This systematic review aimed to explore the effects of acupuncture on parameters related to PSD. RCTs on acupuncture interventions for patients with acute or recovering PSD without language restrictions were included. To maintain the rigor of this study, pseudo-randomization, quasi-randomization, experimental studies, reviews, case reports, letters, studies with incomplete data, and publications with duplicate data (data extracted from the most recent literature) were excluded. Studies that did not report any of the parameters of interest were excluded.

All qualifying studies included a control group as a minimum requirement. Furthermore, patients with PSD who underwent acupuncture in the included studies had positive outcomes compared to controls.

PSD is the most frequent condition in which stroke damages the swallowing network, leading to dysphagia (5). Patients with acute (within 7 days) and recovery (within 6 months) PSD were included. Patients with functional dysphagia and dysphagia due to esophageal cancer, laryngeal cancer, or inflammation were excluded. No restrictions based on the type of stroke (hemorrhage or ischemia), location of the lesion (brainstem or non-brainstem), side of the lesion (unilateral or bilateral), age, sex, or geographic region imposed.

Acupuncture therapies in the intervention group included manual acupuncture, electroacupuncture, and conventional physical therapy techniques, whereas those in the control group included rehabilitation, sham acupuncture, placebo acupuncture, or drugs. Studies using moxibustion, acupoint burrowing, auricular acupuncture, and non-invasive interventions (Chinese herbs and tuina) were excluded. No clinical trials comparing different acupuncture methods or acupoint-taking protocols were considered.

Study inclusion in the analysis depended on providing at least one of the following standardized and validated dysphagia scales: the Penetration Aspiration Scale, Functional Oral Intake Scale, Water Swallow Test, Standardized Swallowing Assessment, Videofluoroscopic Swallowing Study, Modified Mann Assessment of Swallowing Ability, Dysphagia Outcome Severity Scale and Swallowing-Quality of Life. Studies that reported only physical or chemical examinations were excluded. The analysis time point was set at the end of all scheduled treatment sessions.

All articles describing acupuncture and PSD were obtained from the Cochrane Library, PubMed, Embase, Web of Science, China National Knowledge Infrastructure, Wanfang Database, Chinese Biomedical Literature, and Chongqing VIP Database between December 2009 and December 2023 in English and Chinese.

The PubMed and Web of Science search strategies according to population, intervention, control, and outcomes (PICO) format are presented in Table 1. For the other electronic databases, the search approach was adjusted as required.

After eliminating duplicates, two independent reviewers (MW and WS) screened the literature against the inclusion eligibility criteria (checking the title, abstract, and full text of the papers to identify eligible trials) and cross-checked the screening results. In cases of disagreement, a third reviewer (XW) made the final decision on whether the study should be included.

Data for further analysis were extracted from the included studies: demographic information (title, authors, and year), sample characteristics (age, sex, and patient volume), interventions, and acupuncture parameters (needle type, frequency of treatment, duration of session, duration of stimulation, and acupuncture points).

To evaluate the data on an intention-to-treat basis, missing data were contacted through the corresponding author’s e-mail addresses. If missing data could not be obtained, the currently available literature was excluded.

Two independent reviewers (MW and WS) autonomously evaluated potential biases using “yes,” “no,” or “unclear,” including selection, performance, detection, attrition, reporting, and other biases.

In total, 3,205 relevant studies were identified. Of these, 925 were duplicates, and 1729 were excluded after screening the titles and abstracts. A total of 512 articles were excluded after full-text evaluation, and 39 articles (11–49) were included for further analysis. The study screening process is visualized using a flowchart (Figure 1).

Detailed characteristics of the included studies are presented in Table 2. Only 3 of the 39 studies were published in English (13, 31, 41), while the rest were published in Chinese. All of these studies were conducted in China over the last 15 years. A total of 3,507 patients with PSD were included, with sample sizes ranging from 21–125 patients. In the treatment group, 27 studies (12–21, 23–25, 27, 28, 30, 31, 33, 35–37, 39, 40, 42, 44, 45, 49) used acupuncture combined with rehabilitation, and seven studies used acupuncture alone (26, 34, 38, 41, 46–48). For the control group, 32 studies used rehabilitation (12–21, 23–25, 27, 28, 30, 31, 33–37, 39–42, 44–49), whereas the remaining studies used sham acupuncture, medication, repetitive transcranial magnetic stimulation, and general nursing care. Regarding numerous outcome indicators, 13 and 27 studies reported the standardized swallowing assessment (SSA) (15–17, 20, 22, 23, 26–28, 31, 40, 46, 47) and water swallow test (WST) (11, 12, 16, 18–21, 23–25, 28–39, 41–43, 45, 47, 49) as valid outcomes, respectively. Nine studies (15, 19, 20, 24, 31, 36, 40, 42, 46) reported swallowing-related quality of life (SWAL-QOL) as an outcome indicator in patients with PSD.

