NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Goyal M, Singh S, Sibinga EMS, et al. Meditation Programs for Psychological Stress and Well-Being [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2014 Jan. (Comparative Effectiveness Reviews, No. 124.)

Cover of Meditation Programs for Psychological Stress and Well-Being

Meditation Programs for Psychological Stress and Well-Being [Internet].

Show details

Results

Results of the Search

Figure 3 summarizes the search results. The literature search identified 17,801 unique citations. During the title and abstract screening, we excluded 16,177 citations, during the article screening, we excluded 1,447 citations, and during Key Question (KQ) applicability screening we excluded an additional 136 articles (Appendix D). In total 41 articles met our inclusion criteria and were included in our review.

Figure 3 describes the flow of articles through the screening and review process. 28,881 unique articles were identified in the search, 17,801 were screened out at the title-abstract level, 1624 were screened out at the article level, 177 were screened out at the Key Question Applicability review level, 136 were excluded on full text review and 41 articles were included in the report.

Figure 3

Summary of the literature search. * Total exceeds the number in the exclusion box because reviewers were allowed to mark more than 1 reason for exclusion

Description of Types of Trials Retrieved

Of the included trials, 32 addressed KQ1 (negative and positive affect), one trial addressed KQ2 (attention), 13 trials addressed KQ3 (health-related behaviors affected by stress), and 14 addressed KQ4 (pain and weight). The majority of trials targeted patient populations with mental health or substance abuse problems (n=20). Other population groups under investigation included individuals with breast cancer (n=2), cardiovascular disease (hypertension and congestive heart failure (CHF) (n=4), chronic pain (n=5), human immunodeficiency virus (HIV) (n=2), diabetes and other metabolic disorders (n=3), respiratory diseases such as chronic obstructive pulmonary disorder (COPD), asthma or history of colds (n=3), tinnitus (n=1), and organ transplant recipients (n=1) (Table 4).

Table 4. Characteristics of included trials.

Table 4

Characteristics of included trials.

The interventions included mindfulness-based stress reduction (MBSR) (n=16), mindfulness-based cognitive therapy (MBCT) (n=4), modified MBSR or similar mindfulness training (n=11), transcendental meditation (TM) (n=7), and other mantra meditation (n=3) (Table 4). The trials took place in various countries: U.S. trials (n=28), Non-U.S. trials (n=13) (Table 4).

Since the amount of training and practice in any meditation program may affect its results, we collected this information and found a fair range in the quality of information reported. Not all trials reported on amount of training and home practice recommended. In general, MBSR programs provided 20–27.5 hours of training over 8 weeks. The modified mindfulness trials generally provided about half this level of training (8–13.5 hours of training over 4–8 weeks) as did other mantra programs (7.5–8 hours of training over 5–8 weeks). TM trials generally provided more training (16–39 hours) over longer periods of time (3–12 months) (Tables 5 and 6).

Table 5. Training dose for included trials over duration of training period (numbers are calculated from information provided in trials).

Table 5

Training dose for included trials over duration of training period (numbers are calculated from information provided in trials).

Table 6. Teacher qualifications for included trials.

Table 6

Teacher qualifications for included trials.

Most trials did not describe the specific expertise of the trainers. Only five of the trials reported the trainers' actual meditation experience (ranging between 4 months to 25 years) and six reported the trainers' actual teaching experience (ranging between 0–15.7 years).

We rated 10 trials as low risk of bias, 20 as medium risk of bias, and 11 as high risk of bias (Table 7).

Table 7. Risk of bias for included trials.

Table 7

Risk of bias for included trials.

Key Question Results

Since there were numerous scales for the different measures of affect, as well as subgroups within each affect, we organized the scales to best represent the clinically relevant aspects of each affect. For this review, the comparisons with nonspecific active controls were the most meaningful as they allowed a consistent comparison with a similar control group across all outcomes (efficacy). Comparisons with specific active controls were more difficult to draw conclusions from due to the large heterogeneity of type and strength of control groups (comparative effectiveness). Therefore, our results are presented first for all the comparisons with nonspecific active controls, and then for the specific active controls. We present summary results for all outcomes in Figure 4a (comparisons with nonspecific active controls) and 4b (comparisons with specific active controls) prior to describing each of the sections in detail. Tables 816 give synthesis summaries of all the trials by outcome.

Figure 4a shows summary across measurement domains of comparisons of meditation with non-specific active controls. We show the meta-analytic effect sizes for each outcome which are listed below. Negative numbers favor meditation, and positive numbers favor control. Anxiety mindfulness: -0.40 (confidence interval (CI) -0.71, -0.08) Anxiety mantra: -0.22 (-.49, +.04)Depression mindfulness: -.32 (-.66, +.01) Depression mantra: -.24 (-.70, +.21)Stress/distress mantra: -.26 (-.55, +.04) Negative affect mindfulness: -.34 (-.53, -.14)Positive affect mindfulness: -.31 (-.71, +.09) Quality of life mindfulness: -.28 (-.62, +.06) Sleep mindfulness: -.12 (-.37, +.14) Pain mindfulness: -.33 (-.62, -.03)

Figure 4a

Summary across measurement domains of comparisons of meditation with nonspecific active controls. [See combined legend for Figures 4a and 4b following the figures for further information, including explanations of symbols and definitions of lettered footnotes] (more...)

Figure 4b shows summary across measurement domains of comparisons of meditation with specific active controls. We show the meta-analytic effect sizes for each outcome which are listed below. Negative numbers favor meditation, and positive numbers favor control. Anxiety mindfulness: +0.06 (-.20, +.32) Depression mindfulness: -.16 (-.36, +.03) Stress/distress mindfulness: -.03 (-.23, +.17) Positive affect mindfulness: +.04 (-.24, +.33) Quality of life mindfulness: -.05 (-.24, +.13) Sleep mindfulness: +.14 (-.09, +.37) Pain mindfulness: +.06 (-.13, +.26)

Figure 4b

Summary across measurement domains of comparisons of meditation with specific active controls. [See combined legend for Figures 4a and 4b following the figures for further information, including explanations of symbols and definitions of lettered footnotes] (more...)

Table 8. Synthesis summary for anxiety.

Table 8

Synthesis summary for anxiety.

Table 9. Synthesis summary for depression.

Table 9

Synthesis summary for depression.

Table 10. Synthesis summary for stress/distress.

Table 10

Synthesis summary for stress/distress.

Table 11. Synthesis summary for negative affect.

Table 11

Synthesis summary for negative affect.

Table 12. Synthesis summary for positive affect (well being and positive mood).

Table 12

Synthesis summary for positive affect (well being and positive mood).

Table 13. Synthesis summary for quality of life/mental component of health-related quality of life.

Table 13

Synthesis summary for quality of life/mental component of health-related quality of life.

Table 14. Synthesis summary for substance use, eating, sleep.

Table 14

Synthesis summary for substance use, eating, sleep.

Table 15. Synthesis summary for pain.

Table 15

Synthesis summary for pain.

Table 16. Synthesis summary for weight.

Table 16

Synthesis summary for weight.

Key Question 1. What are the efficacy and harms of meditation programs on negative affect (e.g., anxiety, stress) and positive affect (e.g., well-being) among those with a clinical condition (medical or psychiatric)?

Key Points and Evidence Grades

Comparisons With Nonspecific Active Controls

Anxiety
  • The strength of evidence is moderate that mindfulness meditation programs result in a small improvement in anxiety among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings for a small positive effect, directness of measures, and precise estimates.
  • The strength of evidence is low that mantra meditation programs do not have an effect on anxiety among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings, directness of measures, and imprecise estimates.
Depression
  • The strength of evidence is moderate that mindfulness meditation programs improve symptoms of depression among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings for a positive effect, directness of measures, and precise estimates. However, since one trial is missing from the meta-analysis and the post-intervention I2 is high, this strength of evidence warrants a cautious interpretation.
  • The strength of evidence is insufficient that mantra meditations have an effect on symptoms of depression among cardiac and HIV populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, inconsistent findings, directness of measures, and imprecise estimates.
Stress/Distress
  • The strength of evidence is low that mindfulness meditation programs result in a small improvement in stress and distress among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, inconsistent findings, directness of measures, and precise estimates.
  • The strength of evidence is low that mantra meditation programs have no effect on stress when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings of a null effect, directness of measures, and imprecise estimates.
Negative Affect
  • The strength of evidence is low that mindfulness meditation programs improve negative affect among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent results, indirect measures of negative affect, and precise estimates.
  • The strength of evidence is insufficient that mantra programs have an effect on negative affect among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, inconsistent results, indirect measures of negative affect, and imprecise estimates.
Positive Affect
  • The strength of evidence is insufficient that mindfulness meditation programs have an effect on positive affect when compared with a nonspecific active control. We based this rating on medium risk of bias, consistent findings, indirect measures, and imprecise estimates.
  • The strength of evidence is insufficient about the effects of TM on positive affect when compared with a nonspecific active control. We based this rating on a single low risk-of-bias study, unknown consistency, indirect measures, and imprecise estimates.
Mental Component of Health-Related Quality of Life
  • The strength of evidence is low that mindfulness meditation programs improve the mental component of health-related quality of life (QOL) in various patients as compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings, direct measures, and imprecise estimates.

Comparisons With Specific Active Controls

Anxiety
  • The strength of evidence is insufficient that mindfulness meditation programs have an effect on anxiety among various clinical populations when compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent findings, directness of measures, and imprecise estimates.
  • The strength of evidence is insufficient about the effects of clinically standardized meditation on anxiety in an anxious population when compared with progressive muscle relaxation. We based this rating on a single study with medium risk of bias, unknown consistency, directness of measures, and imprecise estimates.
Depression
  • The strength of evidence is insufficient that mindfulness meditation programs have an effect on depressive symptoms among various clinical populations compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent results, direct measures, and imprecise estimates.
  • The strength of evidence is insufficient that clinically standardized meditation has an effect on depressive symptoms in an anxious population compared with progressive muscle relaxation. We based this rating on a single study with medium risk of bias, unknown consistency, direct measures, and imprecise estimates.
Stress/Distress
  • The strength of evidence is insufficient that mindfulness meditation programs affect distress among those with mood disturbance or symptoms of anxiety compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent results, direct measures, and imprecise estimates.
Positive Affect
  • The strength of evidence is insufficient that mindfulness meditation programs have an effect on positive affect among those with a mood disturbance or symptoms of anxiety when compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent findings, indirect measures, and imprecise estimates.
Mental Component of Health-Related Quality of Life
  • The strength of evidence is insufficient that mindfulness meditation programs have an effect on the mental component of health-related QOL among various clinical populations when compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent findings, direct measures, and imprecise estimates.
Harms
  • Four studies reported on adverse events, but participants experienced no adverse events and 28 studies did not report on adverse events.

Trial Characteristics

We included 32 trials for this KQ, of which 19 took place in the United States. Three trials took place in Canada. Seven trials took place in Europe, including Belgium, the United Kingdom (two trials), Spain, Denmark, Italy, and Germany. The remaining three trials were done in Hong Kong, South Korea, and Iran. Twenty-two of the trials took place in an outpatient setting, two in a university setting, and one in multiple settings; the remaining trials did not report the setting or it was unclear.

Nine trials explicitly reported the time period of recruitment. The year when recruitment started ranged from 1998 to 2010 in these trials. Twenty-five trials reported the trial duration, which ranged from 5 weeks to 9.3 years. All trials reported the length of treatment. The length of additional followup after treatment ranged from none (i.e. treatment assessed at its end) to over 9 years.

Eleven trials excluded patients with past or present substance abuse, 20 trials had exclusion criteria related to psychiatric conditions or treatment, and 20 trials excluded patients according to some medical diagnostic criteria (Appendix E, Evidence Table E2). Most trials (N=18) were of medium risk of bias, five were of high risk of bias, and nine were of low risk of bias.

Population Characteristics

The majority of trials recruited populations with chronic medical conditions, anxiety, or depression. Information was not available for the majority of trials on racial, ethnic, education, or gender composition.

The sample size of the trials ranged from 23–201, with a median sample size of 83. In eight trials the participants were from populations with psychiatric disorders, and in 16 trials the participants were from medical populations, including substance abuse, chronic pain, and fibromyalgia. Of the trials in medical populations, three trials were of subjects with acute or chronic pain or fibromyalgia;69,70,74 seven trials were of subjects with anxiety disorders, anxiety trait, or worry;51,58,60,68,79,83,88 three trials were of subjects with depression;59,71,86 and 13 trials were of subjects with chronic medical conditions, including metabolic syndrome, COPD, HIV, asthma, and CHF.53-55,57,65-67,75,78,81,88-90. Twenty-eight trials provided information on the gender characteristics of the participants. In five trials, the population was 100 percent female.51,53,57,66,70 The mean percentage of female participants in the remaining trials was 56 percent.

Thirty trials provided information on the age distribution of the trial population. The mean age in these trials ranged from 21.8–67.4 years (median=47). Only 16 trials provided information on racial or ethnic characteristics of their trial population. The proportion of white subjects among these populations ranged from 0 percent (in trials of African Americans with CHF) to 99 percent.78 Twenty trials provided information on the level of completed education among trial participants (Appendix E, Evidence Table E3).

Intervention Characteristics

In the intervention arms, 14 trials administered MBSR, four administered MBCT, eight administered a mindfulness variant, four administered TM, and two administered other mantra meditations.

Mindfulness Trials

The mindfulness trials conducted a weekly training session that typically ran for 6–8 weeks. Exceptions include one mindfulness meditation trial that ran for 5 weeks on high worriers,51, another that ran for 12 weeks with stressed employees 73, and one that ran for 10 weeks on alcohol-dependent people.52

Twelve of the 14 MBSR trials provided training that generally ranged from 20–27.5 hours; two trials did not clearly specify training time. Of those two, one used MBSR as a control group for a spirituality intervention; we estimated the maximal training time for that trial at 12 hours.63 All MBSR trials, except two,56,57 noted that they provided homework. Seven MBSR trials specified the amount of homework, which ranged from 24–42 hours over an 8-week period. Eleven of 14 MBSR trials noted that the teachers were trained, two noted they were certified, and three trials noted that their teachers had between 5–15.7 years of teaching experience. Three trials did not report on teacher qualifications. Seven of the MBSR trials used a nonspecific active control and seven used a specific active control.

For the four MBCT trials, the amount of meditation training ranged from 16–24 hours over an 8-week period. All but one of the trials86 recommended home practice, and only two specified the amount, which ranged from 28–37.5 hours over the 8-week period. One reported the teacher was trained, and three reported the teachers were trained and certified. None gave details on amount of meditation or teaching experience. One used a nonspecific active control and three used a specific active control (Table 5).

Among the remaining eight mindfulness-variant trials, the amount of training ranged from 8–13.5 hours over 5–12 weeks. All except one recommended home practice and two trials specified the amount of home practice, which ranged from 17.5–18.5 hours over the training period. Seven of eight trials reported that their teachers were trained, and two noted that the amount of teaching experience ranged from 3–5 years. One trial did not report anything regarding teacher qualifications. Five used a nonspecific active control and three used a specific active control (Table 6).

