Identifying the most effective Traditional Chinese Medicine treatment modalities for premature ovarian insufficiency: a systematic review and network meta-analysis [version 1; peer review: awaiting peer review]

Background: This study aimed to systematically evaluate and compare the clinical efficacy of different modalities of Traditional Chinese Medicine (TCM) for the treatment of premature ovarian insufficiency (POI) and identify the most effective treatment modality. Methods: Six electronic databases (PubMed, OVID, Scopus, CNKI, VIP and WanFang) were screened from their inception until 20/01/2022. This network meta-analysis (NMA) incorporated hormone replacement therapy (HRT) as the placebo and six different TCM treatment modalities as interventions. The primary outcome was total effective rate; the secondary outcomes were improvement in serum follicle-stimulating hormone (FSH), estradiol (E 2 ), luteinizing hormone (LH) and anti-Müllerian hormone (AMH) levels. The Cochrane risk of bias tool was used to appraise the quality of the included studies. Random effect model was used for pairwise meta-analysis using RevMan 5.4. NMA was performed in STATA. The surface under the cumulative ranking curves (SUCRA) and mean rank (MR) were used as evaluation indexes. Results: Overall, 37 randomized clinical trials involving 2,597 subjects that conformed to the inclusion criteria were included. Herbal


Introduction
Premature ovarian insufficiency (POI) is a pathological state of a woman below the age of 40 years suffering from irreversible diminished ovarian function in her reproductive age, it mainly manifests as oligomenorrhoea, hypomenorrhea, amenorrhea, elevated follicle-stimulating hormone (FSH) level (>25 U/L) and short or long term effects of estrogen deficiency. 1 The ovarian reserve can be measured by serum anti-Müllerian hormone (AMH) levels. According to the Bologna criteria, a low ovarian reserve is defined as an AMH level below 0.5-1.1 ng/mL (3.6-7.9 pmol/L). 2 Review data from 2020 reported that the prevalence of POI is about 10% among women below 40 years old. 3 The occurrence of this disease has been increasing among populations living in urban cities. This has captured the attention of public health policy makers and a timely and appropriate preventive measure is urged to be implemented.
Hormone replacement therapy (HRT) is the most recommended medical intervention for patients with POI, referring to the European Society of Human Reproduction and Embryology (ESHRE) and International Menopause Society (IMS) guidelines, as well as the Chinese Medical Experts' Consensus on Clinical Diagnosis and Treatment of POI, stipulated by the Obstetrics and Gynecology Society of Chinese Medical Association. 1,[4][5][6] However, there are risks of adverse reactions and associated diseases after HRT application. 7, 8 Hence, patients often seek out complementary medicine to improve their ovarian functions and increase chances of conception.
Various Traditional Chinese Medicine (TCM) treatments such as herbal medicine, acupuncture, moxibustion and catgutembedding therapy have been applied as the treatment for patients with POI. These treatment modalities have been proven to be effective and safe to improve ovarian functions among patients with POI through numerous clinical trials and direct comparison by using meta-analysis. [9][10][11] However, there are inadequate direct and indirect comparisons between different TCM treatment modalities that contribute to clinical health promotion and selection of the optimal therapeutic procedure. Therefore, there is a need to systematically evaluate and analyze the efficacy of these alternatives in order to provide a strong theoretical basis and scientific evidence for clinical practice and research. In this study, we compared the efficacy of different TCM treatment modalities for patients with POI by conducting network meta-analysis (NMA) to rank the therapeutic measures in line with their treatment effect through direct and indirect comparisons of the means, with the prospective to provide an evidence-based guidance for future clinical guidelines.