All 39 included RCTs reported positive outcomes, supporting the therapeutic effectiveness of acupuncture for patients with PSD.

All studies reported randomized allocation methods and no pseudo- or quasi-RCTs. Eight studies reported allocation concealment (13, 15, 31, 35, 39–42), except for one, which had a high risk of bias (46), whereas the remaining studies did not report allocation concealment, giving an unclear risk of bias. The acupuncturists were almost impossible to blind, and the remaining literature had a high risk of bias, except for two English-language papers, which were unclear about the risk of bias (13, 31). Seven studies reported a detection bias (13, 14, 31, 40–42, 46), and the remaining studies reported an unclear risk of bias. In the incomplete outcome and selective reporting section, the remaining studies gave a low risk of bias, except for one that gave unclear selective reporting (41). Regarding other biases, 12 studies reported an unclear risk of bias (11, 12, 24, 27, 29, 30, 35, 38, 43–45, 47), while the remaining studies reported a low risk of bias (Figure 2).

Eleven needle types were analyzed in 39 studies with a frequency of 27 uses. However, 21 studies did not report the needle type, which may have affected the RCT replication. Notably, one study reported needle-type unavailability, which was 0.20–0.25 mm × 25–75 mm. The most used needle type in these studies was 0.25 mm × 40 mm with a frequency of eight uses, accounting for 30% of the total frequency. Six needle types were used at a frequency of 1 and 6 for almost more than half of the needle types (Table 3).

Five retention times were observed in 39 studies, with the longest being 30 min and the shortest being 5 min. The 30 min retention time was used as frequently as 24 times or 71%. In clinical practice, the choice of 30 min of acupuncture intervention for PSD may have the intended effect. Although only 5 min occurred once, the study yielded positive results. Notably, retention time was not available in six studies (Table 4).

The results of the frequency analysis of treatments in the 39 studies showed four types, which were 5, 6, 7, and 3 days per week. The highest frequency was 5 days per week, and the lowest frequency was 3 days per week. Except for 3 days per week, the frequency share of the other treatments was approximately 0.3. However, the frequency of treatment was not available in 3 out of 39 studies (Table 5).

A total of 39 studies included five acupuncture interventions for PSD, all of which reported the total treatment time. Four weeks of intervention were used as frequently as 22 times, which was more than half of all frequencies. Six and 8 weeks total intervention time frequency of 6 times is nearly 20%. This may indicate that the PSD recovered better and for a shorter duration with acupuncture interventions (Table 6).

Forty PSD-eligible intervention prescriptions, 71 acupoints, and 322 occurrences were identified in 39 studies. The top 15 acupoints included Fengchi (GB20), Lianquan (RN23), Jinjin (EX-HN12), Yuye (EX-HN13), Tunyan (new acupoint), Yifeng (SJ17), Yiming (EX-HN14), Wangu (GB12), Gongxue (new acupoint), Zhiqiang (new acupoint), Fengfu (DU16), Anterior oblique line of vertex-tempora (MS6), Baihui (DU20), Wai jinjin (Wai EX-HN12), and Wai yuye (Wai EX-HN13) (Table 7 and Figure 3A).

Of the 40 acupoint prescriptions, 46 were used by 14 conventional meridians. Specifically, there were 12 conventional meridians, 1 Conception Vessel (RN), 1 Governor Vessel (DU), and nonconventional meridians. Notably, nonconventional meridian acupoints accounted for the highest number of acupoints (19) and were used as frequently as 80 times. Although not conventional, the high frequency of new acupoints indicates that clinical experience should not be ignored. In the meridian analysis, LU, LR, and HT meridians were all used twice, and all were one acupoint. Although GB had a frequency of 45 usages, only two acupoints (GB20 and GB12) may be potentially valuable for acupuncture interventions in PSD. Moreover, the SJ, ST, LI, BL, PC, SP, KI, LU, LR, and HT meridians were used <5% of the total frequency, and these meridians may be less appropriate for acupuncture intervention in PSD (Table 8).