Mantra Trials

The four TM trials generally had a format generally consistent with TM training.78,81,83,90 There was an initial period of daily training for 1–1.5 hours for about 1 week, followed by periodic checks lasting 30–60 minutes over the followup period. One TM trial did not give enough information to calculate a training amount. All trials recommended daily homework, with the two 6-month trials recommending approximately 90 hours. The TM trials all use trained and certified teachers, although none specified the amount of meditation or teaching experience these teachers had. All four trials used a nonspecific active control.

Two trials used a mantra and were not of the TM tradition. Bormann et al. used mantras representing various spiritual traditions, based on the Easwaran approach.75,90 Lehrer et al. used a clinically standardized meditation program.79 Both trials consisted of no more than 7.5 hours of training over a 5-week period, with instructions to practice at home. Both studies reported that teachers were trained. The teachers for clinically standardized meditation were undergraduate and graduate students who had 4 months of training and had no prior meditation teaching experience.

Outcomes

Comparisons With Nonspecific Active Controls

Anxiety

Seven mindfulness meditation programs and three mantra meditation programs trials examined the effect of the meditation program on anxiety as compared with a nonspecific active control.53,54,57,60,70,72,75,81,83,86,91 The trials included in this analysis used three measures of anxiety. We selected measures that are widely used in trials of anxiety, giving preference to those that most of the other trials in their comparison group used. This was to maintain as much homogeneity in the outcome scale as possible (Appendix E).

One mindfulness meditation program trial found nonsignificant results for its anxiety measure and did not report the data.57

Mindfulness Meditation Programs Versus Nonspecific Active Controls

Seven trials compared mindfulness meditation programs to nonspecific active controls for this outcome, and tended to show a small effect (Table 8, Figure 5). Five were MBSR trials, one was MBCT, and one was a modified version of MBSR. Four trials used the state trait anxiety inventory (STAI), while others used the brief symptom inventory anxiety subscale 18 or Beck anxiety inventory (BAI) scale. The five MBSR trials gave an equivalent amount of training, ranging from 23–27 hours, while the modified mindfulness trial gave 8 hours of training. The trials did not give enough information on the amount of home practice recommended or completed.

Figure 5 shows relative difference between groups in the changes in measures of general anxiety, in the mindfulness versus non-specific active control studies. 7 trials are shown with an overall small but consistent effect favoring a reduction in anxiety symptoms in the mindfulness groups.

Figure 5

Relative difference between groups in the changes in measures of general anxiety, in the mindfulness versus nonspecific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline (more...)

Among the trials that reported scores, a difference-in-change calculation shows that all had a 0.3–44 percent improvement post intervention (8 weeks), and a −2.3 to +6.8 percent improvement at the end of the trial (3–6 months). The trial conducted in Korea showed statistically significant results by the end of treatment, and the results reached statistical significance at the end of the study period for two other trials.

Gross et al. randomized patients with an organ transplant (n=138) to 8 weeks of MBSR or health education arms.54 Anxiety was a primary outcome measure and it saw nonsignificant changes at 8 weeks and 6 months. Schmidt et al. randomized women with fibromyalgia (n=177) to one of three arms: (1) MBSR, (2) a nonspecific active control, or (3) a wait list.70 The anxiety scale was a secondary outcome. The MBSR group showed a statistically significant 4.6 percent decrease in STAI trait score at 4 months (p=0.02) compared with the nonspecific active control. Gaylord et al. randomized women to an MBSR program adapted for individuals with irritable bowel syndrome (IBS) or a nonspecific active control (n=97).53 The MBSR group showed a 6.8 percent change over baseline at 3 months (p=0.02). In a three-arm randomized clinical trial of women with early stage breast cancer, Henderson et al.57 examined the effect of MBSR (n=100). They found no differences in scores of the BAI or the symptom checklist 90 (SCL-90) phobic anxiety scores, and did not report either set of scores.

Lee et al. randomized patients with anxiety disorders (n=46) recruited from a psychiatric hospital or its clinics in South Korea, to either an 8-week mindfulness-based stress management program or nonspecific active control (anxiety disorder-based education).60 It was the only trial to use anxiety patients. The Korean meditation program did not appear to be a direct derivative of MBSR as most other trials in this review are, but shared overlapping features of mindfulness meditation. Outcome measures included both self-report measures (State-Trait Anxiety Inventory, State and Trait subscales; SCL-90 anxiety subscale; and a clinician-rated measure Hamilton psychiatric rating scale for anxiety. The trial standardized all of the self-report measures in Korean. The program provided 8 hours of training targeted towards anxiety reduction, with unspecified amount of home practice. At the end of 8 weeks of treatment, the meditation group showed a significantly greater improvement (p <.05) in all outcome measures compared with the education group, with relatively large effects (15–43 percent overall reduction on the measures compared with the education group). Of note, the trial saw the largest reduction (43 percent) on the clinician-rated Hamilton anxiety rating scale. This trial had a medium risk of bias.

Whitebird et al. randomized patients who were caregivers of family members with dementia (n=78) to either MBSR or education support group. This trial did not specify primary or secondary outcomes, but categorized anxiety as a primary focus of the study. The MBSR group showed no difference in the STAI state scores as compared with the education support group. We rated this trial as medium risk of bias. It provided 25 hours of training over 8 weeks by a trained teacher, with an average of 26.7 hours of homework completed by the participants.

Chiesa et al. randomized patients with major depression (n=18) who failed to achieve remission after at least 8 weeks of antidepressant therapy, to either MBCT or nonspecific active control. The trial found a nonsignificant 44 percent reduction in the BAI, which was a secondary outcome. We rated this trial as medium risk of bias. It provided 16 hours of training over 8 weeks by a trained and certified teacher, with unspecified home practice.

We conducted two meta-analyses, one of post-intervention outcomes at 8 weeks and one of end of study outcomes at 3–6 months (Figures 67). Both showed small and significant effect sizes favoring meditation, generally consistent with the difference-in-change analysis (Figure 5). Since the I2 on the post-intervention meta-analysis was large and significant, we conducted a sensitivity analysis by removing the outlier trial by Lee et al. The effect size dropped to −0.24 (−0.44, −.04) with an I2 of 0 percent (p=.49), and did not change our conclusions. Of note, these effect sizes do not account for the baseline differences and therefore may not be entirely consistent with the difference-in-change graphs.

Figure 6 shows meta-analysis of the effects of meditation programs on anxiety with up to 8 weeks of followup. Mindfulness shows a statistically significant effect size of -.40 compared with nonspecific active controls, and a nonsignificant effect size of +.06 compared with specific active controls. Negative numbers favor meditation, while positive numbers favor the control group.

Figure 6

Meta-analysis of the effects of meditation programs on anxiety with up to 12 weeks of followup. Notes: AC = Active Control; BAI = Beck Anxiety Inventory; BSI = Brief Symptom Inventory; CBGT = Cognitive Behavioral Group Therapy; HE=Health Education; MM (more...)

Figure 7 shows meta-analysis of the effects of meditation programs on anxiety after 3-6 months of followup. Mindfulness shows a statistically significant effect size of -.22 compared with nonspecific active controls and a nonsignificant effect compared with specific active controls. Mantra shows a nonsignificant effect compared with nonspecific active controls.

Figure 7

Meta-analysis of the effects of meditation programs on anxiety after 3–6 months of followup. Notes: AC = Active Control; BSI = Brief Symptom Inventory; CSM = Clinically Standardized; HE = Health Education; MBCT=Mindfulness-based Cognitive Therapy; (more...)

In summary, the Korean meditation trial used an anxious population and showed large effect sizes on all measures of anxiety. The remaining trials used diverse clinical populations; among these, two trials showed small significant effects at 3–4 months. There was general consistency among all three measures of anxiety. All seven trials had a medium risk of bias.

The strength of evidence is moderate that mindfulness meditation programs result in a small improvement in anxiety among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings for a small positive effect, directness of measures, and precise estimates (Table 17).

Table 17. Grade of trials addressing the efficacy of mindfulness meditation program on anxiety compared with nonspecific active controls among various populations.

Table 17

Grade of trials addressing the efficacy of mindfulness meditation program on anxiety compared with nonspecific active controls among various populations.

Mantra Mindfulness Programs Versus Nonspecific Active Controls

Two trials of TM and one trial of another mantra meditation programs evaluated an anxiety outcome (Table 8).

Bormann et al. randomized HIV-infected adults (n=93) to a mantra meditation or an education group. The intervention was 10 weeks with a 22-week followup, and provided 7.5 hours of training and unspecified amount of home practice over 10 weeks.75 At 10 weeks, the difference-in-change score on the STAI trait scale was 6.1 percent favoring the mantra group; however, this was not statistically significant. This difference reduced to 2.1 percent at 22 weeks. This trial had a medium risk of bias. It listed anxiety as one of seven primary outcomes.

Smith et al. randomized university students (n=100) interested in an anxiety reducing technique to either TM or a sham meditation program to match expectations, time, and attention.83 This trial had 59 percent attrition and was also categorized as high risk of bias. The trial did not report on amount of meditation training given but it estimated a maximum home practice of 87.5 hours over 6 months. STAI trait score was a primary outcome, and at 6 months, the difference-in-change scores were not different between the two groups.

Paul-Labrador et al. randomized participants with stable coronary heart disease (n=103) to 16 weeks of either TM or health education.81 The STAI measured anxiety as a secondary outcome. The program provided up to 39 hours of training over 16 weeks with an unspecified amount of home practice. At 16 weeks of followup, the difference-in-change between the two groups was only 2.8 percent favoring the control, and was nonsignificant. This was a well-designed trial with a low risk of bias and relatively large sample size.

Overall, two TM trials had point estimates favoring the null, including one for which anxiety was a primary outcome. The largest and highest quality trial using cardiac patients showed no effect of TM compared with a nonspecific control trial.81 The other mantra trial among HIV patients had similarly null effects on anxiety. The difference-in-change graphs showed consistent results favoring a null effect (Figure 8). The meta-analysis of mantra meditation programs on anxiety was also nonsignificant (Figure 7).

Figure 8 shows relative difference between groups in the changes in measures of general anxiety, in the mantra versus non-specific active control/ specific active control studies. It shows four trials with consistently null results.

Figure 8

Relative difference between groups in the changes in measures of general anxiety, in the mantra versus nonspecific active control/specific active control studies. Relative difference between groups in the change score. This is a relative percent difference, (more...)

The strength of evidence is low that mantra meditation programs do not have an effect on anxiety among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings, directness of measures, and imprecise estimates (Table 18). An evaluation of TM programs only does not change this conclusion.

Table 18. Grade of trials addressing the efficacy of mantra meditation programs on anxiety compared with nonspecific active controls among various populations.

Table 18

Grade of trials addressing the efficacy of mantra meditation programs on anxiety compared with nonspecific active controls among various populations.

Depression
Mindfulness Meditation Programs Versus Nonspecific Active Control

Five trials compared MBSR with a nonspecific active control (Table 9).53,54,57,70,72 All were rated as medium risk of bias and had sample sizes ranging from 75–137. These five trials provided between 23–27 hours of training with unclear amounts of home practice. In addition, three trials compared a modified MBSR program with a nonspecific active control.60,61,66 These three were at medium to low risk of bias, provided 8–9 hours of training with unclear amounts of home practice, and had sample sizes ranging from 19–186. One trial compared MBCT with a nonspecific active control.86 This trial had medium risk of bias, provided 16 hours of training with unclear amounts of home practice. These nine trials included diverse populations. Five trials used the Center for Epidemiologic Studies Depression Scale (CES-D), two used the Symptom Checklist 90 (SCL90) Depression subscale, and one used the Brief Symptom Inventory 18 depression subscale.

Henderson et al. randomized patients with early-stage breast cancer (n=100) to MBSR or a nutrition education program. They used two scales to measure depression. They found nonsignificant results on their main measure of depression, the Beck Depression Inventory (BDI), and did not report values. However, this trial measured numerous outcomes and did not correct for multiple comparisons. A difference-in-change estimate revealed a 49 percent improvement on the SCL-90 depression subscale (p<.05). Gaylord et al. randomized women with IBS (n=75) to MBSR versus a support program for women with IBS and showed no significant difference between trial arms at 2 or 3 months.53 Schmidt et al. randomized women with fibromyalgia (n=109) to MBSR or nonspecific active control. The MBSR arm showed no changes at 8 weeks but showed a 12.4 percent nonsignificant improvement in the CES-D at 4 months compared with the control arm.70 Gross et al. randomized solid organ transplant patients, post-surgery, (n=137) to MBSR versus health education. A difference-in-change calculation showed that MBSR participants had 25.8–31.8 percent reductions in the CES-D that were consistently maintained between 2–12 months. However, these changes did not reach significance (p=0.10).54 Whitebird et al. randomized patients who were caregivers of family members with dementia (n=78) to either MBSR or education support group. This trial did not specify primary or secondary outcomes, but categorized depression as a primary focus of the study. The MBSR group showed a 29.1 and 10.6 percent reduction in CES-D scores at post intervention and 6 months, respectively, (p=.07 for overall reductions) as compared with the education support group. We rated this trial as medium risk of bias. It provided 25 hours of training over 8 weeks by a trained teacher, with an average of 26.7 hours of homework completed by the participants.

Chiesa et al. randomized patients with major depression (n=18) who failed to achieve remission after at least 8 weeks of antidepressant therapy, to either MBCT or nonspecific active control. The trial found a 51.6 percent reduction (p=.04) in the Hamilton rating scale for depression. We rated this trial as medium risk of bias. It provided 16 hours of training over 8 weeks by a trained and certified teacher, with unspecified home practice.

Three trials evaluated other mindfulness programs against a nonspecific active control. Oken et al. randomized people who take care of elderly relatives with dementia (n=19) to mindfulness meditation program or a nonspecific active control.66 This trial found a nonsignificant 10.1 percent improvement on CES-D favoring the mindfulness group. This trial had a medium risk of bias, provided 9 hours of training over 7 weeks by a trained teacher and an unspecified amount of home practice.

Malarkey et al. randomized people, who either had or were at risk for cardiovascular disease due to elevated C-Reactive protein levels (n=186), to mindfulness meditation or nonspecific active control.61 It provided 9 hours of abbreviated MBSR training at work with approximately 18.5 hours of homework over 8 weeks. At 8 weeks, the trial found no differences between the groups, but did not provide data for comparisons of the size of effect. This trial had a low risk of bias.

Lee et al. randomized 46 patients with anxiety disorders recruited from a psychiatric hospital or its clinics in South Korea, to either an 8-week mindfulness-based stress management program or nonspecific active control group (anxiety disorder-based education).60 The Korean meditation program did not appear to be a derivative of MBSR or MBCT as most other trials in this review are, but shared some overlapping features of mindfulness meditation. It found nonsignificant 30.3 percent reduction in the BDI and 17.4 percent reduction in SCL-90 depression scores. The trial standardized all of the self-report measures in Korean. The program provided 8 hours of training targeted towards anxiety reduction, with unspecified amount of home practice. This trial had a medium risk of bias.

