Effects of Acupuncture-Related Therapies in the Rehabilitation of Patients with Post-Stroke Aphasia—A Network Meta-Analysis of Randomized Controlled Trials

Objective: The purpose of this study was to evaluate the rehabilitation effects of four common interventions (BA: body acupuncture, SA: scalp acupuncture, TA: tongue acupuncture, SLT: speech and language training) used singly or in combination with language function in patients with post-stroke aphasia (PSA). Design: We systematically searched PubMed, EMBASE, Cochrane Library, Ovid, Web of Science, CNKI, VIP, and Wanfang from inception to 4 April 2022. Only randomized controlled trials that met the eligibility criteria were included. The risk of bias of studies included was assessed using the RoB-2 tool. The effects of different interventions for PSA patients were analyzed and ranked according to the surface under the cumulative ranking (SUCRA) analysis. Results: A total of 69 RCTs were included, including 5097 total participants. According to the results of the SUCRA curves, TA ranked highest in improving overall efficacy (SUCRA = 86%) and oral expression score (SUCRA = 86%). BA + TA ranked highest in increasing the comprehension score (SUCRA = 74.9%). BA + SA ranked highest in improving aphasia patients’ repetition (SUCRA = 89.2%) and denomination scores (SUCRA = 93%). Conclusions: Results of our network meta-analysis and SUCRA ranking showed that tongue acupuncture, body acupuncture + tongue acupuncture, and body acupuncture + scalp acupuncture seem to offer better advantages than other interventions for improving the language function in PSA patients. Moreover, it is noteworthy that our results are limited to the Chinese population, since all eligible studies are from China. Future well-designed studies with larger sample sizes and more ethnic groups are required to further verify these findings.


Introduction
Today, stroke is the third most common cause of death and the leading cause of permanent adult disability in the world. It is estimated that there are about 13 million new stroke cases worldwide each year, with over 5.5 million people suffering from strokes in China [1]. Approximately 21-38% of stroke survivors suffer post-stroke aphasia (PSA) [2]. PSA, a common language disorder after stroke, is characterized by impairment of partial or all language functions, including oral expression, comprehension, and reading and writing [3]. It is mainly caused by strokes in the left hemisphere and severe trauma or neurodegenerative illness affecting the brain's language network pathways, primarily Broca and Wernicke areas [4]. Speech is a crucial instrument for thinking and communicating in human social activities. Aphasia not only impairs the ability of patients to participate in social activities and career development and reduces their quality of life, but also causes

Assessment of the Risk of Bias
Two reviewers (X. Liu and S. Chen) independently used the Cochrane RoB−2 tool to assess the risk of bias in each study. The assessment includes evaluation in the following seven domains: random sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and other possible biases. The studies were classified into low risk, high risk, and unclear risk. Discrepancies between the reviewers were solved through discussions with a third reviewer (Y. Feng) [36].

Statistical Analysis
In this study, odds ratios were used for dichotomous outcomes, and continuous variables in the study were reported as mean difference (MD = the difference in means between the treatment and control groups and calculated using the same scale) or standardized mean difference (SMD = mean difference in outcome between groups/standard deviation of outcome between subjects, used to combine data when trials have different scales) with estimated 95% confidence intervals (95% CI). To take into account differences between studies, we synthesized data using a random-effects model [37].
We performed a random-effects network meta-analysis within a frequentist framework using Stata/SE (version 15.0), and the results are presented with the mvmeta package [38]. A frequentist framework allowed us use p-values to make statistical comparisons [39]. Meanwhile, it was simpler in model specification and more useful for doctors to solve clinical practical problems compared with Bayesian framework [40]. Inconsistency in the network meta-analysis involved the design-by-treatment inconsistency model and the node-splitting approach [41]. The network evidence plots were drawn to summarize the geometry of the network of evidence. In the network evidence plot, each dot represents a different intervention, and the lines represent head-to-head comparisons. The size of each node and the width of the connecting lines are proportional to the number of studies. Simultaneously, the P rank score of the surface under the cumulative ranking curve (SURCA) was calculated to compare the rank of effect estimation for each treatment. SUCRA was expressed as a percentage (range, 0-100%): a larger value indicated a better intervention grade in sort results, and when the SUCRA value was close to 100%, it indicated that the treatment was the best one [42]. Efficacy of different treatments was displayed in a league table. Treatments were ranked from best to worst along the leading diagonal. Data presented are the ORs and MD (with 95% CI) in the column-defining treatment compared to the row-defining treatment for total clinical effective rate and ABC scale. OR above 1 indicated higher efficacy in improving language function, and MD above 0 favored the treatment in column in improving ABC scores. Effects were considered significant when OR and MD did not contain 1 and 0. Significant results are showed in bold. In addition, a "comparison-correction" funnel plot was used to assess the publication bias of the included studies [43]. Notably, the present results are based on the pooled data from each study included. Clinicians should interpret them cautiously and make reasonable clinical decisions according to the specific clinical situation [44].