Of the 71 acupoints, 44 were specific, and 10 contained more than one attribute. Specifically, PC6, LU7, SJ5, and SP4 were Luo-connecting points and eight confluent points. RN12 possesses the attributes of eight influential points: the front Mu and crossing points. The frequency, percentage, and highest number of acupoints used as crossing points in the specific-point analysis were 123, 38%, and 15, respectively. Acupuncturists should combine the crossing points in clinical practice to achieve better therapeutic outcomes. The Yuan primary point and Luo connecting point also had a moderate frequency in this study: 13 and 9 times, respectively. Notably, the frequencies of the eight influential points, lower He-Sea points, front Mu points, and back Shu points were <0.01, indicating that clinical acupuncturists seldom considered these specific acupoints for therapeutic PSD (Table 9).

In the acupoints distribution analysis, the head, face, and neck were the most selected areas, with frequencies, percentages, and acupoints of 279, 87%, and 47, respectively. This area is consistent with the clinical symptoms of patients with PSD, indicating that acupuncturists tend to consistently select the acupoint locations. The chest, abdomen, and back had very few acupoints, nine in total, and these locations were not considered, possibly because of disease specificity (Table 10).

Association rule analysis of the Apriori algorithm was performed using R software (version 4.3.0) to obtain 72 association rules. A grouping matrix plot was used to visualize the 72 association rules (Figure 3B). Darker purple circles indicate higher degrees of lift, whereas larger circles indicate higher support.

Based on the minimum support threshold of 20%, the minimum confidence threshold of 90%, and an uplift factor of>1, 20 pairs of acupoint combinations with the highest support thresholds in the PSD prescriptions were summarized. The first 12 pairs included {EX-HN12} ≧ {EX-HN13}, {EX-HN13} ≧ {EX-HN12}, {EX-HN12} ≧ {GB20}, {EX-HN13} ≧ {GB20}, {EX-HN12, EX-HN13} ≧ {GB20}, {EX-HN12, GB20} ≧ {EX- HN13}, {EX-HN13, GB20} ≧ {EX-HN12}, {EX-HN12, RN23} ≧ {EX-HN13}, {EX-HN13, RN23} ≧ {EX-HN12}, {GB12} ≧ {GB20}, {Zhiqiang} ≧ {Tunyan}, and {SJ17} ≧ {GB20}. The top 20 pairs of acupoint combinations had confidence, support, and lift values (Table 11), and more than half of the acupoint combinations had confidence values of 100%. Notably, all 20 pairs of acupoint combinations had lift values >1, with nine acupoint combinations >2.

A complex network analysis was performed using Cytoscape to derive 38 acupoints (nodes) and 210 edges, which were classified into four layers by degree scores: the darker the color and the higher the frequency. Hub acupoints were obtained using the maximal clique centrality (MCC), maximum neighborhood component (MNC), and molecular complex detection tool (MCODE). Specifically, the top 10 acupoints and 45 edges were obtained using MCC in the Cytoscape tool, with the acupoints in descending order as follows: GB20, RN23, EX-HN12, EX-HN13, Tunyan, EX-HN14, Zhiqiang, Gongxue, MS6, and DU20. The top 10 acupoints and 42 edges were obtained using MNCs, including GB20, RN23, EX-HN12, EX-HN13, Tunyan, SJ17, DU16, EX-HN14, MS6, and DU20. Additionally, nine acupoints and 32 edges were obtained using MCODE, including DU20, Fayin, Tunyan, EX-HN14, Zhiqiang, Wai EX-HN12, Gongxue, Wai EX-HN13, and Tiyan. Notably, these hub acupoints were consistent with the results of association rule analysis (Figure 4).

The 20 most frequently used acupoints were clustered and correlated into six clusters. Cluster 1 includes EX-HN14, Wai EX-HN12, Wai EX-HN13, Gongxue, Tunyan, and Zhiqiang. Cluster 2 comprised EX-HN12 and EX-HN13. Cluster 3 included strains SJ17 and GB12. Cluster 4 includes RN23, GB20, MS6, and MS7. Cluster 5 included DU20, Pang RN23, and RN22. Cluster 6 included DU16, LI4, and PC6 (Figure 5).