Summary of main results

In this review, we sought to explore evidence from randomised controlled trials (RCTs) to support or dispute whether Tai Chi practice is effective in ameliorating dyspnoea and facilitating exercise capacity, as well as to show its impact on psychosocial functions among people with chronic obstructive pulmonary disease (COPD). We included 12 studies, regardless of language, which compared the immediate effects of a Tai Chi programme (Tai Chi alone or in addition to another intervention) versus a control (i.e. usual care or other interventions identical to that adopted in the Tai Chi group, such as breathing exercise or exercise). This systematic review involves a total of 984 participants. Of these, 811 were randomly allocated to a Tai Chi group (with Tai Chi alone or added to another intervention) or a control group (i.e. usual care or other interventions identical to those adopted in the Tai Chi group) for minimum training of six weeks (Ng 2014) to maximum training of one year (Zhang 2012; Zhang 2014). The quality of evidence for the main outcomes of interest ranged from very low to moderate (summary of findings Table for the main comparison; summary of findings Table 2; summary of findings Table 3).

When Tai Chi was compared with usual care only, post‐programme data revealed that the Tai Chi group attained better functional capacity (as reflected by the lengthened six‐minute walk test (6MWT)) and improved pulmonary function (as reflected by the level of forced expiratory volume in one second (FEV₁)) when compared with the control group given usual care only. For quality of life, we observed no significant between‐group differences in total scores and domain scores for health‐related quality of life in three studies using St George's Respiratory Questionnaire (SGRQ) and in two studies using the Chronic Respiratory Questionnaire (CRQ), although we noted a better domain score for mastery of CRQ in the Tai Chi group (mean difference (MD) 0.89 unit, 95% confidence interval (CI) 0.31 to 1.47 unit; Analysis 1.21). Given the very low to moderate level of the quality of evidence, participants in the Tai Chi group, when compared with those in the usual care alone group, may have showed better functional capacity and pulmonary function.

However, comparison of Tai Chi in addition to other interventions (i.e. breathing exercise or exercise) revealed no between‐group differences in post‐programme data for level of dyspnoea and functional or psychosocial outcomes. Current data suggest that Tai Chi does not show a superior effect to other interventions and shows no additional effects when used with other interventions in terms of promoting physical and psychosocial functions.

Evidence from one study suggests that a Tai Chi programme may improve endurance capacity, with low quality of evidence noted in people with COPD. However, additional studies are needed to examine these effects. Currently, available data are insufficient for evaluation of the impact of Tai Chi on maximal exercise capacity, dynamic balance and muscle strength in people with COPD. In addition, evidence is insufficient to show whether Tai Chi style or duration of a Tai Chi programme will affect overall outcomes in COPD.

Overall completeness and applicability of evidence

The studies included in this review recruited participants with mild to very severe forms of COPD according to the Global Initiative for Chronic Lung Disease (GOLD) classification (Vestbo 2013) or that of the Chinese Thoracic Society (Chinese Thoracic Society 2013). Only one study specified recruiting people with a very severe form of COPD (Ng 2014). As disease severity may be one potential factor affecting response to treatment outcomes, we originally planned to perform a subgroup analysis to examine the impact of disease severity. However, in the light of overlapping of disease severity as reported in different studies, we found that the subgroup analysis based on disease severity was not feasible. When the results of individual studies were reviewed, differences in magnitude change for each of the outcome measures among studies with different levels of severity recruited were not apparent. Thus, additional studies are needed to evaluate the treatment effect in relation to disease severity. Lack of adverse events in any of the included studies despite various severity levels reported among people with COPD may imply that Tai Chi is safe to practise, and that it could have potential beneficial effects in this disease population.

As discussed, exertional dyspnoea is one of the major complaints among people with COPD that may prevent physical activity, thereby reducing their level of functional capacity. In this review, we observed a trend toward reduced dyspnoeic levels in modified Borg Scale scores (Yao 2004; Chan 2010) and in other dyspnoeic parameters (i.e. MMRC) (Du 2013; Wang 2014). However, pooled results suggest that evidence was insufficient to show beneficial effects of Tai Chi in reducing the level of dyspnoea.

