This meta-analysis were conducted strictly in accordance with the protocol registered in PROSPERO (CRD42021220780) and PRISMA guidelines (see Supplementary Materials).

We searched Embase, PubMed, The ClinicalTrials.gov, VIP Database for Chinese Technical Periodicals, Wanfang Database on Academic Institutions in China, China National Knowledge Infrastructure (CNKI), Cochrane Library, China Biology Medicine (CBM), MEDLINE Complete, and Web of Science with the retrieval time up to December 2020. The search strategy is shown in Supplementary Table S1.

In the first stage of literature screening, two reviewers independently conducted manual screening of the initially included clinical research literature. First, two reviewers read the title, abstract, and keywords and then exclude non-RCT, duplicate or identical papers, papers with obscure data, and papers whose full text cannot be obtained. In the second stage, after reading the full text of the literature, two reviewers further screened the literature that could eventually be included in the meta-analysis based on selection criteria. Finally, two reviewers independently extracted data from the literature. The differences between the two reviewers in the selection of literature and data extraction were resolved through consultation with the third reviewer.

The “risk of bias” assessment tool recommended by the Cochrane Collaboration (Deeks et al., 2021a) is used to assess the risk of bias. The tools include 1) generation of random sequence; 2) allocation concealment; 3) blinding of subjects and intervention providers; 4) blinding of outcome evaluators; 5) incomplete outcome data; 6) selective outcome report; and 7) other sources of bias.

RevMan 5.3 software was used for risk-of-bias assessment and meta-analysis. First, the heterogeneity test was performed. The subgroup analysis was carried out based on the intervention measures of the control group. When the heterogeneity among the included studies was low (p > 0.1 and I² ≤ 50%), the fixed-effect model was used; otherwise, the random-effect model was used (Deeks et al., 2021b). For continuous variables, when the measurement data units are different, the values differ greatly, or the measurement methods are different between different studies, the standardized mean difference (SMD) is used as the effect size indicator, and the other uses the mean difference (MD) as the effect size indicator. For dichotomous variables, the risk ratio (RR) is used as the effect size indicator. Each effect size is expressed with a 95% confidence interval (95% CI). The publication bias was detected by STATA 15 with the Egger method (continuous variable) and Harbord method (dichotomous variable) for primary outcomes. p > 0.1 is considered to have no publication bias. The GRADE tool was utilized to rate the quality of the evidence (GRADEpro 2015) according to the GRADE handbook (Schünemann et al., 2013).

The total records identified through database searching and other sources were 500. According to the search strategy, a total of 45 articles were obtained through preliminary search. By eliminating duplicate documents, carefully reading the title and abstract, a total of 456 articles were excluded. After carefully reading the full text and comparing the selection criteria, 33 RCTs were screened out and finally included (Shi et al., 2015; Lu 2014; Duan et al., 2015; Xiong et al., 2020; Xu et al., 2015; Ding et al., 2019; Deng et al., 2017; Tian et al., 2020; Xie et al., 2018; Xu et al., 2017; Bi et al., 2019; Yan et al., 2018; Tian et al., 2017; Chen et al., 2018; Xiong et al., 2015; Mo et al., 2015; Fan et al., 2020; Zhao et al., 2018; Li et al., 2020; Li et al., 2019; Meng et al., 2017; Li et al., 2016; Xia et al., 2020; Meng et al., 2015; Mo et al., 2018; Meng et al., 2016; Wang et al., 2019; Qi et al., 2019; Ju et al., 2020; Zhao et al., 2016; Hara et al., 2014; Ishiguro et al., 2013; Xia et al., 2016) (Figure 1). The main reason why 12 studies were excluded is that eight studies are not RCTs (Yoshioka et al., 2016; Wang et al., 2017; Suto et al., 2019; Okamura et al., 2015a,b; Wang 2017; Meng et al., 2016; Gu et al., 2020), and the other four studies are not comparing IGU + MTX with MTX (Liu 2016; Xu et al., 2017; Lu and Liu 2018; Zhu 2019).

