Eight electronic databases and three clinical trial registries were searched from their inception to May 2, 2023, including PubMed, Cochrane Library, Embase, Web of Science, China National Knowledge Infrastructure (CNKI), Chongqing Chinese Science and Technology Journal Database (VIP), China BioMedical Literature Service System (SinoMed), Wanfang Database, World Health Organization International Clinical Trials Registry Platform (https://www.who.int/clinical-trials-registry-platform), ClinicalTrials.gov, and the Chinese Clinical Trial Registry (ChiCTR, http://www.chictr.org.cn/index.aspx). There were no language and publication restrictions. The references of reviews or clinical trials were searched as supplements. Search terms, such as “hyperthyroidism”, “Graves’ disease”, “Jiawei-Xiaoyao-San”, and “Danzhi-Xiaoyao-San” were identified according to previous systematic reviews, clinical practice guidelines, MeSH terms and Emtree. Detailed search strategies are presented in Supplementary Material S2 ( Table 1 ).

The screening was conducted independently by two authors (GW and MYX) according to the eligibility criteria. Retrieval results were imported into NoteExpress (Beijing Aegean Software company, v3.8.0.9492) and automatic duplicate removal was conducted. The first screening was conducted according to the titles and abstracts of all potential studies, and then the full texts were obtained for the second screening. Two authors (RTZ and WXM) independently extracted data from all eligible trials. A data extraction spreadsheet was designed and modified for this review. If the required data were not available in the trial, further information was sought by contacting the original researchers. Data from trials published in duplicate were extracted only once. Discussions and a third reviewer’s (ZLL) proposal were required if there was any discrepancy.

Two authors (RTZ and WXM) used the Cochrane Risk-of-Bias (ROB) 1.0 tool in Review Manager (RevMan) 5.4 software to evaluate the quality of the included RCTs with reference to the following aspects: (a) random sequence generation (selection bias), (b) allocation concealment (selection bias), (c) blinding of participants and personnel (performance bias), (d) blinding of outcome assessment (detection bias), (e) incomplete outcome data (attrition bias), (f) selective reporting (reporting bias), (g) other bias. Each aspect was judged as low, unclear or high. Discrepancies were resolved by discussion or involving a third author (ZLL). For case reports, we used the Joanna Briggs Institute (JBI) tool 2020 (28) for critical appraisal.

For RCTs, data analyses were performed using RevMan 5.4. Dichotomous variables were presented as risk ratio (RR) with 95% confidence interval (CI). Continuous variables were presented as weighted mean difference (MD) with 95% CI if the measurement units of outcomes were the same in different trials, and the standardized mean difference (SMD) was used when the measurement units of outcomes differed from one another.

Trial characteristics were summarized in tables by Microsoft excel 2021. For RCTs, the meta-analysis was performed using RevMan 5.4. Forest plots were used to describe the results. For the comparison of different interventions and outcomes, the results were synthesized respectively. We intended to pool the collected data at similar time points, with data from one trial being synthesized only once. If the standard deviation was not reported by means, it was calculated from the information reported (29). If the required data were unavailable in the trial and we failed to contact the original researchers, we only used the available data in the analyses.

The heterogeneity test was performed by chi-square test and presented as the value of I² statistics. According to the Cochrane handbook, more than 25%, 50%, and 75% values were considered as mild, moderate, and severe heterogeneity, respectively (25). The fixed-effects model (FEM) was applied when I² < 50%; otherwise, the random-effects model (REM) was applied. The Mantel-Haenszel method was used for meta-analyses of dichotomous outcomes. The inverse-variance method was used for meta-analyses of continuous outcomes. The subgroup analysis was conducted to explore any factors that might explain the heterogeneity. Overall effects with a p-value below 0.05 were considered statistically significant.

Sensitivity analysis was performed using STATA 14.0 software for the primary outcome to determine whether the conclusion was robust. When severe heterogeneity existed, sensitivity analysis was conducted for potential sources of heterogeneity.