In summary, these nine trials used diverse populations of patients, with only one of them overtly depressed. The difference-in-change graphs showed generally consistent findings favoring an improvement in depressive symptoms across studies. Two of the four trials in which depression was a primary outcome showed statistically significant results (Figure 9). The study by Malarkey had nonsignificant results, but also started out with much lower CES-D scores as compared with the other trials. We performed two meta-analyses, one of 2-month outcomes and the other of 3–6 month outcomes. The meta-analyses at 2 months found small and marginally nonsignificant effects of mindfulness meditation programs on depressive symptoms, while the meta-analysis at 3–6 months found small but significant effects (Figures 10 and 11). The 2-month meta-analysis also had a high I2 (p=.012). These meta-analysis do not take into account the baseline differences, while the difference-in-change analysis do take the baseline differences into account.

Figure 9 shows relative difference between groups in the changes in measures of depression, in the mindfulness versus non-specific active control studies. It shows nine trials with consistently small positive effects.

Figure 9

Relative difference between groups in the changes in measures of depression, in the mindfulness versus nonspecific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline (more...)

Figure 10 shows meta-analysis of the effects of meditation programs on depression with up to 3 months of followup. Mindfulness shows an effect size of -.32 (-.66, +.01) compared with nonspecific active controls, and -.16 (-.36, +.03) compared with specific active controls.

Figure 10

Meta-analysis of the effects of meditation programs on depression with up to 3 months of followup. Notes: AC = Active Control; BDI = Beck Depression Inventory; BSI = Beck Stress Inventory; CES-D = Center for Epidemiological Studies Depression Scale; CBGT (more...)

Figure 11 shows meta-analysis of the effects of meditation programs on depression after 3-6 months of followup. Mindfulness shows an effect size of -.23 (-.42, -.05) compared with nonspecific active controls, and a nonsignificant effect compared with specific active controls. Mantra also shows a nonsignficant effect compared with nonspecific active controls.

Figure 11

Meta-analysis of the effects of meditation programs on depression after 3–6 months of followup. Notes: AC = Active Control; BDI = Beck Depression Inventory; BSI = Beck Stress Inventory; CES-D = Center for Epidemiological Studies Depression Scale; (more...)

The strength of evidence is moderate that mindfulness meditation programs improve symptoms of depression among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings for a positive effect, directness of measures, and precise estimates (Table 19). However, since one trial is missing from the meta-analysis and the post-intervention I2 is high, this strength of evidence warrants a cautious interpretation.

Table 19. Grade of trials addressing the efficacy of mindfulness meditation programs on symptoms of depression compared with nonspecific active controls among clinical populations.

Table 19

Grade of trials addressing the efficacy of mindfulness meditation programs on symptoms of depression compared with nonspecific active controls among clinical populations.

Mantra Meditation Programs Versus Nonspecific Active Control

Three trials of TM assessed a depression outcome among cardiac patients. One trial of other mantra assessed depression as an outcome among HIV patients. We rated all three TM trials as low risk of bias and the other mantra trial as medium risk of bias. The TM studies ranged from 22–39 hours of training over 16–25 weeks, although one trial lasted on average 5.4 years with an estimated training time of 78 hours and 1,310 homework hours. The other mantra trial in HIV patients provided 7.5 hours of training over 5 weeks (Table 9).

Paul-Labrador et al. randomized 103 participants with stable coronary heart disease to 16 weeks of either TM or health education.81 The team measured depression as a secondary outcome using the Center for Epidemiologic Studies depression scale (CES-D). They provided up to 39 hours of training over 16 weeks with an unspecified amount of home practice. At 16 weeks of followup, the difference-in-change between the two groups was 19.1 percent favoring the control, and was nonsignificant. This trial had a low risk of bias.

Jayadevappa et al. randomized CHF patients (n=23) to either 3 months of TM or health education and used the CES-D scale to assess depression as a secondary outcome.78 Post-intervention, difference-in-change point estimates were 46.1 and 49 percent at 3 and 6 months respectively. The trial reported these results as nonsignificant. This trial had a low risk of bias, and provided 22.5 hours of training over 6 months by trained and certified teachers. It recommended up to 90 hours of home practice during this time.

Schneider et al. randomized 201 patients with coronary artery disease to either TM or nonspecific active control. The study followed patients on average for 5.4 years. It found a nonsignificant 6.8 percent improvement in the CES-D score compared with control. This trial had a low risk of bias, and provided an estimated 78 hours of training over the study period by trained and certified teachers.

Bormann et al. randomized HIV-infected adults (n=93) to mantra meditation or an education group with primary outcomes related to the reduction of intrusive thoughts and improvement in QOL and well-being.75 The intervention was 10 weeks with a 22-week followup, and provided 7.5 hours of training and unspecified amount of home practice.75 At 10 weeks, the difference-in-change score on the center for epidemiologic studies depression scale was 1.6 percent and was not statistically significant. This difference increased to 20.1 percent at 22 weeks favoring the control (p=.07). This trial had a medium risk of bias. It listed depression as one of seven primary outcomes.

In summary, the difference-in-change graphs showed inconsistent results (Figure 12). All three of the TM trials were low risk of bias, conducted in cardiac patients, and depression was a secondary outcome. Only two of the four trials provided data to conduct a meta-analyses at 4–6 months of followup (Figure 11), showing a small nonsignificant effect size.

Figure 12 shows relative difference between groups in the changes in measures of depression, in the mantra versus non-specific active control / specific active control studies. It shows inconsistent results.

Figure 12

Relative difference between groups in the changes in measures of depression, in the mantra versus nonspecific active control/specific active control studies. Relative difference between groups in the change score. This is a relative percent difference, (more...)

The strength of evidence is insufficient that mantra meditation programs have an effect on symptoms of depression among cardiac and HIV populations when compared with a nonspecific active control. While two of the TM trials did not have data to be included in the meta-analysis, due to conflicting results in the difference-in-change analysis, we do not believe that data would change our conclusions. We based this rating on overall medium risk of bias, inconsistent findings, directness of measures, and imprecise estimates (Table 20).

Table 20. Grade of trials addressing the efficacy of mantra meditation program on symptoms of depression compared with nonspecific active controls among cardiac and HIV populations.

Table 20

Grade of trials addressing the efficacy of mantra meditation program on symptoms of depression compared with nonspecific active controls among cardiac and HIV populations.

Stress and Distress
Mindfulness Meditation Programs Versus Nonspecific Active Control

Eight trials compared mindfulness meditation programs with nonspecific active controls, and evaluated stress or distress as an outcome (Table 10).52,53,61,65,66,72,88,89 Four used MBSR and four used an abbreviated version of MBSR. We rated two as low risk of bias, four as medium risk of bias, and two as high risk of bias. These trials involved diverse patient groups including patients suffering from IBS, lung disease, and HIV, as well as alcoholics and caregivers of family members with dementia. The trial sizes ranged from 19–186. Six trials used a measure of stress and two used a measure of distress.

Oken et al. randomized people who take care of elderly relatives with dementia (n=19) to mindfulness meditation or a nonspecific active control.66 The purpose of this trial was to see if mindfulness meditation would decrease stress in caregivers of relatives with dementia. For inclusion, participants had to endorse greater than 9 points on the perceived stress scale (PSS). Although stress was a primary outcome, the PSS was a secondary measure for this trial. This trial found a nonsignificant 14.1 percent improvement on the PSS favoring the mindfulness meditation group. This trial had a medium risk of bias, provided 9 hours of training over 7 weeks by a trained teacher, and an unspecified amount of home practice.

Garland et al.52 assessed the effects of a modified MBCT for alcoholics versus a nonspecific active control on alcohol dependent adults (n=37) to assess whether mindfulness meditation could disrupt the risk chain of stress-precipitated alcohol relapse. The intervention lasted 10 weeks and did not specify information on the amount of training provided, although participants could have done a maximum of 17.5 hours of home practice over the 10 weeks. This trial had a medium risk of bias and found a statistically significant 21.2 percent reduction in the PSS favoring the mindfulness meditation group (p=.03). This trial studied mostly African American males.

Mularski et al. randomized elderly patients, predominantly men, with moderate to severe chronic obstructive pulmonary disorder (n=49) to MBBT or an active support group.65 It found no difference in perceived stress scores between the two arms of the trial after 2 months. This trial suffered from a 42 percent attrition rate and had a high risk of bias.

Malarkey et al. randomized people, who either had or were at risk for cardiovascular disease due to elevated C-reactive protein levels (n=186), to mindfulness meditation or nonspecific active control.61 It provided 9 hours of abbreviated MBSR training at work with approximately 18.5 hours of homework over 8 weeks. At 8 weeks, the trial found no differences between the groups, but did not provide data for comparisons of the size of effect. This trial had a low risk of bias.

Gaylord et al. randomized women with IBS (n=75) to MBSR versus support program for women with IBS, and showed no significant difference (3.6 percent favoring MBSR) between trial arms at 2 months on the BSI 18.53 At 6 months this had increased slightly to 5.2 percent (p=.049). The trial provided 23 hours of training and unspecified amount of home practice. It had a medium risk-of-bias.

Whitebird et al. randomized patients who were caregivers of family members with dementia (n=78) to either MBSR or education support group. This trial did not specify primary or secondary outcomes, but categorized stress/distress as a primary focus of the study. The MBSR group showed a 19.3 and 12.7 percent reduction in perceived stress scores at post intervention and 6 months respectively (p=.01 for overall reductions), as compared with the education support group. This trial provided 25 hours of training over 8 weeks by a trained teacher, with an average of 26.7 hours of homework completed by the participants. It had a medium risk of bias.

Seyedalinaghi et al. randomized HIV positive patients in Iran to MBSR or nonspecific active control (n=171). The trial did not specify primary or secondary outcomes, but stress/distress was a primary focus of the study. This trial provided approximately 25 hours of training over 8 weeks by trained teaches, and unspecified amount of homework, and had a high risk of bias. The trial found a 11 percent improvement in the SCL-90 revised at the end of the intervention, and a 4.9 percent worsening at 12 months compared with control. The overall effect was significant at p<.001.

Pbert et al. randomized asthmatics to MBSR or education control (n=82), and found 16.2 percent (p=.055) and 26 percent (p=.001) improvement at 10 weeks and 12 months, respectively. The trial provided 26 hours of training over 8 weeks with approximately 24 hours of recommended home practice, and did not provide information about the teachers. It had a low risk of bias.

The difference-in-change graphs generally showed consistent effects on measures of stress and distress favoring a reduction in the mindfulness groups (Figure 13). The effect size calculations included six trials and excluded two (Figure 14). However, we felt an overall meta-analysis of this data would be biased since the largest included trial89 had a high risk of bias and carried nearly 40 percent of the statistical weight, while an even larger trial with null results that had a low risk of bias was excluded.61 Therefore, we did not present an overall effect size. Because the largest (and lowest risk-of-bias) trial by Malarkey et al.61 was inconsistent with the others trials on stress/distress, we rate the overall evidence as inconsistent. In the absence of an overall effect size, we rate the precision of the group of studies as precise due to the majority of trials (5 of 8) finding statistically significant results.

Figure 13 shows relative difference between groups in the changes in measures of stress/distress, in the mindfulness versus non-specific active control studies. It shows a small consistent effect.

Figure 13

Relative difference between groups in the changes in measures of stress/distress, in the mindfulness versus nonspecific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline (more...)

Figure 14 shows meta-analysis of the effects of meditation programs on stress/distress with up to 14 weeks of followup. It shows nonsignificant effects for mindfulness and mantra.

Figure 14

Meta-analysis of the effects of meditation programs on stress/distress with up to 16 weeks of followup. Notes: AC = Active Control; AE = Aerobic Exercise; ASG = Alcohol Dependence Support Group; BSI = Beck Stress Inventory; CBGT = Cognitive Behavioral (more...)

The strength of evidence is low that mindfulness meditation programs result in a small improvement in stress and distress among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, inconsistent findings, directness of measures, and precise estimates (Table 21).

Table 21. Grade of trials assessing the efficacy of mindfulness programs on stress and distress compared with nonspecific active controls among various populations.

Table 21

Grade of trials assessing the efficacy of mindfulness programs on stress and distress compared with nonspecific active controls among various populations.

Mantra Meditation Programs Versus Nonspecific Active Control

Three trials of mantra meditation programs evaluated stress as an outcome for cardiac patients (Table 10). Two were TM and one used another mantra meditation program. Both TM trials studied cardiac patients and both had a low risk of bias. One used the Life Stress Instrument Questionnaire and the other used the PSS. The other mantra meditation trial studied HIV patients and used the PSS.

Paul-Labrador et al. randomized patients with stable coronary heart disease (n=103) to 16 weeks of either TM or health education.81 Stress was a secondary outcome measured by the Life Stress Instrument Questionnaire. The program provided up to 39 hours of training over 16 weeks with an unspecified amount of home practice. At 16 weeks of followup, the difference-in-change between the two groups was 5.9 percent favoring the control, and was nonsignificant. This trial had a low risk of bias.

Jayadevappa et al. randomized CHF patients (n=23) to either 3 months of TM or health education, assessing stress as a secondary outcome using the PSS scale.78 With 100 percent trial completion and a 95 percent compliance rate among the originally randomized subjects, there was no difference in perceived stress scores between the two groups at 3 or 6 months. Difference-in-change point estimates were 0.9 and 1.3 percent at 3 and 6 months, respectively. These were reported as nonsignificant. This trial provided 22.5 hours of training over 6 months by trained and certified teachers and recommended up to 90 hours of home practice during this time. It had a low risk of bias.

Bormann et al. randomized adults with HIV (n=93) to mantra meditation or an education group with primary outcomes related to the reduction of intrusive thoughts and improvement in QOL and well-being.75 The intervention was 10 weeks with a 22-week followup, and provided 7.5 hours of training and unspecified amount of home practice over 10 weeks.75 The difference-in-change score on the PSS was 1.2 and 3 percent at 10 and 22 weeks, respectively, favoring the null, and was not statistically significant. This trial had a medium risk of bias. Stress was one of seven primary outcomes.

The difference-in-change graphs showed consistent findings of a null effect of mantra meditation programs on stress (Figure 15). A meta-analysis of two of the trials suggested a small nonsignificant effect (Figure 14).

Figure 15 shows relative difference between groups in the changes in measures of stress, in the mantra versus non-specific active control studies. It shows consistently null results for three trials.

Figure 15

Relative difference between groups in the changes in measures of stress, in the mantra versus nonspecific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline mean in (more...)

The strength of evidence is low that mantra meditation programs have no effect on stress when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings of a null effect, directness of measures, and imprecise estimates (Table 22).

Table 22. Grade of trials addressing the efficacy of mantra meditation programs on stress compared with nonspecific active controls among cardiac and HIV patients.

Table 22

Grade of trials addressing the efficacy of mantra meditation programs on stress compared with nonspecific active controls among cardiac and HIV patients.

Negative Affect
Mindfulness Meditation Programs Versus Nonspecific Active Control

Thirteen trials compared mindfulness meditation programs with nonspecific active controls, and evaluated a negative affect outcome (Table 11). Since some trials reported on more than one outcome, for these trials we prioritized anxiety over depression and depression over stress/distress as indirect measures of negative affect. None of the trials used a direct measure of negative affect. Seven trials reported on anxiety, two on depression, and four on stress/distress. The trials included diverse populations, ranging in sample size from 18–186. Two trials had a low risk of bias, nine had a medium risk of bias, and two had a high risk of bias. For five of the trials the outcome was a primary outcome. We previously described these trials, and displayed them in graphical form in Figure 16. The difference-in-change graphs showed a consistent improvement in negative affect when we compared mindfulness meditation programs to a nonspecific active control. Two trials showed small nonsignificant effects, which became significant at the end of study, and four trials showed significant effects post-intervention. A meta-analysis of these trials showed a small statistically significant effect size of 0.34 favoring meditation (Figure 17). We conducted a sensitivity analysis reversing our prioritization order, prioritizing stress/distress over depression and depression over anxiety, to see if this would change our conclusions (Figures 1819). Both analyses gave similar results.