Methods
The study method is compliant with the Cochrane Handbook for Systematic Reviews of Interventions 12 and the methods outlined in Lei Zhang et al. 13 It followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extension statement for network meta-analyses (PRISMA-NMA). 14,57 It was registered in the International Prospective Systematic Registration Review (PROSPERO) with registration number CRD42020163873 on 28 April 2020.
Search strategy and literature screening Records identified from six electronic databases, which included three English databases (PubMed (RRID: SCR_004846), OVID, Scopus (RRID:SCR_022559)) and three Chinese databases (CNKI, VIP, WanFang). Reported studies from the inception of each database until 20 January 2022 were screened. Search terms used were a combination of text words and medical subject headings (MeSH), including "Premature Ovarian Insufficiency", "POI", "Premature Ovarian Failure" or "Primary Ovarian Insufficiency". Table 1 shows the search strategy details for each database. Only English and Chinese articles were included. Additionally, the reference lists of all eligible studies were screened to avoid

Types of Intervention (I):
The treatment group received TCM treatment modalities, such as herbal medicine, acupuncture, moxibustion and catgut embedding therapy alone or in combination.

Comparisons (C):
The control group received HRT or one of the aforementioned TCM treatments.
Total effective rate = (number of recovery cases + number of effective cases + number of efficient cases)/total number of cases Â 100%.
(i) Recovery: clinical manifestations disappeared, menstrual cycle, serum FSH and E 2 levels returned to normal, total effective rate ≥95%; (ii) Effective: clinical manifestations were significantly improved, menstrual cycle, serum FSH and E 2 levels were significantly improved, total effective rate ≥70% and <95%; (iii) Efficient: clinical manifestations were slightly improved, menstrual cycle, serum FSH and E 2 levels were slightly improved, total effective rate ≥30% and <70%; (iv) Invalid: there was no improvement in clinical signs and symptoms, no improvement in menstrual cycle, no improvement in serum FSH and E 2 levels, total effective rate <30%.

Exclusion criteria
Studies were omitted if the studies did not meet the inclusion criteria, for instance animal studies, review articles, case reports, editorials, letters and comments.

Data collection
The name of authors, year of publication, subject age group, course of disease, types of TCM treatment modalities, number of subjects in intervention and comparison groups, were extracted into a spreadsheet. Two researchers (SYS and CM) performed the data extraction individually. In case of any discrepancies, another researcher (SKS) were asked to resolve the discrepancies via discussion.

Literature quality evaluation
The Cochrane risk of bias tool was used to appraise the quality and risk of bias of the included studies. The assessment was done independently by two researchers (SYS and CM). 15 In this tool, the random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome reporting, incomplete outcome data, selective reporting, and other items were being evaluated in all the included studies. Low, unclear and high risk of bias were used to evaluate each element. In case of discrepancies, discussion among the researchers (SYS, CM and SKS) were done to resolve it.

Statistical analysis
For pairwise meta-analysis of each intervention pair, a random effect model was used, with RevMan (RRID: SCR_003581) 5.4 (Nordic Cochrane Centre, Cochrane Collaboration, 2020) being employed. Continuous outcomes between intervention groups can be compared by using standard mean differences (SMDs), while dichotomous outcomes can be analyzed using odds ratios (OR). The relative treatment effectiveness between each intervention pair were estimated using 95% confidence intervals.
A NMA was performed in Stata (RRID:SCR_012763) 14.0 (free alternative, RStudio) and the network evidence graph was generated using the 'netplot' command. To compare direct and indirect comparisons of each intervention, I 2 and P-value of the global consistency were evaluated, and then the node-splitting analysis method was applied to detect inconsistencies. The inconsistency factor within each closed loop of evidence was used to detect inconsistency among studies. The evaluation indexes included surface under the cumulative ranking curves (SUCRA) and mean rank (MR). The SUCRA value reflected the possibility of the intervention measures, and the intervention was considered more effective if the MR was closer to 0. P<0.05 was considered statistically significant. 11 To determine if there was a small sample effect or publication bias in the network, a comparison-adjusted funnel plot was used.

Study selection
The selection procedure was recorded in Figure 1 according to the most updated PRISMA 2020 flow diagram for new systematic review. 16 A total of 2,602 studies were firstly identified through six databases, including three English databases (PubMed, OVID and Scopus), as well as three Chinese databases (CNKI, VIP and WanFang). Of those, 586 duplicate records were eliminated using EndNote X9 (RRID:SCR_014001) (Clarivate Analytics, USA). By assessing the title and abstract, 1,940 studies were subsequently removed. Full articles of the remaining 76 studies were screened, of which 43 were removed for not reporting relevant outcome data, non-RCT, no full-text access, not relevant to  TCM intervention, or intervention only included in one trial, which may lead to small study effects in a NMA. Eventually, 33 RCTs involving 2,597 subjects that conformed to the inclusion criteria were selected into this study and proceed with data extraction and analysis. Among them, all RCTs were Chinese articles. This NMA incorporated HRT as the placebo and six TCM treatment modalities. The characteristics and details of the included 33 RCTs are listed in Table 2.

