We registered this systematic review and meta-analysis at PROSPERO: https://www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=329925 (Registration ID: CRD42022329925). This systematic review and meta-analysis was conducted according to A Measurement Tool to Assess Systematic Reviews (AMSTAR 2) (Shea et al., 2017) and reported in light of the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) 2020 statement (Page et al., 2021).

We included studies that met all of the following criteria:

Studies were excluded if they met one of the following conditions:

The following databases were searched from their inceptions until 27 October 2022: PubMed, EMbase, the Cochrane Library, the Chinese Biomedical Literature Database (CBM), the China National Knowledge Infrastructure (CNKI), the Chinese Science and Technology Periodical Database (VIP), and the Wanfang database (Wanfang Data). We utilized Medical Subject Headings (MESH) and free terms related to acupuncture and CTS to build search strategies. The search strategies for the above databases are provided in Supplementary material 1. We manually searched gray literature, reference lists of relevant reviews, and trial registers (ClinicalTrials.gov and the Chinese Clinical Trials Registry). Meanwhile, relevant experts were consulted for potentially eligible studies.

EndNote X9 was used to manage the literature. Two independent reviewers (PY and GC) conducted the study selection. After removing duplicates, irrelevant records which were screened according to titles or abstracts were excluded. Then, the rest records with full text were scrutinized to identify eligible studies. The two reviewers cross-checked their identified studies and discussed any disputes.

The data on the following aspects were extracted by two reviewers (JL and JD) independently:

If there are multiple-arm RCTs, we included only data from the arms with interventions relevant to this study. Two reviewers cross-checked the extracted information. Any discrepancy was resolved through discussion. The authors would be contacted if there was missing information.

Two independent reviewers (QD and XL) assessed the risk of bias in the included studies using version 2 of the Cochrane risk-of-bias tool for randomized trials (ROB 2). According to ROB 2, five domains of bias were evaluated: the randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Each domain of individual study and all included studies were rated as “low risk,” “some concerns,” or “high risk.” Any disagreements were resolved with a third reviewer (RJ).

Two independent reviewers (HF and QD) utilized the Revised Standards for Reporting Interventions in Clinical Trials of Acupuncture (STRICTA) to evaluate the reporting quality of interventions for each included study based on the following six items (17 sub-items) (MacPherson et al., 2010): acupuncture rationale, details of needling, treatment regimen, other components of treatment, practitioner background, and control or comparator interventions. The third reviewer (RJ) participated in resolution of discrepancies.

Two independent reviewers (FW and YY) assessed the certainty of the evidence with the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) system. Each outcome was assessed based on five aspects: limitations, inconsistency, indirectness, imprecision, and publication bias, and categorized as high, moderate, low, or very low evidential certainty. GRADEpro (Version 3.6) software was used to evaluate the evidence and summarize the findings.

We evaluated acupuncture’s effects as monotherapy and adjunctive treatment, respectively. If feasible, meta-analyses were conducted using post-intervention data when clinical homogeneity existed between studies. We calculated the mean difference (MD) with 95% confidence intervals (CIs) for continuous data measured by uniform standards. Otherwise, standardized mean differences (SMDs) and 95% CIs were evaluated. For dichotomous data (e.g., responder rate), we calculated the risk ratios (RRs) and 95% CIs. The Chi-square test with a significance level of P < 0.10 and I² statistic were used to detect and quantify heterogeneity, respectively. The random-effects model (REM) was applied in meta-analyses if there was substantial heterogeneity (P < 0.1 or I² value >50%). Otherwise, the fixed-effects model (FEM) was used. We conducted descriptive analyses when meta-analyses were not appropriate or possible. Review Manager software (version 5.4.1) was used for data synthesis.

