The Effectiveness and Safety of Manual Acupuncture Therapy in Patients with Poststroke Cognitive Impairment: A Meta-analysis

Background. Poststroke cognitive impairment (PSCI) is a common cause of disability among patients with stroke. Meanwhile, acupuncture has increasingly been used to improve motor and cognitive function for stroke patients. The aim of the present study was to summarize and evaluate the evidence on the effectiveness of acupuncture in treating PSCI. Methods. Eight databases (PubMed, The Cochrane Library, CNKI, WanFang Data, VIP, CBM, Medline, Embase databases) were searched from January 2010 to January 2020. Meta-analyses were conducted for the eligible randomized controlled trials (RCTs). Assessments were performed using Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Barthel Index (BI), or modified Barthel Index (MBI). Results. A total of 657 relevant RCTs were identified, and 22 RCTs with 1856 patients were eventually included. Meta-analysis showed that acupuncture appeared to be effective for improving cognitive function as assessed by MMSE (mean difference ðMDÞ = 1:73, 95% confidence interval (CI) (1.39, 2.06), P < 0:00001) and MoCA (MD= 2:32, 95% CI (1.92, 2.73), P < 0:00001). Furthermore, it also suggested that acupuncture could improve the activities of daily life (ADL) for PSCI patients as assessed by BI or MBI (SMD = 0:97, 95% CI (0.57, 1.38), P < 0:00001). Conclusions. Compared with nonacupuncture group, acupuncture group showed better effects in improving the scores of MMSE, MoCA, BI, and MBI. This meta-analysis provided positive evidence that acupuncture may be effective in improving cognitive function and activities of daily life for PSCI patients. Meanwhile, long retention time of acupuncture may improve cognitive function and activities of daily life, and twist technique may be an important factor that could influence cognitive function. However, further studies using large samples and a rigorous study design are needed to confirm the role of acupuncture in the treatment of PSCI.


Introduction
Stroke ranks only second to ischemic heart disease as the leading cause of death and the third leading cause of disability-adjusted life-years (DALYs) lost worldwide [1]. In China, with over 2 million new cases annually, stroke has a close relationship with the highest DALYs lost of any disease [2]. Cognitive decline is a major cause of disability in stroke survivors [3]. It is estimated that 11.8 million patients who have had a stroke, 9.5 million of whom have had cognitive impairments after their stroke [4].
Poststroke cognitive impairment (PSCI) contains two different degrees of cognitive impairment, including poststroke cognitive impairment with no dementia (PSCIND) and poststroke dementia (PSD) [5]. In the study [6] which enrolled 620 patients in 12 hospitals with ischemic stroke, of the 506 patients who were followed-up at 3 months after stroke, 353 patients (69.8%) suffered cognitive impairment as measured by the Korean Vascular Cognitive Impairment Harmonization Standards neuropsychological protocol (K-VCIHS-NP). In America, the study on 212 subjects from the Framingham Study suggested that 19.3% of cases developed into dementia in 10 years after stroke [7]. PSCI is strongly related to a higher risk of mortality [8], poor functional outcome [9], and poor quality of life [10]. Identifying patients at risk of cognitive impairment is, therefore, important as well as targeting interventions to this group.
Unfortunately, treatment of PSCI has not been standardized [11]. So far, there are many drugs to improve cognitive, including acetylcholinesterase inhibitors, memantine [12], and nicergoline [13]. But because of the unclear efficacy and side effect, until now, there is a none drug that has been approved by the Food and Drug Administration (FDA) to treat vascular cognitive impairment [14].
In recent years, the spectrum of diseases suitable for acupuncture abroad has been significantly broader, like nervous system, muscular tissue, skeletal system, connective tissue, mental and behavioral disorders, digestive system, and respiratory system [15]. The World Health Organization has also recommended acupuncture as an alternative and complementary strategy for stroke treatment and improvement [16]. In the treatment of apoplexy sequelae, acupuncture is mainly used to treat dyspraxia [17], enhance life quality [18], improve cognition [19], and deal with depression and anxiety [20]. Animal experiments [21] showed that acupuncture could improve cognitive function by stimulating cholinergic enzyme activity and regulating brain-derived neurotrophic factor (BDNF) and cAMP-response elementbinding protein (CREB) expression in the rats' brain; Cai et al. [22] demonstrated that electroacupuncture can improve cognitive impairment in Alzheimer's disease mice by inhibiting synaptic degeneration and neuroinflammation.
After searching the database, we found that there were few systematic reviews or meta-analysis focused on the effectiveness of acupuncture in treating PSCI recently, although the number of papers related to this area has an upward trend in the last five years. For instance, the latest meta-analyses on acupuncture treating PSCI were published in 2014 [23] and 2016 [24]. But due to the limitation of sample size and the quality of trials included in the former one, the effectiveness of acupuncture in treating PSCI has not been fully determined. Meanwhile, the later one which also lacked quality randomized controlled trials (RCTs) was merely focused on the effectiveness of scalp acupuncture. Therefore, the purpose of this study is to evaluate the clinical efficacy of acupuncture in the treatment of PSCI by meta-analysis and provide evidence for its rational clinical application.