Risk of bias of the included studies
For the random sequence generation element, 21 studies reported an appropriate method for randomization, thus "low risk". There were two studies allocated participants to interventions based on participants' preference and hospital record number therefore "high risk". The rest only reported random, so it is "unclear".
In terms of allocation concealment, 30 studies were judged as having "high risk" of bias as there were inadequate concealment of allocations prior to assignment. Only one study reported clear allocation concealment by using sequentially numbered, opaque and sealed envelope, therefore it was judged as "low risk". There were two studies reported using assignment envelopes but sufficient details such as whether the envelopes were sequentially numbered, opaque and sealed were not mentioned, thus "unclear".
In terms of blinding of participants and personnel, only one study was stated as single-blinded design but did not address the outcome, therefore "unclear". The rest of the studies were judged as "high risk" as there were no blinding and due to different treatment modalities, participants might have the wrong belief that the intervention was more effective for their treatment.
In terms of outcome assessment, all studies were judged as "unclear" due to inadequate information for evaluation. In terms of incomplete outcome data, only one study was judged as "high risk" of bias as there were missing outcome data across the intervention groups. The rest were "low risk". In terms of selective reporting, only one study was judged as "high risk" of bias due to incomplete outcome of interest in the study. The rest were "low risk". In terms of other bias, all studies were "low risk" except two, in which there were incorrect reported data and discrepancies found in the studies. The graph and summary of the risk of bias are listed in Figure 2.
The forest plot of AMH indicates that herbal medicine (MD 0.68, 0.22 to 1.14, I 2 =99%) showed better efficacy compared to only HRT in improving AMH levels.

NMA on total effective rate
The results of the NMA showing comparisons of different treatments based on total effective rate are shown in Figure 3A.
The total effective rate are presented in the form of a league table can be found in Figure 4A. A total of 33 RCTs (2,862 participants) with seven different treatment modalities were analyzed. The results of NMA indicate that herbal medicine combined with catgut-embedding therapy have the highest efficiency. According to the SUCRA, which presents treatment ranking to offer the most important benefit outcome (the highest total effective rate), herbal medicine combined with catgut-embedding therapy (87.8%) had the highest effective rate, followed by combination of HRT with herbal  (11.9%). This also showed that combination of TCM treatments with HRT are more effective than HRT alone. Table 4 (part A) and Figure 5A depict the rankings of these seven treatments.

NMA on FSH levels
The results of the NMA showing comparisons of different treatments based on improvements of FSH levels are shown in Figure 3B. The league table results of the NMA on FSH levels are presented in Figure 4B. A total of 28 RCTs (2,384 participants) with seven different treatment modalities were analyzed. The results of NMA indicate that the combination of HRT with herbal medicine and acupuncture achieved the highest efficiency in improving the FSH levels. According to SUCRA, the combination of HRT with herbal medicine and acupuncture (99.9%) had the highest effective rate, followed by HRT combined with herbal medicine (67.4%), acupuncture combined with moxibustion (54.9%), herbal medicine combined with acupuncture (49.6%), herbal medicine combined with catgut-embedding therapy (49.3%), herb alone (23.1%) and HRT only (5.8%). Table 4 (part B) and Figure 5B depict the rankings of these seven treatments.