We obtained a total of 1,486 records in the literature search. After removing 550 duplicates, we excluded 880 irrelevant records based on their title and abstract. The full text of 56 remaining records was then evaluated, and 16 eligible studies (Kumnerddee and Kaewtong, 2010; Jin and Lang, 2011; Li, 2011; Yang et al., 2011; Yao et al., 2012; Ramin, 2013; Xiang et al., 2014; Hadianfard et al., 2015; Chung et al., 2016; Maeda et al., 2017; Ural and Özturk, 2017; Xie et al., 2018; Tezel et al., 2019; Bahrami-Taghanaki et al., 2020; Xiong, 2020; Huang and Lin, 2022) were included in the final analysis. A list of excluded records with reasons is provided in Supplementary material 2. The PRISMA flow chart presents the selection procedure (Figure 1). Of the included studies, eight were conducted in China (Jin and Lang, 2011; Li, 2011; Yang et al., 2011; Xiang et al., 2014; Chung et al., 2016; Xie et al., 2018; Xiong, 2020; Huang and Lin, 2022), two in the USA (Yao et al., 2012; Maeda et al., 2017), three in Iran (Ramin, 2013; Hadianfard et al., 2015; Bahrami-Taghanaki et al., 2020), two in Turkey (Ural and Özturk, 2017; Tezel et al., 2019), and one in Thailand (Kumnerddee and Kaewtong, 2010). The sample size of the studies ranged from 27 to 181, with a total of 1,025 participants. The mean age of participants varied between 36.4 and 53.6 years. Fifteen studies included patients with mild to moderate CTS, and one study (Jin and Lang, 2011) did not specify the severity of CTS. Five studies used acupuncture as monotherapy, while 11 studies investigated its adjunctive effect. Table 1 show the characteristics of the included studies.

During the randomization process, 11 studies specified the randomization method (Kumnerddee and Kaewtong, 2010; Yang et al., 2011; Yao et al., 2012; Xiang et al., 2014; Hadianfard et al., 2015; Chung et al., 2016; Maeda et al., 2017; Ural and Özturk, 2017; Xie et al., 2018; Tezel et al., 2019; Bahrami-Taghanaki et al., 2020). Two studies (Yang et al., 2011; Chung et al., 2016) implemented appropriate methods to conceal the allocation sequence. All studies reported that there were comparable baselines between groups. Two studies (Yao et al., 2012; Maeda et al., 2017) blinded patients by conducting sham comparisons between the groups. Additionally, several outcomes, including symptom severity, functional status, and pain intensity, were participant-reported outcomes, which meant outcome assessors were blinded in the studies (Yao et al., 2012; Maeda et al., 2017). Seven studies (Kumnerddee and Kaewtong, 2010; Yang et al., 2011; Yao et al., 2012; Chung et al., 2016; Maeda et al., 2017; Tezel et al., 2019; Bahrami-Taghanaki et al., 2020) described dropouts rate with 1.6–18.3%; among these studies, four trials (Chung et al., 2016; Maeda et al., 2017; Tezel et al., 2019; Bahrami-Taghanaki et al., 2020) did not give the detailed reason of dropouts, and three studies (Yang et al., 2011; Yao et al., 2012; Chung et al., 2016) used intent-to-treat analysis. Four trials (Yang et al., 2011; Chung et al., 2016; Maeda et al., 2017; Bahrami-Taghanaki et al., 2020) provided the registration number or published protocol, and all of them reported planned outcomes. Overall, 1 RCT (Yang et al., 2011) was rated as having a low risk of bias, 14 (Kumnerddee and Kaewtong, 2010; Jin and Lang, 2011; Li, 2011; Yao et al., 2012; Ramin, 2013; Xiang et al., 2014; Hadianfard et al., 2015; Chung et al., 2016; Maeda et al., 2017; Ural and Özturk, 2017; Xie et al., 2018; Tezel et al., 2019; Xiong, 2020; Huang and Lin, 2022) had some concerns, and 1 (Bahrami-Taghanaki et al., 2020) was a high risk of bias. The results of the risk of bias in individual studies and the overall risk of bias are shown in Figure 2.