Figure 16 shows Relative difference between groups in the changes in measures of negative affect, in the mindfulness versus non-specific active control studies. It shows small consistently positive results.

Figure 16

Relative difference between groups in the changes in negative affect, in the mindfulness versus nonspecific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline mean (more...)

Figure 17 shows Meta-analysis of the effects of meditation programs on negative affect. Shows a statistically significant effect size of -.34.

Figure 17

Meta-analysis of the effects of meditation programs on negative affect—main analysis (mindfulness meditation versus nonspecific active control interventions). Notes: AC = Active Control; ASG = Alcohol Dependence Support Group; BAI=Beck Anxiety (more...)

Figure 18 shows relative difference between groups in the changes in measures of negative affect, in the mindfulness versus non-specific active control studies (sensitivity analysis). This shows a consistent effect favoring mindfulness.

Figure 18

Relative difference between groups in the changes in measures of negative affect, in the mindfulness versus nonspecific active control studies (sensitivity analysis). Relative difference between groups in the change score. This is a relative percent (more...)

Figure 19 shows Meta-analysis of the effects of meditation programs on negative affect - sensitivity analysis (Mindfulness meditation versus non-specific active control interventions). This shows a statistically significant effect size of -.31.

Figure 19

Meta-analysis of the effects of meditation programs on negative affect—sensitivity analysis (mindfulness meditation versus nonspecific active control interventions). Notes: AC = Active Control; ASG = Alcohol Dependence Support Group; BSI = Beck (more...)

The strength of evidence is low that mindfulness meditation program improve negative affect among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent results, indirect measures of negative affect, and precise estimates (Table 23).

Table 23. Grade of trials addressing the efficacy of mindfulness meditation programs on negative affect compared with nonspecific active controls among diverse populations.

Table 23

Grade of trials addressing the efficacy of mindfulness meditation programs on negative affect compared with nonspecific active controls among diverse populations.

Mantra Meditation Programs Versus Nonspecific Active Control

Five trials compared mantra meditation programs with nonspecific active controls, and evaluated a negative affect outcome (Table 11). Four were TM trials and one was other mantra meditation program. Three trials reported on anxiety and two on depression. The difference-in-change graphs show inconsistent results (Figure 20). We conducted a sensitivity analysis reversing the order of prioritization, prioritizing stress/distress over depression and depression over anxiety, to see if this would change our conclusions. The difference-in-change graph now showed consistently null results (Figure 21). A meta-analysis of the main outcomes for negative affect among mantra studies only replicated the anxiety meta-analysis (Figure 7) due to missing data on two of the trials that had a depression outcome. The meta-analysis of the sensitivity analysis showed a small nonsignificant overall effect (Figure 22).

Figure 20 shows Relative difference between groups in the changes in measures of negative affect, in the mantra versus non-specific active control studies. This shows inconsistent results.

Figure 20

Relative difference between groups in the changes in measures of negative affect, in the mantra versus nonspecific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline (more...)

Figure 21 shows Relative difference between groups in the changes in measures of negative affect, in the mantra versus non-specific active control studies (sensitivity analysis). This shows inconsistent results.

Figure 21

Relative difference between groups in the changes in measures of negative affect, in the mantra versus nonspecific active control studies (sensitivity analysis). Relative difference between groups in the change score. This is a relative percent difference, (more...)

Figure 22 shows Meta-analysis of the effects of mantra meditation programs on negative affect - sensitivity analysis (mantra versus non-specific active control interventions). This shows an effect size of -.26 (-.53, +.01)

Figure 22

Meta-analysis of the effects of mantra meditation programs on negative affect—sensitivity analysis (mantra vs. nonspecific active control interventions). Note: AC = Active Control; HE=Health Education; GA = General Anxiety; mos = months; PSS = (more...)

The strength of evidence is insufficient that mantra programs have an effect on negative affect among various clinical populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, inconsistent results, indirect measures of negative affect, and imprecise estimates (Table 24).

Table 24. Grade of trials addressing the efficacy of mantra meditation programs on negative affect compared with nonspecific active controls among diverse populations.

Table 24

Grade of trials addressing the efficacy of mantra meditation programs on negative affect compared with nonspecific active controls among diverse populations.

Positive Affect
Mindfulness Meditation Programs Versus Nonspecific Active Control

Three trials compared mindfulness meditation programs with nonspecific active controls, and evaluated positive affect as an outcome. They used differing populations, included a range of 18–137 patients, and were all of medium risk of bias (Table 12).

Henderson et al. randomized women with early-stage breast cancer (n=100) to MBSR or nonspecific active control.57 The study used the Sense of Coherence Meaningfulness Subscale to measure subjective well-being as a secondary endpoint. At 4 months there was a statistically significant 6.8 percent improvement in mean Sense of Coherence Meaningfulness Subscale scores in the MBSR group as compared with the control group (p <0.05). However, this trial measured numerous outcomes and did not make any corrections for multiple comparisons. This trial had a medium risk of bias, provided 25 hours of training over 8 weeks, and did not specify whether it recommended home practice or not.

Gross et al. randomized solid organ transplant patients, post-surgery, (n=137) to MBSR versus health education.54 The study used the Short Form-36 (SF-36) vitality score to measure improvement in positive mood as a secondary outcome. There were no differences between the groups at end of treatment. This trial provided 27 hours of training by a trained teacher, and unspecified amount of home practice over 8 weeks.

Chiesa et al. randomized patients with major depression (n=18) who failed to achieve remission after at least 8 weeks of antidepressant therapy, to either MBCT or nonspecific active control. The trial found a 54.6 percent reduction (p=.05) in the Psychological General Well-being Index. This trial had a medium risk of bias. It provided 16 hours of training over 8 weeks by a trained and certified teacher and had unspecified home practice.

Overall, the difference-in-change graphs show a small consistent effect of the mindfulness meditation programs on positive mood with one trial showing a small significant effect that diminishes with time, and another trial showing a large significant effect (Figures 2324).

Figure 23 shows Relative difference between groups in the changes in measures of positive affect, in the mindfulness versus non-specific active control / specific active control studies. This shows a consistent small effect compared with nonspecific active controls, and inconsistent results compared with specific active controls.

Figure 23

Relative difference between groups in the changes in measures of positive affect, in the mindfulness versus nonspecific active control/specific active control studies. Relative difference between groups in the change score. This is a relative percent (more...)

Figure 24 shows Meta-analysis of the effects of meditation programs on positive affect with up to 4 months of followup. This shows nonsignificant results compared with nonspecific and compared with specific active controls.

Figure 24

Meta-analysis of the effects of meditation programs on positive affect with up to 4 months of followup. Notes: AE = Aerobic Exercise; HE = Health Education; HLC = Healthy Living Course; HAM-D = Hamilton Psychiatric Rating Scale for depression; MBSR = (more...)

The strength of evidence is insufficient that mindfulness meditation program have an effect on positive affect when compared with a nonspecific active control. We based this rating on medium risk of bias, consistent findings, indirect measures, and imprecise estimates (Table 25).

Table 25. Grade of trials addressing the efficacy of mindfulness meditation programs on positive affect compared with nonspecific active controls among organ transplant recipients and breast cancer patients.

Table 25

Grade of trials addressing the efficacy of mindfulness meditation programs on positive affect compared with nonspecific active controls among organ transplant recipients and breast cancer patients.

Transcendental Meditation Versus Nonspecific Active Control

Jayadevappa et al. randomized CHF patients (n=23) to either 3 months of TM or health education, assessing positive mood as a secondary outcome using the SF-36 vitality subscale.78 With 100 percent trial completion and a 95 percent compliance rate among the originally randomized subjects, this trial found no differences at 3 and 6 months (Figure 25). This trial had a low risk of bias, and provided 22.5 hours of training over 6 months by trained and certified teachers. It recommended up to 90 hours of home practice during this time (Table 12).

Figure 25 shows Relative difference between groups in the changes in measures of positive affect, in the mantra versus non-specific active control studies. It shows null results for one trial by Jayadevappa.

Figure 25

Relative difference between groups in the changes in measures of positive affect, in the mantra versus nonspecific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline (more...)

The strength of evidence is insufficient about the effects of TM on positive affect when compared with a nonspecific active control. We based this rating on a single low risk-of-bias study, unknown consistency, indirect measures, and imprecise estimates (Table 26).

Table 26. Grade of trials addressing the efficacy of transcendental meditation on positive affect compared with nonspecific active controls among cardiac patients.

Table 26

Grade of trials addressing the efficacy of transcendental meditation on positive affect compared with nonspecific active controls among cardiac patients.

Mental Component of Health-Related Quality of Life
Mindfulness Meditation Programs Versus Nonspecific Active Control

Pbert et al. randomized asthmatics to MBSR or education control (n=82), and specified asthma QOL as a primary outcome. It found a 6.2 percent (ns) and 26 percent (p=.002) improvement at 10 weeks and 12 months, respectively, in the emotional function domain of asthma quality of life. The trial provided 26 hours of training over 8 weeks with approximately 24 hours of recommended home practice, and had a low risk of bias. There was no information about the teachers (Table 13).

Whitebird et al. randomized patients who were caregivers of family members with dementia (n=78) to either MBSR or education support group. This trial did not specify primary or secondary outcomes, but the short form-12 (SF-12) mental component score was categorized as a primary focus of the study. The MBSR group showed a 28.4 and 24.3 percent reduction in perceived stress scores post-intervention and 6 months, respectively, (p<.001 for overall reductions) as compared with the education support group. This trial had a medium risk of bias, provided 25 hours of training over 8 weeks by a trained teacher, and had an average of 26.7 hours of homework completed by the participants.

Gross et al. randomized solid organ transplant patients, post-surgery, (n=137) to MBSR versus health education.54 The trial used the SF-12 mental component score to measure improvement in the mental component of health-related QOL as a secondary outcome. There were no differences between the groups at end of treatment (p=.29). This trial provided 27 hours of training by a trained teacher, and had an unspecified amount of home practice over 8 weeks. This trial had medium risk of bias.

Mularski et al. randomized elderly patients, predominantly men, with moderate to severe chronic obstructive pulmonary disorder (n=49) to a mindfulness-based breathing therapy or a support group.65 The trial used the Veterans Rand-36 to measure QOL as a secondary outcome. There was a nonsignificant 8.3 percent improvement in the Veterans Rand-36 scores in the MBBT group after 2 months. This trial suffered from a 42 percent attrition rate and had a high risk of bias.

The difference-in-change graphs suggested a small improvement for mindfulness meditation programs in the mental component of QOL when compared with nonspecific active controls (right side of Figure 26). The meta-analysis suggests a small nonsignificant effect (Figure 30)

Figure 26 shows Relative difference between groups in the changes in measures of studies mental component of health-related quality of life, in the mindfulness versus non-specific active control / specific active control studies. This shows small consistently positive results compared with nonspecific active controls, and inconsistent results compared with specific active controls.

Figure 26

Relative difference between groups in the changes in measures of mental component of health-related quality of life, in the mindfulness versus nonspecific active control/specific active control studies. Relative difference between groups in the change (more...)

Figure 30 Shows meta-analysis of the effects of meditation programs on the mental health component of health-related quality of life with up to 3 months of followup. This shows nonsignificant results compared with both nonspecific and specific active controls.

Figure 30

Meta-analysis of the effects of meditation programs on the mental health component of health-related quality of life with up to 3 months of followup. Notes: HE = Health Education; HLC = Healthy Living Course; MBSR = Mindfulness Based Stress Reduction; (more...)

The strength of evidence is low that mindfulness meditation programs improve the mental component of health-related QOL in various patients as compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent findings, direct measures, and imprecise estimates (Table 27).

Table 27. Grade of trials addressing the efficacy of mindfulness meditation programs on the mental component of health-related quality of life compared with nonspecific active controls among various patients.

Table 27

Grade of trials addressing the efficacy of mindfulness meditation programs on the mental component of health-related quality of life compared with nonspecific active controls among various patients.

Comparisons With Specific Active Control

Anxiety
Mindfulness Meditation Programs Versus Specific Active Control

Nine trials evaluated a mindfulness meditation program against a specific active control for the outcome of anxiety (Table 8). Six trials used MBSR, one used MBCT, and two used mindfulness meditation. The control groups were heterogeneous including medications, spirituality interventions, exercise, and group therapies. Sample sizes ranged from 25–110. Two trials had a high risk of bias, five had a medium risk of bias, and two had a low risk of bias.

Wong et al.74 randomized Chinese-speaking participants with chronic pain (n=99) to an 8-week MBSR program or a multidisciplinary pain intervention. The trial saw nonsignificant changes at 2 and 6 months post-intervention in the STAI state and trait scores. The profile of mood states (POMS) tension difference-in-change score showed the greatest change (11.5 percent) favoring MBSR, but was also nonsignificant.

Gross et al. randomized adults with primary chronic insomnia (n=30) to an 8-week MBSR program or an 8-week course of pharmacotherapy with eszopiclone.55 At 2 and 5 months post-intervention, there were no significant changes in STAI state or trait scores in either group, but the directionality of difference-in-change point estimates favored the MBSR group.

Moritz et al. randomized people with mood disorders (n=165) recruited from primary care clinics to 8 weeks of either MBSR or an 8-week audio taped spirituality home trial program.63 This trial evaluated the superiority of a spirituality program to MBSR, as opposed to other trials, using a comparative effectiveness design. MBSR was used as the control. They utilized a POMS score of 40 or greater as inclusion criteria, indicating a moderate degree of mood disturbance, and as a main outcome measure. Although groups appeared matched for amount of training (12 hours over 8 weeks), the spirituality group received up to 42 hours of home practice over that time and it is unclear whether the MBSR group received the same. At 8 weeks, the difference in the MBSR group from baseline was 39 percent lower than that in the spirituality group (p=0.007).

Koszycki et al.58 randomized patients with generalized social anxiety disorder (n=53) to an 8-week course of MBSR or a 12-week course of group cognitive behavior therapy. MBSR received a maximum of 27.5 hours of training and a maximum of 28 hours of home practice over 8 weeks. Outcome measures included four scales of social anxiety, which favored group cognitive behavior therapy over MBSR: Liebowitz social anxiety-fear scale (p=.09), social anxiety-avoidance scale (p=.009), social phobia scale (p=.006), and social interaction scale (p=.057). Although the groups cognitive behavior therapy group ran for 4 weeks longer than MBSR, the total dose was similar (27.5 hours of training for MBSR vs. 30 hours for group cognitive behavior therapy). It remains unclear if it was the effect of the training over a longer period of time in the group cognitive behavior therapy arm that accounted for the differences. The analysis appeared to compare post-treatment scores only, and it was unclear whether they accounted for baseline differences in the analysis, given that there were large baseline differences between the groups.

Barrett et al.85 randomized patients with a history of upper respiratory infections to MBSR or exercise (n=98). The trial provided about 20 hours of training by trained teachers, and approximately 42 hours of recommended homework over the 8-week training period. The STAI state score was a secondary outcome. The trial found no significant differences between the two arms.