Funnel plot and Egger’s test (30) was performed to assess potential publication bias in a single meta-analysis involving ten or more trials. Two authors (RTZ and WXM) conducted the certainty assessment of each outcome using the GRADEpro Guideline Development Tool (GRADEpro GDT) developed by the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) working team. Discrepancies were resolved by discussion or involving a third author (ZLL).

From 8 databases and 3 clinical trial registries, 157 records were identified. After removing 81 duplicates, 14 records irrelevant to this review were excluded, leaving 62 records for full-text screening. Of these, 60 reports were obtained in full text. For RCTs, 18, 1, 3, and 4 reports were excluded due to lack of randomization, duration of intervention less than 4 weeks, no available data, and stable thyroid function of participants, respectively. Nine case reports were excluded due to the intervention including other therapies. Three reports contained data from the same study. Finally, 25 reports (15 reports including 13 RCTs, 3 case reports including 3 cases and 7 reports including 3 experiments for mechanism research) were assessed to meet the eligibility criteria. The detailed screening process is shown in the PRISMA flow diagram ( Figure 1 ) (26).

In 13 RCTs from 15 reports, a total of 979 participants were enrolled, age ranged from 16 to 76 years old, with 485 in the control group and 494 in the intervention group. The data in three reports (31–33) came from the same trial. In 9 trials, hyperthyroidism in patients was explicitly described as being caused by GD, which was not reported in the other trials. Treatment duration ranged from 4 weeks to 6 months, with the most common being 12 weeks (considered equivalent to 3 months). Therefore, when outcomes at multiple time points were reported, we extracted outcomes at or closest to 12 weeks. In terms of intervention measures, 10 trials compared JWXYS plus ATDs with the same dose of ATDs (22, 24, 31–43), 2 trials compared JWXYS plus half-dose ATDs with normal dose ATDs (21, 43), and only one trial compared JWXYS with ATDs (44). Two trials reported hepatic dysfunction (21, 32), one reported heart disease (40) and one reported Hashimoto’s thyroiditis (44) as comorbidity. JWXYS was administered as a modified decoction in all trials. The most common TCM syndromes are “Liver depression inducing fire” (in 5 trials), “Prosperity of liver fire” (in 3 trials), “Liver depression with fire and spleen deficiency” (in 2 trials) and “Hyperactivity of heart-liver fire” (in 1 trial). The detailed characteristics of the included trials are shown in Table 1 . The detailed components of the prescriptions used in each study are summarized in the Supplementary Material S2 ( Table 2 ).

The risk of bias on primary outcomes in 13 RCTs was summarized in Figure 2 . Nine trials reported the use of random number table as the random sequence generation method, which was considered to have a “low risk of bias”. Five trials just mentioned “randomization” without detailed methods (21, 37, 40, 41, 44) and were considered to have an “unclear risk of bias” in random sequence generation. None of the trials mentioned the details of allocation concealment of randomization methods and implementation of blinding. All trials were considered to have an “unclear risk of bias” in allocation concealment. The majority of the trials were considered to have a “high risk of bias” in performance bias due to apparently different intervention without placebo. Four trials (21, 22, 32, 40) were considered to have an “unclear risk of bias” in performance bias and a “low risk of bias” in detection bias because they only reported physiological indicator as outcomes. In the aspect of “incomplete outcome data”, two trials (31–33, 44) that did not report the cause of the missing data were considered to have an “unclear risk of bias”. In two trials (32, 40), the proportion of male and female participants did not correspond to the clinical ratio of men to women, therefore, these trials were considered to have a “high risk of bias” in other bias.

Funnel plots were drawn to explore the possibility of publication bias for the ten trials comparing JWSYS plus ATDs with ATDs on the primary outcomes of FT3 and FT4. The scattered points distribution in the funnel plots was basically symmetrical. Egger’s tests were performed for further verification. Egger’s test of funnel plot asymmetry indicated almost no publication bias in the included studies (P = 0.997, P = 0.290), suggesting reliable results ( Figure 10 ).

