The effectiveness of yoga in modifying risk factors for cardiovascular disease and metabolic syndrome: A systematic review and meta-analysis of randomized controlled trials
Background Yoga, a popular mind-body practice, may produce changes in cardiovascular disease (CVD) and metabolic syndrome risk factors.
Design This was a systematic review and random-effects meta-analysis of randomized controlled trials (RCTs).
Methods Electronic searches of MEDLINE, EMBASE, CINAHL, PsycINFO, and The Cochrane Central Register of Controlled Trials were performed for systematic reviews and RCTs through December 2013. Studies were included if they were English, peer-reviewed, focused on asana-based yoga in adults, and reported relevant outcomes. Two reviewers independently selected articles and assessed quality using Cochrane’s Risk of Bias tool.
Results Out of 1404 records, 37 RCTs were included in the systematic review and 32 in the meta-analysis. Compared to non-exercise controls, yoga showed significant improvement for body mass index (−0.77 kg/m2 (95% confidence interval −1.09 to −0.44)), systolic blood pressure (−5.21 mmHg (−8.01 to −2.42)), low-density lipoprotein cholesterol (−12.14 mg/dl (−21.80 to −2.48)), and high-density lipoprotein cholesterol (3.20 mg/dl (1.86 to 4.54)). Significant changes were seen in body weight (−2.32 kg (−4.33 to −0.37)), diastolic blood pressure (−4.98 mmHg (−7.17 to −2.80)), total cholesterol (−18.48 mg/dl (−29.16 to −7.80)), triglycerides (−25.89 mg/dl (−36.19 to −15.60), and heart rate (−5.27 beats/min (−9.55 to −1.00)), but not fasting blood glucose (−5.91 mg/dl (−16.32 to 4.50)) nor glycosylated hemoglobin (−0.06% Hb (−0.24 to 0.11)). No significant difference was found between yoga and exercise. One study found an impact on smoking abstinence.
Conclusions There is promising evidence of yoga on improving cardio-metabolic health. Findings are limited by small trial sample sizes, heterogeneity, and moderate quality of RCTs.
Cardiovascular disease (CVD) and metabolic syndrome are major public health problems in the USA and worldwide.1,2 Metabolic syndrome is defined as having at least three metabolic risk factors – increased blood pressure, high blood sugar level, excess body fat, and abnormal cholesterol levels – and greatly increases the chance of future cardiovascular problems.3 Lifetime risk of CVD is substantial as estimated through risk functions like those from the Framingham Heart Study,4 underlining the need for prevention and control of risk factors.
CVD and metabolic syndrome share many of the same modifiable risk factors. Several guidelines name physical inactivity, the fourth leading risk factor of global mortality,5 as an important modifiable risk factor for CVD and metabolic syndrome.6⇓–8 They state that regular and adequate levels of physical activity in adults can reduce the risk of hypertension, coronary heart disease, stroke, diabetes, and can help maintain a healthy weight. Yoga, an ancient practice from India that incorporates physical, mental, and spiritual elements, may be an effective form of physical activity.
In recent years, clinical literature has reported cardiovascular health benefits from mind-body therapies.9⇓–11 Yoga, one type of mind-body therapy, has been increasing in popularity in the USA and in many parts of the world. Yoga, meaning “union” in Sanskrit, incorporates physical, mental, and spiritual elements. In the West, Hatha yoga, one style of yoga, has been most commonly practiced. Hatha yoga consists of a series of physical exercises that focus on stretching and stimulating the spine and muscles in coordination with breath control, thought to stabilize the hypothalamic-pituitary-adrenal axis and sympathoadrenal activity.12⇓–14 According to the 2007 National Health Interview Survey, about 20% of the US population used some form of mind-body practice.15 Another study estimates that about 15 million adults in America report having practiced yoga at least once in their life,16 seeking wellness or treatment for specific health conditions.
A 2005 Cochrane study reviewed the evidence of yoga for secondary prevention of coronary heart disease on mortality, cardiovascular events, hospital admissions, and quality of life and found no randomized controlled trials (RCTs) meeting its inclusion criteria.17 Another review done in 2005 examined CVD clinical endpoints and insulin resistance with observational studies, uncontrolled trials, and nonrandomized controlled trials and found improvements in insulin resistance syndrome with yoga.13 Other reviews have shown yoga to be beneficial in treatment of coronary heart disease, post-myocardial infarction rehabilitation, and hypertension.11,13,18⇓⇓⇓–22 Since this time, several new RCTs have been published. We sought to comprehensively review recent RCT evidence of the effectiveness of yoga on these risk factors and provide a pooled quantitative measure.