There were different acupuncture techniques, among which manual acupuncture was applied in seven studies (Yang et al., 2011; Yao et al., 2012; Hadianfard et al., 2015; Ural and Özturk, 2017; Tezel et al., 2019; Bahrami-Taghanaki et al., 2020; Huang and Lin, 2022) and electroacupuncture in nine (Kumnerddee and Kaewtong, 2010; Jin and Lang, 2011; Li, 2011; Ramin, 2013; Xiang et al., 2014; Chung et al., 2016; Maeda et al., 2017; Xie et al., 2018; Xiong, 2020), respectively. All studies reported the selected acupoints, and the frequency of all acupoints is shown in Figure 3. The most used acupoints were Daling (PC7, 100%), Neiguan (PC6, 75.0%), Hegu (LI 4, 50.0%), Quchi (LI 11, 50.0%), and Laogong (PC 8, 37.5%). Thirteen included studies (Kumnerddee and Kaewtong, 2010; Jin and Lang, 2011; Yang et al., 2011; Yao et al., 2012; Ramin, 2013; Xiang et al., 2014; Hadianfard et al., 2015; Chung et al., 2016; Ural and Özturk, 2017; Tezel et al., 2019; Bahrami-Taghanaki et al., 2020; Xiong, 2020; Huang and Lin, 2022) applied the fixed acupoint protocol and 3 (Li, 2011; Maeda et al., 2017; Xie et al., 2018) used individualized acupoint protocol (fixed main acupoints plus acupoints based on syndrome differentiation). In addition, the retention time was mainly 20 or 30 min, and the total sessions ranged from 6 to 36 sessions within 20 days to 17 weeks of treatment duration.

According to the STRICTA checklist, the items with more than 70% of reporting rates were item 3a (number of treatment sessions, 100%), item 2e (needle stimulation, 100%), item 3b (frequency and duration of treatment sessions, 93.8%), item 6b (precise description of the control or comparator, 87.5%), item 2d (response sought, 81.3%), item 2f (needle retention time, 81.3%), and item 2g (needle type, 81.3%). Item 4b (setting and context of treatment) and 1c (the extent to which treatment was varied) were not reported in the included studies. Detailed information on the STRICTA checklist is provided in Supplementary material 3.

The results for electrophysiological parameters are shown in Table 2.

Three studies (Jin and Lang, 2011; Xie et al., 2018; Xiong, 2020) provided the responder rate. Jin and Lang (2011) reported a comparable responder rate between the acupuncture group and the medicine group. Xie et al. (2018) reported a superior responder rate in the acupuncture plus medicine group than the medicine group. Xiong (2020) observed that acupuncture plus ultrashort wave therapy had no better than ultrashort wave therapy in responder rate.

Four studies (Jin and Lang, 2011; Yang et al., 2011; Yao et al., 2012; Tezel et al., 2019) found no adverse events or serious adverse events related to acupuncture treatment occured. Kumnerddee and Kaewtong (2010) observed 6 of 30 cases in electroacupuncture group experienced skin bruises but no serious complication took place. Another study (Chung et al., 2016) reported electroacupuncture-related adverse events, including bruises at acupoints (4/90), mild local dermatitis around acupoints (3/90), increased pain (2/90), and numbness and tingling after electroacupuncture treatment (2/90), and the above adverse events disappeared within a week. The rest of the 10 studies provided no information on the adverse events.

There was low and very low certainty of evidence attributed to some concern risk of bias, imprecision, and strongly suspected publication bias. A summary of the finding table is provided in Supplementary material 4.

We included 16 RCTs with 1,025 subjects and explored the effect of acupuncture as monotherapy and adjunctive therapy on CTS. Compared with night splints, acupuncture alone was more effective in relieving pain, but there were no differences in symptom severity and functional status. Acupuncture had no advantage over medicine in improving symptom severity or electrophysiological parameters. As an adjunctive treatment, acupuncture might effectively alleviate symptom severity, functional status, pain intensity, and electrophysiological parameters.

Meanwhile, acupuncture as adjunctive therapy was more effective than medicine to ameliorate symptom severity, functional status, pain intensity, and electrophysiological parameters. According to narrative analysis, acupuncture as monotherapy or adjunctive therapy, showed no superiority to sham acupuncture. Few acupuncture-related adverse events were reported. The above evidence had low or very low certainty.

