Efficacy and safety of iguratimod combined with methotrexate vs. methotrexate alone in rheumatoid arthritis

The inclusion criteria included the following: 1) studies conducted in a human population aged >18 years; 2) studies in which all patients were diagnosed with RA based on the ACR or the European League Against Rheumatism (EULAR) classification criteria; 3) studies that evaluated the combination of IGU and MTX therapy in RA; 4) studies with outcome data including ACR20, ACR50, ACR70, and adverse events; 5) randomized placebo-controlled clinical trials or trials in which the treatment arm was compared with a control arm; and 6) studies that were available in all languages. The full text of potentially related studies was extracted, and the methods and results of the trial were reviewed. When necessary data could not be determined, every effort was made to contact the study authors.

The exclusion criteria included the following: 1) nonhuman studies (animal studies) and studies in children; 2) studies in which patients diagnosed with RA were treated with a combination of other DMARDs; 3) studies with a duration <3 months; 4) non-RCT studies and studies without a clear control arm or placebo arm; and 5) studies without end-of-trial outcomes (ACR20, ACR50, ACR70, and adverse events). The participant, intervention, comparison, outcome, and study design (PICOS) data are presented in Table 1.

Two independent investigators searched for original RCT studies related to the combination of IGU and MTX therapy in RA published before November 1, 2019, in PubMed, Cochrane Library, Embase, the China National Knowledge Infrastructure (CNKI), the Chinese Biomedical Literature Database (CBM), and WanFang Data using the medical subject header (MeSH) terms “Iguratimod” or “T-614” and “Methotrexate” or “Amethopterin” or “MTX” and “Rheumatoid Arthritis” or “Arthritis, Rheumatoid” or “RA.” Additionally, we searched clinical trial registry websites (http://​www.​ClinicalTrials.​gov and http://​www.​chictr.​org.​cn) to identify trials that had been completed but not yet published, as well as all ongoing trials with available results and data. When multiple publications were found regarding the same trial, we used the latest or the most complete report of the trial. If it included complete information about the study design, participant characteristics, interventions, and outcomes, we also considered conference summaries. Additional RCTs were identified from the reference lists of relevant full-text articles retrieved.

Data were extracted independently by two investigators (LJ‑C and YJ-Z). A third investigator (JY-L) resolved the differences between the two investigators to reach a consensus. Only the results of outcomes that had been prespecified in the protocol were extracted from the studies. The following study characteristics were extracted: study title, unique study name or ID, primary author, contact information, publication type, funding and conflicts of interest, year the study was initiated and completed, country of origin of the study, study setting, study design (e.g., multicenter or single center), and study inclusion and exclusion criteria. The following study patient characteristics were extracted: total number of participants screened for the study, number of participants randomized into each study arm, and participant age, sex, and race/ethnicity. The following intervention and comparison group characteristics were collected: all information provided by the study for the regimen (therapeutic agent name, dose, route, frequency, and total duration of administration). The study outcomes were summarized. The adverse event definitions and the total number and individual number of events experienced were also extracted. Screening of the data extraction forms was completed to identify the missing or extraneous fields and to facilitate the most effective extraction process (LJ‑C and YJ-Z).

We assessed the methodological quality of the included trials using the Cochrane Collaboration tool. Studies were graded as having a “low risk”, “high risk,” or “unclear risk” of bias across the seven specified domains [10]. We also used the seven-point Jadad scale, which includes randomization, double-blinding, and sequential removal of one study at a time, a practice that is in agreement with the methods of other meta-analyses performed in this context [11]. Assessments were stored and managed in RevMan 5.3 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Danmark; 2014).