Our objectives were (a) to identify and systematically evaluate the evidence on the effectiveness of yoga for modifying risk factors for CVD and metabolic syndrome in adult populations using published systematic reviews, (b) to update the evidence by conducting a systematic review of recent RCTs and (c) to estimate a summary measure of effectiveness by conducting a meta-analysis of the evidence of yoga’s effectiveness versus no-exercise and exercise controls.
Data sources and search terms
The protocol for this review has been published on the PROSPERO website (http://www.crd.york.ac.uk/PROSPERO) with the registration number CRD42013006375. An amendment was added to the protocol including an exercise control group and published in an online revision note. Articles in this review were identified by accessing the following biomedical electronic databases with the assistance of a medical librarian: MEDLINE, CINAHL, Cochrane Central Register of Controlled Trials (CENTRAL), Cochrane Database of Systematic Reviews, EMBASE, and PsycINFO. Using existing published systematic reviews (SRs) as a starting point for gathering evidence, SRs and/or meta-analyses were searched through December 2013. To collect any recent data that may have been missed, we supplemented the search by searching for RCTs published in the last three years through December 2013. Citations were also retrieved by manually searching reference lists of relevant articles. The databases were searched using the keywords “yoga” and “systematic review” for published SRs and “yoga” and “randomized controlled trials” for recent RCTs (see online Supplementary Table S1 for search strategies).
Study selection and inclusion process
Records were pooled from the various databases. Titles and abstracts of SRs that appeared to meet the inclusion criteria were retrieved for further evaluation. Systematic reviews were defined as articles that included an explicit and repeatable literature search method and had explicit and repeatable inclusion and exclusion criteria for studies. RCTs included in the SRs were then retrieved. The process was repeated for the supplementary search of RCTs.
For inclusion in our SR, the studies had to be published in English in a peer-reviewed journal, be conducted in adults (18+ years) who were either healthy, at risk, or with a history of CVD or metabolic syndrome and no other major comorbidities, test an asana- (or posture-) based intervention, and report relevant outcomes. We focused only on SRs that included at least one randomized controlled trial with yoga therapy as a trial arm. No restrictions were placed on style of yoga practiced, frequency, or duration. Articles were excluded if we were unable to isolate the effect of yoga (i.e. yoga was part of a multimodal intervention whose non-yoga components were given to the active intervention group but not to the control group), outcomes reported only psychosocial risk factors or psychological outcomes like stress and anxiety, and the population treated focused on other conditions or comorbidities (e.g. women with breast cancer, populations with renal disease). Two investigators (PC and RG) independently selected studies for inclusion; disagreements were resolved by discussion.
The outcomes of interest were changes in the levels of modifiable risk factors for CVD and metabolic syndrome. Particularly, we were interested in measures of body composition, blood pressure, lipid panel, glycemic control, heart rate, and smoking status. Primary outcomes include body mass index (BMI), systolic blood pressure (SBP), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C). Other outcomes – body weight, diastolic blood pressure (DBP), total cholesterol (TC), triglycerides (TG), fasting blood glucose (FBG), glycosylated hemoglobin (HbA1c), heart rate, and smoking status – were considered secondary outcomes. Outcomes were kept in their natural units.
Data extraction and quality assessment
From each eligible study we extracted the characteristics of the participants, intervention description (type, length of session, frequency), control group description, duration of follow-up, number of patients randomized at baseline and number at follow-up, and effect measures (pre- and post-mean and standard deviations in intervention and control arms, mean change scores and standard deviations if reported). Data from the longest follow-up was extracted. Data extraction was performed by one investigator (PC) and checked for accuracy and completeness by a second reviewer (RG). Any discrepancies were resolved by discussion.
RCTs were appraised using the Cochrane Collaboration’s Risk of Bias (ROB) tool, a commonly used tool to assess risk of bias.23 Trial quality was evaluated by using categories of high, low, or unclear risk in regards to randomization method, allocation concealment, blinding of study personnel and outcomes assessment, attrition, and reporting methods. Two reviewers (PC and RG) independently evaluated RCT quality and resolved any discrepancies by discussion.
Change scores, mean differences (MDs) between treatment arms, and sample sizes reported were on an intention-to-treat basis. MDs were calculated by subtracting the change score in the control group from the change score in the yoga group. Where MDs and standard deviations were not reported, standard deviations were calculated using a conservative correlation coefficient of 0.5 for within-patient correlation from baseline to follow-up. MDs between groups and 95% confidence intervals (CIs) were calculated for each outcome.
