatic drugs (DMARDs). is study is distinguished as a NMA which includes new and recent studies to evaluate the efficacy and safety of commonly used conventional synthetic DMARDs as both monotherapy and combination therapy for treatment of RA, including TwHF, methotrexate (MTX), leflunomide (LEF), sulfasalazine (SSZ), cyclosporine A (CsA), tacrolimus (FK506), minocycline (MINO).

Study
(i) e subjects were diagnosed with RA in accordance with the 1987 Guidelines of the American Rheumatology Association [7] or the 2010 ACR/European League against Rheumatism (EULAR) Criteria [8]; without diagnosis of other autoimmune diseases or serious cardiovascular and cerebrovascular diseases; no restrictions on age, sex, race, or nationality.

Types of intervention:
(i) TwHF, MTX, LEF, SSZ, CsA, FK506, and MINO used singly or as a two-drug combination in the treatment of RA. TwHF includes both tripterygium glycoside tablet and tripterygium tablet, the two root preparations of TwHF that have shown therapeutic promise [9,10]. e time limit for intervention was ≥12 weeks. Use of nonsteroidal antiinflammatory drugs, folic acid, vitamins, calcium tablets, and low-dose hormones as adjuvant therapy during the treatment was not limited.
Types of outcome measures: (i) Primary outcome: the American College of Rheumatology (ACR) response criteria ACR20 [11]. (ii) Secondary outcomes: ACR50, ACR70, and incidence of adverse events. All literature studies on adverse events were included, inclusive of all types of adverse events; (iii) e analyses of outcomes were conducted on an intent-to-treat (ITT) basis, or modified ITT (number actually receiving treatment at baseline) if the number randomized to treatment was not reported.

Exclusion Criteria
(i) Publications where full text literature cannot be obtained; (ii) Studies where research data are incomplete or cannot be extracted for analysis; (iii) Interventions as herbs containing TwHF.

Data Extraction and Quality Assessment.
e literature screening and extraction were carried out by two researchers, respectively, according to the inclusion criteria for literature retrieval. After the preliminary screening of titles and abstracts, the full text was screened, and the literature inclusion and data extraction were carried out based on intentionality analysis. Finally, the data extracted was compared and sorted. Two authors independently evaluated the methodological quality of eligible publications by using the Cochrane Collaboration's tool for assessing the risk of bias [12] (random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other sources of bias). If there were differences, a third-party researcher was invited to assist the ruling.

Data Synthesis and Analysis.
e primary outcome of this analysis was the American College of Rheumatology (ACR) response criteria: ACR20. e ACR20 is defined as a reduction by 20% or more, in the number of tender and swollen joints plus 20% improvement in at least three of the following five measures: pain, patient global assessment, physician global assessment, a score of physical disability, and blood acute-phase reactants. e secondary outcomes were ACR50, ACR70, and adverse events. e ACR50 is defined as an improvement of 50% or more in the number of tender and swollen joints, plus 50% improvement in at least three of the aforementioned five measures. e ACR70 is defined as an improvement of 70% or more in the number of tender and swollen joints, plus 70% improvement in at least three of the aforementioned five measures.

Network Meta-Analysis.
Results are reported as odds ratios (ORs) with 95% confidence intervals (CI) for all comparisons of interventions. Initially, traditional pairwise meta-analysis was performed by using a random-effects model. en network meta-analysis was performed to compare different therapies by using a frequentist approach. We included multi-arm trials in the analysis by breaking multi-arm trials into separate two-arm trials. We employed a multi-variate random-effects meta-analysis model for each outcome separately, combining direct evidence for each comparison [13,14].
For each "loop" of treatment comparisons from three or more independent sources and for each outcome, we computed the difference between estimates from direct and indirect evidence on the log OR scale. Inconsistency was defined as disagreement between direct and indirect evidence with a 95% CI excluding 0. For each outcome, we estimated the probability of which intervention was the best for each outcome, the second best, the third best, and so on, from the ranked order of the treatments at each interaction. ese ranking probabilities were used to calculate the surface under the cumulative ranking curve (SUCRA), which is expressed as percentage (100% for the best intervention, 0% for the worst intervention, and approximately 50% for equivalent interventions) [15].

Funnel Plot and Publication Bias.
e difference between the observed effect size and comparison-specific summary effect for each study was calculated. is variable was then regressed on the standard error (SE), thus adding a simple linear regression line in the funnel plot. is method could help to visually determine if there is a publication bias in the results between small and large studies. We performed traditional and network meta-analysis by using Stata software (version 12.0, the StataCorp, College Station, Texas, USA).

