Acupuncture Methods for Primary Trigeminal Neuralgia: A Systematic Review and Network Meta-Analysis of Randomized Controlled Trials

Background Primary trigeminal neuralgia (PTN) is a clinical refractory disorder characterized by excruciating pain that severely impacts the quality of life. Several studies have shown that acupuncture can improve PTN pain. However, the comparative efficacy and safety of acupuncture are unknown. Herein, a systematic review was conducted to compare the efficacy and safety of various acupuncture methods for PTN treatment. Methods Relevant randomized controlled trials (RCTs) published up to 1 August 2021 were obtained from PubMed, Embase, Cochrane Central Register of Controlled Trials, Web of Science Core Collection, Chinese National Knowledge Infrastructure, Chinese Biomedical Literature Database, CQVIP Database, Wanfang Database, Allied and Alternative Medicine Database, and related registration platforms. Two authors independently selected the studies and obtained data. Cochrane Handbook was used to assess the methodological quality. We put the pain relief as the primary outcome and the response rate and adverse events as the secondary outcomes. Review Manager v5.3, ADDIS v1.16.8, and STATA v15.0 software were used for data analysis. The intraclass correlation coefficient was used to assess the consistency of the two investigators. Results A total of 58 RCTs with 4,126 participants were obtained. The meta-analysis indicated that five acupuncture methods were superior to conventional medicine (carbamazepine) in pain reduction intensity and response rate. Meanwhile, electronic acupuncture plus manual acupuncture was the most effective therapy since it reduced pain intensity in 11 methods and improved the response rate in 10 interventions. Moreover, six interventions had acceptable adverse events, and none of the included studies reported severe adverse events. However, most pieces of evidence were ranked as critically low. Conclusion These findings show that acupuncture methods can be effective and safe for PTN. Moreover, electronic acupuncture plus manual acupuncture maybe the best acupuncture treatment for PTN and should be administered to PTN patients. However, additional well-designed and high-quality RCTs should be conducted to verify the above findings in the future. The systematic review is registered with CRD42020221456.


Introduction
Primary trigeminal neuralgia (PTN) is a common neuralgia caused by the compression of an aberrant tissue associated with the trigeminal nerve [1,2]. e latest classification system identifies TN as either classical or idiopathic TN based on the degree of neurovascular contact or secondary TN caused by pathology other than neurovascular contact [1]. Meanwhile, it is severe, unilateral, paroxysmal, and recurring pain [3] which can severely impact the quality of life [4]. Al-Quliti and colleagues illustrated that 26.8 people per 100,000 suffer from PTN, usually occurring in middleaged and elderly people [5][6][7]. PTN is a global public health issue [8]. Several interventions have been used to control PTN, and carbamazepine is the common treatment method since the 1960s [9]. However, many patients do not tolerate pharmacological therapies over a long period due to the carbamazepine side effects [8,10], thus seeking nonpharmacological treatments.
As an ancient nonpharmacological therapy, acupuncture has been used for PTN treatment in China [3,[11][12][13][14][15] for a long period [16]. Numerous studies have also shown that acupuncture can be used for PTN treatment [3,[11][12][13][14][15][16][17]. Furthermore, the acupuncture analgesic effect is associated with substance P (SP) and β-endorphin. Several trials have demonstrated that acupuncture can decrease SP content and enhance the expression of β-endorphin [18][19][20][21]. Although previous systematic reviews [3] have shown that acupuncture methods have good efficacy and safety in improving PTN symptoms, they only focused on a unitary acupuncture method against antiepileptic drugs (carbamazepine)/sham acupuncture (SA). Meanwhile, there are various acupuncture therapies. For instance, the PTN acupuncture methods include manual acupuncture (MA) therapy, acupuncturemoxibustion (AM) therapy, electronic acupuncture (EA) therapy, and fire acupuncture (FA) therapy. erefore, it is necessary to determine the most appropriate acupuncture methods for PTN therapy.
Network meta-analysis (NMA) analyzes diversified methods from different trials and calculates their relative effects [22][23][24]. It was applied to compare various interventions simultaneously in a unitary analysis by combining indirect and direct evidence in a network of trials. It may also assist to determine suitable therapeutic methods and illustrate their efficacy and safe application in clinical operations, thus guiding health policies [25]. us, this study aimed to compare and rank the efficacy and safety of all known acupuncture therapies on PTN via NMA.

Methods
e study was registered on PROSPERO (registration number: CRD42020221456) and was conducted following the Preferred Reporting Items for Systematic Review and Meta-Analysis-Network Meta-Analysis (PRISMA-NMA) [26] and the accompanying checklist (Appendix 1).