Results of Study Identification and Selection
A total of 1803 studies were included via the original search terms, and 526 articles were excluded due to duplication. After screening for title and abstract, 462 articles that failed the inclusion criteria were deleted, and after reading the whole text, 746 studies (84 non-RCT studies, 307 studies that did not meet the diagnosis of post-stroke aphasia, 266 studies whose interventions and controls were inconsistent with the criteria, 51 studies with duplicate, missing, or incalculable data, and 38 studies without targeted outcome indicators) were again removed. Finally, 69 studies were included into the risk of bias assessment and NMA. The process is presented in Figure 1.
non-RCT studies, 307 studies that did not meet the diagnosis of post-stroke aphasia, 266 studies whose interventions and controls were inconsistent with the criteria, 51 studies with duplicate, missing, or incalculable data, and 38 studies without targeted outcome indicators) were again removed. Finally, 69 studies were included into the risk of bias assessment and NMA. The process is presented in Figure 1.

Quality Assessment of the Included Studies
We assessed the risk of bias for each study using the Cochrane RoB−2 tool (Supplementary Table S3). In 69 included trials, 40 trials were at low risk of selection bias related to sequence generation because they used appropriate methods for random sequence generation, such as computer randomization or a random numbers table. Of the included trials, 26 that only mentioned random were evaluated as unclear risk. The remaining three trials used the order of consultation to generate random sequence and thus were rated as high risk of bias. We assessed all trials as unclear bias due to insufficiently reported allocation concealment methods. For blinding of subjects and outcome assessment, we evaluated almost all articles as high or unclear risk of bias. Only one article, Wang 2011, was accessed as at low risk of bias in blinding outcomes. All included trials carried a high risk for the blinding of subjects and outcome assessment. Regarding the risk of incomplete outcome data, 69 articles included in the current study showed an overall low risk of bias. Except for He A 2014, which was at high risk of bias, the remaining 68 studies were at low risk of selective reporting. Only one paper, Wei 2021, which reported no conflicts of interest, was rated as low risk of other bias. The remaining studies were considered to have an unclear risk of bias. Details of the risk of bias are presented in Supplementary Figure S1  [45,48,51,[57][58][59]61,65,71,73,75,81,82,85,95,100,103,107,109,[111][112][113]. A total of 21 studies with 1497 participants reported the denomination score, including 13 interventions: BA, BA + SA, BA + SA + TA, BA + TA, SA, SLT, SLT + BA, SLT + BA + SA, SLT + BA + SA + TA, SLT + BA + TA, SLT + SA, SLT + SA + TA, and SLT + TA [45,48,[57][58][59]61,65,71,73,75,81,82,85,95,100,103,107,109,[111][112][113]. The details of the NMA figures were shown in Figures 2-5 and 6A.
Our NMA results showed that all the intervention measures were statistically significant, indicating that acupuncture has a significant effect on improving language function. The details are shown in Table 2. The SUCRA rankings of all treatments for the clinical effective rate are shown in Figure 2B. According to the results of SUCRA analyses, TA (SUCRA, 86%) was likely to be the most effective intervention to improve post-stroke aphasia, followed by BA + SA (SUCRA, 75    Our NMA results showed that BA (21.02; 95% CI, 0.12 to 42.32) and BA + SLT (15.09; 95% CI, 0.21 to 29.96) were better than SLT. The other different treatments showed no significant difference. The details are shown in Table 3. According to the results of SU-CRA analyses, BA + TA (SUCRA, 74.9%) had the highest probability of being the best treatment for increasing comprehension score, followed by BA + SA (SUCRA, 72.     Efficacy of interventions are displayed in a league table. Treatments are ranked from best to worst along the leading diagonal. Data presented are the MD with 95% CI in the column-defining treatment compared to the row-defining treatment. For ABC sores, MD above 0 favors the treatment in column in improving ABC scores. Significant results are shown in bold. A: body acupuncture, B: body acupuncture + scalp acupuncture, C: body acupuncture + scalp acupuncture + tongue acupuncture, D: body acupuncture + tongue acupuncture, E: scalp acupuncture, F: speech and language training, G: speech and language training + body acupuncture, H: speech and language training + body acupuncture + scalp acupuncture + tongue acupuncture, I: speech and language training + body acupuncture + tongue acupuncture, J: speech and language training + scalp acupuncture, K: speech and language training + scalp acupuncture + body acupuncture, L: speech and language training + scalp acupuncture + tongue acupuncture, M: speech and language training + tongue acupuncture, N: tongue acupuncture.   Our NMA results showed that BA + SA ( Table 5