For continuous data, the mean difference (MD) was determined using the DerSimonian and Laird method if the same scale was used across studies [12]. Results for dichotomous data were provided as relative risks and analyzed using the Mantel–Haenszel method. We examined heterogeneity in results across studies using Cochrane’s Q statistic, and inconsistency was quantified with the I² statistic (100% × (Q-df) /Q), which represents the percentage of total variation across studies that is attributable to heterogeneity rather than chance [13]. We considered a p-value of less than 0.10 as indicative of substantial heterogeneity. When substantial heterogeneity was not observed, the pooled estimate calculated with the fixed-effects model was reported using the inverse variance method. When substantial heterogeneity was observed, the pooled estimate calculated with the random-effects model was reported using the DerSimonian and Laird method [12], which considers both within-study and between-study variation. Publication bias was evaluated via funnel plots (i.e., plots of study results against precision) and quantified with Begg’s and Egger’s tests [14, 15]. However, if there were fewer than 10 studies included in the systematic review, funnel plot analysis was not performed. A two-tailed p-value of less than 0.05 was considered statistically significant. Statistical analyses were performed using Review Manager (RevMan) 5.3. Meta-regression and publication bias analyses were performed using Stata version 14 software (StataCorp., College Station, TX, USA).

All titles and abstracts found in the systematic review were independently screened by two researchers (LJ‑C, YJ-Z). A third investigator (JY-L) resolved the differences among two investigators to reach a consensus. Our search strategy yielded a total of 320 publications potentially related to the combination of IGU and MTX therapy for RA. After removing duplicates, a total of 203 studies were identified. Following screening of the titles and abstracts, 163 manuscripts did not fulfill our inclusion criteria and were excluded, leaving 40 selected manuscripts. After subsequent screening, an additional 33 were excluded: 30 randomized controlled trials (RCTs) did not include the ACR20/50/70, 2 RCTs did not report adverse events, and 1 RCT was repeated in more than one bibliographic source. After the selection process, a total of 7 RCTs consisting of 665 participants with 368 participants in the active arm and 297 in the placebo arm were included in the meta-analysis [1, 16‐21]. A Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram is included in Fig. 1 to illustrate the number of publications included and excluded during each phase of screening.

The included studies were published between 2013 and 2019. All included trials were randomized, with six phase II trials and three phase I trials. All seven trials were published as full manuscripts. Using the Cochrane Collaboration tool for risk of bias classification, we found the quality of the included studies to be generally good and fair (Figs. 2 and 3). Trials were also ranked for Jadad score (Table 2). The Jadad score of the included studies ranged from 3 to 4. The overall quality of the included studies was high.

Table 2 summarizes the baseline characteristics of the included studies. All the studies were RCTs; five studies did not report the blinding methods, one reported a double-blind, placebo-controlled design [16], and one study applied blinding methods to the investigators [1]. Six trials were from China and were single-center studies, and the remaining trial was from Japan and was a multicenter study [16]. All studies were conducted between 2013 and 2018, often for more than a year. Patients participating in these trials followed eligibility criteria determined by each unique trial, which usually included active RA patients older than 18 years and were based on ACR or EULAR criteria. Three trials [16‐18] reported that the primary efficacy endpoint was the rate at which patients (the full analysis set) achieved ACR20 at week 24 or last observation carried forward (LOCF), and the rate of ACR20/50/70 was reported in the other four studies. Other outcomes are detailed in Table 2.

The characteristics of the participants, interventions, and comparator details in the included studies are presented in Table 3. One study [1] did not report age and sex characteristics for each group (only total age averages and sex ratios were reported), and complete age and sex characteristics were obtained for one study [21] by contacting the authors. The average age of the participants was similar across all studies. The iguratimod dose used in the study [16] was 25 mg once daily (QD) (0–4 weeks) and 25 mg twice daily (BID) (5–24 weeks), and the remaining six studies were 25 mg BID. The use of the MTX dose in three studies was phased using 10 mg/week (0–4 weeks) and 12.5 mg/week (5–24 weeks) [17, 20], and 7.5–10 mg/week (0–4 weeks) and 10 mg/week (5–24 weeks) [18]. However, there was no phased administration in four studies, including 6 or 8 mg/week [16], 10 mg/week [1, 21], and 15 mg/week [19]. It is worth noting that four studies added complementary drugs. Two studies [1, 20] allowed the use of one NSAID (0.2 g celecoxib capsule, two times a day, oral) and (or) a small dose of a glucocorticoid (prednisone 7.5 mg/d or 10 mg/d), and two studies allowed the use of folic acid 5 mg/week [16] or folic acid 10 mg/week [18].