Eligibility Criteria and Exclusion Criteria.
e PICOS (participant, intervention, control, outcome, and study design) criteria were used for inclusion and exclusion analysis.

Types of Studies.
All RCT studies reported in English/ Chinese were included, while quasi and cluster RCTs, non-RCTs, case reports, and studies with no detailed data were excluded.

Types of Participants.
All PTN patients, regardless of gender, race, and age, were included. Similarly, all types of PTN, including classical TN and idiopathic TN [1] regardless of their etiology, severity, or specific criteria diagnoses, such as diagnostic criteria or references, were included. TN patients without an association with the distinct primary disease were excluded.

Types of Interventions.
Herein, only trials using acupuncture as monotherapy or alternative treatments were included. According to Revised Standards for Reporting Interventions in Clinical Trials of Acupuncture (STRICTA) [27], the expected acupuncture methods included MA, EA, AM, FA, and the like. Moreover, articles describing combinations of these acupuncture methods with conventional medicine (CM) were included. e studies without a clear description of the acupuncture process, such as disinfection and sterilization, acupuncture manipulation, and posttreatment processes, were excluded. Besides, studies using acupoint embedding, acupoint application, acupoint injection, bee venom acupuncture, and other therapies were excluded because of using related drugs. Trials combining acupuncture and herbal medicine, cupping, and blood-letting puncture therapies were also excluded.

Type of Controls.
e comparison groups, such as the placebo group (SA), and conventional medicine (carbamazepine, the dosage of carbamazepine should be well documented) were based on different acupuncture methods.

Types of Outcomes.
e following studies were included: ose that compared and ranked the efficacy and safety of all acupuncture methods used in PTN. Pain intensity reduction determined using the Visual Analogue Scale (VAS) and Numerical Rating Scale (NRS) was the acceptable primary outcome. e VAS and NRS were converted to the 11-point NRS (0 indicates no pain, and 10 shows the most severe pain) due to their similarity [28]. e response rate and adverse events (AEs) were the secondary outcomes. Reduction of pain intensity and response rate were used to evaluate the efficacy of intervention methods, while AE was used to assess the safety of intervention methods. Outcome measures that were not relevant to PTN were excluded.

Study Selection and Data Extraction. Two investigators
independently selected the studies and extracted data. We first read the study title and abstract to identify duplicate research and then uploaded the remaining part to Note-Express V.3.0. e two reviewers (ZY and FW) preliminary screened the titles/abstracts to find suitable articles and then downloaded and read the studies. Finally, the 2 researchers each read the residual full-text studies to select those that meet the inclusion criteria. A 3rd party (LZ or FL) resolved any controversy.
A standardized Microsoft Excel 2010 sheet was used to extract data, such as study identity (first author, publication year, and country), study design, diagnostic criteria, characteristics of patients (age, gender, and sample size), details of intervention group and control group, treatment courses, outcomes (primary and secondary outcomes), and results. e corresponding/first author of the article was requested for additional information in ambiguous or insufficient detail cases. Besides, a description was added to the final result when the data details were not available. e selection procedure was outlined following the PRISMA flowchart.

Study Quality Assessment.
Two authors assessed the risk of bias (ROB) in all eligibility trials using the Cochrane Collaboration ROB assessment tool 2.0 [29]. Risk levels of randomization process, deviations from intended interventions, missing outcomes data, measurement of the outcome, and selection of the reported results were indicated as "low," "some concerns," or "high." A 3rd party (LZ or FL) resolved any misunderstanding.

Statistical
Analysis. RevMan 5.3 software was used for data analysis. e 3-arm trials were divided into 2-arm trials for all possible combinations. A pooled mean difference (MD) for continuous outcomes or relative risk (RR) of dichotomous variable data with 95% confidence intervals (CI) was presented for each intervention. e random effects model adopted by the DerSimonian-Laired method was used as a conservative estimate [30]. e I 2 statistic and p values were determined for statistical heterogeneity analysis. p < 0.05 and I 2 >75% indicated significant heterogeneity. e Bayesian network analysis framework and Markov Chain Monte Carlo (MCMC) method [31] were used to assess and process a priori data via the Aggregate Data Drug Information System (ADDIS V.1.16.8 software, Drugis, Groningen, NL). e parameters were 4 Markov chains for simulation at 50,000 simulation iterations to determine their posterior distributions. Moreover, the first 20,000 simulation iterations were used to eliminate the initial value's impacts and the last 30,000 were used for sampling. e node-splitting method was used to integrate direct and indirect multiple-treatment comparisons of the RCTs [32]. STATA version 15.0 software (StataCorp LP, Texas, USA) was used for network plot analysis, where node sizes indicated the number of study patients and connection sizes showed the number of trials in each treatment. Finally, the node-splitting method was also used to assess the local inconsistency [33]. p < 0.05 was considered a significant statistical difference between indirect and direct multiple-treatment comparisons. Either the inconsistency model or consistency model was used. e model convergence is the potential scale reduced factor (PSRF), and PSRF close to 1 indicates successful convergence [34].