At similar baseline levels, the Tai Chi group (when compared with the group given usual care alone) in six of the included studies showed a significantly longer 6MWT during post‐programme measurement than was reported in the control group, by 29.64 metres (95% CI 10.52 to 48.77 metres; Analysis 1.4). Such a difference falls within the range of minimal clinically important differences (MCIDs) (i.e. 25 to 35 metres) in 6MWT suggested for people with COPD (Holland 2013). In addition, one study (Leung 2013) evaluated the change in endurance shuttle walk test (ESWT) and reported an increase of 373 seconds. This change was greater than the MCID reported in ESWT by 65 seconds after use of bronchodilator (Pepin 2011), or by 165 seconds after an eight‐ to 10‐week ground walking programme (Ng 2013). These findings suggest that a Tai Chi programme may be beneficial in promoting functional capacity. However, available data were insufficient to show changes in aerobic capacity as reflected by changes in the incremental shuttle walk test (ISWT) (Leung 2013) and in peak oxygen consumption (Yeh 2010) probably associated with the nature of low and moderate intensity of Tai Chi and the small sample size used (n = 38 in Leung 2013 and n = 10 in Yeh 2010). In contrast, when comparing the additional effects of Tai Chi on top of other interventions (i.e. breathing exercise (Zhang 2012) or exercise (Ng 2014)) versus other interventions, we found no between‐group differences in 6MWT, suggesting that Tai Chi did not show a superior effect nor an additional effect when provided with other interventions to induce extra health benefits.

As Tai Chi involves a series of rhythmic movements conducted in a semi squatting position, it may promote muscle endurance after training, thereby prolonging exercise tolerance as reflected by longer distance capacity in 6MWT and longer duration in ESWT. However, only one study (Leung 2013) examined effects of Tai Chi on other physical functions and demonstrated a significant improvement in dynamic balance but an insignificant change in muscle strength. Thus, currently available evidence is insufficient to prove the effect of Tai Chi in this area among people with COPD.

Pooled values of post‐programme FEV₁ in the Tai Chi group were significantly greater than those in the control group given usual care only by 0.11 L (95% CI 0.02 to 0.20 L; Analysis 1.9) in four studies (Zhou 2009; Chan 2010; Niu 2014; Zhang 2014). As Tai Chi incorporates deep breathing into a body movement, this improvement may be associated with the repetitive deep breathing training provided during the practice of Tai Chi. The commonly reported MCID of FEV₁ in a clinical trial ranged from 0.045 to 0.18 L (Donohue 2005). Although the difference appears to fall within the reported range, interpretation of pulmonary function (e.g. FEV₁) should be reviewed carefully, as the MCID reported could be influenced by the baseline level of FEV₁, diurnal variability, decline in baseline over time, baseline reversibility and repeatability of the data (Donohue 2005). As discussed, when Tai Chi and breathing exercise was compared with breathing exercise alone (Li 2012; Zhang 2012), we observed no between‐group differences in post‐programme data (MD 0 L, 95% CI ‐0.11 to 0.12 L). Such findings could probably be explained by the similarity of breathing components involved in both Tai Chi and the corresponding control. Specifically, the Tai Chi group incorporated deep breathing into each form of body movement, and the breathing exercise group emphasised deep breathing exercise and/or pursed‐lip breathing as the routine control training (i.e. breathing exercise). Hence, similar effects on promoting pulmonary function could be induced. Additional studies are needed to support this hypothesis.

Quality of life is another important domain of disease management for people with COPD. Most studies used the health‐related quality of life (HRQoL) questionnaire, which consisted of individual component scores and total score, to evaluate disease impact on overall quality of life. In this review, six studies adopted SGRQ and two used CRQ to evaluate HRQoL. Of these, three studies compared effects of a Tai Chi programme versus usual care (Chan 2010; Du 2013; Wang 2014), and one study compared effects of Tai Chi and exercise (in pulmonary rehabilitation) versus exercise (in pulmonary rehabilitation) (Ng 2014). This demonstrated an insignificant improving trend, while minimal or even no changes were observed when Tai Chi and breathing exercise were compared against control adopting breathing exercises (Li 2012; Zhang 2012). Two studies used CRQ to evaluate the change in quality of life (Yeh 2010; Leung 2013) in people receiving Tai Chi against people receiving usual care. A significant between‐group difference was demonstrated in the component scores of mastery by 0.89 unit (95% CI 0.30 to 1.47 units), which was greater than the reported MCID of 0.5 unit in people with COPD (Schunemann 2005). Pooled data may be related to higher weighting (85.7%) in one study, which involved a greater number of participants (i.e. 19 participants per group) (Leung 2013) than the other (i.e. five participants per group) (Yeh 2010). Thus, the conclusion in CRQ is mainly related to the findings of one study, and interpretation should be carefully reviewed.

