Effects of traditional Chinese exercise on patients with cognitive impairment: A systematic review and Bayesian network meta‐analysis

Abstract Objective To systematically review the effectiveness of four types of traditional Chinese exercise (TCE) on patients with cognitive impairment (CI) and to rank these four TCE types. Design A Bayesian network meta‐analysis. Methods Four English databases, including PubMed, EMBASE, Cochrane Library and Web of Science, and three Chinese databases, including CNKI, VIP and Wanfang, were searched from their inception to December 2019. Randomized control trials conducted to verify the effects of TCE on patients with CI were included. We used network meta‐analysis to evaluate the relative effects and rank probabilities of the four types of TCE. Results The results of the network meta‐analysis indicated that baduanjin (N = 9), tai chi (N = 11), liuzijue (N = 2) and qigong (N = 1) all had significant benefits compared with control conditions. According to the ranking probabilities, baduanjin was most likely to be associated with substantial improvement in cognition, followed by tai chi, liuzijue and qigong. Conclusion Our study revealed the effectiveness of TCE for improving global cognition in adults with cognitive impairment. Baduanjin may be the most effective exercise type. The evidence summarized in our study still contains bias, and more research should be carried out to verify the validity of TCE.

(MCI) was defined by Petersen in the 1980s. As a transitional stage between age-matched normal cognition and dementia, MCI is associated with a high risk of progressing to dementia Tschanz et al., 2006). Because the relevant symptoms are easily overlooked, MCI may affect many more people than has been reported (Petersen & Morris, 2005). Some studies have found that more than 60% of elderly individuals with MCI progress to dementia in the next 5-10 years (Hebert et al., 2003). CI severely affects functioning and quality of life, increases economic and psychological burden and increases stress in caregivers and families. Therefore, preventing the progression of cognitive decline and living well with CI is a priority for older people and the general healthcare system (Britain, 2009).
The fundamental way to enable people with dementia and MCI to live with it is to slow and reverse the progression of CI.

| BACKG ROU N D
We hope that the efforts we made can delay or reverse the progress of cognitive decline by applying various pharmacological and nonpharmacological interventions. Evidence from systematic reviews does not support the utilization of pharmacological treatments for cognitive protection in patients with MCI due to their adverse effects, which has increased the attention being paid to nonpharmacological interventions (Fink et al., 2018). Physical activity has been shown to improve cognitive function and delay the onset of cognitive impairment (Gheysen et al., 2018). Traditional Chinese exercise (TCE), as a type of exercise, is characterized by its low intensity, high level of safety, and ease of learning and by the fact that no special equipment is required, and it has shown its advantage in recent years. Recent randomized controlled trials (RCTs) and reviews have explored the effects of various types of TCE, such as tai chi (Chang et al., 2011), baduanjin (Zhu et al., 2015, qigong (Cai & Zhang, 2018) and liuzijue (Zheng et al., 2013). One systematic review in 2019 (Zhang et al., 2019) included five RCTs that shared the same usual care control groups without head-to-head comparisons of different TCE types. Due to the lack of pairwise randomized controlled trials (RCTs) and the lack of common control groups, we cannot compare alterations and rank the probabilities of these TCEs producing these changes in cognitive function in patients with CI based on the traditional pairwise meta-analysis method. At the same time, some contradictory findings remain among studies. One review  also reported that tai chi was beneficial for ameliorating cognitive function in older people. However, for those with CI, a subgroup analysis in a review in 2018  showed that tai chi was not an effective method. As mentioned above, inconsistent results should be clarified to draw more solid conclusions.
Overall, this study provides practical recommendations for healthcare professionals and offers more options for patients with CI. This systematic review and Bayesian network meta-analysis were performed to collect all of the direct and indirect evidence (Hoaglin et al., 2011) and to compare and rank the efficacy of different types of TCE in improving cognitive function in patients with CI.

| Aims
A Bayesian network analysis was used to compare and rank the efficacy of four TCE types on cognitive function in elderly individuals diagnosed with CI and to provide appropriate recommendations and possible directions for future clinical practice and research.

| Design
The review was conducted according to the guidelines of the Cochrane Collaboration and the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (Hutton et al., 2015).