Inclusion/exclusion Criteria.
Relevant clinical trials were included if the following criteria were met: (1) they were randomized controlled trials (RCTs); (2) they included patients diagnosed with poststroke cognitive impairment; (3) they use cognitive function as an outcome measure; (4) the difference of interventions between experimental and control groups is whether they use acupuncture or not. To be more precise, the intervention of experimental group is acupuncture therapy plus another therapy or standard treatment; the intervention of control group is standard treatment or the therapy which also be used in the experimental group except acupuncture therapy. If trials which met above criteria contained more than two groups, the group receiving acupuncture was chosen as the experimental group, and the nonacupuncture treatment group was chosen as the control group.
Trials were excluded if they met any of the following criteria: (1) acupuncture were used in the control group; (2) neither the Mini-Mental State Examination (MMSE) nor the Montreal Cognitive Assessment (MoCA) was used as cognitive function evaluation scale; (3) specific type of acupuncture treatment was used in the experimental group, such as electroacupuncture and laser needles for acupuncture.

Data Extraction.
Two reviewers (L.W. and R.C.) independently extracted the general information of the included trials and reached consensus on all items. Extracted data included authors, year of publication, sample size, source of diagnosis, interventions, main outcomes, and information about acupuncture treatment (including course, frequency, and retention time).
Measures of the outcome evaluation that were reported in the included studies were Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Barthel Index (BI), or modified Barthel Index (MBI). MMSE which contains domains of orientation, memory, attention, language, and visuospatial ability is the most common used screening scale [25,26]. MoCA, which is also a cognitive screening and tracking tool, is more useful for the mild stages of the cognitive impairment [27]. It is composed of several cognitive domains such as memory, executive function, attention, language, abstraction, naming, delayed recalls, and orientation [28]. BI is one of the most widely used outcome measures to assess functioning for the patients who have neurological disorders [29,30]. It consists of 10 others' rated questions which are for the purpose of evaluating the ability of daily life [31].
2.4. Quality Assessment. The methodological quality and the risk of bias of the included studies were evaluated by the risk of bias 2.0 (ROB 2.0) tool. One author assessed the risk of bias of included studies by using ROB 2.0, and the other author confirmed the judgment. The following items were categorized as having high, low, or unclear risk of bias: 2 Neural Plasticity randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result and overall.
2.5. Statistical Analysis. The Review Manager software (version 5.3 Cochrane Collaboration, Oxford, United Kingdom) was used to perform most of the statistical analysis. The mean difference (MD) which was used as the effect analysis statistics for continuous data was calculated with a 95% confidence interval (CI). The I 2 statistic was used to analyze the heterogeneity between the data of included trials. If the figure of I 2 was above 50%, which means significant heterogeneity, sensitivity analysis would be performed to analysis the source of heterogeneity. Random-effects model was applied to calculate the study results with significant heterogeneity, while a fixed effect model is used if the statistical heterogeneity was inapparent. Publication bias was detected using a funnel plot.

Study
Quality. The risk of bias assessment of the included RCTs is illustrated in Figure 2. With regard to randomization process, 4 studies had a low ROB, and 18 studies had an unclear ROB. With regard to deviations from intended inter-ventions, all RCTs had a low ROB. With regard to missing outcome data, 16 studies had a low ROB, and 6 studies had a high ROB. With regard to measurement of the outcome, 2 studies had a low ROB, and 20 studies had an unclear ROB. With regard to selection of the reported result, 1 study had a low ROB, and 21 studies had an unclear ROB.
In terms of acupoint selecting, single area acupuncture points were used in five studies [39,40,45,49,50], including the eye, ear, and scalp acupuncture. Special acupuncture therapy, Jin three-needle therapy, and tri-jiao therapy were also used in three studies [46,47,53]. Apart from that, a total of 33 acupoints were selected in other 14 studies. Acupoints used for cognitive impairment in most trials were DU 20 (Bai Hui) and DU24 (Shen Ting) ( Figure 3).       [40,48,51] of them were removed for incorrect using of MMSE. The score of MMSE should be related to education level when detecting cognitive impairment [54], but these studies used same scoring criteria for people with different educational levels in the inclusion criteria which could cause that patients without cognitive impairment, but low educa-tional level were included in the studies. In addition, another two [46,53] of them that we removed could significantly vary the direction of the combined estimates (test for subgroup differences: chi 2 = 187:08, df = 1 (P < 0:00001), I 2 = 99:5%) ( Figure 4). After studying the full articles carefully, we found that the difference between these two and other article is that they both used Jin 3-needle technique which has a unique twist technique. This indicated that the source of heterogeneity might be related to the frequency and duration of twist technique. It also illustrated that twist technique may influence the effect of improving cognitive function, as the studies used Jin 3-needle technique (MD = 11:82, 95% CI (10.41, 13.23), P < 0:00001, I 2 = 0%) had better outcomes than the studies did not (MD = 1:73, 95% CI (1.39, 2.06), P < 0:00001, I 2 = 50%). However, further research is still needed due to the small number of included studies using Jin 3needle technique.

