Effects of Meditative Movements on Major Depressive Disorder: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Background: Tai Chi, Qigong, and Yoga are recognized as the most popular complementary approaches for alleviating musculoskeletal pain, improving sleep quality, and reducing blood pressure. The therapeutic effects of these meditative movements for treating major depressive disorder (MDD) is yet to be determined. Therefore, we examined whether meditative movements (Tai Chi, Qigong, and Yoga) are effective for treating MDD. Seven electronic databases (SPORTDiscus, PubMed, PsycINFO, Cochrane Library, Web of Science, CNKI, and Wanfang) were used to search relevant articles. Randomized controlled trials (RCT) using Tai Chi, Qigong or Yoga as intervention for MDD were considered for the meta-analysis (standardized mean difference: SMD). Results: Meta-analysis on 15 fair-to-high quality RCTs showed a significant benefit in favor of meditative movement on depression severity (SMD = −0.56, 95% CI −0.76 to −0.37, p < 0.001, I2 = 35.76%) and on anxiety severity (SMD = −0.46, 95% CI −0.71 to −0.21, p < 0.001, I2 = 1.17%). Meditative movement interventions showed significantly improved treatment remission rate (OR = 6.7, 95% CI 2.38 to 18.86, p < 0.001) and response rate (OR = 5.2, 95% CI 1.73 to 15.59, p < 0.001) over passive controls. Conclusions: Emphasizing the therapeutic effects of meditative movements for treating MDD is critical because it may provide a useful alternative to existing mainstream treatments (drug therapy and psychotherapy) for MDD. Given the fact that meditative movements are safe and easily accessible, clinicians may consider recommending meditative movements for symptomatic management in this population.


Data Items and Collection Process
Two review authors (L.Z. and C.L.) independently extracted data using a pre-created table. Discrepancies in the data extraction were discussed or resolved by a third reviewer (A.Y.). Data items in the extracted table included references (the first author, year of publication, and country), participant characteristics (initial sample and attrition rate, the number of male/female, mean age/age range, diagnostic criteria, percentage of female, and predominant ethnicity), intervention protocol (weekly training dosage, type of meditative movement, qualification of instructors, total training time, training mode, intervention duration follow-up assessment, and co-intervention), outcome measured, and safety. To calculate the pooled effect size, we also extracted the mean and standard deviation of the outcomes, along with the number of participants in each group.
In order to accurately outline the strengths of each selected study and to correctly interpret the data, we contacted the corresponding author for extra information regarding the details of the study, which may not have been clear in the published paper. We requested the details of 11 selected studies, including the percentage of patients with MDD, eligibility criteria (DSM-IV) administered by psychiatrist(s), percentage of ethnicity, qualification of instructors and/or therapists in the meditative movement and control groups, co-intervention, and/or quantitative data for calculating the pooled effect size of the effect of meditative movement versus control group on depression severity. In addition, given that some researchers may not have clearly described the methodology in their published papers that had been actually used in their studies, we also emailed the author to confirm the information, such as randomization procedure(s) and allocation concealment, blinding of assessor(s), and/or use of intent-to-treat analysis. Luckily, 90.9% (n = 10) of the authors emailed us back in response to our questions accordingly.

Assessment of Methodological Quality for Selected Trials
Methodological quality of the selected trials was assessed using the Physiotherapy Evidence Database (PEDro) scale [30]. The original PEDro scale consists of 11 items, including eligibility criteria, random allocation, concealment of allocation, baseline equivalence, blinding of stakeholders (participants, instructors, and assessors), retention rate of more than 85%, intention-to-treat analysis, between-group statistical comparisons, and point measures, and measures of variability. Given the fact that blinding of participants and instructors are impractical during a meditative movement intervention, these items were removed from the original scale, leading to a total of nine items. Points were only awarded if a criterion was clearly satisfied, with high scores indicating better methodological quality.