Jazaieri et al.87 randomized patients with social anxiety disorder to MBSR or exercise (n=56). The trial provided about 25 hours of training by trained teachers, and their participants performed an average of 28.3 hours of homework over the 8-week training period. Although they did not specify primary or secondary outcomes, the study characterized the Liebowitz social anxiety scale as a primary outcome since it was a primary focus of the study. The trial found a nonsignificant improvement of 6.2 percent, which worsened over time in the MBSR group as compared with exercise. This trial had a high risk of bias.

Philippot et al. randomized patients with tinnitus (n=30) to a 6-week modified MBCT program or progressive muscular relaxation training.67 This trial used the STAI (unspecified) and found no statistically significant differences between-groups. It provided 13.5 hours of training and an unspecified amount of home practice. We rated it as medium risk of bias.

Delgado et al. randomized worriers (n=36) to 5 weeks of mindfulness meditation or progressive muscular relaxation, providing 10 hours of training and unspecified amount of home practice.51 They found no significant differences in the STAI trait score, and had a medium risk of bias. Piet et al. randomized 26 patients with social phobia to MBCT or group cognitive behavior therapy.68 They provided 16 hours of training and up to 28 hours of home practice over an 8-week period. This trial found no difference between the groups on the BAI However, the cognitive behavior therapy group was provided nearly double the amount of group training, 28 hours over 14 weeks, and this increased time and attention in the control group may not allow appropriate comparisons between the groups. This trial had a medium risk of bias.

The difference-in-change graphs showed inconsistent results (Figure 27). A meta-analysis of these trials showed nonsignificant effects around the null at end of treatment and end of study time points (Figures 6 and 7).

Figure 27 shows Relative difference between groups in the changes in measures of general anxiety, in the mindfulness versus specific active control studies. This shows inconsistent results.

Figure 27

Relative difference between groups in the changes in measures of general anxiety, in the mindfulness versus specific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline (more...)

The strength of evidence is insufficient that mindfulness meditation programs have an effect on anxiety among various clinical populations when compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent findings, directness of measures, and imprecise estimates (Table 28).

Table 28. Grade of trials addressing the efficacy of mindfulness meditation programs on anxiety compared with specific active controls among diverse populations.

Table 28

Grade of trials addressing the efficacy of mindfulness meditation programs on anxiety compared with specific active controls among diverse populations.

Other Mantra Meditation Versus Specific Active Control

Lehrer et al. assigned anxious participants to clinically standardized meditation (n=23) or progressive muscular relaxation (n=19).79 The program provided 7.5 hours of training and unspecified amount of home practice over 5 weeks (Table 8). Undergraduate and graduate students, with 4 months of training in the technique and no prior teaching experience, provided the training. Results on all four anxiety measures favored the progressive muscular relaxation group over the clinically standardized meditation group. For measures it used institute for personality and ability testing anxiety inventory, symptom checklist 90 anxiety subscale, and state trait anxiety index state and trait scales. At 6 weeks the differences were all nonsignificant, but ranged from 6–21 percent favoring the progressive muscular relaxation group (Figure 8).

The strength of evidence is insufficient about the effects of clinically standardized meditation on anxiety in an anxious population when compared with progressive muscular relaxation. We based this rating on a single study with medium risk of bias, unknown consistency, directness of measures, and imprecise estimates (Table 29).

Table 29. Grade of trials addressing the efficacy of clinically standardized meditation programs on anxiety compared with progressive muscle relaxation among anxious participants.

Table 29

Grade of trials addressing the efficacy of clinically standardized meditation programs on anxiety compared with progressive muscle relaxation among anxious participants.

Depression
Mindfulness Meditation Programs Versus Specific Active Control

Eleven trials evaluated a mindfulness meditation programs against a specific active control for the outcome of depression (Table 9). Five trials compared MBSR to various specific active controls in diverse populations. Four trials compared MBCT to either antidepressant among depressed patients, cognitive behavior therapy among anxious patients, or progressive muscle relaxation among those suffering from tinnitus. One trial compared a mindfulness meditation program to progressive muscular relaxation and one trial compared a mindfulness meditation program to viniyoga. Four trials had a low risk of bias, five had a medium risk of bias, and two had a high risk of bias. Sample sizes ranged from 25–186.

Wong et al. randomized patients with chronic pain (n=99) in Hong Kong to MBSR or a multidisciplinary pain intervention.74 The study used two scales to assess depression. It found a nonsignificant 10.7 percent improvement on the POMS-depression at 2 months, which maintained to 6 months. However, it found no difference in the Center for Epidemiologic Studies depression scale at 2 or 6 months. This trial had a low risk of bias, provided 27 hours of training, and an unspecified amount of home practice over 8 weeks. Its teachers were trained and had 5 years of experience teaching meditation.74

Gross et al. randomized people with insomnia (n=27) to MBSR or eszopiclone.55 They found a 25.4 percent change in Center for Epidemiologic Studies depression scale favoring the drug at the end of 2 months, which increased to 42.2 percent at 5 months. Although these appeared to be large effects, the study reported the differences as not significant. This trial provided 26 hours of training and up to 36 hours of home practice over 8 weeks.

Koszyki et al. randomized patients with social anxiety disorder (n=53) to MBSR or group cognitive behavior therapy. The trial had a high risk of bias. They found a nonsignificant 5.3 percent difference favoring the cognitive behavior therapy group on the BDI II.58

Moritz et al. randomized patients with mood disorders (n=110) to a spirituality program versus MBSR.63 In this trial, MBSR was the active control for the spirituality intervention. The spirituality intervention included a meditative component. It provided about 12 hours of training in both interventions over an 8-week period, with unspecified amount of home practice in the MBSR group. It provided up to 42 hours of home practice in the spirituality group. There was no information on teacher qualifications for MBSR. There was a significant 31.7 percent improvement on the POMS-depression scale in the spirituality program as compared with MBSR (p<0.013). This trial had a low risk of bias.

Jazaieri et al.87 randomized patients with social anxiety disorder to MBSR or exercise (n=56). The trial provided about 25 hours of training by trained teachers, and their participants performed an average of 28.3 hours of homework over the 8-week training period. Although they did not specify primary or secondary outcomes, the study identified depression as a primary focus of the study. The trial found nonsignificant improvements of 22.8 and 14.2 percent at 8 weeks and 5 months, respectively, in the MBSR group as compared with exercise. The trial had a high risk of bias.

Segal et al. randomized depressed patients in acute remission to MBCT with tapering of antidepressant or maintenance antidepressant medication (n=53) to assess depression relapse. Relapse rates by 600 days were 46 percent for the antidepressant group and 38 percent for MBCT. This absolute 8 percent difference did not reach statistical significance. This trial had a low risk of bias. It provided 23 hours of training by trained and certified teachers, and recommended an unspecified amount of home practice.71

Kuyken et al. randomized patients with recurrent depression (n=123) who were in full or partial remission to either maintenance anti-depressant medication or MBCT with support to taper medication.59 After 15 months, 60 percent of the antidepressant group had relapsed as compared with 47 percent in the MBCT group. This 13 percent absolute difference did not reach statistical significance. They also measured the Hamilton depression rating scores, which were 31.7 percent lower in the MBCT group at 3 months and 26.7 percent lower at 15 months (p=.02). On a third measure, the BDI II, the MBCT group showed a 14.6 percent reduction at 3 months and 15 percent reduction at 15 months compared with the antidepressant group. These differences did not reach statistical significance. Of note, 75 percent of the MBCT had discontinued their antidepressant by 6 months. This was a low risk-of-bias trial. It provided 24 hours of training and recommended up to 37.5 hours of home practice over an 8-week period. The teachers were trained and certified.

Piet et al. randomized young adults with social phobia (n=26) to either MBCT or group cognitive behavioral therapy in a crossover design with participants receiving both treatments.68 We evaluated comparisons after the first intervention period only, before any crossover. They provided 16 hours of training and up to 28 hours of home practice over an 8-week period. This trial found a 24.3 percent nonsignificant change favoring the cognitive behavioral therapy group on the BDI-II. However, the cognitive behavioral therapy group received nearly doubles the amount of group training, 28 hours over 14 weeks, and this increased time and attention in the control group may not allow equivalent comparisons between the groups. This trial had a medium risk of bias.

Philippot et al. randomized patients with tinnitus (n=25) to a 6-week modified MBCT program or progressive relaxation training.67 This trial used the BDI and found an insignificant 8.7 percent differences between groups at 6 weeks favoring mindfulness meditation. At 18 weeks this effect disappeared. This trial had a medium risk of bias and provided 13.5 hours of training with an unspecified amount of home practice.

Delgado et al. randomized female university students (n=32) who were worriers to 5 weeks of mindfulness meditation or progressive muscular relaxation, providing 10 hours of training and unspecified amount of home practice.51 The study found a nonsignificant 13.3 percent improvement in the BDI in the mindfulness meditation group as compared with progressive muscular relaxation. This trial had a medium risk of bias.

Wolever et al. randomized stressed employees (n=186) to a mindfulness-at-work program or viniyoga for 12 weeks. Participants received 14 hours of training by trained teachers and unspecified amount of homework. Depression was a secondary outcome. The trial found a nonsignificant 8.5 percent improvement in the mindfulness group compared with control. This trial had a medium risk of bias.

The difference-in-change graphs show significant inconsistency (Figure 28). Two meta-analyses of results at the end of treatment and end of study show small nonsignificant effects slightly favoring meditation (Figures 10 and 11).

Figure 28 shows Relative difference between groups in the changes in measures of depression, in the mindfulness versus specific active control studies. This shows inconsistent results.

Figure 28

Relative difference between groups in the changes in measures of depression, in the mindfulness versus specific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline (more...)

The strength of evidence is insufficient that mindfulness meditation programs have an effect on depressive symptoms among various clinical populations compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent results, direct measures, and imprecise estimates (Table 30).

Table 30. Grade of trials addressing the efficacy of mindfulness meditation programs on depressive symptoms compared with specific active controls among diverse populations.

Table 30

Grade of trials addressing the efficacy of mindfulness meditation programs on depressive symptoms compared with specific active controls among diverse populations.

Other Mantra Meditation Versus Specific Active Control

Lehrer et al. assigned anxious participants to clinically standardized meditation or progressive muscular relaxation (n=42).79 The program provided 7.5 hours of training and an unspecified amount of home practice over 5 weeks (Table 9). The trainers were undergraduate and graduate students with 4 months of training in the technique and no prior teaching experience. symptom checklist-90 depression scores favored the progressive muscular relaxation group over the clinically standardized meditation group. The difference-in-change scores were all nonsignificant, but ranged from 27.8 percent at 6 weeks to 7.8 percent at 6 months favoring the progressive muscular relaxation group (Figure 12).

The strength of evidence is insufficient that clinically standardized meditation has an effect on depressive symptoms in an anxious population compared with progressive muscular relaxation. We based this rating on a single study with medium risk of bias, unknown consistency, direct measures, and imprecise estimates (Table 31).

Table 31. Grade of trials addressing the efficacy of clinically standardized meditation programs on depression compared with progressive muscle relaxation among anxious participants.

Table 31

Grade of trials addressing the efficacy of clinically standardized meditation programs on depression compared with progressive muscle relaxation among anxious participants.

Stress and Distress
Mindfulness Meditation Programs Versus Specific Active Control

Six mindfulness trials evaluated stress/distress as an outcome among populations with some form of emotional distress (Table 10). Delgado et al. randomized female university students (n=32) who had high scores on the Penn State worry questionnaire to 5 weeks of mindfulness meditation or progressive muscular relaxation, providing 10 hours of training and unspecified amount of home practice.51 Scores on the positive and negative affect scale-negative mood were a primary focus of the trial, and were relatively unchanged at 5 weeks of intervention, and there was no difference between the two groups at the end of treatment. This trial had a medium risk of bias.

Moritz et al. randomized patients with mood disorders (n=110) to a spirituality program versus MBSR.63 In this trial, MBSR was the active control. It provided about 12 hours of training in both interventions over an 8-week period. It provided up to 42 hours of home practice in the spirituality group and an unspecified amount of home practice in the MBSR group. There was no information on teacher qualifications for MBSR. This trial used two scales that assessed distress, which was a primary outcome for the trial. They found a 23.8 percent change favoring spirituality at 8 weeks (p=.034) on the POMS total mood disturbance score, and a 22.4 percent change favoring spirituality at 8 weeks (p=.0.34) on the SF-36 mental health subscale score. This trial had a low risk of bias. It is notable that this intervention included a meditative component, as well as breathing exercises that may resemble features of MBSR.

Piet et al. randomized young adults with social phobia (n=26) to MBCT or group cognitive behavioral therapy in a crossover design with participants receiving both treatments.68 We evaluated comparisons after the first intervention period only, before any crossover. They provided 16 hours of training and up to 28 hours of home practice over an 8-week period. This trial found a 13.2 percent nonsignificant change favoring the cognitive behavior therapy group on the symptom checklist 90 global severity index. However, the cognitive behavior therapy group received nearly twice the amount of group training, 28 hours over 14 weeks, and this increased time and attention in the control arm may not allow equivalent comparisons between the groups. This trial had a medium risk of bias.

Jazaieri et al.87 randomized patients with social anxiety disorder to MBSR or exercise (n=56). The trial provided about 25 hours of training by trained teachers, and their participants performed an average of 28.3 hours of homework over the 8-week training period. Although they did not specify primary or secondary outcomes, stress was identified as a primary focus of the study. The trial found a nonsignificant improvement of 17.6 percent in the perceived stress scale at 8 weeks in the MBSR group as compared with exercise. This trial had a high risk of bias.

Barrett et al.85 randomized patients with a history of upper respiratory infections to MBSR or exercise (n=98). The trial provided about 20 hours of training by trained teachers, and approximately 42 hours of recommended homework over the 8-week training period. The perceived stress scale was a secondary outcome. The trial found no significant differences between the two arms. This trial was rated as medium risk of bias

Wolever et al. randomized stressed employees (n=186) to a mindfulness at work program or viniyoga for 12 weeks. Participants received 14 hours of training by trained teachers and unspecified amount of homework. Perceived stress was a primary outcome. The trial found no significant differences in the mindfulness group compared with control. This trial had a medium risk of bias.

The difference-in-change graphs showed inconsistent results (Figure 29). A meta-analysis suggested a nonsignificant null effect (Figure 14).

Figure 29 shows Relative difference between groups in the changes in measures of depression, in the mindfulness versus specific active control studies. This shows inconsistent results.

Figure 29

Relative difference between groups in the changes in measures of distress, in the mindfulness versus specific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline mean (more...)

The strength of evidence is insufficient that mindfulness meditation programs affect improve distress among those with mood disturbance or symptoms of anxiety compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent results, direct measures, and imprecise estimates (Table 32).

Table 32. Grade of trials addressing the efficacy of mindfulness meditation programs on distress compared with specific active controls among populations with emotional distress.

Table 32

Grade of trials addressing the efficacy of mindfulness meditation programs on distress compared with specific active controls among populations with emotional distress.