NMA on E 2 levels
The results of the NMA showing comparisons of different treatments for the changes of E 2 levels are shown in Figure 3C. The league table results for the E 2 comparison are shown in Figure 4C. A total of 26 RCTs (2,108 participants) with seven different treatment modalities were analyzed. The results of NMA indicate that the combination of HRT with herbal medicine and acupuncture had the highest efficiency. It was also found that the combination of different treatment modalities pairing with herbal medicine were more effective than HRT alone. According to SUCRA, the combination of HRT with herbal medicine and acupuncture (92.6%) had the highest effective rate, followed by herbal medicine combined with catgut-embedding therapy (80.1%), herbal medicine combined with acupuncture (66.7%), HRT combined with herbal medicine (51.6%), herbal medicine alone (40.8%), HRT alone (15.1%) and acupuncture combined with moxibustion (3.2%). Table 4 (part C) and Figure 5C depict the rankings of these seven treatments.

NMA on LH levels
The results of the NMA showing treatment comparisons based on the difference of LH levels are shown in Figure 3D. The league table results of the NMA for LH levels are shown in Figure 4D. A total of 19 RCTs (1,505 participants) with six different treatment modalities were analyzed. The results of NMA indicate that the combination of HRT with herbal medicine and acupuncture have the highest efficacy. It was also discovered that treatment modalities combined with herbal medicine are more useful than HRT alone. According to SUCRA, the combination of HRT with herbal medicine and acupuncture (99.9%) had the highest effective rate, followed by HRT combined with herbal medicine (71.0%), herbal medicine combined with catgut-embedding therapy (53.9%), herbal medicine alone (40.4%), herbal medicine combined with acupuncture (22.3%) and HRT alone (12.3%). Table 4 (part D) and Figure 5D presents the ranking of these six treatments.

NMA on AMH levels
The results of the NMA showing treatment comparisons based on the improvement of AMH levels in five treatment modalities are shown in Figure 3E. The league table results regarding the comparison of the AMH levels are presented in Figure 4E. A total of 11 RCTs with reported AMH levels (1,030 participants) were analyzed. The results of NMA indicate that herbal medicine combined with catgut-embedding therapy had the highest efficiency and treatment modalities with the presence of herbal medicine are more effective compared to treatment without herbal medicine. According to SUCRA, herbal medicine combined with catgut-embedding therapy (67.2%) had the highest effective rate, followed by HRT combined with herbal medicine (66.7%), herbal medicine alone (65.8%), acupuncture combined with moxibustion (33.6%) and HRT alone (16.7%). Table 4 (part E) and Figure 5E depict the rankings of these five treatments.

Evaluation of statistical inconsistency
Although global inconsistency showed there that were no significant statistical differences for the outcomes, the node-splitting method was employed to further evaluate on the local inconsistency (Table 5). Overall, no statistical inconsistency for the outcomes were seen from the node-splitting outcome, except between direct and indirect comparisons of HRT with combination of HRT and herbal medicine, HRT with herbal medicine, HRT with combination of herbal medicine and acupuncture, herbal medicine with combination of HRT and herbal medicine, herbal medicine with combination of herbal medicine and acupuncture, as P<0.05 indicates for local inconsistency between the local comparisons. These might be due to ineffective sample size for the particular treatment comparisons.

Small-study effects
There was a lack of strong evidence on the small-study effects across outcomes based on total effective rate and FSH level ( Figure 6A and B).
However, among the outcome indicators of E 2 , LH and AMH ( Figure 6C, D and E), it could be inferred that heterogeneity between studies were one of the reasons for those studies located on the lateral side of the funnel plot. Comparisoncorrection funnel plots based on E 2 , LH and AMH levels were made upon 26, 19 and 11 included studies, respectively (involving seven, six and five interventions, respectively). The asymmetrical funnel plots suggest probability of smallstudy effects in this study.

Sensitivity analysis
For the paired studies with significant heterogeneity (I 2 > 50%), a sensitivity analysis was done by removing any single study one by one. Results showed that there were no significant changes, stipulating that the pooled results were consistent.