Data Synthesis and Analysis
Based on the random-effects model, effect sizes (standardized mean difference, SMD) across individual studies were pooled in Comprehensive Meta-Analysis Software. The forest plots were automatically generated for the severity of depression by entering the quantitative data (mean, standard deviation, and the number of participants in each group). The magnitude of SMD was classified according to the following cut-off values: (i) 0-0.19 = negligible effect, (ii) 0.2-0.49 = small effect, (iii) 0.5-0.79 = moderate effect, and (iv) 0.8 or above = large effect [31]. I-squared was also computed to determine the degree of homogeneity of effect sizes across the selected individual trials: (i) 25% = small, (ii) 50% = medium, and (iii) 75% = large. The funnel plot was used to visually assess publication bias, along with the Egger's regression intercept test. We also calculated the odd ratio (OR) for two dichotomous outcomes (response rate and remission rate).
To decrease the unit-analysis error, if there were trials with more than one control groups, the sample size of the meditative movement group was equally divided for two comparisons, with means and standard deviations of the meditative movement group remaining unchanged [32]. For those studies with two measuring scales on the depression severity, we selected the clinician-administered scale [33]. For moderator analysis, subgroup meta-analysis and meta-regression were individually performed for categorical variables and continuous variables, based on mixed-effects models. The categorical moderators included attrition rate ≥ 15% (Yes vs. No), 100% of MDD (Yes vs. No), predominant ethnicity (Chinese/Indians vs. Caucasian/Hispanic), control type (Yes vs. No), type of meditative movement (Tai Chi/Qigong vs. Yoga), intervention duration (<12 weeks vs. ≥12 weeks), training mode (group vs. mixed), and concomitant drugs/psychotherapy (Yes vs. No). Continuous moderators included the mean age of participants and total minutes of meditative movement intervention. In our meta-analysis, subgroup analysis based on study quality was not performed because none of the selected trials scored 5 (low quality) or below.

Trial Selection
Both electronic and manual searches resulted in 814 records in total. Thirty-one full-text articles were assessed according to the pre-determined inclusion criteria, leading to a final number of 16 eligible RCTs. The detailed process of trails selection is showed in Figure 1.
ial Selection oth electronic and manual searches resulted in 814 records in total. Thirty -one full-text articles wer ed according to the pre-determined inclusion criteria, leading to a final number of 16 eligible RCTs. Th ed process of trails selection is showed in Figure 1. udy Characteristics he study characteristics of selected trials are summarized in Table 1. Eleven trials were conducted in th 4,26,27,[34][35][36][37][38][39][40][41], two in India [42,43] and China (Hong Kong), [25,28] and one in Germany [44]. All trial

Study Quality Assessment
Study quality of all the selected trials is summarized in Table 2. The selected trials demonstrated fair-to-high (sum scores ranged from 6 to 9) study quality, with a mean value of 7.6 and a median of 8. It was noted that allocation concealment was not used in half of the selected trials [25,27,[36][37][38][42][43][44]. Three trials did not use intention-to-treat analysis for missing data [28,36,39]. Note: Item 1 = eligibility criteria; Item 2 = randomization; Item 3 = concealed allocation; Item 4 = similar baseline; Item 5 = blinding of assessors; Item 6 = more than 85% retention; Item 7 = missing data management (intent-to-treat analysis); Item 8 = between-group comparison; Item 9 = point measure and measures of variability; 1 = explicitly described and present in details; 0 = absent, inadequately described, or unclear.

Effects of Meditative Movements on Dichotomous and Continuous Outcomes
Based on control type (active and passive conditions), two sub-analyses were performed to determine the effects of meditative movements on remission rate (8 arms) [26,27,37,40,42,43] and response rate (six arms) [26,27,37,40,44] To detect the consistency of the effects of meditative movements on depression severity, a sensitivity analysis was performed by removing a trial with an outlying effect size (SMD = −2.37) [25], based on the visually asymmetrical Funnel plot ( Figure 2) and the Egger's Regression Test (Egger's regression intercept = −2.172, p = 0.01). While this outlier was removed for further analysis, the funnel plot of remaining trials showed a symmetrical Funnel plot (Egger's regression intercept = −1.71, p = 0.07). For the meta-analysis in 15 trials with 19 treatment arms, compared with the control group, the aggregated result showed a significant benefit in favor of meditative movements on depression severity (SMD = −0.56, 95% CI −0.76 to −0.37, p < 0.001, I 2 = 35. 76%; Figure 3). Because the anxiety severity was not considered as the primary outcome in a small number of trials, a pooled effect size was calculated to determine the magnitude of treatment effects on the anxiety severity with no subgroup analysis of control type and other potential moderators. The results of the meta-analysis indicated a significant benefit in favor of meditative movements on anxiety severity (SMD = −0.46, 95% CI −0.71 to -0.21, p < 0.001, I 2 = 1.17%; Figure 4).