Positive Affect
Mindfulness Meditation Programs Versus Specific Active Control

Four trials evaluated the effect of mindfulness meditation programs compared with a specific active control on the outcome of positive affect (Table 12). Delgado et al. randomized female students (n=32) with high scores on the Pittsburgh sleep quality index (PSQI) to 5 weeks of either mindfulness training or progressive muscle relaxation training, providing 10 hours of training and unspecified amount of home practice.51 The trial did not detect any within or between-group effects on the positive and negative affect schedule. This trial had a medium risk of bias.

Moritz et al. randomized patients with mood disorders (n=110) to a spirituality program versus MBSR.63 In this trial, MBSR was the active control for the spirituality intervention they were testing. The trial selected participants with high scores on the POMS scale. The spirituality program had meditative components in it. It provided about 12 hours of training in both interventions over an 8-week period, with an unspecified amount of home practice in the MBSR group. It provided up to 42 hours of home practice in the spirituality group. There was no information on teacher qualifications for MBSR. The study used the SF-36 vitality score to measure improvement in positive affect as a secondary outcome. The SF-36 vitality scores were 45 percent greater for the spirituality group (p=.024). This trial had a low risk of bias.

Jazaieri et al.87 randomized patients with social anxiety disorder to MBSR or exercise (n=56). The trial provided about 25 hours of training by trained teachers, and their participants performed an average of 28.3 hours of homework over the 8-week training period. Although they did not specify primary or secondary outcomes, positive affect was identified as a primary focus of the study. The trial found a nonsignificant improvement of 10.2 percent in the satisfaction with life scale at 8 weeks in the MBSR group as compared with exercise. This trial was rated as high risk of bias.

Barrett et al.85 randomized patients with a history of upper respiratory infections to MBSR or exercise (n=98). The trial provided about 20 hours of training by trained teachers, and approximately 42 hours of recommended homework over the 8-week training period. The positive and negative affect scale was a secondary outcome. The trial found no significant differences between the two arms in the positive portion of this scale at 9 weeks and 5 months. This trial had a medium risk of bias.

The difference-in-change graphs showed inconsistent results (Figure 23). A meta-analysis showed a nonsignificant and null effect (Figure 24).

The strength of evidence is insufficient regarding the effect mindfulness meditation programs have on positive affect among those with a mood disturbance or symptoms of anxiety when compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent findings, indirect measures, and imprecise estimates (Table 33).

Table 33. Grade of trials addressing the efficacy of mindfulness meditation programs on positive affect compared with progressive muscle relaxation or spirituality among various patients.

Table 33

Grade of trials addressing the efficacy of mindfulness meditation programs on positive affect compared with progressive muscle relaxation or spirituality among various patients.

Mental Component of Health-Related Quality of Life
Mindfulness Meditation Programs Versus Specific Active Control

Six trials evaluated the effect of mindfulness meditation programs compared with a specific active control on the outcome of the mental component of health-related QOL (Table 13). Five were MBSR trials and one was an MBCT trial. Three trials were low risk of bias, two medium, and one high. They used a variety of patient populations and specific active controls. Sample sizes ranged from 15–123.

Wong et al. randomized chronic pain patients (n=99) to an 8-week program in MBSR or multidisciplinary pain intervention.74 The study used the validated Chinese SF-12 mental component subscale to measure QOL as a secondary outcome. There was no significant change in the scores between groups at 2 or 5 months. This trial had a low risk of bias, provided 27 hours of training and an unspecified amount of home practice over 8 weeks. Its teachers were trained and had 5 years of experience teaching meditation.74

Gross et al. randomized people with insomnia (n=27) to 8 weeks of MBSR versus pharmacotherapy for sleep (eszopiclone).55 The trial used the SF-12 mental summary score to measure QOL as a secondary outcome. There was no significant change in SF-12 scores between the two groups. This trial provided 26 hours of training and up to 36 hours of home practice over 8 weeks. Its teachers were trained and certified.

Moritz et al. randomized patients with mood disorders (n=110) to a spirituality program versus MBSR.63 In this trial, MBSR was the active control. It provided about 12 hours of training in both interventions over an 8-week period, with unspecified amount of home practice in the MBSR group. It provided up to 42 hours of home practice in the spirituality group. There was no information on teacher qualifications for MBSR. The trial used the SF-36 mental component survey to measure QOL as a secondary outcome. They found a 23 percent change favoring spirituality at 8 weeks (p=.029). This trial had a low risk of bias. It is notable that this intervention included a meditative component, as well as breathing exercises that may resemble features of MBSR.

Plews-Ogan et al. randomized people with chronic musculoskeletal pain (n=15) to 8 weeks of MBSR training or weekly massage.69 The trial used the SF-12 mental health score to measure QOL as a primary endpoint. The difference-in-change point estimates were 10.8 percent favoring massage at 8 weeks and 18.4 percent favoring MBSR at 12 weeks. The trial did not calculate significance for difference-in-change estimates. This trial provided 20 hours of training over 8 weeks, and unspecified amount of home practice. There was no information on teacher qualifications. It had a high risk of bias.

Kuyken at al. randomized depressed patients at risk for relapse (n=123) to 8 weeks of MBCT and antidepressant tapering or maintenance antidepressant therapy.59 The trials used the World Health Organization quality of life instrument psychological subscale to measure QOL as a secondary outcome. At 3 months it found a 9.2 percent improvement in the MBCT group, which maintained at 15 months (p=.01). This trial provided 24 hours of training over 8 weeks by trained and certified instructors, and recommended up to 37.5 hours of home practice during that time. This trial had a low risk of bias.

Barrett et al.85 randomized patients with a history of upper respiratory infections to MBSR or exercise (n=98). The trial provided about 20 hours of training by trained teachers, and approximately 42 hours of recommended homework over the 8-week training period. QOL was a secondary outcome. The trial found no significant differences between the two arms in the SF-12 mental component at 9 weeks and 5 months. This trial was rated as medium risk of bias.

The difference-in-change graphs showed inconsistent results (Figure 26). Meta-analysis showed a null and nonsignificant effect (Figure 30).

The strength of evidence is insufficient that mindfulness meditation program have an effect on the mental component of health-related quality of life among various clinical populations when compared with a variety of specific active controls. We based this rating on overall medium risk of bias, inconsistent findings, direct measures, and imprecise estimates (Table 34).

Table 34. Grade of trials addressing the efficacy of mindfulness meditation programs on the mental component of health-related quality of life compared with specific active controls among various populations.

Table 34

Grade of trials addressing the efficacy of mindfulness meditation programs on the mental component of health-related quality of life compared with specific active controls among various populations.

Applicability

Most of the trials that we included for this KQ took place in outpatient settings in the United States or Europe; two trials took place in Asia, and one in the Middle East. Almost all the trials listed some exclusion criteria which would apply to a large number of patients in an everyday internal medicine or primary care practice, including substance abuse, psychiatric disorder, or various medical disorders.

Regarding the population characteristics of the trials for this KQ, most of the trials did not specify the racial or ethnic characteristics of the included population. While about half the trials specified the educational characteristics of the study populations, the trials did not report other measures of socioeconomic status.

Although some of the trials for this KQ addressed a number of chronic medical conditions, including metabolic syndrome, chronic obstructive pulmonary disease, HIV, and CHF, the trials did not address a number of common medical conditions frequently found in medical practice, and often associated with anxiety, depression, stress, and distress, including diabetes, IBS, and opiate dependence. While half of the trials included patients with some form of mental health issue, a large number of them excluded patients with serious mental health conditions.

Thus, the findings for this KQ may be least applicable to patients with substance abuse, dementia, or other medical or psychiatric conditions excluded from the study populations. Given that the trials only substantially represented two continents, and the trials did not always specify the racial and ethnic makeup of the populations, it's unclear whether these findings would be applicable to more diverse patient populations.

Regarding the applicability of an intervention to a medical practice, both TM and mindfulness trials involved training for about 10–40 hours over several weeks, which makes them fairly practical in a typical outpatient setting.

Key Question 2. What are the efficacy and harms of meditation programs on attention among those with a clinical condition (medical or psychiatric)?

Key Points and Evidence Grades

  • The strength of evidence is insufficient that mindfulness meditation programs have an effect on measures of attention among older caregivers compared with a nonspecific active control due to medium risk of bias in a single trial, unknown consistency, directness of measures, and imprecise estimates.

Harms

  • We found no studies that reported on harms

Trial Characteristics

One RCT assessed the efficacy of a meditation program on attention as a component of their study. Oken et al. assessed the effects of a 7-week mindfulness meditation program on stress among caregivers of close relatives with dementia.66 The study did not report the specific period of recruitment. The trial took place in the United States in an outpatient setting among a stressed population. The trial included participants with a score greater than 9 on the perceived stress scale and excluded individuals who were medically unstable, had significant cognitive dysfunction, significant visual impairment, or previous experience with stress-reduction classes66 (Appendix E, Evidence Table E2).

Population Characteristics

The trial enrolled 31 dementia caregivers with a mean age range in the 60s.66 Participants were predominantly female and greater than 90 percent were white (Appendix E).

Intervention Characteristics

The trial included three arms: a composite intervention based on MBSR/MBCT, which was compared with education (nonspecific active control), and to respite care.66 The trial delivered all meditation interventions in a group format. The maximal training dose for was 9 hours delivered over 7 weeks. The trial used trained teachers but did not specify the amount of training or meditation experience. The trial recommended practice at home but did not specify the total duration and did not record actual amounts of training or home practice by the participants66 (Appendix E).

Outcomes

The trial used the attentional network test as the measure of attention. The attentional network test is a computerized task for assessing various attention networks. This test requires participants indicate the direction of a target arrow that is accompanied on each side by two additional arrows. Occasionally, the target arrow is preceded by cues. The trial used a shortened version of this test that included only two attention conditions: cued/noncued and congruent/incongruent conditions, which present companion arrows in the same or opposite direction as the target arrow. The results included a conflict score, calculated as the reaction time difference between the incongruent and congruent conditions; and the alerting score, calculated as the reaction time difference between the noncued and cued conditions.

Attention

Mindfulness Meditation Programs Versus Nonspecific Active Control

The attentional network test alerting score for the meditation group was worse than for the education group at baseline. At 8 weeks post-intervention the meditation group improved its performance by doubling its score, resulting in an 81 percent increase from baseline compared with education.66 This suggests an appropriate use of a cue by the meditation group to improve their performance from baseline to post-intervention. However, the data were highly skewed, and it is not apparent that the differences between meditation and education arms were statistically significant. There was a 15 percent nonsignificant difference among the groups on the attentional network test conflict score favoring mindfulness meditation (p=0.14).66

In summary, this trial had a medium risk of bias due to several factors including high attrition, allocation to groups was not concealed, and there was no intention-to-treat analysis. Overall, the strength of evidence is insufficient to comment on whether mindfulness meditation interventions improve attention among an older population compared with a nonspecific active control due to medium risk of bias, unknown consistency, directness of measures, and imprecision (Table 35).

Table 35. Grade of trial addressing the efficacy of a meditation program on a measure of attention compared with a nonspecific active control among older caregivers.

Table 35

Grade of trial addressing the efficacy of a meditation program on a measure of attention compared with a nonspecific active control among older caregivers.

The trial did not report on harms.

Applicability

The trial took place in the United States in an outpatient setting with predominantly female and predominantly white participants. The trial studied an older population of dementia caregivers without direct complaints of cognitive difficulties (i.e., attention). Therefore, these findings may not be applicable to other clinical populations where cognitive function is a reported concern (e.g., attention-deficit/hyperactivity disorder), and improvement (or lack thereof) on cognitive measures could provide more useful clinical information.

Key Question 3. What are the efficacy and harms of meditation programs on health-related behaviors affected by stress, specifically substance use, sleep, and eating, among those with a clinical condition (medical or psychiatric)?

Key Points and Evidence Grades

Comparisons With Nonspecific Active Controls

  • The strength of evidence is insufficient about the effects of mindfulness meditation programs on sleep quality among a variety of populations when compared with a nonspecific active control. We based this rating on overall medium risk of bias, inconsistent findings, direct measures, and imprecise estimates.

Comparisons With Specific Active Controls

  • The strength of evidence is insufficient that mindfulness programs have an effect on sleep when compared with exercise or eszopiclone. We based this rating on overall medium risk of bias, inconsistent results, directness of measures, and imprecise estimates.
  • The strength of evidence is insufficient that mindfulness programs have an effect on eating compared with specific active controls. We based this rating on overall high risk of bias, consistent results, directness of measures, and imprecise estimates.
  • The strength of evidence is insufficient that mindfulness meditation programs have an effect on substance use among smoking and alcoholic populations when compared with certain specific active controls. We based this rating on overall high risk of bias, inconsistent findings, direct measures, and imprecise estimates.
  • The strength of evidence is low that mantra meditation programs do not reduce alcohol use among alcohol abusing populations when compared with intensive running or biofeedback. We based this rating on overall medium risk of bias, consistent findings, direct measures, and imprecise estimates.
Harms
  • Four trials reported that they evaluated harms; none found any adverse events.

Trial Characteristics

Of the 13 trials that we included for this KQ,49,50,54-56,61,62,66,70,73,80,84,85 12 took place in the United States, while the other took place in Germany. Seven of these trials took place exclusively in an outpatient setting, two took place in an inpatient setting, and the remaining two trials had multiple locations. Only two of these trials explicitly reported the year of recruitment, and none of the trials reported the time period of recruitment.

All but two of these trials explicitly stated the length of treatment and timing of subsequent followup. Treatment ranged from 4–15 weeks, and followup ranged from none (i.e. treatment assessed at its end) to 18 months.

All 13 trials reported inclusion and exclusion criteria. One trial excluded individuals with chronic substance dependence. Five trials excluded subjects if they had unstable medical conditions. Eight other trials excluded patients due to psychiatric criteria. Three trials excluded due to severe cognitive dysfunction. Most trials excluded people with prior or recent experience in meditation.

Four of the 13 trials that we included in this review evaluated the effects of meditation on substance use: one related to cigarette smoking,50 and three related to alcohol and drug use.49,80,84 Two trials considered the effect of meditation on eating behaviors.56,62 The remaining seven of the 13 included trials examined the effect of meditation programs on sleep54,55,61,66,70,73,85 (Appendix E).

Population Characteristics

Seven trials took place in populations with chronic medical conditions;54-56,61,62,70,85 four trials took place in populations with substance abuse;49,50,80,84 and two trials targeted a population of caregivers under stress.66,73 The percentage of female subjects totaled 30 percent or greater in 10 of the 13 trials,50,54-56,61,62,66,70,73,85 with two of the 13 trials including female subjects exclusively.56,70 The mean age of trial participants ranged from 24–67 years. Two of the 13 trials exhibited significant racial diversity in the subject populations.50,84 Ten of the 13 trials provided information on the level of education completed by trial participants (Appendix E).

Intervention Characteristics

Of the four trials assessing the effects of meditation on substance use, two used mindfulness meditation based on mindfulness-based relapse prevention, with 9–12 hours of training over 4–9 weeks. Training and experience ranged from 12 years to greater than 13 years in mindfulness experience and social work, although there was no explicit mention of centralized training or certification. Another trial used clinically standardized meditation, a mantra-based concentrative meditation intervention taught by “experienced meditators,” after which the group meditated together 3 times per week for the 8-week intervention.80 One trial used a TM intervention taught by a certified instructor. Instruction used a seven-step process, followed by group meditations.84 For the substance-use trials, comparisons included cognitive behavioral therapy treatment,49 biofeedback,84 smoking-cessation education,50,76 and exercise.80

Two trials assessed eating in response to a mindfulness intervention. Hebert et al.,56 assessing eating in breast cancer patients, compared a nutrition education program with a mindfulness-based program adapted from MBSR, while Miller et al.62 assessed a mindfulness based eating program among diabetics.