Publication Bias.
A funnel plot was used to assess reporting bias since over 10 studies were evaluated in the meta-analysis.

Quality of Evidence.
e Grades of Recommendations, Assessment, Development, and Evaluation (GRADE) was used to assess the overall quality of the evidence [35,36] and ranked it as "high," "moderate," "low," and "critically low." 2.8. Assessing Reviewer Agreement. Since the two reviewers independently selected and extracted data and evaluated the quality of RCTs, the intraclass correlation coefficient (ICC) [37] was determined to assess their consistency (ICC score, 0.95).

Study Selection.
A total of 1187 studies were obtained, and only 58 RCTs [11,17, with 4126 patients were selected for the systematic review. e selection process is presented in Figure 1.
e response rate was the frequently used outcome measures in most studies. e characteristics of the included studies are shown in Table 1.

Study Quality Assessment.
e methodology quality of the included RCTs was assessed. (1) Randomisation process: only two trials [45,58] were ranked as "low risk" because they had detailed information, while 56 trials were ranked as "unclear" due to insufficient information. (2) Deviations from intended interventions: all trials had insufficient information on intended interventions and were ranked as "some concerns." (3) Missing outcomes data: all RCTs were ranked as "low risk" due to the complete implementation plan. (4) Measurement of the outcome: a total of 56 trials were ranked as "some concerns" and only two [17,73] as Evidence-Based Complementary and Alternative Medicine "low risk." (5) Selection of the reported results: all the RCTs were ranked as "some concerns" due to insufficient information. e quality assessment detail of the RCTs is shown in Figure 2.

Reduction in Pain Intensity.
A total of 14 pairwise meta-analyses were conducted to compare the pain reduction intensity of various therapies (         pain intensity reduction (Figure 3(a)), while 53 RCTs with 3,744 participants using 10 therapies showed a response rate (Figure 3(b)).

Statistical Inconsistency Analysis.
e node-splitting method was used for local inconsistency analysis. e direct and indirect effects had no statistically different reduction of pain intensity and response rate (p ≥ 0.05).

Reduction in Pain
Intensity. STATA 15.0 was used for network plot analysis of the 11 treatments (Figure 3(a). e Bayesian network meta-analysis was conducted using the consistency model to produce the ranking probability plot since the PSRF scores were close to 1 (Appendix 3(a)) and p values were higher than 0.05 (Appendix 3(b)). MA + EA, FA, and AM + CM significantly reduced pain intensity (Figure 4(a)). However, MA + EA was the optimum intervention method. e meta-analyses details are shown in Table 4. MA + EA was more effective than EA, MA, MA + CM, AM, CM, SA, and SA + CM. Furthermore, MA + EA and FA were more effective than MA, SA + CM, SA, and CM. MA + EA, FA, AM + CM, EA, EA + CM, MA, and MA + CM were more effective than CM. AM + CM was significantly effective than SA.

Response
Rate. STATA 15.0 was also used for network plot analysis of 10 interventions (Figure 3(b)). Similarly, the Bayesian meta-analysis was conducted using the consistency model to produce the ranking probability plot (Figure 4(b)) since the PSRF scores were close to 1 (Appendix 4(a)) and p values were higher than 0.05 (Appendix 4(b)). MA + EA, FA, and AM had the most significant response rates (Figure 4(b)). However, MA + EA/FA maybe the optimum intervention method. e meta-analyses details are shown in Table 5. MA + EA was more effective than 4 interventions (EA, MA, CM, and SA). MA + EA, MA + CM, MA, EA + CM, and AM + CM were more effective than CM, and MA + EA, FA, AM, EA, MA, EA + CM, and AM + CM were highly effective than SA. ree studies reported 52 MA + CM-related adverse events, one study showed seven EA + CM-related adverse events, and one study reported two AM + CM-related adverse events. e details of adverse events between acupuncture and the CM group are shown in Table 6.

3.7.
Heterogeneity. RevMan 5.3 was used for sensitivity analysis to evaluate the stability and reliability of the results. e results were shown to be considerably reliable and stable   16 Evidence-Based Complementary and Alternative Medicine     Evidence-Based Complementary and Alternative Medicine

Publication Bias.
A funnel graph was used to determine the reporting bias assessment ( Figure 5). e graph indicated a low risk of publication bias, as shown in the comparisonadjusted funnel plots for reducing pain intensity ( Figure 5(a)). However, the funnel plots of response rate indicated a potential reporting bias ( Figure 5(b)).