Moderator Analysis
There were no significant moderator effects on both categorical and continuous variables using mixedeffects model analyses (Table 3). When moderator analysis was performed using the fixed -effects model, a significant difference between two control conditions existed (Q = 4.

Moderator Analysis
There were no significant moderator effects on both categorical and continuous variables using mixed-effects model analyses (Table 3). When moderator analysis was performed using the fixed-effects model, a significant difference between two control conditions existed (Q = 4.2, df = 1, p = 0.04): (1) meditative movements vs. active control (SMD = −0.42); (2) meditative movements vs. passive control (SMD = −0.76). For training mode, although no significant difference was observed, there was an increasing trend in terms of the magnitude of effects of meditative movement on the depression severity: Individual (SMD = −0.28) < group (SMD = −0.55) < mixed mode (SMD = −0.84). it was not included for data synthesis and the total treatment arms was 18; ®: A trial by Sshuver et al. [36] used the individual training mode (DVD-guided, selfpractice) found a treatment effect (SMD = 0.28). Thus, although no significant differences among the three training modes, the magnitude of the treatment effects showed an increased trend: individual (SMD = 0.28) < group (SMD = 0.55) < Mixed (SMD = 0.84).
: A trial by Field [34] used a mixed meditative movement of Tai Chi and Yoga, so it was not included for data synthesis and the total treatment arms was 18; ®: A trial by Sshuver et al. [36] used the individual training mode (DVD-guided, self-practice) found a treatment effect (SMD = 0.28). Thus, although no significant differences among the three training modes, the magnitude of the treatment effects showed an increased trend: individual (SMD = 0.28) < group (SMD = 0.55) < Mixed (SMD = 0.84).