The study of Hara et al. (2014) (Ishiguro et al., 2013; Hara et al., 2014) was conducted in Japan. The participants in the other RCTs were from China. There are 10 RCTs with more than 100 participants (Hara et al., 2014; Xia et al., 2016; Meng et al., 2017; Tian et al., 2017; Chen et al., 2018; Li et al., 2019; Ju et al., 2020; Tian et al., 2020; Xia et al., 2020; Xiong et al., 2020); there are 20 RCTs ranging from 50 to 100 participants (Duan et al., 2015; Meng et al., 2015; Mo et al., 2015; Shi et al., 2015; Xu et al., 2015; Li et al., 2016; Meng et al., 2016; Zhao et al., 2016; Deng et al., 2017; Xu et al., 2017; Mo et al., 2018; Xie et al., 2018; Yan et al., 2018; Xiong et al., 2015; Zhao et al., 2018; Bi et al., 2019; Ding et al., 2019; Qi et al., 2019; Wang et al., 2019; Fan et al., 2020). The interventions in the control group were mostly MTX alone. The control group of Deng et al. (2017), Tian et al. (2020), Bi et al. (2019), and Meng et al. (2015) used MTX + leflunomide; the control group of Li et al. (2019), Xia et al. (2020), and Mo et al. (2018) used MTX + tripterygium; and the control group of Li et al. (2020) used MTX + adalimumab. Subgroup analysis would be based on the treatment of the control group. The details of study characteristics are presented in Table 1.

The secondary outcomes include 1) symptom-related outcomes: HAQ, morning stiffness time (min), number of swollen joints, and number of tender joints; 2) bone protection-related outcomes: N-MID, total T-PINP levels, 25(OH)D, and β-CTX; 3) inflammation and immune response-related outcomes: RF, ESR, CRP, anti-CCP, and TNF-α; and 4) angiogenesis-related outcomes: VEGF. The results are shown in Table 2.

Twenty-four RCTs reported the adverse events. According to the intervention characteristics, they were subdivided into three subgroups (MTX-only subgroup, MTX + leflunomide subgroup, MTX + tripterygium subgroup). There was low heterogeneity in the most subgroup (MTX only: I² = 30%, p = 0.11; MTX + leflunomide: I² = 0%, p = 0.58; MTX + tripterygium: I² = 0%, p = 0.90) among the RCTs. The fixed-effect model was used. According to Figure 8, the adverse events in the IGU + MTX group were lower than those in the control in the MTX + leflunomide subgroup (RR 0.74, 95% CI 0.62–0.88, p = 0.0009; fixed-effect model). However, the difference is of no statistical significance in the MTX-only subgroup (RR 0.99, 95% CI 0.87–1.13, p = 0.90; fixed-effect model) and MTX + tripterygium subgroup (RR 0.73, 95% CI 0.29–1.85, p = 0.50; fixed-effect model). The summary result also showed that the difference is of no statistical significance between the two groups (RR 0.90, 95% CI 0.81–1.00, p = 0.06; fixed-effect model).

This study also conducted a subgroup analysis of primary outcomes based on duration and drug dose (Table 3). For the drug dose, since the dose of IGU is basically the same, the subgroup analysis is performed based on the starting dose of MTX. The results show that there is no obvious rule for the outcomes of different starting doses. For the duration of the intervention, the results showed that DAS28 improved after the intervention lasted at least 12 weeks.

The publication bias of the primary outcomes was detected by STATA 15.0. 1) ACR20: the publication bias detection suggests that the possibility of publication bias was small (p = 0.355) (Figure 9A). 2) ACR50: the publication bias detection suggests that the possibility of publication bias was small (p = 0.837) (Figure 9B). 3) ACR70: the publication bias detection suggests that the possibility of publication bias was small (p = 0.699) (Figure 9C). 4) DAS28: the publication bias detection suggests that there may be publication bias (p = 0.097) (Figure 9D). 4) Adverse events: the publication bias detection suggests that the possibility of publication bias was small (p = 0.234) (Figure 9E).

The evidence was judged to be high to very low (Table 2). The quality of ACR70 was high; the quality of adverse events and ACR50 test was moderate; the quality of ACR20 was low; and the quality of DAS28 was very low (Table 4).

This systematic review and meta-analysis included 31 RCTs involving 2,776 participants. The number of study participants was mostly between 30 and 240, and the interventions in the control group were mostly MTX alone. The control group of Deng et al. (2017), Tian et al. (2020), Bi et al. (2019), and Meng et al. (2015) used MTX + leflunomide; the control group of Li et al. (2019), Xia et al. (2020), and Mo et al. (2018) used MTX + tripterygium; and the control group of Li et al. (2020) used MTX + adalimumab. Sixteen RCTs failed to describe their random sequence generating method. Only Tian et al. (2020) described the allocation concealment methods. Twenty-nine RCTs failed to utilize blinding. Li et al. (2020) and Xia et al. (2016) have incomplete outcome, and Mo et al. (2015) mentioned the morning stiffness time, number of tender joints, and number of swollen joints but did not report in the article. The overall risk of bias is higher.