Results
e flow chart of studies considered for inclusion is shown in Figure 1. On the basis of the title and abstract, 113 publications were selected and analyzed in full text versions. Eventually, 21 publications were included in the systematic review, and the characteristics of the literature were extracted as Table 1. Figure 2 shows the network of all treatment comparisons analyzed according to ACR 20, 50, 70, and adverse events. All reviews followed the methods in the Cochrane Handbook, including standardized searches, inclusion criteria, and outcomes. Table 1 summarizes the clinical and methodological characteristics as well as the main outcomes of each trial. A total of 21 trials (5,039 participants) were identified, and the characteristics of the literature were extracted as shown in Table 1. e risk of bias assessments for the included trials is illustrated in Figure S2 and Figure S3. Most of the evidence was of moderate-togood quality. All 21 RCTs mentioned the word "randomization". Over half of the studies did not report adequate information about allocation sequence generation and allocation sequence concealment. Unblinded designs were used in over half of the trials included.

ACR20.
In the evaluation of the ACR20 response, 21 studies were included, involving a total of 5039 patients, including a total of 12 kinds of interventions. e interventions were MTX, TwHF, TwHF combined with MTX, LEF, TwHF combined with LEF, SSZ, SSZ combined with MTX, CsA, CsA combined with LEF, FK506, MINO, and placebo ( Figure S4A). Efficacy was evaluated by drawing cumulative probability diagram, probability efficacy ranking table (Table 2), and inverted triangle table (Table 3). According to the analysis results, TwHF combined with MTX had the greatest probability of the best efficacy among the treatments involved, and TwHF used singly ranked second ( Figure S5A).

ACR50.
In the evaluation of ACR50 response, 15 literature studies were included, involving 2,968 patients, including 11 interventions: MTX, TwHF, TwHF combined with MTX, LEF, TwHF combined with LEF, SSZ combined with MTX, CsA, CsA combined with LEF, FK506, and placebo ( Figure S4B). e efficacy was evaluated by drawing a cumulative probability diagram, a probability efficacy ranking table (Table 2), and an inverted triangle table (Table S1). According to the analysis results, the efficacy of TwHF combined with MTX ranked second only to CsA combined with LEF in all treatment schemes, while TwHF alone ranked third in all treatment schemes, second only to TwHF combined with MTX ( Figure S5B).

ACR70.
In the evaluation of ACR70 response, 10 literature studies were included, involving 2,374 patients, including 11 interventions: MTX, TwHF, TwHF combined with MTX, LEF, TwHF combined with LEF, SSZ, SSZ combined with MTX, CsA, CsA combined with LEF, FK506 and placebo ( Figure S4C). e efficacy was evaluated by drawing a cumulative probability diagram, a probability efficacy ranking table (Table 2) and an inverted triangle table (Table S2). According to the analysis results, CsA combined with LEF ranked first, TwHF combined with LEF ranked second, TwHF combined with MTX ranked third, and TwHF used singly ranked fourth ( Figure S5C).

Adverse Events.
In the analysis of incidence of adverse events, 13 literature studies were included, involving a total of 3,415 patients, including 11 interventions: MTX, TwHF, TwHF combined with MTX, LEF, TwHF combined with LEF, SSZ combined with MTX, CsA, CsA combined with LEF, FK506 and placebo ( Figure S4D). Incidence of adverse events was evaluated by drawing a cumulative probability diagram, a probability efficacy ranking table (Table 2), and an inverted triangle table (Table S3). According to the analysis results, TwHF and TwHF combined with MTX, ranked first and second, respectively ( Figure S5D).

Forest Plots.
In this study, a forest plot was drawn to assess for inconsistency, as shown in Figure S6A through to S6D. With exception of the M-S-T closed loop with ACR20 as the endpoint, there was no obvious inconsistency in all other closed loops. After analyzing the literature included in the M-S-T closed loop with ACR20 as the endpoint, it is considered that the sources of inconsistency may include different treatment time, different drug doses, heterogeneity caused by allowable adjuvant drugs.

Publication
Bias. In addition, this study also evaluated publication bias with funnel plots ( Figure S7A through to S7D). e scatters in the 4 funnel plots were almost symmetrical visually, and occasionally a small number of scatters were slightly less symmetrical, indicating that the publication bias in the included studies was overall satisfactory.