Certainty of evidence and the reasons for the upgrade and downgrade are presented in Table 2 . The certainty of evidence for each outcome was downgraded considering its risk of bias, inconsistency, indirectness, imprecision and other potential biases. The evidence for all outcomes were assessed as moderate to very low certainty. Moderate-certainty evidence showed that JWXYS plus ATDs resulted in fewer AEs. Low-certainty evidence suggests that JWXYS plus ATDs is more effective than ATDs in controlling thyroid functions (FT3, FT4, and TSH), alleviating TCM syndrome, and reducing goiter score and TRAb. The evidence of all other comparisons were assessed as low to very low certainty.

Three case reports containing three patients were included, and the quality of the reports was evaluated as high by the JBI tool ( Table 3 ).

Reasons for patients seeking TCM-only treatment include suffering from or being afraid of severe AEs following ATDs and multiple relapses. Complete control of hyperthyroidism can be achieved with JWXYS therapy alone for 3 months to 3 years. The detailed components of the prescriptions used in each study are summarized in Table 4 .

Three experiments in seven reports were included (45–51), and the characteristics and main findings were summarized in Table 5 . Wu XY’s study (49) showed that JWXYS plus Anemone flaccida Fr. Schmidt (Diwu) regulated T cell differentiation and restored Th17/Treg cell balance by inhibiting the JAK1-STAT3 pathway of CD4⁺ T cells. Bao CY et al. (45) found that modified JWXYS plus PTU reduced the incidence of adverse pregnancy in female mice, as well as lowered thyroid function and activity of deiodinases of offspring mice. Tan HZ et al. (46–48, 50, 51) found that modified JWXYS could regulate the proliferation and apoptosis of thyroid cells and protect liver function by alleviating oxidative stress. The detailed components of the prescriptions used in each study are summarized in the Supplementary Material .

Thirteen RCTs involving 979 participants on JWXYS for hyperthyroidism were included. We found that compared with ATDs, the complementary treatment of JWXYS could help control thyroid functions, alleviate TCM syndrome and goiter, and reduce TRAb and AEs. Meanwhile, when combined with JWXYS, half-dose of ATDs could ideally control thyroid function, thus significantly reducing the occurrence of AEs, and were also effective in alleviating TCM syndrome and goiter. But the effectiveness of mono-JWXYS in controlling thyroid function was significantly inferior to that of ATDs. In cases of severe adverse reactions following ATDs, alternative therapy with JWXYS may be attempted, which may allow complete control of hyperthyroidism. Sensitivity analyses were conducted by removing each trial, which indicated that the pooled effects of primary outcomes were robust. The risk of bias was high among the studies, mainly due to the lack of description of the randomization process and the absence of blinding. Funnel plots and Egger’s tests indicated no existence of publication bias. The majority of the evidence were assessed as moderate to very low certainty.

JWXYS can be used to treat a variety of thyroid disorders, such as Hashimoto’s thyroiditis (52), subacute thyroiditis (53), thyroid-related eye diseases (54), nodular goiter (55), etc., especially for the treatment of hyperthyroidism. In TCM theory, patients with hyperthyroidism are usually diagnosed as the syndrome of liver depression inducing fire, the syndrome of hyperactivity of fire due to yin deficiency or the syndrome of hyperactivity of heart-liver fire. As a classical CHM prescription, JWXYS is commonly used in the treatment of the syndrome of liver depression inducing fire and spleen deficiency. It is often used to relieve various symptoms, such as anxiety, depression, hot flashes, night sweating, palpitations, headache, dry eyes, sleep disturbance, rash, and irregular menstruation, all of which are just in line with the characteristics of hyperthyroidism.