Quality of Evidence.
e GRADE criteria were used to compare the direct and indirect evidence. e quality of outcomes was critically low to moderate. However, most evidence were critically low (Appendixes 5 and 6), mainly due to the ROB, inconsistency, and imprecision.

Discussion
Although CM (carbamazepine) is used for PTN treatment, it has some inevitable adverse effects [94]. In this study, alternative PTN treatments were explored. Acupuncture has been successfully adopted for PTN treatment for a long time. However, several acupuncture therapies used are not regulated and standardized. NMA is used when there is no possibility of head-to-head or direct comparison of the intervention of interest versus control. us, it is applied to analyze several RCTs with various acupuncture therapies and rank these interventions [95]. is is the first study to conduct a Bayesian network meta-analysis to assess the efficacy of different acupuncture methods for PTN treatment based on the PRISMA-NMA.

Summary of Evidence.
In this study, the efficacy of various therapies for PTN treatment was compared and reliable results were obtained [96,97]. For reduction of pain intensity, EA + MA efficacy was statistically different compared with EA and MA efficacies. MA/EA efficacy was also statistically significant compared with CM efficacy. Furthermore, MA/EA/ AM combined with CM showed statistically significant efficacy compared with CM. FA and SA also had different efficacies. EA + MA was the optimal acupuncture method for pain intensity reduction. EA + MA significantly reduced pain intensity compared with EA, MA, MA + CM, AM, CM, SA, and SA + CM. MA + EA, FA, AM + CM, EA, EA + CM, MA, and MA + CM were more effective than CM. For the response rate, MA and CM had different efficacies. MA + EA, MA + CM, MA, EA + CM, and AM + CM were more effective than CM. MA + EA, FA, AM, EA, MA, EA + CM, and AM + CM were highly effective than SA. EA + MA/FA was the optimal acupuncture treatment for improved response rate. However, most evidence had a critically low quality. In safety, 24 RCTs (41.38%) showed adverse events and 22 RCTs (37.93%) reported carbamazepine-related adverse events (dizziness, drowsiness, and gastrointestinal reaction). However, 15 RCTs (25.86%) reported acceptable acupuncture-related adverse events (dizziness, drowsiness, and fatigue). No study reported severe adverse events.

Strengths and Limitations.
is study has several strengths. (a) is is the first network meta-analysis to include 58 RCTs and compare different acupuncture methods, such as AM, EA, FA, MA, and their combinations with CM, or acupuncture versus sham acupuncture and CM groups for PTN treatment; (b) this systematic review was registered on PROSPERO and followed the PRISMA-NMA guidelines to reduce ROB; (c) ICC was used to assess the reviewers' agreements to ensure the reliability of the assessment results; (d) GRADE was used to assess the quality of evidence.
However, this study has some limitations. First, the evidences were of low quality, and several limitations, including study design, implementation, analysis, and publication, influence ROB. Second, while there were strict criteria to control the quality of evidences, the included studies had several therapies and can affect the evidence quality. ird, the results had a statistically significant heterogeneity possibly due to age, gender, treatments, details of acupuncture, carbamazepine dose, or other factors from the PTN patients. Besides, most included RCTs had no follow-up time and the long-term effect of acupuncture could not be determined. Finally, all the included studies were from China and the search strategies were limited to English and Chinese, thus regional limitation.

Suggestions to Future
Research. In this study, the methodological quality of all trials was moderate, but the quality of evidences was critically low, possibly due to the exclusion of several details (randomization, deviations from intended interventions, outcome measurement, and selective bias). erefore, the project design, implementation, analysis, and writing of studies should strictly follow the latest edition of Cochrane Handbook for Systematic Reviews [98], the Consolidated Standards of Reporting Trials (CONSORT) [99], and STRICTA [27] in the future. Second, this study did not show the long-term efficacy and safety of acupuncture methods. erefore, it is important to consider longitudinal trials to explore long-term efficacy and safety. Moreover, several factors, such as acupoint selection, treatment duration, intervention time, and dose of carbamazepine, impact heterogeneity. erefore, future studies should standardize and regulate the acupuncture and CM details for PTN treatment. Finally, other core outcomes, such as changes in the psychological situation and quality of life, should be considered [100].

Conclusion
In conclusion, acupuncture reduces pain intensity, improves response rate, and has less adverse events on PTN patients. Moreover, five acupuncture therapies are superior to conventional medicine (carbamazepine). In this study, EA + MA was regarded as the optimal therapy for PTN. However, the overall quality of evidences from included studies was critically low. erefore, well-designed and high-quality clinical trials are needed to confirm the abovementioned findings in the future.