| Search strategy
Four English databases, including PubMed, Cochrane Library, EMBASE and Web of Science, and three Chinese databases, including the China National Knowledge Infrastructure (CNKI), Weipu (VIP) and Wanfang Data, were searched from their inception to 01 December 2019. A search was also performed on search engines, including Google Scholar. The following keywords were chosen to screen studies: cognitive impairment, mild cognitive impairment, dementia, baduanjin, tai chi, qigong, liuzijue and cognition. The PubMed search strategy is presented in Appendix S1. We also reviewed the references of the included studies and reviews related to this field to avoid omission.

| Quality assessment
The methodological quality of the included studies was appraised by using The Cochrane Collaboration "risk of bias" tool version 5.1.0.
Six domains of bias were used to fully evaluate the quality of the included studies, including selection bias, performance bias, detection bias, attrition bias, reporting bias and other bias. The quality was rated by two reviewers as low, moderate or high. All included studies were independently assessed, and scores were obtained through final consensus. Data abstraction involved two reviewers used a standardized form to abstract the relevant data, including the author, year of publication, control group details (e.g. length of intervention, type of intervention, frequency), country, baseline characteristics of participants, diagnosis, intervention, cognition function outcomes and measurement tools. Any disagreement was resolved by discussion or consultation with a third reviewer.

| Statistical methods
Initial pairwise meta-analyses were performed with RevMan 5.3 software. The end point of the primary outcome was extracted to evaluate the effectiveness of TCE. For different scales used in the cognitive function assessment, standardized mean differences (SMDs) and 95% confidence intervals (CIs) were computed to obtain pooled effect sizes. The I 2 statistic was used to rate heterogeneity as low (<25%), moderate (25%-50%) or high (>50%) (Higgins et al. 2003). A random effects method was chosen if there was high level of heterogeneity in the pooled results; otherwise, a fixed effects method was chosen. Stata 12.0 and d GeMTC 0.14.3 were used to perform the network meta-analysis. A consistency test was performed by using inconsistency factors and node-splitting analysis. When the 95% credible intervals (CrIs) for an inconsistency factor contained 0 and the p value of node-splitting analysis exceeded .05, the consistency model was used if there was no significant inconsistency to calculate the effect size of four TCE types and evaluate the rank probabilities (Higgins et al., 2012).
F I G U R E 1 Flow diagram of studies included and excluded at each stage of review Otherwise, the inconsistency model would be used. At the same time, the potential scale reduction factor (PSRF) was evaluated to manifest the convergence of the model. The closer the PSRF value is to 1, the better the convergence of the model is indicated (Brooks & Gelman, 1998). The range of rank probabilities was from 0%-100%, and probabilities closer to 100% imply better outcomes of the intervention.  Table 1 demonstrates the characteristics of the included studies.  Table 2. Overall, all included studies showed a relatively moderate risk of bias. All studies reported randomization, but allocation concealment details were not found in most studies. To some extent, potential selection bias may have influenced the results.

| Analyses of outcomes
The pairwise comparisons of the four types of TCE are shown in In the network meta-analysis, 23 two-arm studies of 27 studies were included. Eleven studies compared tai chi with control conditions, 9 studies compared baduanjin with control conditions, 2 studies compared liuzijue with control conditions, and only 1 study compared qigong with control conditions. As shown in Figure 3, a consistency test was performed by using node-splitting analysis.
The results showed that the 95% CI of the inconsistency factor contained 0, and the PSRF value was 1, indicating that the consistency model could be selected. The relative effects of the four types of TCE are presented in Table 3 The rank probabilities of the four types of TCE are shown in Table 4 and Figure 4. Baduanjin was most likely to rank first (53%), tai chi was most likely to rank second (40%), liuzijue was most likely to rank third (26%), and qigong was most likely to rank fourth (26%).
The results indicated that baduanjin produced more positive outcomes than the other three TCE types.

| Adverse events
Most of the included studies reported adverse events, but no events related to TCE had occurred. One study (Sungkarat et al., 2018) reported that one participant from each group experienced a fall with a bone fracture during the intervention period. The cause of the fall was not related to the interventions and occurred outside of the exercise space. One study (Li, 2016) reported the number of falls in the past 12 months, and the average for the intervention group was 0.4, while that for the control group was 0.6; there was no difference between the two groups.

| Main findings and interpretation
Our analysis was based on 27 studies including 2,414 individuals randomly assigned to 4 different TCE, tai chi, baduanjin, qigong and liuzijue. This review showed that significant benefits were found with improved cognitive function in elderly individuals with CI, as measured by MMSE and MoCA scales. Bayesian network analysis including the four types of TCE showed their respective advantages in promoting cognition. According to the analysis of the rank probabilities, baduanjin was the most advantageous in terms of promoting cognitive functions.