JWXYS formula contains ten herbs: Paeonia suffruticosa Andr. (Danpi), Gardenia jasminoides Ellis (Zhizi), Angelica sinensis (Oliv.) Diels (Danggui), Paeonia lactiflora Pall. (Baishao), Bupleurum chinense DC. (Chaihu), Poria cocos (Schw.) Wolf (Fuling), Atractylodes macrocephala Koidz. (Baizhu), Glycyrrhiza uralensis Fisch (Gancao), Mentha haplocalyx Briq. (Bohe), Zingiber officinale Rosc. (Shengjiang). In clinical practice, JWXYS is often modified to target different symptoms of patients. In the included studies, Prunella vulgaris (Xiakucao), Scutellariae Radix (Huangqin), Lycii Cortex (Digupi), Ziziphi Spinosae Semen (Suanzaoren), and Figwort Root (Xuanshen) are the most commonly added herbs, while Zingiber Officinale Roscoe (Shengjiang) and Menthae Herba (Bohe) are often removed. In case of significant goiter, Fritillariae Thunbrgii Bulbus (Beimu), Xuanshen, Ostrea gigas Thunberg (Muli), and particularly Xiakucao, are often added (16, 23, 37, 43). Xiao-Luo-Wan, which consists of Beimu, Xuanshen and Muli, is also one of the classic TCM prescriptions and is mainly used to treat a variety of proliferative diseases, including goiter and thyroid nodule. Xiakucao is one of the most commonly used Chinese herbal medicines in the treatment of GD, which can be used either alone or in a Chinese herbal formula. Many clinical trials have verified their effectiveness, and several pharmacological studies have been conducted (56, 57). When the symptoms of fear of heat, dry mouth and hyperhidrosis are obvious, huangqin or Digupi is often added to treat TCM syndrome of fire of excessive type and deficiency type, respectively. Suanzaoren is often added to treat insomnia.

In addition to the studies and outcomes included in this systematic review, the existing literature also reported other applications and effects of JWXYS in the treatment of hyperthyroidism. Liu XX et al. (58) reported that after microwave ablation, the application of JWXYS can significantly relieve clinical symptoms such as palpitations and irritability, and help control thyroid functions. Liu HJ et al. (59) have shown that the application of JWXYS to patients with hyperthyroidism during pregnancy can improve pregnancy outcomes, such as reducing postpartum blood loss, natural labor rate, and postpartum infection rate. Furthermore, studies have shown that JWXYS can reduce the Hamilton Depression Scale score (31) and bilateral superior thyroid artery systolic peak flow velocity (60) in patients with hyperthyroidism. The above studies were not included in this systematic review for meta-analysis because they did not meet the eligibility criteria and the frequency of outcome reporting was low.

Hyperthyroidism is mainly caused by GD, and the core of its pathogenesis is the loss of self-tolerance to TSHR, resulting in the production of stimulating thyrotropin receptor antibodies. Three experiments explored the mechanism of JWXYS in the treatment of hyperthyroidism. All three studies were conducted in animals to simulate the onset of GD by injecting antigen (thyroid membrane protein, adenovirus or recombinant adenovirus packaging TSHR plasmid), and all used a three-dose procedure. The key to the mechanism study is the success of the disease model, and the modeling rate and the maintenance of the disease state seriously affect the results of the intervention experiment. All the studies evaluated the success of the modeling, however, none of the three studies reported the modeling rate. The recombinant adenovirus packaged with the TSHR289 plasmid in Tan HZ et al.’s study (46–48, 50, 51) was from our team. In recent years, our team has further optimized the modeling method. The improved single injection method has shortened the modeling time, and the modeling rate can reach 100% within 4 weeks after injection (61).

When GD occurs, the inhibitory function and/or proportion of regulatory T cells (Treg) is decreased, and helper T cells (Th) are abnormally activated. The increased ratio of Th1 and Th17 secretes cytokines such as interferon gama (IFN-γ) and interleukin (IL) 17, thus promoting recruitment and differentiation of B cells and plasma cells. Wu XY’s study (49) shows that JWXYS derived formula can regulate T cell differentiation and restore Th17/Treg cell balance in the treatment of GD by inhibiting JAK1-STAT3 pathway of CD4⁺T cells.

Abnormal thyroid function is an important cause of adverse pregnancy outcomes and has adverse effects on the health of offspring (62). Bao CY’s study shows that modified JWXYS reduced the incidence of adverse pregnancy in female mice, and meanwhile JWXYS reduced the secretion of thyroid hormones and activity of deiodinases (type I, II, III) in the offspring mice.