Results of the Denomination Score
A total of 21 studies were aggregated to compare the effect of each intervention on the denomination score [45,48,[57][58][59]61,65,71,73,75,81,82,85,95,100,103,107,109,[111][112][113]. All p-values of the two methods for accessing inconsistency (Supplementary Figure S6) Table 6. According to the results of SUCRA analyses, BA + SA (SUCRA, 93%) was likely to be the most effective intervention to improve post-stroke aphasia, followed by BA + TA (SUCRA, 88.  Figure 6B.    Efficacy of interventions are displayed in a league table. Treatments are ranked from best to worst along the leading diagonal. Data presented are the MD with 95% CI in the column-defining treatment compared to the row-defining treatment. For ABC sores, MD above 0 favored the treatment in column in improving ABC scores. Significant results are shown in bold. A: body acupuncture, B: body acupuncture + scalp acupuncture, C: body acupuncture + scalp acupuncture + tongue acupuncture, D: body acupuncture + tongue acupuncture, E: scalp acupuncture, F: speech and language training, G: speech and language training + body acupuncture, H: speech and language training + body acupuncture + scalp acupuncture, I: speech and language training + body acupuncture + scalp acupuncture + tongue acupuncture, J: speech and language training + body acupuncture + tongue acupuncture, K: speech and language training + scalp acupuncture, L: speech and language training + scalp acupuncture + tongue acupuncture, M: speech and language training + tongue acupuncture.

Presence of Adverse Effects
None of the included studies reported adverse effects, so the network meta-analysis could not be further performed.

Publication Bias and Consistency Assessment
We constructed a comparison-correction funnel plot of the primary outcome of the clinical effective rate by Stata/SE 15.0 for evaluation. As shown in Figure 7, the funnel plot shows that the scattered points are symmetrically distributed within the funnel borders, suggesting limited publication bias.

Presence of Adverse Effects
None of the included studies reported adverse effects, so the network meta-analysis could not be further performed.

Publication Bias and Consistency Assessment
We constructed a comparison-correction funnel plot of the primary outcome of the clinical effective rate by Stata/SE 15.0 for evaluation. As shown in Figure 7, the funnel plot shows that the scattered points are symmetrically distributed within the funnel borders, suggesting limited publication bias.

This Work
With the continuous progress of diagnostic techniques and medical treatments, the