The main findings of this study can be summarized as the following: 1) compared with MTX alone, the ACR20, ACR50, and ACR70 of IGU + MTX are higher, while DAS28 is lower; and the symptom-related outcomes (HAQ, morning stiffness time, number of tender joints, number of swollen joints), bone protection-related outcomes [N-MID, T-PINP, β-CTX, 25(OH)D], inflammatory and immune response-related outcomes (RF, ESR, CRP, anti-CCP, TNF-α), and angiogenesis outcomes (VEGF) in the IGU + MTX group improved better. In terms of safety, there was no significant difference in the incidence of adverse events between the IGU + MTX group and the MTX-only group. 2) Compared with MTX + leflunomide, IGU + MTX has no significant difference in improving ACR20, ACR50, and ACR70, but IGU + MTX improves DAS28, HAQ, and anti-CCP more significantly. In terms of improving number of tender joints, number of swollen joints, RF, ESR, and CRP, the difference between the two groups was not statistically significant. However, the incidence of adverse events in the IGU + MTX group was lower. Our previous RCT also showed that compared with MTX + leflunomide, IGU + MTX has no statistical significance in improving ACR20, ACR50, ACR70, ESR, and CRP (Tian et al., 2020). Our RCT also shows that IGU + MTX is safer than MTX + leflunomide (Tian et al., 2020). 3) Compared with MTX + tripterygium, IGU + MTX has no significant difference in improving DAS28 and anti-CCP, but IGU + MTX improves morning stiffness time, number of tender joints, number of swollen joints, RF, ESR, and CRP more significantly. In terms of safety, there was no significant difference in the incidence of adverse events between the IGU + MTX group and the MTX + tripterygium group. 4) Compared with the IGU + adalimumab group, the DAS28 in the IGU + MTX group is higher. 5) For the duration of the intervention, DAS28 improved after the intervention lasted at least 12 weeks. 6) The publication bias test of the primary outcomes showed that there was no publication bias in ACR20, ACR50, ACR70, and adverse events, while DAS28 may have publication bias.

In summary, this systematic review and meta-analysis provide a new treatment strategy for the combination of IGU and MTX in the treatment of RA. When the intervention method is (IGU 25 mg Bid, MTX 10–25 mg once a week), and the intervention lasts for at least 12 weeks, the curative effect may be achieved without obvious adverse events.

RA is a common autoimmune disease characterized by painful and swollen joints, which severely impairs the patient’s physical function and quality of life. The treatment of RA has made great progress in the past 10 years. Relevant therapeutic drugs have been continuously introduced, which has improved the treatment and management of RA patients. In recent years, standard treatment has been used in the management of many chronic diseases (Jin et al., 2020). Especially considering the complex pathogenesis of RA, a single drug often fails to effectively achieve the treatment goal. Therefore, international and domestic guidelines recommend the use of combination therapy when a single DMARD treatment fails to meet the standard (Jiang et al., 2020; Jin et al., 2020). Combination therapy can improve the efficacy and reduce the incidence of adverse reactions, which is the main trend of RA treatment (Xie et al., 2020).

MTX is a first-line drug for the treatment of RA. Its main mechanism of action is as follows: inhibiting the formation of tetrahydrofolate by inhibiting dihydrofolate reductase, blocking DNA synthesis, and achieving anti-inflammatory and anti-immune effects (Lu et al., 2009; Mimori et al., 2019). Recent studies have shown that MTX is the “anchoring drug” in the treatment of most RA patients. The European Federation for the Prevention and Treatment of Rheumatism (EULAR) recommends that the efficacy of MTX be evaluated after 3 months of treatment. If there is no improvement within 3 months of the first treatment or the treatment goal is not reached within 6 months, the treatment plan should be reconsidered (GRADEpro 2015). The 2018 Chinese Rheumatoid Arthritis Diagnosis and Treatment Guidelines also mentioned that if the standard treatment of MTX still fails to reach the target, it is recommended to combine another synthetic DMARD for treatment (Schünemann et al., 2013). However, in clinical practice, there are a variety of prescriptions for combination drugs for the treatment of RA, such as IGU + MTX, MTX + leflunomide, and islammod + diacerein (Hao and Li 2014; Yoshikawa et al., 2018; Dai et al., 2019).