The BODE Index predicts risk of mortality in people with COPD (Celli 2004). A one‐point increase may increase the hazard ratio to 1.34 for all cause‐mortality and to 1.62 for respiratory‐related mortality (Celli 2004). In this review, two studies (Du 2013; Wang 2014) demonstrated significantly lower post‐programme data in the Tai Chi group by 1.04 points (95% CI ‐1.69 to ‐0.40 point). However, high heterogeneity (I² = 95%) was reported because one study favoured Tai Chi (Du 2013), while the other favoured control (Wang 2014) with similar weightings. Thus, interpretation of these findings may confer potential bias.

Variation in Tai Chi training duration and protocol is another factor that may influence the results. All included studies adopted the practice of Tai Chi for 30 minutes or longer per session. The study that compared effects of Tai Chi and exercise (i.e. 40 minutes of aerobic exercise in pulmonary rehabilitation) against exercise (i.e. 40 minutes of aerobic exercise in pulmonary rehabilitation) (Ng 2014) reported that Tai Chi practice lasted for 15 minutes and emphasised repeating the practice of five of the whole 21 forms of short‐form Sun‐style Tai Chi. Hence, such a difference in practice duration and in the number of forms practiced may explain the non‐superiority or the additional effect observed. Given the limited studies retrieved, the impact of training protocol variation on outcomes could not be analysed.

Apparently, people in the Tai Chi group demonstrated significantly better functional capacity and pulmonary function, although with very low to moderate quality of evidence. When outcomes are interpreted, other factors may be found to influence the results. As discussed in the Background section, different styles and numbers of forms of Tai Chi may affect the amount of energy expended during practice (Lan 2001). Thus, different styles of Tai Chi training may yield varied outcomes (i.e. styles that involve a greater angle of squatting and a greater number of forms may involve greater energy expenditure, which may trigger a relatively larger cardiopulmonary and/or neuromuscular effect and in turn may induce a better therapeutic response). Thus, these differences may add bias to the current interpretation of results. In this review, practice styles and numbers of forms reported in the included studies were scattered and diverse; this made it not feasible to perform subgroup analyses based on styles and numbers of forms of Tai Chi. Future studies are needed to explore these topics.

Quality of the evidence

The overall risk of bias of studies included in this review ranged from very low to moderate. Four studies (Yeh 2010; Leung 2013; Ng 2014; Niu 2014) had four or more categories ranked as having low risk of bias; four studies were rated as having one or two categories of low risk (Chan 2010; Li 2012; Du 2013; Zhang 2014); and the remaining studies provided insufficient information and therefore were graded as having unclear risk (Yao 2004; Zhou 2009; Zhang 2012; Wang 2014). All studies reported the use of randomisation, but eight reported the randomisation methods used (Chan 2010; Yeh 2010; Li 2012; Du 2013; Leung 2013; Ng 2014; Niu 2014; Zhang 2014). Three of these studies were graded as having low risk of allocation concealment (Leung 2013; Ng 2014; Niu 2014), and the others had unclear risk because of insufficient data reporting. As Tai Chi is a physical exercise that consists of different forms and postures, participants could not be blinded about group allocation; therefore, blinding of participants is not practical. Lack of group blinding of participants may influence the performance of participants during post‐programme assessment, leading to performance bias. Thus, all studies were graded as having 'high risk' in this category. Half of the studies reported blinded assessors of outcome measures (Chan 2010; Yeh 2010; Du 2013; Leung 2013; Ng 2014; Niu 2014). The overall dropout rate ranged from 2.5% to 28.1%. Three studies reported use of intention‐to‐treat (ITT) analysis (Chan 2010; Yeh 2010; Ng 2014), while other studies used per‐protocol analysis or did not mention the number of participants who completed the trial. High dropout rates may also have influenced overall effects. Among the 12 included studies, three studies had pre‐registered a protocol and were then graded as having low risk in the categories of selective reporting. More than half of the included studies were written in Chinese and reported insufficient information;therefore, they were graded as having unclear or high risk of bias.