F I G U R E 2 The direct comparisons of different types of TCE and the control
Our results showed that baduanjin and tai chi improved cognitive function in various populations, consistent with previous studies (Chang et al., 2011;Tao et al., 2017;Wayne et al., 2014). As types of mind-body interventions, baduanjin and tai chi decrease the speed of cognitive decline by training slow motions that allow elderly individuals to avoid the consequences and complications of CI. Regarding the opposite conclusions of the two reviews Zhang et al., 2019), this may be possibly be due to the sample size, the number of included studies and the types of TCE, which may have partly led to differences in the findings.
We included two studies examining the efficacy of liuzijue and only one examining qigong, and positive outcomes were shown. The study of qigong (Ladawan et al., 2017) that investigated enhanced global cognitive functions in individuals also revealed significant improvements in global cognitive functions following aerobic exercise. However, regarding studies investigating liuzijue, most studies have focused on chronic obstructive pulmonary disease (Wu et al., 2018). In our study, the evidence provides little support for cognitive improvement with liuzijue. Therefore, more RCTs related to qigong and liuzijue should be carried out in older individuals with CI to confirm their efficacy.
Thus, changes in cognitive assessment scores do not equal a clinically relevant change (Cohen Mansfield & Billig, 1986).
Two included studies (Li, 2017;Xia, 2017)  TCE is composed of simple, brief movements with low physical and cognitive demand and can be self-learned and practised . One meta-analysis included 180 reviews and suggested that exercise increases the odds of adverse events but not of serious adverse events, such as falls (Niemeijer et al., 2019). Our study also Exergaming is a feasible and relatively safe intervention that offers an environment in which physical and cognitive exercise is combined (Colombo et al., 2012). One study (Hsieh et al., 2018) revealed that VR tai chi had a significant protective effect on cognitive and physical outcomes. The application of VR programmes with tai chi also showed high attendance (Lan et al., 2013). One review showed that hardly any robust scientific research has been conducted on exergaming and dementia (Van Santen et al., 2018), and further studies should adopt TCE with advanced technology to adapt this intervention to home-based applications in individuals with cognitive impairment. To ameliorate cognitive functions, individuals with CI should be encouraged to practise TCE during daily life, and Baduanjin is recommended as the first choice. Furthermore, as TCE is easy to learn and practise, it might be a useful strategy to decrease caregivers' burden. Therefore, TCE should be recommended as an integral part of treatment for patients and family caregivers.

| LI M ITATI O N S
First, and perhaps most notably, neither the methodological quality of nor the heterogeneity among the studies included in the meta-analysis was strong enough to draw a solid conclusion. Most studies did not report the allocation details. Although it may have been hard to blind the experimental group, the assessors could have been blinded; to some extent, measurement bias could have been reduced. Most of the included studies were performed in China, which may result in the results suffering from certain degrees of selection bias. To make clearer recommendations, future studies should include larger sample sizes with rigorous study designs and provide more information on the potential mechanisms of the applied TCE.

| CON CLUS IONS
For patients with CI, our network meta-analysis suggests that four different types of TCE have potential therapeutic use in improving general cognitive function compared with control conditions.
Baduanjin may be the most effective, followed by tai chi, liuzijue and qigong. Multi-arm RCTs are necessary to confirm the effects of TCE on additional aspects of CI to provide more options for healthcare professionals.

ACK N OWLED G M ENTS
The authors thank Mingyue Hu from the School of Nursing, Xiangya School of Medicine, for her guidance on the development of this article.

CO N FLI C T O F I NTE R E S T
The authors state no conflicts of interest.

Chen Li and Dongxiang Zheng searched articles, Chen Li and Jinglan
Luo performed the data extraction and analysis, and Chen Li wrote this paper.

E TH I C A L A PPROVA L
There is no ethical statement for this trial.

S TATEM ENTS
The authors contributed to this review equally and approved the final version of the manuscript, which has not been previously published.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

F I G U R E 4 Rank probability of the efficacy of different interventions
O RCI D Dongxiang Zheng https://orcid.org/0000-0002-8320-4409