This Work
With the continuous progress of diagnostic techniques and medical treatments, the condition of most stroke patients has been effectively controlled in the acute and subacute phases but still leaves a series of sequelae. PSA is one of the most severe and persistent symptoms in cerebrovascular sequelae [114][115][116][117]. Losing language functions in PSA patients leads to a decline in quality of life and a massive psychological burden on patients and their families [118][119][120]. Over the past decades, many research teams have devoted substantial effort to studying PSA's potential mechanisms and therapeutic approaches. It is generally accepted that PSA's pathogenesis was related to functional impairment, hypoperfusion, and hypometabolism of cerebrocortical areas after stroke [121][122][123][124]. The clinical strategy for PSA focuses on improving specific language function deficits of PSA patients by using various approaches to enhance functional communication. However, these treatments show no significant differences in efficacy, and there is no universally recognized standard treatment scheme for PSA patients [125]. Their therapeutic effects are also affected by many factors, such as exposure, the level of medical service, dose, course of treatment, patient compliance, sample size, and so on. Therefore, given the unsatisfactory efficacy of current treatments, it is crucial to find alternative or complementary therapies. Compared with many rehabilitation therapies, acupuncture has the advantages of safety, simple operation, non-toxic side effects, and relatively low cost. It has more than 3500 years of medical practice history and has been widely used in the rehabilitation of cerebrovascular diseases in Chinese hospitals. Many scholars have verified the effectiveness of acupuncture in treating aphasia by conducting clinical trials, meta-analyses, and other brain functional studies. Nevertheless, few studies have compared the efficacy of different intervention measures in treating aphasia and given the best treatment protocol. Therefore, choosing the best combination of acupuncture interventions has become a hot issue. Based on this, we searched the relevant literature and performed a network meta-analysis.
In this study, we compared the effects of different acupuncture interventions on improving language function in people with post-stroke aphasia. Our analysis showed that TA (SUCRA = 86%) was the best treatment for improving global efficiency (8.03; 95% CI, 3.62 to 17.81). At the same time, TA (SUCRA = 86%) also showed the highest probability of being the best treatment in oral expression according to the results of SUCRA. However, compared to SLT, TA has no significant statistical difference in improving expression ability (Table 4). Our result was in disagreement with the previous results by Tang et al. [126]; hence, further studies are still needed to identify the possible reasons for this discrepancy. Notably, although TA + BA ranked second with 83.2%, it showed statistically significant differences relative to SLT (18.08; 95% CI, 1.46 to 34.69). Therefore, these results indicated that BA + TA could also be served as an effective alternative treatment for PSA. BA + TA (SUCRA = 74.9%) showed the best outcome in increasing the comprehension score, but again, this was not statistically significant compared to SLT (Table 3). Even if BA (SUCRA = 67.1%, Rank 3) was superior to SLT with a significant difference (21.22; 95% CI, 0.12 to 42.32), we still recommend BA + TA as a treatment modality to improve speech comprehension in patients with PSA, because combined therapy was more consistent with the actual clinical situation, which might provide a potential therapeutic benefit for PSA patients. This was also confirmed by the review of Sun et al., which indicated that combination acupuncture therapies such as body acupuncture plus tongue acupuncture were more effective than body acupuncture in improving language ability [127]. BA + SA was the best treatment to improve aphasia patients' repetition (25.55; 95% CI, 3.59 to 47.25) and denomination functions (29.47; 95% CI, 4.32 to 54.63). Although the curative effects of various measures discussed above are different under the evaluation of multiple indicators, TA, BA + TA, and BA + SA always ranked at the top of the SUCRA values. According to the above analysis, we inferred that TA, BA + TA, and BA + SA could be employed as complementary or alternative therapies to promote the functional recovery of PSA pa-tients. Moreover, clinicians can make reasonable choices according to the characteristics of functional impairment in patients with PSA.
In the current literature, there are only a few reports on the mechanisms of tongue for treating neurological diseases such as visual impairment, cerebral palsy, and poststroke dysphagia. Previous studies had shown that tongue acupuncture could improve the visual status of children with visual impairment and had a potential therapeutic effect on neural plasticity [128]. Another study showed that tongue acupuncture could enhance glucose metabolism in the frontal lobe, parietal lobe, temporal lobe, occipital cortex, and cerebellum of children with cerebral palsy in a short time and could be used as a candidate non-pharmacological intervention to improve functional recovery [129]. Although we were not able to find clinical studies on tongue treatment for PSA, it is apparent from the present studies that tongue acupuncture has substantial therapeutic potential as adjuvant therapy for post-stroke rehabilitation. An excellent recent review by Bono et al. describing the sensory motor the of tongue attracted our attention. They not only illustrated the importance of the extrinsic lingual muscles in language but also showed that different tongue regions could be somatotopically arranged in the cerebral cortex by electrocorticography studies, which had a high potential correlation with speech production and could enhance the control of fine movement required by language [130]. Thus, we speculate that the mechanism of tongue acupuncture improving aphasia symptoms may be related to the following reasons. First, tongue surfaces are innervated by numerous mechanoreceptors and nerves, which have high sensitivity for responding to multiple different stimuli. Second, tongue acupuncture may promote the regeneration of neurons and synapses in peri-infarcted regions, attenuate tissue damage, and finally improve overall neurological function. Lastly, long-term tongue acupuncture treatment may significantly improve the coordination and flexibility of the tongue muscle, reactivate primary language pathways, and enhance signal transduction between the tongue and the cerebral cortex to improve language abilities in PSA patients.
Body acupuncture often used with other alternative therapies is one of the most critical interventions in acupuncture therapies. Among the three treatment regimens we recommend, all the others included body acupuncture except for one using tongue acupuncture alone. The therapeutic benefits of acupuncture for PSA patients may be related to the activation of multiple language-related brain regions such as the frontal lobe, temporal lobe, and parietal lobe during treatment [131]. The study by Xie et al. showed that acupuncture could enhance the interregional interaction between the anterior cingulate cortex (ACC) and posterior cingulate cortex (PCC) to improve memory and cognitive ability [132]. In addition, stroke patients showed an increased level of mean flow velocity (MFV) [133], and the hemorheological parameters such as the index of erythrocyte aggregation and blood sedimentation were decreased after acupuncture treatment [134]. This may be an essential physiological mechanism by which acupuncture can treat cerebral ischemia and promote the recovery of neural function. There are few clinical trials and animal studies on treating aphasia with body acupuncture combined with tongue acupuncture. The combination of two interventions may play a synergistic role in improving language understanding and cognition, thus accelerating the rehabilitation of aphasia patients. The frequency, intensity, and duration of two interventions combined are further explored to verify this assumption.
Scalp acupuncture is a kind of acupuncture therapy that stimulates different brain functional areas to active neural issues and improves reflexivity for achieving the goals of treating diseases [135]. Some clinical studies have shown that scalp acupuncture has certain benefits in enhancing the language function and behavior of autistic children [136][137][138]. However, its related mechanism still needs to be further studied and explained. In recent years, RS-fMRI has been used widely in studying cerebral nerve activity and rehabilitation mechanisms of stroke patients treated with acupuncture. Liu et al. [139] performed RS-fMRI scanning on 30 patients with acute ischemic stroke who received scalp acupuncture treatment. The result showed that compared with the control group, the activity of neurons in the left angular gyrus, middle temporal gyrus, superior temporal gyrus, and fusiform gyrus were significantly increased, which were indispensable components of the language function network involved in language input and cognitive processing. Li et al., in their recent review, described in detail that scalp acupuncture and body acupuncture caused activity changes in the common brain regions during the treatment of aphasia patients, such as the right fusiform gyrus, the left superior frontal gyrus, and the left inferior temporal gyrus [140], which may be an important reason for the combination of the two acupuncture therapies to enhance the curative effect. Although few studies have directly revealed the mechanism of scalp acupuncture and body acupuncture in improving the repetition and naming of patients with post-stroke aphasia, these studies are of great significance for us to understand the mechanism of scalp acupuncture combined with body acupuncture in the treatment of post-stroke aphasia.
In summary, our study has made some efforts in rehabilitating patients with PSA. We found that tongue acupuncture, body acupuncture combined with tongue acupuncture, and body acupuncture combined with scalp acupuncture have a good therapeutic effect on aphasia after stroke, which can provide a reference for clinical practice. In addition, compared with expensive drugs and rehabilitation treatment, clinicians can try to take acupuncture as a good alternative or complementary therapy to promote aphasia rehabilitation after stroke.