Sim et al. (2011) included six RCTs and published the first systematic review of acupuncture for CTS, but three of the RCTs they identified were excluded from our study because the participants received other traditional Chinese medicine in one RCT (Shi et al., 2006) and cointerventions between groups were not comparable in the other two RCTs (Hu et al., 2000; Cai, 2007). Limited by insufficient RCTs, Sim et al. (2011) summarized the evidence of acupuncture for CTS as encouraging but not convincing. Choi et al. (2018) also found insufficient evidence to assess the effect of acupuncture and related interventions on CTS with 12 identified RCTs. Wu et al. (2020) conducted the latest systematic review involving 10 RCTs. Except for manual and electroacupuncture, they included laser acupuncture, moxibustion, and transcutaneous electrical nerve stimulation treatment. They drew the conclusion that acupuncture and related therapies appeared to be effective in improving symptoms, function, and pain in CTS, and emphasized that the validity of such a conclusion was limited. We included 16 RCTs to update the evidence and investigate the effect of acupuncture as monotherapy or adjunctive therapy for CTS.

The narrative analysis showed that neither acupuncture alone nor acupuncture as a adjunctive treatment had superiority over sham acupuncture. However, these results were derived from two independent studies, respectively (Yao et al., 2012; Maeda et al., 2017). Given the limited studies and risk of the underrated effect of acupuncture in sham-control trials with a small sample size (Lundeberg et al., 2008; Birch et al., 2022a,b), we failed to identify the advantage of acupuncture over sham acupuncture for CTS, which should continue to be explored in future studies.

One included RCT showed that 10-session acupuncture alone might be more effective than night splints in relieving pain intensity but not symptom severity or functional status. Night splints are recommended for CTS to improve short-term symptoms and function (Erickson et al., 2019). Whether there is a different long-term effect between acupuncture and night splints is unknown. No clear advantages of acupuncture as a monotherapy were observed compared with medicine. Among comparative medicines used in included studies, such as prednisolone (Yang et al., 2011; Ramin, 2013) and oral vitamin B12 (Jin and Lang, 2011), only prednisolone was recommended by the AAOS. Yang et al. (2011) found that patients with CTS who received acupuncture had a lower recurrence rate than those who received prednisolone in the 1-year follow-up period, which indicated acupuncture might have a better long-term effect than prednisolone. Due to insufficient studies, we were unable to compare the effect of acupuncture with other active treatments. More relevant head-to-head trials should be conducted in the future to focus on clinical and cost effects.

Patients with CTS who received acupuncture plus other treatment(s) showed more improvement in symptoms, function, or pain. However, these positive findings of acupuncture as adjunctive therapy came from open-label RCTs, which could be influenced by the patients’ subjective intentions. Based on the results of electrophysiological parameters, we found acupuncture combined with medicine could improve median nerve function better than medicine alone, which provided objective evidence for the adjunctive effect of acupuncture. However, the adjunctive effect of acupuncture should be further investigated in clinical trials with objective outcomes.

The outcomes that were measured by subjective tools, such as CTQ, GSS, DASH, and VAS, relied on participants’ self-reports, which might induce measurement bias favoring acupuncture in open-label studies. Therefore, studies using objective outcomes are vital to build convincing evidence of acupuncture for CTS. According to the ROB 2 assessment, allocation concealment and advanced registration, or protocol, should be improved to enhance the credibility of the evidence. Meanwhile, in compliance with the STRICTA, authors should take care to report the details of the intervention, especially in items of acupuncture rationale, cointerventions, practitioner background and control or comparator interventions.

Our systematic review and meta-analysis included the latest RCTs and assessed the effect of acupuncture on CTS. However, several limitations exist and should be considered. In the present review, the small sample size, substantial heterogeneity, and potential risk of bias of the included studies reduced the certainty of the evidence. Thus, the findings should be treated with caution. Owing to limited RCTs and data, we failed to investigate the advantages of different acupuncture techniques, identify the optimal parameters of the acupuncture protocol, or explore the follow-up effect of acupuncture.