Discussion
Previous studies 9,10,17 have demonstrated the efficacy of herbal medicine and acupuncture in treating POI. In this study, we utilized the advantages of NMA method to compare the efficacy of various TCM treatment modalities for this condition. Our findings revealed the comparative efficacy of six treatment modalities: HRT combined with herbal medicine and acupuncture, HRT combined with herbal medicine, herbal medicine combined with acupuncture, herbal medicine combined with catgut-embedding therapy, acupuncture combined with moxibustion and herbal medicine alone. Table 6 shows the compilation of SUCRA rank for each outcome measurement in NMA. Our results indicated that the combination of HRT with herbal medicine and acupuncture was the most effective treatment modality for patients with POI in the present study. These findings suggest that TCM treatment approaches could be considered as a promising therapeutic option for women with POI.
To the best of our knowledge, this study is the first systematic evaluation covering all treatment modalities in TCM to provide reliable evidence to determine the efficacy and feasibility of TCM treatment for patients with POI. There are a variety of different treatment modalities in TCM and all seem effective in treatment. 11,18 However, until now there have been no consensus or standard procedure of TCM treatment recommended for POI. This scientific direct and indirect comparison can provide important guidance on the pros and cons of each intervention and their treatment priority.
There is no record about POI found in the TCM literature. However, comparing its clinical manifestations, POI can be categorized as "amenorrhoea", "oligomenorrhoea" and "infertility" based on TCM perspective. The pathogenesis is mainly related to malnourishment of the thoroughfare and conception vessels, as well as the uterus, which might be due to congenital insufficiency, physical and mental overwork, or sexual overstrain. TCM treats a disease based on syndrome differentiation. Different treatment modalities such as oral administration of herbal decoction, acupuncture, catgutembedding therapy, moxibustion, electroacupuncture, ear acupressure and herbal enema, are suitable for the treatment of POI. In our study, 33 clinical trials using TCM treatment were screened and the therapeutic effect of seven treatment modalities were analyzed. It has been proven that comprehensive treatment models are more effective in improving ovarian function, compared to just using HRT as sole treatment. This would provide clinical practitioners with strong and reliable evidence, as well as a useful alternative during clinical decision-making.
We also found that the integration of herbal medicine into POI treatment, such as its combination with HRT, acupuncture or catgut-embedding therapy, can significantly improve the ovarian functions by observing on indicators including total effective rate, FSH, E 2 , LH and AMH levels. Therefore, herbal medicine may be considered as an intervention when HRT alone, the standard procedure recommended by recognized guidelines worldwide, 1,4,5 cannot effectively cure POI. This study possesses significance in providing clinical guidance. Acupuncture has been widely applied on POI treatment particularly in China. 10,11,18 It has been demonstrated that acupuncture is safe and effective. [19][20][21] Through early intervention of acupuncture, it has also been found that acupuncture can increase pregnancy rate significantly. 22,23 Our study has demonstrated that application of acupuncture in the treatment of POI is also useful to observe clinical improvement. However, it is suggested that acupuncture must be combined with herbal medicine and even HRT in order to optimize the therapeutic effects. This recommended treatment modality not only demonstrates the advantage of TCM treatment, it also provides a potential practical, safe and low-cost treatment option for both the patients and clinicians.
In this study, we found a shortage of rigorous clinical trials adhering to the international requirements, as well as a paucity of heterogeneity of study candidates, interventions, controls, and outcome measures, which challenge the process of systematic review and NMA. Due to inadequate data from the included studies, the safety of each intervention could not be evaluated. The risk of bias assessment in this study indicates that most of the studies did not clearly state the blinding of outcome assessors, which might impact on the reliability of data analysis outcome.

Conclusions
In conclusion, herbal medicine combined with catgut-embedding therapy was shown to be the most effective treatment model to improve total effective rate and AMH levels in the present study. The combination of HRT, herbal medicine and acupuncture was demonstrated to be the most effective treatment model for improving serum FSH, E 2 and LH levels. An overview on SUCRA analysis outcomes showed that comprehensive treatment models with the integration of herbal medicine seemed more prone to improve ovarian functions. Monotherapies, such as solely using HRT or herbal medicine treatment, were found to have no obvious advantages on their curative effect for POI. The integration of herbal medicine to conventional HRT may be a favorable approach for enhancing the clinical utility of HRT in the management of POI. RCTs involving multiple centers and different countries are essential to further verify the conclusions of this study, which can also create international awareness on the efficacy of TCM treatments for POI.

Data availability
Underlying data All data underlying the results are available as part of the article and no additional source data are required.