IGU is a new type of immunosuppressant used in RA in recent years. It can not only inhibit intracellular inflammatory factors and suppress immunity but also significantly inhibit bone resorption and bone destruction (Jiang et al., 2020; Xie et al., 2020). Joint damage is one of the important characteristics of RA, caused by abnormal bone metabolism. The balance of bone metabolism is controlled by bone resorption and bone formation, and they can be regulated by a variety of cytokines and signaling pathways. With regard to the combination of IGU and MTX, recent studies have also shown the complementarity of the two. Research by Wang et al. showed that both IGU and MTX can significantly inhibit the high expression of RANKL mRNA (p < 0.01), and the combination of the two drugs showed a stronger inhibitory effect (compared with MTX, p < 0.01; compared with IGU, p < 0.05). In addition, the combined drug group showed a more significant difference in inhibiting the ratio of RANKL mRNA/OPG mRNA than the single drug (p < 0.05) (Wang 2017). Yan and Wang (2018) also found that IGU combined with MTX can synergistically reduce the inflammatory response in patients with RA, better play the role of bone formation, and antagonize bone resorption.

This study found that the ACR20, ACR50, and ACR70 of IGU + MTX were not statistically different from that of MTX + leflunomide. However, the results of DAS28 show that the effect of IGU + MTX is better than that of MTX + leflunomide. This may be due to the fact that the MTX + leflunomide subgroup in ACR20, ACR50, and ACR70 only contains one RCT with extractable data, and the conclusion is unstable. In the future, more relevant research is needed to revise or verify this result.

The summary results show that the heterogeneity of ACR20 and DAS28 is high, while the heterogeneity of ACR50 and ACR70 is low. After subgroup analysis (based on the control group interventions, drug dosage and intervention duration), the heterogeneity of ACR50 and ACR70 is still low, while the heterogeneity of DAS28 and ACR20 is still high. As for the heterogeneity after subgroup analysis, we considered that the heterogeneity may be related to the following points: 1) it may be related to the patient’s baseline state of illness activity, and the baseline patient’s illness activity is not clearly stated in each study, so it cannot be further analyzed; 2) the heterogeneity of DAS28 may also be related to publication bias. In DAS28, IGU + MTX compared with IGU + adalimumab showed the opposite result from the comprehensive result. However, since only one study reported this, no definite conclusion can be drawn, and more relevant RCTs are needed to modify or verify this result. Most RCTs reported adverse reactions, but no deaths were reported. There was no difference in adverse reactions between the IGU combined with MTX and MTX alone in the blood system (leukopenia), liver function abnormalities, and gastrointestinal adverse reactions (malignant, vomiting). This shows that IGU combined with MTX will not cause additional infection or gastrointestinal adverse reactions to patients.

Safety analysis showed that compared with MTX + tripterygium and MTX only, there was no statistically significant difference in the incidence of adverse events in IGU + MTX. Compared with MTX + leflunomide, the incidence of adverse events in IGU + MTX was lower. Current research also shows that compared with other DMARDs, IGU is safe and suitable for long-term use. For example, a multicenter, randomized, double-blind, controlled trial showed that the IGU group had a lower incidence of adverse reactions than the MTX group after 24 weeks of treatment (Lu et al., 2009). The long-term application of IGU has relatively mild adverse reactions and a low incidence. The most common adverse reactions are gastrointestinal reactions and elevated liver enzymes. Tsuneyo Mimori et al. conducted a 52-week multicenter, prospective observational, phase IV clinical study in Japan, which proved that the incidence of IGU adverse reactions reached a peak after about 4 weeks of treatment. Subsequently, the incidence of various adverse reactions did not increase over time (Mimori et al., 2019). Our previous RCT showed that during 52 weeks of treatment, IGU combined with MTX therapy is safer than traditional leflunomide combined with MTX therapy. The incidence of overall adverse events, alanine aminotransferase (ALT)/aspartate aminotransferase (AST) that reflect liver function increased, and the rate of leukopenia was lower (Tian et al., 2020). Our previous RCT showed that during 52 weeks of treatment, IGU combined with MTX therapy is safer than MTX + leflunomide therapy. The incidence of overall adverse events, the incidence of ALT/AST elevation, and leukopenia decrease are lower (Tian et al., 2020).