Under normal conditions, the proliferation and apoptosis of thyroid cells are balanced to maintain the morphological stability. Under the stimulation of TRAb, the proliferation and apoptosis of thyroid cells were unbalanced, resulting in hyperplasia and proliferation. Tan HZ et al.’s studies (46) showed that modified JWXYS regulated the imbalance of a pair of markers of proliferation and apoptosis (Bcl-2 and BAX). Oxidative stress is a characteristic of hyperthyroidism and can lead to hepatic damage. JWXYS reduced the concentration of aspartate aminotransferase (AST), alanine transaminase (ALT) and alkaline phosphatase (ALP) in serum (47, 48) and 8-hydroxy-2 deoxyguanosine (8-OHDG) expression in the liver (51). Meanwhile, modified JWXYS reduced the levels of malondialdehyde (MDA), superoxide dismutase (SOD) and catalase (CAT) in the liver of GD mice (46–48, 51), indicating the restoration of the balance of oxidation/antioxidant system, which is superior to methimazole. Modified JWXYS could regulate the expression of mRNA and protein of nuclear factor erythroid 2-related factor 2(Nrf2) and heme oxygenase-1(HO-1), the key factors of oxidative stress pathway (46). Therefore, modified JWXYS could protect liver function by alleviating oxidative stress in GD mice.

A network pharmacological study (49) predicted that the main active ingredients of JWXYS in GD treatment include (+)-catechin, beta-sitosterol, formononetin, kaempferol, licochalcone A, naringenin, paeoniflorin, quercetin, shinpterocarpin, sitosterol, stigmasterol and sudan III. JWXYS may intervene in inflammation, immunity, apoptosis and other mechanisms to treat Graves’ disease through IL-17 signaling pathway, TNF signaling pathway, cytokine - cytokine receptor interaction and other signaling pathways. Validation is still needed in vivo and in vitro.

To our knowledge, this is the first systematic review to evaluate the effectiveness and safety of JWXYS in the treatment of hyperthyroidism. A comprehensive search of eight databases and three trial registries was conducted to ensure that all RCTs were included, and the methodology was reported in detail and transparently to allow for replication. At the same time, we also searched for the case reports of JWXYS alone for the treatment of hyperthyroidism, and summarized the reasons for patients seeking this treatment and its effectiveness, providing a possible alternative therapy for the treatment of hyperthyroidism. In addition, we also searched and summarized the pharmacological mechanism of JWXYS in the treatment of hyperthyroidism. This systematic review provides an effective and safe TCM complementary and alternative therapy for the treatment of hyperthyroidism.

However, there are several limitations in this systematic review. The included studies were all published in Chinese and only Chinese participants were recruited, so the generalizability of the findings was limited. All of the included studies were of low quality, with small sample sizes and without rigorous experimental protocols and follow-up. None of the studies reported relapse rates. Due to the insufficient number of trials included, some comparisons could not be pooled for meta-analysis, and subgroup analyses classified by course of intervention did not show significant differences in primary outcomes. But some trials showed that the adjuvant of JWXYS achieved quicker effectiveness, especially in alleviating symptoms (41). In addition, all studies administered modified JWXYS with different drugs and dosages, and the symptoms and signs included in various studies were not exactly the same, resulting in clinical heterogeneity. Publication bias assessment was performed only for primary outcomes in the JWXYS plus ATDs and ATDs comparisons, as few studies could be included in other outcomes and comparisons.

Future studies with appropriate randomization and double-blind methods are warranted to confirm these findings. The CONSORT checklist (63) is recommended for researchers when drafting RCTs. In the literature screening, it was found that some studies only took clinical effective rate as the primary outcome, and did not report thyroid function data. Single objective outcome is recommended as the primary outcome, rather than the compound outcome of clinical effectiveness. Thyroid volume data obtained from thyroid color ultrasound should be used to explore the improvement of goiter, rather than visual judgment of appearance. It is necessary to establish a core outcome set and promote its application to standardize the trial design, so as to reduce the clinical heterogeneity of meta-analysis. More studies are needed to explore the effect of JWXYS on TRAb, and follow-up should be conducted after treatment to obtain recurrence rates. When conducting a pharmacological mechanism study, it is essential to analyze the components of the drugs utilized. A more robust modeling approach should be employed and the success rate should be confirmed in future animal experiments to ensure the stability and reliability of the results.