All seven trials evaluating meditation for sleep evaluated either MBSR or an abbreviated derivative of MBSR.54,55,61,66,70,73,85 Comparison treatments included pharmacotherapy for sleep55, exercise73,85 programs in relaxation,70 or health education matched for time and attention.54,66

Only three of the 13 trials investigating stress-related behaviors measured adherence to home meditation practice50,54,55 (Appendix E).

Outcomes

Mindfulness Meditation Programs Versus Nonspecific Active Control

Sleep

Four trials compared a mindfulness meditation program with a nonspecific active control on the outcome of sleep quality (Table 14). All four used the PSQI. Three had a medium risk of bias and one had a low risk of bias. Gross et al. randomized solid organ transplant recipients, post-surgery, (n=137) to either 8 weeks of MBSR or nonspecific active control.54 The trial used the PSQI to measure sleep quality as a primary outcome. In a difference-in-change analysis, the MBSR group showed a 24.1 percent improvement in PSQI at 8 weeks, which further improved to 30.1 percent at 1 year (p=.02). This trial had a medium risk of bias. It provided 27 hours of training by a trained teacher and an unspecified amount of home practice over 8 weeks.

Schmidt et al. randomized women with fibromyalgia (n=109) to 8 weeks of MBSR or a nonspecific active control.70 The study used the PSQI to measure sleep as a secondary endpoint and showed no difference between the arms. This trial provided 27 hours of training over 8 weeks by trained and certified teachers, and recommended up to 42 hours of home practice. It had a medium risk of bias.

Oken et al. randomized people who take care of elderly relatives with dementia (n=19) to 6 weeks of mindfulness meditation or a nonspecific active control.66 The trial used the PSQI and Epworth sleepiness scale to measure sleep as a secondary outcome. This trial showed a 12.8 percent change on the Epworth sleepiness scale and a 3.4 percent change on the PSQI, both were nonsignificant and favored the control group. This trial had a medium risk of bias. It provided 9 hours of training over 7 weeks by a trained teacher and an unspecified amount of home practice.

Malarkey et al. randomized people who either had or were at risk for cardiovascular disease due to elevated C-Reactive protein levels (n=186) to mindfulness meditation or nonspecific active control.61 It provided 9 hours of abbreviated MBSR training at work with approximately 18.5 hours of homework over 8 weeks. It measured sleep as a secondary outcome. At 8 weeks, the trial found no differences between the groups, but did not provide data for comparisons of the size of effect. This trial had a low risk of bias.

The difference-in-change graphs showed inconsistent results (Figure 31). A meta-analysis showed a small nonsignificant effect around the null (Figure 32). The strength of evidence is insufficient that mindfulness meditation programs have an effect on sleep quality among a variety of populations when compared with a nonspecific active control. We based this rating on trials of medium bias, inconsistent findings, direct measures, and imprecise estimates (Table 36).

Figure 31 shows Relative difference between groups in the changes in measures of sleep, in the mindfulness versus non-specific active control studies and compared with specific active control studies. This shows inconsistent results for both.

Figure 31

Relative difference between groups in the changes in measures of sleep, in the mindfulness versus nonspecific/specific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline (more...)

Figure 32 shows Meta-analysis of the effects of meditation programs on sleep with up to 3 months of followup. This shows nonsignificant results compared with both nonspecific and specific active controls.

Figure 32

Meta-analysis of the effects of meditation programs on sleep with up to 3 months of followup. Notes: AC = Active Control; HE = Health Education; MBSR = Mindfulness Based Stress Reduction; PSQI = Pittsburgh Sleep Quality Index

Table 36. Grade of trials addressing the efficacy of mindfulness meditation program on sleep quality among various populations compared with a nonspecific active control.

Table 36

Grade of trials addressing the efficacy of mindfulness meditation program on sleep quality among various populations compared with a nonspecific active control.

Mindfulness Meditation Programs Versus Specific Active Control

Sleep

Three trials evaluated the effects of mindfulness meditation programs against a specific active control on the outcome of sleep (Table 14).55,73,85 Gross et al. randomized people with insomnia (n=27) to MBSR or eszopiclone.55 Sleep was a primary outcome. The study measured sleep time by wrist actigraphy. It measured overall sleep quality by the PSQI and insomnia severity index. Total sleep time and wake after sleep onset were not different between the groups, although it favored the eszopiclone group. The PSQI indicated a 14.7 percent improvement favoring the MBSR group, while the insomnia severity index showed a 15.5 percent improvement favoring the eszopiclone group. Both were nonsignificant. This trial provided 26 hours of training and up to 36 hours of home practice over 8 weeks. It had a medium risk of bias.

Barrett et al.85 randomized patients with a history of upper respiratory infections to MBSR or exercise (n=98). The trial provided about 20 hours of training by trained teachers, and approximately 42 hours of recommended homework over the 8-week training period. Sleep quality was a secondary outcome. The trial found no significant differences between the two arms in the PSQI at 9 weeks and 5 months. This trial had a medium risk of bias.

Wolever et al. randomized stressed employees (n=186) to a mindfulness-at-work program or viniyoga for 12 weeks. Participants received 14 hours of training by trained teachers and unspecified amount of homework. Sleep quality was a secondary outcome. The trial found no significant differences in the PSQI in the mindfulness group compared with control. This trial had a medium risk of bias.

The difference-in-change graphs showed inconsistent results (Figure 31). A meta-analysis showed a nonsignificant result around the null (Figure 32). The strength of evidence is insufficient that mindfulness programs have an effect on sleep when compared with exercise or eszopiclone. We based this rating on overall medium risk of bias, inconsistent results, directness of measures, and imprecise estimates (Table 37).

Table 37. Grade of trials addressing the efficacy of mindfulness meditation programs on sleep quality compared with specific active controls in various populations.

Table 37

Grade of trials addressing the efficacy of mindfulness meditation programs on sleep quality compared with specific active controls in various populations.

Eating

Two trials evaluated the effects of mindfulness meditation programs against a specific active control on the outcome of eating (Table 14).56,62 Hebert et al. evaluated the effects of MBSR compared with a nutrition education program among women with stage I or II breast cancer (n=106).56 Ninety-five percent of the participants had complete diary data post-intervention (at 4 months) and 93 percent at 1 year. Women in the nutrition group had a significant 19.1 percent reduction in fat consumption at 4 months (p<.05) and 11.3 percent reduction at 1 year (p<.05) compared with MBSR. There were no differences in overall caloric consumption between groups at 4 months or 1 year. This trial had a medium risk of bias

Miller et al.62 randomized diabetics to a mindfulness eating program versus the Smart Choices diabetes group self-management education program (n=52). This mindfulness program provided about 25 hours of training over 12 weeks and unspecified amount of homework. Total caloric consumption was a secondary outcome. The trial found a nonsignificant reduction of 14.9 and 10.4 percent at 3 and 6 months, respectively, compared with control. This trial had a high risk of bias.

The difference-in-change graphs show no improvement in the meditation arm in either trial compared with control (Figure 34). The strength of evidence is insufficient that mindfulness programs have an effect on eating compared with specific active controls. We based this rating on overall high risk of bias, consistent results, directness of measures, and imprecise estimates (Table 38).

Figure 34 shows Relative difference between groups in the changes in measures of substance use, in the mantra versus non-specific/specific active control studies. This shows inconsistent results.

Figure 34

Relative difference between groups in the changes in measures of substance use, in the mantra versus nonspecific/specific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the (more...)

Table 38. Grade of trials addressing the efficacy of mindfulness meditation programs on eating compared with specific active controls in diabetics and breast cancer populations.

Table 38

Grade of trials addressing the efficacy of mindfulness meditation programs on eating compared with specific active controls in diabetics and breast cancer populations.

Substance Use

Two trials evaluated the effects of mindfulness meditation programs against a specific active control on the outcome of substance abuse (Table 14).49,50

Brewer et al. randomized smokers (N=71) to an 8-session, 4-week program of mindfulness meditation compared with a specific active control, the American Lung Association's freedom from smoking program.50 The mindfulness meditation program is based on mindfulness-based relapse prevention and MBSR, and provided up to of 12 hours of meditation training by a single therapist with 13 years of experience with mindfulness meditation. While the freedom from smoking group reduced their cigarette use by 12 cigarettes/day, mindfulness meditation participants smoked 4.2 cigarettes/day less than the freedom from smoking program in a difference-in-change calculation (p=.008) at the end of the 4-week program. Mindfulness meditation participants had an absolute 21 percent higher levels of 1-week point-prevalence abstinence from smoking at 4 weeks (p=.06) and absolute 25 percent higher abstinence at 17-week followup (p=0.012). Additionally, within the mindfulness meditation group, both formal (p=0.019) and informal (p=0.01) mindfulness practice resulted in less cigarette use. This trial had a high risk of bias.50 Overall, the strength of evidence is low to conclude that a 4-week mindfulness meditation program has an effect on smoking compared with a freedom from smoking program among smokers, due to high risk of bias, unknown consistency, directness of measures, and precise results.

Brewer et al. conducted a separate trial in which they randomized individuals with alcohol and/or cocaine abuse that were seeking outpatient treatment (n=24) to mindfulness meditation that consisted of mindfulness-based relapse prevention with cognitive behavioral therapy.49 Following the treatment programs, there were no statistically significant differences in alcohol (p=.17) or cocaine (p=.09) use between groups. This trial provided 9 hours of training over 9 weeks by a teacher with 12 years of meditation experience, and did not report on whether it recommended any home practice. It had a 61 percent attrition rate and a high risk of bias.

The differences-in-change graphs showed inconsistent results (Figure 33). The strength of evidence is insufficient that mindfulness meditation programs have an effect on substance use among smoking and alcoholic populations when compared with certain specific active controls. We based this rating on overall high risk of bias, inconsistent findings, direct measures, and imprecise estimates (Table 39).

Figure 33 shows Relative difference between groups in the changes in measures of substance use/ eating, in the mindfulness versus specific active control studies. This shows inconsistent results.

Figure 33

Relative difference between groups in the changes in measures of substance use/eating, in the mindfulness versus specific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the (more...)

Table 39. Grade of trials addressing the efficacy of mindfulness meditation programs on substance use compared with specific active controls in smoking and alcoholic populations.

Table 39

Grade of trials addressing the efficacy of mindfulness meditation programs on substance use compared with specific active controls in smoking and alcoholic populations.

Mantra Meditation Programs Versus Specific Active Control

Substance Use

Two trials used a mantra meditation programs to assess the effects on alcohol consumption against either an intensive running program among college students or biofeedback among recovering alcoholics (Table 14).80,84 Murphy et al. randomized male college students who were heavy social drinkers (n=27) to an 8-week treatment programs in clinically standardized meditation or running.80 The running group consumed 99.3 mL of ethanol less than the meditation group (p=.35). The meditation group received 8 hours of training over 8 weeks by a teacher with some experience in meditation, and up to 37.5 hour of home practice. The running group received 28 hours of training. This trial had a high risk of bias.

Taub et al. randomized alcoholics (n=87) in residential treatment program to TM or two different specific active control arms: biofeedback or neurotherapy. There was no difference in the percent of days abstinent from alcohol between the TM group and biofeedback. The TM group provided up to 19 hours of training over 4 weeks by certified teachers, and did not specify whether it recommended any amount of home practice. This trial had medium risk of bias.84

The difference-in-change graphs showed consistent results favoring control (Figure 34). The strength of evidence is low that mantra meditation programs do not reduce alcohol use among alcohol abusing populations when compared with intensive running or biofeedback. We based this rating on overall medium risk of bias, consistent findings, direct measures, and imprecise estimates (Table 40).

Table 40. Grade of trials addressing the efficacy and harms of mantra meditation programs on alcohol use among heavy alcohol drinkers compared with intensive running program or biofeedback.

Table 40

Grade of trials addressing the efficacy and harms of mantra meditation programs on alcohol use among heavy alcohol drinkers compared with intensive running program or biofeedback.

Applicability

Twelve of 13 trials took place in the United States, so other regions might not find these findings applicable. Most of the trials took place in outpatient settings, so applicability to the inpatient setting is limited.

Regarding the population characteristics of the trials for this KQ, only two of 13 trials exhibited significant racial diversity in the study populations. Most of the trials excluded subjects from groups who might commonly be found in a medical practice, such as those with unstable medical conditions and psychiatric disorders.

Characteristics of the interventions represented in this KQ were diverse, making it difficult to foresee how these findings would be applicable to a similarly wide array of mindfulness practices under everyday clinical situations. For example, the trials did not specify the certification and training of instructors, and only a few trials specified the time spent in home training.

Key Question 4. What are the efficacy and harms of meditation programs on pain and weight among those with a clinical condition (medical or psychiatric)?

Key Points and Evidence Grades

Comparisons With Nonspecific Active Controls

  • The strength of evidence is moderate that mindfulness meditation programs have a small improvement in pain severity among a variety of populations when compared with a nonspecific active control. We based this rating on trials with medium bias, consistent findings for a small positive effect, direct measures, and precise estimates.
  • The strength of evidence is low that mantra meditation programs have no effect on pain severity among those with CHF when compared with a nonspecific active control. We based this rating on a single trial of low risk of bias, unknown consistency, direct measures, and imprecise estimates.
  • The strength of evidence is low that mantra meditation programs do not have an effect on weight among diabetics, hypertensives, or those with coronary disease when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent null findings, directness of measures, and imprecise estimates.

Comparisons With Specific Active Controls

  • The strength of evidence is low that mindfulness has no effect on pain severity among those with chronic musculoskeletal pain or mood disturbance, compared with a specific active control. We based this rating on overall medium risk of bias, consistent null results, direct measures of pain, and imprecise estimates.
  • The strength of evidence is low that mindfulness meditation programs do not have an effect on weight among breast cancer and chronic pain patients when compared with a specific active control. We based this rating on overall medium risk of bias, consistent results, directness of measures, and imprecise estimates.
Harms
  • Four trials reported that they evaluated harms; none found any adverse events.

Trial Characteristics

We found 14 RCTs on this KQ. Eleven RCTs took place in the United States, one in Canada, one in Germany, and one in Hong Kong. All involved outpatients. Six trials did not report recruitment periods, the others were between 2000 and 2009. Trial duration ranged from 3 months to 9.3 years. All trials recruited only adults. Two recruited only females53,70 (Appendix E).

Population Characteristics

Five of the trials recruited participants who reported a chronic pain condition53,64,69,70,74 while nine used non-pain populations.54,56,62,63,73,76-78,90 Two trials used general chronic pain patients of whom more than 95 percent had musculoskeletal pain,69,74 while the other three used women with IBS (visceral pain), women with fibromyalgia (musculoskeletal pain), and patients with low back pain (also musculoskeletal pain). The sample size in trials that used a pain population ranged from 30–177. Two included only women.53,70 The mean age was around 40–60 for these trials except for a trial that studied elderly low-back-pain patients,64 who were, on average, 75 years old. Four trials reported ethnicity. In two, the majority of participants were white, and in the other two the entire population was black. Five trials reported education level. The percent that had completed high school ranged from 11–72 percent.53,62,70,73,74 The majority of participants in the IBS trial had a college or graduate level education.53 Among the non-pain population trials, participants were either solid organ transplant recipients,54 patients, post-surgery, with psychological distress,63 or African Americans with CHF.78 Sample sizes ranged from 23–186 and included 30–80 percent women (Appendix Evidence Table E3).