Effects of
Finally, after sensitivity analysis, a total of 12 studies were included for evaluating the effect of acupuncture treatment according to MMSE. The fixed effect model was used for analysis on the basis of heterogeneity test (I 2 = 50%). The results of meta-analysis showed that the MMSE score of the acupuncture treatment group was higher than that of the control group (MD = 1:73, 95% CI (1.39, 2.06), P < 0:00001). According to subgroup analysis based on treatment duration, meta-analysis results showed that MMSE score of the acupuncture treatment group was higher than that of the treatment group in ≤7-week subgroup (MD = 1:63, 95% CI (1.28, 1.99), P < 0:00001, I 2 = 60%). MMSE score in the >7-week subgroup was higher than that in the treatment group (MD = 2:36, 95% CI (1.43, 3.29), P < 0:00001, I 2 = 0%) ( Figure 5). It illustrated that the duration of acupuncture treatment was also a factor that could influence heterogeneity but have few effect on the outcomes. We also conducted subgroup analysis based on retention time and frequency of acupuncture, but those did not have much influence on heterogeneity or outcomes.

Neural Plasticity
After using sensitivity analysis of studies, we found one article [49] that had relatively large impact on statistical heterogeneity. What distinguishes it from other studies is the retention time of acupuncture, so we conducted subgroup analysis based on retention time ( Figure 6). It shows that the score of MoCA in the acupuncture group was observed to be higher than that   However, further research is still needed due to the small number of included studies with long retention time.
3.6. Effects of Acupuncture Treatment According to Activities of Daily Living. There are nine [35, 42-44, 46, 47, 49, 50, 52] trials that used Barthel Index or modified Barthel Index to compare ADL between acupuncture group and nonacupuncture group. The other two trials [48,53] also mentioned they evaluated ADL but did not specified what kind of scale they used; therefore, we did not include them in this analysis. A high statistical heterogeneity was observed (I 2 = 83% > 50%). Therefore, random-effects model analysis  10 Neural Plasticity was chosen to conduct meta-analysis. After using sensitivity analysis of studies, we found one article [49] that had relatively large impact on statistical heterogeneity. What distinguishes it from other studies was the retention time of acupuncture, so we conducted subgroup analysis based on retention time (Figure 7). Therefore, we divided the 9 articles into 2 groups according to the acupuncture retention time. It shows that the score in the acupuncture group was observed to be higher than that in the nonacupuncture group (SMD = 0:97, 95% CI (0.57, 1.38), P < 0:000 01); the difference is statistically significant. It also showed that longer acupuncture retention time may lead to better result in activities of daily life (0.5 hours: SMD = 0:79, 95% CI (0.53, 1.05), P < 0:000 01; 6 hours subgroup: SMD = 2:58, 95% CI (1.97, 3.19), P < 0:000 01).
Six [32,33,37,39,43,48] of the articles reported no adverse events related to acupuncture therapy happened during the trials. One article [50] reported one case of adverse event but did not mention the reason. One article [42] reported one case of subcutaneous hematoma in the experiment group and three cases of fainting in the control group. Another article [44] reported one case of fainting during acupuncture and two cases of subcutaneous hematoma.

Publication Bias.
We used STATA V.15.1 to evaluate publication bias. For MMSE, publication bias was assessed using Begg's test (Figures 8 and 9) and Egger's test (Figures 10 and 11), which did not show significant publication bias in the included studies. For MoCA and Barthel, the publication bias could not be assessed as less than ten articles were included.