Discussion
As the first systematic review and meta-analysis synthesize the evidence of the effects of meditative movements (specifically Tai Chi, Qigong, and Yoga) on MDD, we found that meditative movements may have positive effects on the treatment of MDD, and importantly, with no occurrence of significant adverse events. This evidence suggests that there is a possibility for using these exercises as an alternative and/or augmentation approach to conventional treatments for MDD. In addition, our findings also suggests that mixed training modes may be an optimal method for treating MDD, as compared to instructor-led group practices or self-practice alone.
The potential mechanism for how meditative movements work to treat MDD symptoms remains elusive. According to the traditional Chinese medicine theories, Qi or life energy circulates via 12 main meridians (pathways) within the human body [45]. Traditional Chinese physicians believe that somatopsychic disorders occur when the flow of Qi becomes stagnant or blocked, whereas a free flowing and balanced Qi is a sign of good health and spirit [45]. Meditative movements as a mind-body healing art, are believed to cultivate the life energy and enhance its flow [45], which may potentially alleviate the progression of depressive symptoms. Qi is usually carried out by regulating breathing. Qigong, Yoga, and Tai Chi practitioners use typical breathing techniques (in a form of abdominal breathing) during the whole practice. Some studies indicated that abdominal breathing could significantly decrease negative affect and reduce cortisol levels [46]. Abdominal breathing might serve as a bridge linking the autonomic nervous system and the central nervous system to mobilize vagal activation of GABA (gamma-aminobutyric acid) pathways from the prefrontal cortex and insula, and to inhibit amygdala overactivity [47]. This might partly explain why meditative movements have a positive influence on depression.
Recently, advances in brain imaging techniques have provided another option for researchers to investigate the possible neurophsyiological mechanisms underlying the beneficial effects of meditative movements for treating MDD [48]. A neurophysiological study by Chan et al. [49] indicates that a holistic intervention (i.e., Chinese Chan-based mind-body movements, plus health education and diet modification for 10 weeks) significantly elevated left-side anterior activation (an index of positive mood) in individuals with MDD, whereas such encouraging findings were not found in either cognitive behavioral therapy nor waitlist groups. The author further explained that the significant elevation in the brain region in the holistic intervention group of MDD was positively associated with decreased depression severity, as measured by the BDI [49]. Thus, the depression-alleviated effect of meditative movements for patients with MDD may be meditated by elevating the left-side anterior activation.
In this study, we searched empirical studies published in both English and Chinese. We feel that it is important to do so because traditionally, research on Tai Chi and Qigong were predominantly performed in China and Asian countries. Surprisingly, no Chinese-language studies were found to investigate the therapeutic effects of meditative movements for MDD. The medium-to-high quality RCTs in this systematic review showed that meditative movements are a safe and effective treatment for patients with MDD. Such result is attributed to data request by contacting the original of all selected trials and to confirm the unclear report of methodological procedures in some published papers. As suggested by Chen et al. the quality of RCTs may directly affect the magnitude of the pooled effect size [50], but the procedure of data request was not employed in the recently published review that evaluates the effects of Yoga on MDD [23]. To accurately address the strengths of selected RCTs and to objectively interpret the study findings, data requests are necessary in future review studies.
Several study limitations of this systematic review should be acknowledged while interpreting our research findings. One of the most important drawbacks is that the concealment of the intervention was difficult as subjects knew whether they received meditative movement intervention, even though a centralized randomization procedure was used. This might lead to subjectivity and expectation bias by the participants. Second, allocation concealment in half of the selected trials was absent in this systematic review. Such inadequate concealment of allocation was associated with exaggerated estimates of meditative movement treatment benefits for MDD [51]. Third, meditative movements were not offered as mono-therapy in half number of studies, but as adjunctive treatments to existing interventions (drug or psychological therapy). It may be difficult to conclude whether the positive outcomes were attributed to the meditative movement alone, a synergetic intervention effect, or to the conventional treatment received by the patients. This was also confounded by the fact that a variety of interventions were received by control groups. When sub-analysis was performed on two dichotomous outcomes using six trials, meditative movements did not have significant effects on remission rate and response rate when more stringent (active) controls were used. Nevertheless, results from our overall analysis provide support for meditative movements as an adjunctive treatment or as a mono-therapy for reducing depressive symptoms. Fourth, the duration of the meditative movement interventions varied a greatly across included studies, leading to difficulties in recommending an optimal dose of intervention. It is possible that in most studies, participants had attained the minimal duration needed to obtain the psychological benefits. Fifth, because the follow-up period varies greatly in a small number of trials on the outcomes, meta-analysis was not performed to investigate the long-term effects of meditative movements for MDD in this present review. Sixth, the findings of this study may not apply to patients with very severe depression (HAMD score ≥23). These patients are usually excluded in the selected studies since many ethics committees require conventional treatment, rather than meditative movements as interventions for this group, considering the severity of their illnesses. Thus, interpretation of this systematic review is limited to the immediate effects of meditative movements on alleviating MDD symptoms.

Conclusions
This systematic review, based on the existing literature, suggests that meditative movements may be effective interventions to alleviate MDD symptoms. Emphasizing the therapeutic effects of meditative movements for treating MDD symptoms is critical because it may provide a useful option to existing mainstream treatments (drug therapy and psychotherapy) for patients with MDD. Given the fact that meditative movements are safe and easily accessible, clinicians may consider recommending meditative movements for patients of MDD. It must be addressed that significant methodological limitations were found in most of the empirical studies to date, which may have impacted the interpretation of these findings. More randomized controlled trials with rigorous research design are warranted to establish the therapeutic effects of meditative movements for MDD, and its potential use for prevention and as an adjunctive treatment for MDD.
Author Contributions: L.Z., A.Y., K.W.C., G.X.W., Z.R. and J.S.M.C. contributed to the conception and design of the review. L.Z. applied the search strategy. L.Z., C.L., A.Y. applied the selection criteria. L.Z., C.L. and A.Y. completed assessment of risk of bias. All authors analyzed and interpreted the data. L.Z., A.Y., Z.R. and C.L. wrote this manuscript. C.L., K.W.C. and P.A.K. edited this manuscript. L.Z. is responsible for the overall project.