Discussion
TwHF is considered one of the most effective traditional Chinese herbal medicines against rheumatoid arthritis. Extracts of TwHF have been used for hundreds of years in China to treat various symptoms and, over the past 30 years, extracts of TwHF have become a standard therapy for rheumatoid arthritis in China. An earlier meta-analysis on treating RA bone destruction with TwHF was conducted by the team, and the results showed that the TwHF group was superior to the positive drugs MTX and SSZ used in the control group in Van der Heijde modified total sharp score (mTSS), joint erosion (JE), and joint space narrowing (JSN) on X-ray films, with statistical differences (P < 0.01). In the aspects of mTSS, joint erosion, and joint space narrowing, TwHF is better than MTX and SSZ. e analysis results showed that TwHF can effectively delay the bone destruction process of RA [5]. Network meta-analysis is a further development and extension of traditional meta-analysis. e biggest advantage of NMA is that it can evaluate different interventions for the treatment of similar diseases for quantitative statistical analysis and comparison. In recent years, the number of NMAs published in various journals and magazines has increased to provide guidance for clinicians in choosing effective interventions. In a previous NMA analysis that was conducted on the efficacy and safety of using DMARDs singly represented by TwHF in the treatment of RA, we found that TwHF is safe and effective [6]. is study provides an updated evaluation based on the     rank MTX, 10 th rank SSZ, and 11 th rank placebo. In the analysis of incidence of adverse events, we found the least possibility of incidence with TwHF used singly, followed by TwHF combined with MTX, ranking first and second, respectively. e details of the interventions are as follows: 1 st rank TwHF, 2 nd rank TwHF combined with MTX, 3 rd rank CsA combined with LEF, 4 th rank SSZ, 5 th rank CsA, 6 th rank MTX, 7 th rank LEF, 8 th rank SSZ combined with MTX, 9 th rank TwHF combined with LEF, 10 th rank FK506, and 11 th rank placebo. In conclusion, from the results of the current analysis it can be considered that compared with the DMARDs currently used to treat RA, TwHF has shown a clear efficacy in treatment of RA, and TwHF combined with MTX performed well under various endpoints. In the ACR20, ACR50, and ACR70 responses, the analysis showed that the efficacy of combination therapy of TwHF was better than its monotherapy. In the ACR20 and ACR50 responses, both monotherapy and combination therapy of TwHF were found to have good efficacy. In analysis of the incidence of adverse events, we found the least possibility of incidence with TwHF used singly. is study also has some limitations. For example, due to the insufficient number of studies, it was not feasible to assess for different dosages of the same drug across different treatment schemes, which may impact the study's results. In addition, some of the included literature studies do not explicitly mention details of randomization method or method of blinding; thus, there is a risk of publication bias. In clinical practice, TwHF is often considered to possess liver and kidney toxicity and to easily cause adverse events. While our study found that TwHF has little possibility of incidence in adverse events, we cannot completely exclude the possibility of publication bias or selective reporting. us further review after the publication of more rigorous, highquality RCTs is warranted.

Conclusions
is NMA found that in assessment of ACR20 response, TwHF combined with MTX had the greatest probability of achieving the best efficacy among the treatment schemes involved, while TwHF used singly was ranked as second best.
In assessment of ACR50 response, the efficacy of TwHF combined with MTX ranked second only to CsA combined with LEF, while TwHF used singly ranked third. In assessment of ACR70 response, CsA combined with LEF ranked first, TwHF combined with LEF ranked second, TwHF combined with MTX ranked third, while TwHF alone ranked fourth. In analysis of incidence of adverse events, the possibility of incidence ranked the lowest with TwHF used singly and the second lowest with TwHF combined with MTX. In conclusion, it can be considered that compared with the DMARDs currently used to treat RA, TwHF has a clear efficacy on RA. Among all treatments, the monotherapy of TwHF and the combination therapy of TwHF and MTX performed well at various endpoints.
In a previous study [6] we conducted an NMA analysis on the efficacy and safety of TwHF and traditional synthetic DMARDs monotherapy in RA. e results indicated that in the direct comparison, TwHF was better than sulphasalazine in ACR 20, ACR 50, and ACR 70 responses; TwHF was superior to placebo in ACR 20 and ACR 50 responses. In indirect comparison, TwHF was superior to MTX, LEF, FK506, MINO, and placebo in ACR 20 response. In the efficacy ranking, TwHF ranked first in ACR 20 and ACR 50 response, and was the preferred treatment. Also, in ACR 70 response, TwHF ranked second (57.8%), second only to LEF (69.6%), which confirmed its efficacy and safety in RA. In clinical practice, combination therapy is also our conventional treatment for RA. erefore, in this study, based on the previous research, we performed an updated NMA on monotherapy and combination therapy of TwHF and conventional synthetic DMARDs in RA. e research results showed that the clinical protocol of TwHF combination therapy for RA is more in line with clinical practice and has more advantages than other clinical protocols of conventional synthetic DMARD drugs in RA. TwHF can be considered as a potential first-line DMARD for the treatment of RA, but high-quality randomized trial data are still needed to guide the use of TwHF in clinical RA treatment.

Ethical Approval
All analyses were based on previously published studies. As a result, ethical approval and patient consent are not relevant.

Disclosure
Chun-ping Liu and Yu-zheng Yang are co-correspondence authors. Hai-long Wang and Qi Zhao are co-first authors.

Conflicts of Interest
e authors declare that they have no conflicts of interest.  Figure S2: Risk of bias graph. Figure S3: Risk of bias summary. Figure S4: e cumulative probability diagram. A. With ACR20 as the endpoint. B. With ACR50 as the endpoint. C. With ACR70 as the endpoint. D. e analysis of adverse events. Figure S5: Forest plots. A. With ACR20 as the endpoint. B. With ACR50 as the endpoint. C. With ACR70 as the endpoint. D. e analysis of adverse events. Figure S6: Inconsistent assessment. A. With ACR20 as the endpoint. B. With ACR50 as the endpoint. C. With ACR70 as the endpoint. D. e analysis of adverse events. Figure S7: e publication bias. A. With ACR20 as the endpoint. B. With ACR50 as the endpoint. C. With ACR70 as the endpoint. D. e analysis of adverse events.