Discussion
After searching PubMed, the Cochrane Library, CNKI, Wan-Fang Data, VIP, CBM, Medline, and Embase databases, 657 relevant RCTs were found in this meta-analysis, and 22 RCTs with 1856 patients were eventually included. According to this study, more evidence was provided to prove that acupuncture treatment is beneficial to PSCI patients in terms of cognitive function and ability of daily life. PSCI, which has high prevalence rate, is a result of mixed damage mechanisms [55]. Jeremy's study [56] suggested that ongoing ischemic vascular processes were the main mechanism which also emphasized the importance of management for vascular risk factor. Apart from that, there is no established therapy for prevention for PSCI so far. Hence, it is of vital importance for us to explore an effective and highly compliant treatment for PSCI patients.
There were a few relevant meta-analyses about acupuncture treatment for cognitive impairment. Liu et al. [57] searched RCTs that used acupuncture in the treatment of PSCI before February 2012 and included 21 RCTs with a total of 1421 patients. Due to the high risk of bias and lack of unified scale for evaluating cognitive function of the included studies, it is hard to reach reliable conclusions on the effect of acupuncture. Furthermore, it did not focus on analyzing the effects of acupuncture alone but on the combined treatment with acupuncture, which made it difficult to objectively evaluate the effect of acupuncture treatment for PSCI. Kim et al. [58] searched the RCTs on mild cognitive impairment (MCI) patients from October 2007 to August 2017 and   11 Neural Plasticity compared the effect of electroacupuncture (EA) to western medication. Their study, which included 5 RCTs with 257 patients, showed that EA had a higher score on MMSE and MoCA than western medications. But the weak methodological quality of the studies and small sample size may affect the reliability of the results.
This present review offered several significant perspectives. Firstly, a common deficiency of previous RCTs and reviews is the lack of attention to ADL. There were only 9 RCTs that we included used BI or MBI to evaluate ADL on PSCI patients. ADL is one of the most important measures to evaluate how serve PSCI is and an essential measure to distinguish PSCIND and PSD [12]. Moreover, PSD has a significant higher fatality rate than PSCIND [55]. Therefore, we strongly recommend future studies to focus on the ability of daily life for PSCI patients and explore if acupuncture has the benefit of preventing PSCIND progressing to PSD whenever possible. Secondly, it may cause false increase in patients' test results by using the same version of scale to evaluate cognitive function before and after the treatment. In order to ensure accuracy of MoCA test, we completed MoCA's official standardized training and certification program online. We learned from the official website (https:// www.mocatest.org/training-certification/) that the delay between administration should be sufficient to decrease the risk of a possible learning effect when administering the MoCA to the same subject; the alternative/equivalent versions of the MoCA should be used to decreased possible learning effects when the MoCA is administered, respectively. Therefore, we suggest using different versions of MoCA, such as 7.1, 7.2, or 7.3 versions at different stages of assessment. Thirdly, more attention should be paid to the standardized use of the scale. For instance, in this review, three articles were found incorrect using of MMSE. The wrong use of the scale may lead to serious errors in screening and evaluation of patients. Furthermore, most of the included RCTs have unclear bias of randomization process, measurement of the outcome, and selection of the reported results, which have a significant impact on the evaluation of the results. It is true that the nature of the acupuncture made it difficult for investigators to blind participants and almost impossible to blind the therapists; all the RCTs we included likewise did not blind the participants or therapists. However, we should attach more importance to the blinding on outcome assessors, especially the scale evaluators whose judgement may have a large influence on the outcome. We also advise that all the RCTs which study the effect of acupuncture should have a protocol with complete information prior to the start of trials. The protocol should provide detailed information according to the CONSORT statement and STRICTA recommendations in order to minimize the performance and assessment bias of RCTs.
This review also had certain limitations. Firstly, although we searched the trials written in Chinese or English, all the trials included were conducted in China. This is related to the less application of using acupuncture treatment on PSCI in countries other than China, and it may limit the universality of the results. Moreover, our study mainly focused on the effect of acupuncture alone but neglected the synergistic effect between acupuncture and other effective treatment, such as medication or cognitive training. In addition, this study did not consider the possible placebo effect of    Figure 11: Egger's test. 12 Neural Plasticity acupuncture as no sham/placebo-controlled trials were included. Furthermore, significant heterogeneities were observed in our study. Experimental design, various acupoint selecting, and therapist skill difference can contribute to the high heterogeneity. Problems about which acupoints are most effective for PSCI and how long the treatment should last need to be resolved in future studies.

Conclusions
Compared with the nonacupuncture group, the acupuncture group showed better effects in improving the scores of MMSE, MoCA, BI, and MBI. This meta-analysis provided positive evidence that acupuncture may be effective in improving cognitive function and activities of daily life for PSCI patients. Meanwhile, long retention time of acupuncture may improve cognitive function and activities of daily life, and twist technique may be an important factor that could influence cognitive function. However, further studies using large samples and a rigorous study design are needed to confirm the role of acupuncture in the treatment of PSCI.