The magnitude of heterogeneity was evaluated using the I2 statistic testing the null hypothesis that all studies are evaluating the same effect.24 I2 values of 25%, 50%, and 75% correspond to low, moderate, and high heterogeneity, respectively. Because meta-analysis pools studies that are clinically and methodologically diverse, data on MDs from trials were statistically pooled using a random effects model.25 We also categorized patients into four subgroups based on patient conditions – healthy, with CVD risk factors, with diabetes or metabolic syndrome, and diagnosed with coronary artery disease (CAD) – to depict heterogeneity in the populations included and their response to treatment. Healthy patients are those free of clinical manifestations of any medical or psychiatric illness including clinically significant CVD and diabetes mellitus. Those with CVD risk factors included patients with hypertension, high cholesterol levels, obesity, and current smokers. Diabetes and metabolic syndrome were diagnosed through medical examination or history, and CAD was confirmed through angiography.
Controls were separated into aerobic exercise (physical training, aerobic exercise, cycling, running, brisk walking) and non-aerobic exercise groups. Yoga was compared to these two control groups separately to obtain an estimate of its effectiveness versus active controls and versus non-active controls (details published in protocol amendment). Reference Manager (RevMan) Version 5.2 software from the Cochrane Collaboration was used for data analysis.26
Publication bias was assessed for each of the primary outcomes by visual inspection of funnel plots generated using RevMan software. The MDs were plotted on the x-axis and the standard errors, a measure of study size, on the y-axis. In the absence of bias, the scatterplot should be approximately symmetrical; the more asymmetry, the more bias is present.
We identified 643 studies from the SR search and 761 studies from the RCT search for a total of 1404 records (Figure 1). After removal of duplicates, a total of 880 titles and abstracts were screened. A total of 37 RCTs (24 RCTs from 18 SRs and 13 additional RCTs) met our criteria for inclusion in the review. Although 37 studies met criteria, five studies did not report exact numbers for our primary or secondary outcomes and could not be included in the meta-analysis,27⇓⇓⇓–31 leaving 32 studies for statistical analysis.
Study quality and description of methodology varied amongst the included studies (seeTable 1). Thirteen studies32⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓–43 provided details on the specific randomization method that was used in the RCT and four31,35,37,38 described treatment assignment. Due to the nature of the intervention, all studies had high risk of bias for blinding of participants; however, three studies reported blinding of the personnel, indicating that technicians were blinded to treatment assignment of individuals.34,44,45 Almost all studies except one34 had unclear risk for blinding of outcome assessment. However, there was generally low risk of bias for incomplete reporting of outcomes and selective reporting of outcomes. A summary of study quality can be seen in online Supplementary Figure S1.
Characteristics of the included studies are listed in Table 1. The included studies comprised a total of 2768 participants, with about an equal mix of men (47%) and women (53%). RCTs included adult participants of all ages with an average age of 50 years. Of these participants, 1287 (47%) were assigned to receive the yoga intervention and 1461 (53%) assigned to the control arm. Altogether 1094 (85%) of yoga participants completed the study while 1301 (89%) of control participants made it to follow-up. Duration of studies varied, with follow-up times ranging from 3 weeks to 52 weeks, with a median of 12 weeks. Dividing into subgroups, 38% (14/37) of studies were conducted in healthy populations, 22% (8/32) of studies in populations with CVD risk factors, 27% (10/32) in populations with diabetes or metabolic syndrome, and 13% (5/32) in populations with CAD.
Control arms included usual care or conventional medical therapy (23%), a form of relaxation (6%), education (11%), diet alone (4%), waiting list or no intervention (32%), cognitive-based therapy (2%), and exercise (21%). Five two-arm RCTs,41⇓–43,46,47three three-arm RCTs,27,40,45 and one four-arm RCT48 used exercise as one of the comparator strategies. Exercise controls consisted of physical training, cycling, running, brisk walking, or resistance training.43 One exercise trial27 was excluded from the meta-analysis due to incomplete reporting of effect measures.
Risk factor outcomes
Yoga versus non-exercise controls
Yoga showed significant improvement of risk factors versus non-exercise controls for each of the primary outcomes: BMI (−0.77 kg/m2 (−1.09 to −0.44)), SBP (−5.21 mmHg (−8.01 to −2.42)), LDL–C (−12.14 mg/dl (−21.80 to −2.48)), and HDL-C (3.20 mg/dl (1.86 to 4.54)) (Figure 2). For the secondary outcomes, significant improvement was seen in all risk factors except FBG (−5.91 mg/dl (−16.32 to 4.50)) and HbA1c (−0.06% Hb (−0.43 to 0.31)) (online Supplementary Figure S2). Improvements reported in secondary outcomes include reductions of body weight (−2.35 kg (−4.33 to −0.37)), DBP (−4.98 mmHg (−7.17 to −2.80)), TC (−18.48 (−29.16 to −7.80), TG (−25.89 mg/dl (−36.19 to −15.60)), and heart rate (−5.27 beats/min (−9.55 to −1.00)) (online Supplementary Figure S2).