The overall quality of evidence for the main outcomes of interest ranged from very low to moderate, with details listed in summary of findings Table for the main comparison for the comparison between Tai Chi and usual care; in summary of findings Table 2 for the comparison of Tai Chi and breathing exercise versus breathing exercise; and in summary of findings Table 3 for the comparison of Tai Chi and exercise (in pulmonary rehabilitation) versus exercise (in pulmonary rehabilitation). We downgraded most of the outcomes because of performance bias in relation to the un‐blinding of participants and/or outcome assessors; selection bias due to inadequate information on random sequence generation or concealment allocation; and imprecision in relation to a small sample size. We downgraded few outcomes as the result of inconsistency in relation to moderate heterogeneity. Thus, interpretation of pooled findings of this review and its included studies should be considered carefully because of potential risks of bias.

Potential biases in the review process

In this review, we followed our pre‐defined protocol (Ngai 2012) with one minor alteration (see Differences between protocol and review) and included all studies regardless of language of publication. We compared Tai Chi alone or in addition to another intervention versus control (usual care or another intervention identical to Tai Chi) to examine effects of Tai Chi and the superiority of its effects when compared with other interventions, or its additional effects alongside other interventions (i.e. breathing exercise or exercise). Most studies included some of our pre‐defined outcome measures as their outcomes of interest. Thus, in the light of the limited number of studies (i.e. fewer than 10 studies for each outcome measure), we did not prepare a funnel plot analysis to explore potential publication bias. In addition, not all of the authors of included studies could be contacted because incomplete contact details were provided in some of the Chinese papers. The limited information available may have affected grading of risk of bias in this review. All studies reported no significant between‐group differences in all pre‐programme data, except Ng 2014, which reported a significant baseline pre‐programme difference in FEV₁ and FVC, which may have affected the results of pooled data on FEV₁ and FVC (as listed in the pulmonary function section of the secondary outcomes of this review). Thus, we removed from the final pooled analysis data on pulmonary function as reported by Ng and colleagues (Ng 2014).

Agreements and disagreements with other studies or reviews

Currently, two systematic reviews and meta‐analytical studies have examined effects of Tai Chi in people with COPD (Yan 2013; Ding 2014). The former review (Yan 2013) included eight studies(four written in English and four in Chinese) that were also included in this current review. The review by Ding 2014 included 10 studies. Of these, only three studies evaluated the effects of Tai Chi (two written in English and one in Chinese ‐ also included in this review), and remaining studies evaluated the use of Qigong. As Tai Chi and Qigong are two different types of interventions, we did not include studies that used Qigong alone. Thus, results as interpreted in this review may be different from the interpretation reported by Ding 2014, and a direct comparison of that review against the current one may not be appropriate.

In accordance with results provided in Yan 2013, this review also showed a significant difference in post‐programme 6MWT between the Tai Chi group and the control group (usual care) by 29.64 metres (vs 34.22 metres in Yan's review) and in FEV₁ by 0.11 L (vs 0.07 L in Yan 2013). Although we noted a similar improvement trend in reduction of dyspnoea, our findings did not show significant changes, as were reported by Yan 2013, possibly because additional studies were included in this review, data extraction was performed using different methods and three types of data analysis were performed (i.e. Tai Chi vs usual care; Tai Chi in addition to breathing exercise and vs breathing exercise; and Tai Chi in addition to exercise and vs exercise).

In this current review, all included studies reported no significant pre‐programme differences except baseline FEV₁ as reported by Ng 2014; thus, we included in the analysis the comparison between Tai Chi and control along with post‐programme data, as suggested in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), while not including data on FEV₁. However, for Yan's paper, differences were evident between pre‐ and post‐programme data used in the imputation process to compare effects of Tai Chi and control. Differences in the imputation method used in this meta‐analysis (i.e. post‐programme data compared in this review and increased number of papers reviewed) may explain in part the discrepancies observed for some outcomes.