A recently published multicenter, prospective, real-world study of IGU in the treatment of RA also showed its better safety. The data come from the participation of 48 hospitals in China, and a total of 1,759 patients with active RA were enrolled. It showed that the incidence of ≥ grade 3 adverse events was 3.4%, and only 0.7% (13/1751) of serious adverse events were related to IGU. Compared with previous studies, no new adverse reactions were seen in this large sample study. Meanwhile, there is no obvious correlation between the patient’s age and the occurrence of adverse reactions, so it can be considered that IGU has good safety in the treatment of elderly RA patients (Mu et al., 2021). In the non-RCT of IGU combined with MTX in the treatment of RA, Okamura et al. (2015) found that the adverse reactions of this combination were mainly interstitial pneumonia and liver dysfunction (alp increase). Ishiguro et al. (2013) found that the adverse reactions were elevated liver enzymes (elevated ALS), stomatitis, pharyngitis, decreased lymphocyte count, decreased white blood cell count, and decreased red blood cell count. It can be seen that the adverse reactions of IGU combined with MTX in the treatment of RA still exist, and more clinical trial studies are needed to improve the efficacy and safety. It is necessary to expand the sample size and prolong the observation time to further evaluate its clinical efficacy and safety.

Compared with a previous review (Chen et al., 2021), the strength of this systematic review and meta-analysis is that this is the newest evaluation of the efficacy and safety of IGU + MTX for RA patients. This study also used subgroup analysis to compare the efficacy and safety of IGU + MTX versus different control groups (MTX only, MTX + leflunomide, MTX + tripterygium, MTX + adalimumab). This study also conducted an evidence quality assessment to provide clinical information on the use of IGU + MTX. This study also adopted a more rigorous risk of bias assessment and a comprehensive analysis of 31 RCTs.

The limitation of this study is that most studies have high or unclear risk of bias in random sequence generation, allocation concealment, blinding, and incomplete outcomes, which affects the accuracy of the results and the evidence’s grade. Meanwhile, most of the outcomes have high heterogeneity, and the heterogeneity does not decrease after subgroup analysis (such as DAS28). This may be related to the fact that the dosage of the medication is not exactly the same; the basic treatment, the treatment process, and the observation time are different; and the baseline data of the patients in different RCTs are different. In the future, more RCTs from different regions and nationalities with clear random sequence generation methods, allocation concealment, and blinding are needed to modify or verify the results. In addition, the country distribution of RCTs will affect the spread of evidence. The RCTs included in the meta-analysis of this study are mainly from China and Japan. The main reason is that as a new drug, IGU has recently been approved for marketing in China and Japan, and clinical researchers and patients have more convenient access to drugs. Especially in 2017, IGU entered the new version of China’s National Medical Insurance Catalogue (Category B) (Jin and Luo 2020), and its reach was further expanded, reducing the personal burden of RA patients. Since 2017, the RCT of IGU combined with MTX in the treatment of RA has increased rapidly, and it mainly comes from China, and the applicability of the evidence is mainly extrapolated to Asian countries.

In future clinical practice, in addition to combining MTX, RCT of IGU combined with other conventional synthetic DMARDs (csDMARDs) can also be carried out, which also shows good clinical efficacy advantages. Dai et al. found that compared with leflunomide treatment, the improvement of DAS28, joint function indexes, serum inflammation indexes, and bone metabolism indexes of combined treatment with leflunomide and IGU was more significant (p < 0.05) (Dai et al., 2019). A retrospective analysis by Li et al. also showed that after 12 weeks of treatment with other csDMARDs (such as sulfasalazine, hydroxychloroquine sulfate, leflunomide) combined with IGU, the DAS28, RF, CRP, ESR, and morning stiffness of RA patients with chronic interstitial pneumonia were significantly improved compared to those before treatment, and the incidence of adverse reactions was low (Hao and Li 2014). In addition, IGU can also be combined with biological disease-improving antirheumatic drugs (bDMARDs) for the treatment of patients with poor efficacy of biological agents (Yoshikawa et al., 2018). For example, in patients with poor response to tocilizumab, combined IGU treatment can significantly improve their disease activity (DAS28, CDAI, and EULAR response criteria), thereby effectively controlling the disease (Ebina et al., 2019). In summary, future studies can explore the efficacy and safety of IGU combined with various DMARDs (such as ABC) or biological agents and provide new reference information for clinical treatment.