Only one trial was found which evaluated the impact of yoga on smoking status.32When twice-weekly Vinyasa-style yoga was given in addition to cognitive behavioral therapy (CBT) for smoking cessation, smokers in the intervention group had higher odds of seven-day and 24-hour abstinence compared to a control group receiving CBT and education at the end of the eight-week study period (seven-day quit odds ratio (OR) 4.56 (95% CI 1.12 to 18.57), 24-hour quit OR 4.19 (1.16 to 15.11). These results did not last, however, when abstinence was measured at six-month follow-up (seven-day quit OR 1.54 (0.34 to 6.92), 24-hour quit OR 1.87 (0.43 to 8.16)).
When yoga is used in addition to medication, significant improvement was found in body weight,49 BMI,35,36 blood pressure,20,50 lipid levels,35,38,49,51 FBG,38,52HbA1c,38,52 and heart rate36 in patients with type 2 diabetes or CAD. As a substitute for medical therapy, results are less definitive. Two RCTs found yoga more effective than drug therapy in controlling blood pressure53 and body weight.53,54 In a three-arm trial in which yoga was directly compared to a group that received antihypertensive treatment and a group receiving no treatment in patients at high risk for CVD, yoga reduced SBP almost three times more than the antihypertensive therapy (MD −29.17 mmHg (−37.75, −20.59) and −9.60 mmHg (−18.78, −0.42), respectively).53 When yoga is included in addition to continued medication in CAD patients, an additional benefit, although smaller, is still observed.35,36 Among CAD patients, yoga is less effective as a substitute for medication such as statins and lipid-lowering drugs in lowering LDL-C;55however, as an adjunct treatment to medication, yoga provides an additional statistically significant benefit.35,49
Yoga versus exercise
Five out of nine trials comparing yoga to exercise were conducted in healthy populations27,41,42,46,47,56 and the remaining were conducted in young patient populations with hypertension,40,48 an elderly female population with CAD,43 and a population with type 2 diabetes mellitus.45
Among the outcomes that were reported by more than one study, there was no significant difference in the effectiveness of yoga versus aerobic exercise in modifying body weight (−0.61 kg (−2.70, 1.49)),41,43,47 SBP (−0.64 mmHg (−6.71, 5.43)),40,42,46⇓–48 DBP (−0.14 mmHg (−5.73, 5.44)),40,42,47,48 and heart rate (−1.42 beats/min (−6.11, 3.27))41,42,46,47,56 (Figure 3). In addition, there was also no difference comparing the two strategies for BMI,43 LDL-C,45 HDL-C,45 TC,45 TG,45 or FBG.45
When all studies were pooled together, all trends remained irrespective of controls. MDs in risk factor reductions changed only slightly (online Supplementary Table S2).
Funnel plots assessing publication bias of the primary outcomes are shown in online Supplementary Figure S3. As the funnel plots are mostly symmetrical, we do not find evidence of strong publication bias.
The review shows that the practice of yoga may be beneficial to managing and improving risk factors associated with CVD and metabolic syndrome. This finding, however, should be cautiously interpreted as the RCTs included were of limited sample size, heterogeneous, and had unclear or high risk of bias on several domains. When trials were pooled, all but two of the outcomes examined in this review showed improvement after a yoga intervention when compared to non-exercise controls.
Compared to traditional aerobic exercise controls, there was no significant difference in how exercise or yoga changed risk factors, suggesting similar effectiveness of the two forms of physical activity and possibly similar underlying mechanisms. The mechanism behind the therapeutic effect of yoga for CVD is still unclear; studies have suggested that yoga may modulate autonomic function and beneficially alter markers of sympathetic and parasympathetic activity.12⇓–14 Through practicing yoga, the effects of stress can be reduced, leading to positive impacts on neuroendocrine status, metabolic and cardio-vagal function, and related inflammatory responses.12⇓–14 The similarity in effectiveness on risk factors between the two forms of exercise suggest that there could be comparable working mechanisms, with some possible physiological aerobic benefits occurring with yoga practice, and some stress-reducing, relaxation effect occurring with aerobic exercise.