All seven studies that evaluated adverse events were included in the meta-analysis (Fig. 6). The heterogeneity was minimal (I² = 0%) without statistical significance (p = 0.92). The results demonstrated that there was no statistical significance in adverse events (1.06 [95% CI 0.92, 1.23]). The pooled RRs were generated by a fixed-effects model (Table 4). Combination therapy reported more leukopenia (16% vs. 13%); respiratory, thoracic, and mediastinal disorders (20% vs. 18%); increased β2-microglobulin (11% vs. 3%); and decreased blood iron (16% vs. 13%), but smaller increases in transaminase (20% vs. 26%) and gastrointestinal disorders (13% vs. 16%). There were no significant differences between the IGU + MTX and MTX /MTX + placebo groups for all adverse events, except increases in β2-microglobulin (RR: 4.31 [95% CI 1.32, 14.01]). One study [1] reported the adverse event of higher White Blood Cell (WBC) (2% vs. 3%; p > 0.05), one study[18] reported the adverse event of headache (3% vs. 3%; p > 0.05), and two studies [17, 20] reported the adverse event of dental ulcer (0% vs. 2%; p > 0.05).

Due to the high heterogeneity of ACR20, we first checked whether the original data included in the study were correct and whether the method of extracting the data was correct. However, the whole process was normal. We compared the results between the fixed-effects model and the random-effects model for ACR20, and the RR was 1.47 (95% CI 1.30, 1.67) vs. 1.40 (95% CI 1.13, 1.74) in the outcome of ACR20. The conclusion of a favorable effect persisted even using different models. Furthermore, sensitivity analysis performed by Stata did not indicate alterations in the results (RR) by sequentially eliminating individual studies, suggesting that no single study significantly contributed to the heterogeneity of ACR20 (Fig. 7). However, we then performed sensitivity analysis of the outcome of ACR20 by eliminating individual studies from the meta-analysis model in RevMan, the most prominent of which was the study by Ishiguro et al. [16]. After excluding this study, the heterogeneity decreased from 74 to 59%, and the RR decreased from 1.40 (95% CI 1.13, 1.74) to 1.30 (95% CI 1.08, 1.57); however, the heterogeneity remained high and the RR was still significant, suggesting that this study may be a cause of the source of heterogeneity. Next, we carried out meta-regression and subgroup analysis of language, ACR standard, MTX phase, and complementary drugs. Meta-regression for language, ACR standard, MTX phase, and complementary drugs was not significant (p = 0.429, p = 0.923, p = 0.114, p = 0.378, respectively), indicating that none of these four covariates were the source of heterogeneity. Next, we carried out subgroup analyses of language, ACR standard, MTX phase, and complementary drugs (Table 5), although the outcome of meta-regression was not significant. Subgroup analysis by language revealed that there was statistical significance in the Chinese group (RR = 1.435, 95% Cl [1.058, 1.947], p = 0.02), but the English group showed the opposite result (RR = 1.162, 95% Cl [0.959, 1.407], p = 0.125). With respect to subgroup analysis by ACR standard, results for both the 1987 ACR standard (RR = 1.611, 95% CI [1.075, 1.917], p = 0.014) and no 1987 ACR standard (RR = 1.250, 95% CI [0.997, 1.569], p = 0.044) groups suggested obvious statistical significance. According to subgroup analysis by MTX phase, significant RR was found in the no MTX phase group (RR = 1.304, 95% CI [1.085, 1.567], p < 0.001), whereas no significant RR was reported in the two MTX phase group (RR = 1.154, 95% CI [0.95, 1.398], p = 0.145). Subgroup analysis by complementary drugs indicated that the used group (RR = 1.154, 95% CI [0.952, 1.398], p = 0.145) suggested no statistical significance, and the not-used group (RR = 1.47, 95% CI [1.205, 1.794], p = 0.011) marked a favorable effect of combination therapy.

No evidence of publication bias was detected for the RR of ACR20, ACR50, or ACR70 by either Begg’s or Egger’s test (RR of ACR20 Begg’s p = 0.51 and Egger’s p = 0.62; RR of ACR50 Begg’s p = 0.29 and Egger’s p = 0.34; RR of ACR70 Begg’s p = 0.176 and Egger’s p = 0.065).