Strengths and Limitations
First, our study included 69 studies and 5097 patients, which is a very large sample size. At the same time, we involved more than 10 kinds of treatment measures. We evaluated the impact of intervention measures from five aspects to provide more comprehensive evidence-based recommendations. Secondly, the research on acupuncture in treating aphasia mainly focuses on clinical trials and meta-analyses. In addition, there are a small number of fMRI imaging studies. Few studies have given recommendations for different acupuncture-related therapies in the treatment of PSA. We carried out the first network meta-analysis of acupuncture therapies for PSA and provided a preliminary experience for further detailed studies of this domain. However, this study also has some limitations, including the following. (I) Many studies did not specifically report the random method, assignment concealment, and reliability of the results. (II) Different treatment times, frequency, and protocols of acupuncture-related therapies included in the study may increase clinical heterogeneity. (III) Many studies only show the assessment of the four functional areas of aphasia, but not the evaluation of writing, reading, and other abilities. We could only analyze and provide general results based on these studies. In the subsequent study, we need to further introduce a variety of evaluation indicators to more comprehensively and objectively evaluate the effects of acupuncture therapies on PSA. (IV) Most of the participants included in this study were Chinese nationals, and the evaluation scale used was the ABC scale. Therefore, the research results were limited in universality and should be interpreted carefully. More well-designed studies with larger sample sizes and more ethnicities are required to minimize ethnic bias. (V) There is no minimum clinically important difference (MCID) available currently for ABC scores in patients with PSA. Future scholars should develop the MCIDs for such patients to better interpret the results of this paper in the clinical context.

Conclusions
Results of our network meta-analysis and SUCRA ranking showed that tongue acupuncture, body acupuncture + tongue acupuncture, and body acupuncture + scalp acupuncture seem to offer better advantages than other interventions for improving the language function in PSA patients. Moreover, it is noteworthy that our results are limited to the Chinese population, since all eligible studies are from China. Future well-designed studies with larger sample sizes and more ethnicities are required to further verify these findings.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/brainsci12101282/s1, Figure S1: The risk of bias evaluation; Figure S2: Consistency assessment of network meta-analysis of total effective rate; Figure S3: Consistency assessment of network meta-analysis of the comprehension score. Figure S4: Consistency assessment of network meta-analysis of the oral expression score; Figure S5: Consistency assessment of network meta-analysis of the repetition score; Figure S6: Consistency assessment of network meta-analysis of the denomination score; Table S1: Search strategy (take PubMed as an example); Table S2: Details of the intervention measures; Table S3: Risk of bias assessments for included studies.