This review helps strengthen the evidence base for yoga as a potentially effective therapy for cardiovascular and metabolic health. Our results support earlier reviews on the positive benefits of yoga on primary and secondary prevention of CVD and metabolic syndrome.11,13,18⇓–20,22,50,57 Two systematic reviews that were recently published find that there is some evidence for yoga having favorable effects on CVD risk factors.58,59 One review, conducted by the Cochrane Collaboration, included 11 trials with its more restrictive inclusion criteria and found significant improvement in DBP, TG, and HDL.59 The second review, with broad inclusion criteria and a wider list of outcomes, included 44 trials and found that yoga improves SBP, DBP, heart rate, respiratory rate, waist circumference, waist/hip ratio, TC, HDL, very low density lipoprotein, HbA1c, and insulin resistance.58 All studies find that published RCTs on yoga are small, of short duration, and heterogeneous, precluding any strong conclusions on the effectiveness of yoga.
Yoga may provide the same benefits in risk factor reduction as traditional physical activity such as cycling or brisk walking, supporting a previous narrative review.22 This finding is significant as individuals who cannot or prefer not to perform traditional aerobic exercise might still achieve similar benefits in CVD risk reduction. Evidence supports yoga’s accessibility and acceptability to patients with lower physical tolerance like those with pre-existing cardiac conditions, the elderly, or those with musculoskeletal or joint pain.28
Lastly, in addition to CVD risk factor improvements, other benefits may result from practicing yoga. For example, yoga may provide health-related quality of life improvements such as reductions in stress and anxiety and better coping mechanisms distinct from other forms of exercise. Yoga may also be practiced in a variety of settings with no special equipment needed, potentially increasing the frequency and ease of practice. These benefits may produce greater willingness to engage in a form of physical activity and better adherence and sustainability, ultimately facilitating greater long-term individual- and population-level CVD and metabolic risk reductions.
There are potential limitations of this review. First, we included only English language articles and articles published in peer-reviewed journals. Second, several outcomes are related to cardiovascular and metabolic health; we focused on the major risk factors and surrogate markers for these conditions, as they are predictive of CVD risk4 and concrete outcomes such as cardiac death and myocardial infarction were not reported in the RCTs. As with all RCTs, findings are applicable to the patient population in which the study was conducted and wide generalizations should be avoided.
There was a great deal of heterogeneity across included studies. Because part of the appeal and feasibility of yoga is the customizability of the practice to individual practitioners, a wide variety of yoga interventions, frequencies and lengths of practice and follow-up were included. To deal with some of this variation, we used random effects in the meta-analysis and divided patient populations into subgroups. Although I2values did drop within subgroups compared to overall, heterogeneity was still present. As more studies are undertaken and published, further division by yoga tradition, duration of follow-up, and other factors can be performed.
Lastly, study quality and assessment could be improved. Many studies had small sample sizes and did not fully report all methods and outcomes, leading to high or unclear ratings in the risk of bias on several domains. On a related note, although the Cochrane Risk of Bias tool is widely used and applicable, the definitions and structure of the rating system can lead to inaccurate estimation of study quality. For example, blinding of participants is not possible in RCTs, automatically leading to a high bias rating in the ‘performance bias’ domain, which assesses blinding of participants and personnel. Study quality could thus be underestimated in many cases. Nevertheless, more complete reporting of methodology and outcomes by authors can help enhance the usefulness and rigor of the trials.
Future research directions
Despite the growing evidence on the health implications of yoga, the physiological mechanisms behind the observed clinical effects of yoga on cardiovascular risk remains unclear. Inquiries into the minimum effective dose of yoga and the dose-response relationship can help elucidate yoga’s potential as a medical therapy. Research is also still lacking on the costs and economic implications; more research can be done comparing the relative costs and benefits of yoga versus traditional methods like exercise or medication. Yoga has the potential to be a cost-effective treatment and prevention strategy given its low cost, lack of expensive equipment or technology, potential greater adherence, health-related quality of life improvements, and possible accessibility to larger segments of the population.
Our review finds emerging evidence to support a role for yoga in improving common modifiable risk factors of CVD and metabolic syndrome. Whereas previous reviews have looked at a single or a few risk factors, our review updates the existing literature and encompasses numerous CVD and metabolic risk factors that can be used to calculate overall CVD risk. We believe that these findings have important implications for the acceptance of yoga as an effective therapeutic intervention. Given the growing popularity of yoga in the US and around the world, there is a need for larger randomized controlled studies that meet explicit, high quality methodological standards to ascertain the effects of yoga. This review demonstrates the potential of yoga to have an impact on concrete, physiological outcomes that represent some of the greatest health burdens today.
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
The authors would like to acknowledge Carol A Mita at Countway Library at Harvard Medical School for her assistance in developing search strategies.
- Received September 12, 2014.
- Accepted November 14, 2014.