This network meta-analysis was conducted based on the preferred reporting items for systematic reviews and meta-analyses (PRISMA-NMA) (14). This study has been registered on PROSPERO, and the registration website is https://www.crd.york.ac.uk/prospero/, (ID: CRD42023410957).

RCTs on acupuncture treatment for post-stroke SHS are retrieved from PubMed, Cochrane, Embase, Web of Science, VIP (Chinese), Wanfang (Chinese), CNKI (Chinese), and Chinese Biomedical (Chinese) databases. There are no restrictions on language and publication years, and the retrieval date is until 12 March 2023. The retrieval is conducted in the form of “subject word + free word,” the search terms for the Chinese databases include “stroke,” “shoulder-hand syndrome,” “painful malnutrition syndrome,” “painful nutritional disorder,” “complex regional pain syndrome,” “reflex sympathetic dystrophy syndrome,” “acupuncture,” “electroacupuncture,” “warming needle,” “fire needling,” “floating needle,” and “catgut embedding “. The search strategy for the English databases is shown in Supplementary Table S1.

The retrieved studies were imported into EndNote, and duplicate publications were excluded. Then, the titles and abstract were checked to preliminarily select the original studies that met the criteria and download the full text. The full texts were read to select the original RCT studies that met the criteria of this systematic review. The study selection was independently conducted by two researchers (XT and XH), and any disputes were arbitrated by a third researcher (HY) after the search was completed.

Data extraction was independently performed by two researchers (NW and YJ), and a third researcher (JH) was asked to arbitrate disputes, if any. A data extraction electronic spreadsheet was developed to extract data, including the title, English abstract, first author, year of publication, study design, author’s country, diagnostic criteria, interventional protocol, type of stroke, treatment time, number of cases, gender, age, disease course, follow-up time, outcome measures, selection bias, implementation bias, measurement bias, loss to follow-up bias, reporting bias, and other biases.

Quality assessment was conducted by two researchers (JH and XH) using Cochrane’s bias risk assessment tool RoB 2.0 (16, 17) to assess the risk of bias in the 50 included original studies. The assessment tool includes the following seven items: randomization process, method of identifying or recruiting participants, bias due to deviation from the expected intervention, missing outcome data, bias in outcome measurement, reporting selection bias, and overall bias. The risk of bias in each domain is classified into three levels: low risk of bias, high risk of bias, and some concerns. After completion, a cross-check would be conducted, and if there was a dispute, a third researcher (TH) was asked to assist in adjudicating. Finally, the risk of bias graph was plotted.

The primary outcome measures are the visual analog scale (VAS) for pain and the Fugl–Meyer upper limb motor function assessment scale. The secondary outcome measures are the Barthel index (BI) and the shoulder-hand syndrome evaluation scale (SHSS). Since our outcome measures are all continuous variables, and there seem to be differences among the studies at baseline, we used the change from baseline to post-treatment as the effect size in the meta-analysis.

This study used a Bayesian random-effects model to compare the effect estimates of various to compare their effectiveness. The Markov chain Monte Carlo method was used for modeling, with four Markov chains running simultaneously, and the annealing times were set to 20,000. The modeling was completed after 50,000 simulation iterations. The deviance information criterion (DIC) was used to compare model fit and global consistency. If there is a closed loop in the network, we would use the node-splitting method to analyze local consistency. In addition, we ranked these interventions based on the surface under the cumulative ranking curve (SUCRA) values, and a league table was generated to compare the differences in effectiveness between various interventions. A funnel plot was used to intuitively reflect the heterogeneity among the studies. Analyses were completed in Stata 15.0 (Stata Corporation, College Station, TX) and R 4.2.0 (R Development Core Team, Vienna, http://www.R-project.org). A p-value of <0.05 indicates a statistically significant difference.

A total of 4,697 relevant articles were retrieved from the databases, including 162 from English databases and 4,535 from Chinese databases. A total of 558 duplicate articles were marked in EndNote and eliminated. After reading the titles and abstracts of the remaining 4,139 articles, we finally determined to download the full texts of 66 articles. According to a full-text review, 50 RCTs (18–67) were eligible and included in this systematic review (as shown in Figure 1; Supplementary Table S2).

This study included 50 articles in total, involving 3,999 patients with SHS after stroke, and 2 of them were multi-center RCTs (41, 42). The included articles were published between 2002 and 2022, covering 19 acupuncture treatment methods. There were 11 types of acupuncture combined with other treatments. None of the original studies we included had conflicts of interest. The basic information of the included studies can be found in Table 1.

This systematic review covers 19 acupuncture treatment methods (routine acupuncture).

Point-toward-point needle insertion, warming needles, electroacupuncture, fire needling, Jin’s three-needle technique, floating needle, eye acupuncture, silver needle, trocar needle, spoon-like needle, thumbtack needle, abdominal acupuncture, scalp acupuncture, contralateral blood-letting and contralateral needling, catgut embedding, wrist-ankle acupuncture, balancing acupuncture, eightfold method of the sacred tortoise, four special acupuncture techniques (18, 44, 55, 56, 64), and four special needle insertion methods (36, 42, 46, 47, 60, 63, 66). Affected limb movement during acupuncture was reported in three studies (24, 44, 67). The detailed intervention methods are provided in Supplementary Table S3.

In our 50 included studies, 43 were assessed as having “some concerns” for overall bias and 7 had a high risk of bias. The high risk of bias in the seven studies was because the intervention implementers knew the intervention. Since there was no detailed information on allocation concealment in these 50 RCTs, 43 studies were classified as “some concerns.” Due to the experimental environment, five studies deviated from the predetermined intervention measures and were assessed as having “some concerns.” During the implementation of the intervention, because the evaluators knew the intervention, seven studies were classified as high risk and four as some concerns. The detailed bias risk assessment results are presented in Figure 2.

We also analyzed the secondary outcome measures of Barthel index and SHSS. Compared to conventional treatment, various interventions did not significantly improve Barthel index or SHSS. The analysis results of secondary outcome measures are presented in Appendix 1.

Our network meta-analysis found that the top five interventions for improving pain according to the VAS scale are floating needle, rehabilitation + catgut embedding, other multi-needle acupuncture combinations, special needle operation and technique, and acupuncture on non-affected limb, which have a significant effect on pain relief. The best five interventions for the Fugl–Meyer score are eye needle, other multi-needle acupuncture combinations, western medicine + acupuncture + traditional Chinese medicine, warming needles, and special needle operation and technique, which could significantly improve upper limb motor function.

We found that rehabilitation + catgut embedding was effective in pain relief. Previous studies have also reviewed this intervention, and the analgesic principle may be a continuous stimulation by absorbable surgical sutures at acupoints, leading to a sterile inflammatory response, and ultimately promoting tissue regeneration (66, 68).

Although previous studies have also attempted to analyze the effects of acupuncture treatment on post-stroke SHS, some of them only explored the relative effectiveness of a single type of acupuncture or a certain acupuncture combination with rehabilitation and the control group (69–72). Post-stroke SHS is caused by multiple etiologies. Single therapy may not achieve the ideal pain relief. The clinical practice tends to combine multiple intervention methods, which can help explore more reasonable and effective treatment options. The number of types of acupuncture and studies included in existing network meta-analyses is limited (73), but there are many types of acupuncture. Furthermore, the above studies did not consider the existing types of acupuncture or combined acupuncture treatment methods to comprehensively compare various acupuncture interventions.

The mainstream treatment for post-stroke SHS is still rehabilitation therapy. As an indispensable intervention for treating this disease, it focuses on function recovery. Meanwhile, the obstacles affecting function recovery should be tackled to reduce the chances of patients refusing rehabilitation. Aerobic exercise of the upper limbs is found to reduce pain during the day and during exercise (74, 75). Some studies have reported that the combination of two or more therapies may be more effective than rehabilitation alone. Acupuncture combined with rehabilitation training can significantly help in reducing pain scores, promoting limb motor function recovery, and improving overall effectiveness (69–71, 76).

Drug therapy includes oral calcium channel modulators, anti-inflammatory, analgesic, antidepressants, or invasive injections of corticosteroids, bisphosphonates, and botulinum toxin (77, 78). Although drug treatment is convenient and quick, it has limited effects, and long-term use can produce adverse reactions such as hepatorenal and cardiac toxicity, excessive sedation, and drug resistance (79). Some drugs also have the possibility of recurrence after withdrawal and cannot fundamentally control and treat this disease (78, 80–83). In contrast, acupuncture has the advantages of high safety and no toxic side effects and is more easily accepted by patients with heart, liver, and kidney damage.

Neural stimulation, nerve block, and ganglion radiofrequency ablation can avoid the adverse systemic effects induced by drug treatment. Nonetheless, invasive surgery has certain adverse reactions and risks, and the implanted stimulation device may also have problems such as wire disconnection, requiring regular battery replacement (78, 84). A study has even found that the effects of standard high-frequency neurostimulation may be worse than acupuncture or other types of electrical stimulation, and acupuncture can become a promising treatment method for patients with drug-resistant neuropathic pain (84). The nerve block needs to be carried out under ultrasound guidance to improve safety. After a single treatment session, 61.8% of patients still have pain recurrence within 3 months. For chronic pain, it cannot achieve fundamental relief (78, 85–87). Ganglion radiofrequency ablation may cause risks of sensory and motor dysfunction and nerve damage, with small clinical improvement in the VAS score of patients with chronic root pain (88). Its long-term therapeutic effects are still controversial.

The treatment of post-stroke SHS is to relieve pain in the early stage of the disease without affecting the motor function and rehabilitation process of the affected limb. Conventional analgesic drugs, such as opioids and anti-inflammatory drugs, have a variety of side effects such as reduced motor function, induced hyperalgesia, and lethargy (89). At present, acupuncture is widely used for stroke rehabilitation and various pain-related diseases in China, the United States, Europe, and North America. Although acupuncture is an invasive treatment, the incidence of acupuncture-induced adverse reactions is significantly lower than that of many drugs or other recognized medical interventions according to the consensus report for the NIH Conference (90). Acupuncture can reduce the excitability of the pain pathway by reducing excitatory neurotransmitters and promoting inhibitory neurotransmitters (91, 92). Electroacupuncture has been shown to induce peripheral opioid receptors in rats, producing long-term analgesic effects for up to 144 h, improving motor function, and normalizing neurotransmitter metabolism (93). A preclinical study has also shown that acupuncture’s antihyperalgesic effects are associated with 10 peripheral receptors and 8 neurotransmitters (94).

This systematic review is the first to focus on the advantages of different acupuncture treatment plans to broaden the selection of feasible treatment options in clinical practice. The advantages of different acupuncture treatments are different. There are currently no studies comparing the effectiveness of various types of acupuncture for post-stroke SHS directly or indirectly. Our network meta-analysis includes a large number of studies, integrating various acupuncture intervention methods and comparing the improvement of different acupuncture therapies for post-stroke SHS based on VAS, Fugl–Meyer, Barthel, and SHSS scales. The aim is to provide more comprehensive clinical practice references for acupuncture treatment for post-stroke SHS.

However, there are also several limitations. Since acupuncture is a traditional Chinese treatment method, its promotion in the world is still limited to some extent. The included studies are mainly from China and lack ethnic diversity. Therefore, research and exploration outside China are required to enhance clinical promotion and applicability. The multi-center studies included in our research are relatively few. To acquire more convincing evidence, future studies are desired to verify these results.