Intervention Characteristics

Six trials used MBSR,54,62,63,69,70,74 three used mindfulness meditation programs,53,64,73 and four used TM.76-78,90 While all trials used active controls, six of the mindfulness trials used a specific active control such as a multidisciplinary pain management program or massage. All others used a nonspecific active control to control for time, attention, and expectation, such as a health education group. All four of the TM programs used a nonspecific active control.

The studies typically conducted the mindfulness programs weekly for 1.5–2.5 hours over 8 weeks, and they ranged from 12–27 total hours of training. Although all of the trials indicated, in some form, that they recommended daily practice, only two of the trials specified the amount, recommending 45 minutes daily.64,70 None of the trials reported on the actual amount of home practice in the meditation arm. Reports on instructor qualifications were lacking for most trials. Six of 10 mindfulness trials indicated that instructors had some training but only two gave enough information to suggest that the instructors had some kind of certification.64,70 Only one trial reported on the personal meditation experience of the instructors,64 and three trials reported an instructor's level of teaching experience.70,74

On average, the TM trials provided 1.5-hour sessions for seven consecutive days, and followup refresher meetings twice monthly for the first 3 months and then once monthly for the next 3 months. The trials did not give details of the followup meetings, but we estimated the duration at approximately 22.5 hours over a 6-month period, assuming the meetings were also 1.5 hours in length (an amount roughly similar to the mindfulness trials). They recommended approximately half-hour daily home practice for 6 months, which calculates to approximately 90 hours of home practice over 6 months. These trials reported a certified trainer without giving details of years of meditation or teaching experience.

Five trials measured weight changes.77,56,62,76,90 Three were TM77,76,90and two used mindfulness meditation.56,62 None of these trials reported details of hours of training, although we estimated the amount of training where some information was given. These trials gave little information on instructor qualifications or whether the participants performed home practice. The TM trials indicated their teachers were either trained or certified, and recommended between 30–40 minutes of daily meditation for the duration of their study, amounting to a total expected home practice dose of 90–120 hours over 6 months. None of the trials reported actual amounts of meditation (Appendix E).

Outcomes

Ascertainment of Outcomes (Scales)

Studies measured pain severity using the 11-point numerical rating scale for pain intensity or unpleasantness, perceived pain scale affective and sensory subscales, SF-36 bodily pain subscale, McGill pain questionnaire, and the IBS abdominal pain severity subscale. Studies measured weight in either pounds, kilograms, or body mass index (BMI) (Table 3).

Pain Severity

Mindfulness Meditation Program Versus a Nonspecific Active Control

Four trials evaluated MBSR against a nonspecific active control and assessed the outcome of pain severity (Table 15).53,54,64,70 One trial evaluated visceral pain while the other three evaluated musculoskeletal pain. One trial had a low risk of bias, the remaining three had a medium risk of bias.

Gaylord et al. randomized women with IBS (n=75) to MBSR versus support program for women with IBS.53 This was the only trial to assess visceral pain, and found a 30.6 percent reduction in abdominal pain severity in the MBSR group compared with control at 8 weeks; this maintained at 6 months (p=.015). This was a medium risk-of-bias trial that provided 23 hours of training and unspecified amount of home practice over 8 weeks.

Schmidt et al. randomized women with fibromyalgia (n=109) to 8 weeks of MBSR or a nonspecific active control.70 The trial used perceived pain scale to measure pain severity as a secondary outcome. The perceived pain scale has affective and sensory subscales; the affective dimension measures the unpleasantness of the pain experience, whereas the sensory dimension measures the intensity of sensory qualities of the pain experience. There were no significant differences between the MBSR and control on either of the subscales (p=.18 for affective subscale, p=.60 for sensory subscale), although the meditation arm was favored by 5.7 percent for the sensory subscale. This trial provided 27 hours of training over 8 weeks by trained and certified teachers, and recommended up to 42 hours of home practice. It had a medium risk of bias.

Gross et al. randomized solid organ transplant patients, post-surgery, (n=137) to MBSR versus health education.54 They found no change in the SF-36 bodily pain subscale within groups or between groups at 2 months or 1 year, although the meditation arm was favored by 5.1 percent. This trial provided 27 hours of training by a trained teacher, and unspecified amount of home practice over 8 weeks. This trial had medium risk of bias.

Morone et al. randomized older adults with chronic low back of moderate intensity (n=35) to MBSR or a health education program for 8 weeks.64 They used two scales to assess pain severity: SF-36 pain subscale and McGill pain questionnaire current pain score. The MBSR group showed a nonsignificant 8.6 percent improvement in the SF-36 pain subscale at 8 weeks compared with control, but these differences disappeared at 6 months. There were no effects seen in the McGill pain questionnaire in a differences-in-change analysis. This trial provided 12 hours of training over 8 weeks by a teacher with 25 years of meditation experience and some teaching experience. The trial recommended up to 42 hours of home practice over the 8 weeks. This trial had a low risk of bias.

The difference-in-change graphs showed consistent small positive effects on pain severity (Figure 35). A meta-analysis showed a small statistically significant effect size favoring mindfulness meditation programs (Figure 36). The strength of evidence is moderate that mindfulness meditation programs have a small improvement in pain severity among a variety of populations when compared with a nonspecific active control. We based this rating on trials of medium bias, consistent findings for a small positive effect, direct measures, and precise estimates (Table 41).

Figure 35 shows Relative difference between groups in the changes in measures of pain, in the mindfulness versus non-specific active control studies. This shows four trials with consistent and small positive effects.

Figure 35

Relative difference between groups in the changes in measures of pain, in the mindfulness versus nonspecific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline mean (more...)

Figure 36 shows Meta-analysis of the effects of meditation programs on pain severity with 8 to 12 weeks of followup. This figure shows a statistically significant effect size of -.33 favoring mindfulness compared with nonspecific active controls, and a nonsignificant finding compared with specific active controls.

Figure 36

Meta-analysis of the effects of meditation programs on pain severity with 8–12 weeks of followup. Notes: Abd = Abdomen; AC=Active Control; HE = Health Education; NRS = Numeric Rating Scale; MBSR = Mindfulness Based Stress Reduction; MM = Mindfulness (more...)

Table 41. Grade of trials addressing the efficacy of mindfulness-based stress reduction on pain severity compared with nonspecific active controls among visceral pain, musculoskeletal pain, and organ transplant patients.

Table 41

Grade of trials addressing the efficacy of mindfulness-based stress reduction on pain severity compared with nonspecific active controls among visceral pain, musculoskeletal pain, and organ transplant patients.

Transcendental Meditation Versus Nonspecific Active Control

One TM trial on African Americans with CHF assessed pain as a secondary outcome using the SF 36 pain subscale (n=23; Table 15).78 With 100 percent trial completion and 95 percent compliance rate among the originally randomized subjects, there were no differences in the pain scores in both groups at 3 months. However, at 6 months the TM group showed an 18.4 percent improvement over health education (p=.08). This trial had a low risk of bias. It provided 22.5 hours of training over 6 months by trained and certified teachers and recommended up to 90 hours of home practice during this time.

The strength of evidence is low that mantra meditation programs have no effect on pain severity among those with CHF when compared with a nonspecific active control. We based this rating on a single trial of low risk of bias, unknown consistency, direct measures, and imprecise estimates (Table 42).

Table 42. Grade of trials addressing the efficacy of TM on pain severity compared with nonspecific active controls among cardiac patients.

Table 42

Grade of trials addressing the efficacy of TM on pain severity compared with nonspecific active controls among cardiac patients.

Mindfulness Meditation Programs Versus a Specific Active Control

Four trials assessed MBSR against a specific active control for the outcome of pain severity (Table 15). Two trials were conducted in chronic pain populations, one in a mood-disturbed population, and one in stressed employees.

Wong et al. randomized patients with chronic pain (n=99) in Hong Kong to MBSR or a multidisciplinary pain intervention.74 The trial included participants who reported greater than or equal to 4/10 pain on the numerical rating scale. The multidisciplinary pain intervention group specifically excluded teaching of any mind-body techniques that might have overlapped with MBSR. Researchers powered this trial to detect a 1-point difference in the numerical rating scale between the two groups. The trial found no statistically significant difference between interventions. Both interventions reduced pain by approximately 0.5 points post treatment and 1 point at 6 months. This trial had a low risk of bias. It provided 27 hours of training and an unspecified amount of home practice over 8 weeks. Teachers were trained and had 5 years of experience teaching meditation.74

Plews-Ogan et al. randomized people with chronic musculoskeletal pain (n=15) to 8 weeks of MBSR training or weekly massage.69 The study used the 11-point numerical rating scale for pain unpleasantness to measure pain as one of two primary endpoints. It found that the massage group improved 2.9 points while the MBSR group improved by 0.7 points at 2 months. The trial did not calculate significance for difference-in-change estimates. This trial provided 20 hours of training over 8 weeks, and unspecified amount of home practice. There was no information on teacher qualifications. It had a high risk of bias.

Moritz et al. randomized patients with mood disorders (n=110) to a spirituality program versus MBSR.63 In this trial, MBSR was the active control. The spirituality intervention included a meditative component. It used the SF 36 bodily pain scale as a secondary outcome. In a difference-in-change estimate it found a nonsignificant 5.8 percent improvement in the spirituality group compared with the MBSR group. This trial provided about 12 hours of training in both interventions over an 8-week period, with unspecified amount of home practice in the MBSR group. It provided up to 42 hours of home practice in the spirituality group. There was no information on teacher qualifications for MBSR. This trial had a low risk of bias.

Wolever et al. randomized stressed employees (n=186) to a mindfulness-at-work program or viniyoga for 12 weeks. Participants received 14 hours of training by trained teachers and unspecified amount of homework. Pain was a secondary outcome. The trial found no improvement in the mindfulness group compared with control. This trial had a medium risk of bias.

The difference-in-change graphs showed consistent results favoring a null effect or the control group (Figure 37). A meta-analysis suggested a null effect (Figure 35). The strength of evidence is low that mindfulness has no effect on pain severity among those with chronic musculoskeletal pain or mood disturbance, compared with a specific active control. We based this rating on overall medium risk of bias, consistent null results, direct measures of pain, and imprecise estimates (Table 43).

Figure 37 shows Relative difference between groups in the changes in measures of pain, in the mindfulness versus specific active control studies. This shows inconsistent results.

Figure 37

Relative difference between groups in the changes in measures of pain, in the mindfulness versus specific active control studies. Relative difference between groups in the change score. This is a relative percent difference, using the baseline mean in (more...)

Table 43. Grade of trials addressing the efficacy of mindfulness-based stress reduction on pain severity compared with specific active controls among chronic pain and mood disturbance patients.

Table 43

Grade of trials addressing the efficacy of mindfulness-based stress reduction on pain severity compared with specific active controls among chronic pain and mood disturbance patients.

Weight

Mindfulness Meditation Programs Versus a Specific Active Control

Hebert et al. randomized women with early stage breast cancer (n=99) to MBSR versus nutrition education for 15 weeks (Table 16).56 This trial found no difference in weight between the three groups at 4 or 12 months. This trial provided approximately 45 hours of training over 15 weeks, did not report on any teacher qualifications, and did not specify whether they recommended any home practice. This trial had a medium risk of bias.

Miller et al.62 randomized diabetics to a mindfulness eating program versus the Smart Choices diabetes group self-management education program (n=52). This mindfulness program provided about 25 hours of training over 12 weeks and unspecified amount of homework. Weight loss was a primary outcome. The trial found no effect at 3 and 6 months compared with control. This trial had a high risk of bias.

The difference-in-change graphs showed consistent results favoring a null effect (Figure 38). The strength of evidence is low that mindfulness meditation programs do not have an effect on weight among breast cancer and chronic pain patients when compared with a specific active control. We based this rating on overall medium risk of bias, consistent results, directness of measures, and imprecise estimates (Table 44).

Figure 38 shows Relative difference between groups in the changes in measures of weight, in the mindfulness /mantra versus specific/non-specific active control studies. This shows consistently null results for two mindfulness trials and three TM trials.

Figure 38

Relative difference between groups in the changes in measures of weight, in the mindfulness/transcendental meditation versus specific/nonspecific active control studies. Relative difference between groups in the change score. This is a relative percent (more...)

Table 44. Grade of trials addressing the efficacy of mindfulness-based stress reduction on weight among breast cancer and chronic pain patients compared with a specific active control.

Table 44

Grade of trials addressing the efficacy of mindfulness-based stress reduction on weight among breast cancer and chronic pain patients compared with a specific active control.

Transcendental Meditation Versus a Nonspecific Active Control

Three trials of TM evaluated weight as an outcome (Table 16). Elder et al. randomized adults with elevated HgA1c (n=54) to a TM program versus diabetes education classes77. There were no differences between the groups in weight loss (p=.26). This trial did not report on the amount of training provided or the duration of the training. It did specify it recommended about 90 hours of home practice over 6 months. The teachers were trained teachers of TM. This trial had a medium risk of bias.

Castillo-Richmond et al. conducted a trial of TM using a subsample from a larger randomized trial of TM on cardiovascular outcomes (n=60 of 170 from the original trial).76 This trial found no difference in weight after 7 months between the groups (p=.48). This trial did not specify the amount of training provided, but did specify it recommended up to 120.6 hours of home practice over 7 months. The teachers had training and certification in the TM tradition. This trial had a high risk of bias, due largely to uncertain sampling methods from the primary trial.

Schneider et al.90 randomized black adults with coronary artery disease to TM or health education. The study followed patients for an average of 5.4 years. The study estimated they received approximately 78 hours of training over this time by trained and certified teachers, along with approximately 1,310 hours of homework. After an average of 5.4 years, there were no differences in BMI between the two groups. This trial had a low risk of bias.

The difference-in-change graphs showed a consistent null effect on weight (Figure 38). The strength of evidence is low that mantra meditation programs do not have an effect on weight among diabetics, hypertensives, or those with coronary disease when compared with a nonspecific active control. We based this rating on overall medium risk of bias, consistent null findings, directness of measures, and imprecise estimates (Table 45).

Table 45. Grade of trials addressing the efficacy of meditation programs on weight among those with a clinical condition.

Table 45

Grade of trials addressing the efficacy of meditation programs on weight among those with a clinical condition.

Assessment of Potential Publication Bias

We could not conduct any reliable quantitative tests for publication bias since few studies were available for most outcomes, and we were unable to include all eligible studies in the meta-analysis due to missing data. Consequently, funnel plots were unlikely to provide much useful information regarding the possibility of publication bias. We reviewed the clinicaltrials.gov registration database to assess the number of trials that had been completed three or more years ago and that prespecified our outcomes but did not publish at all or did not publish all outcomes that were prespecified. We found 5 trials on clinicaltrials.gov that appeared to have been completed before Jan 1, 2010 that were published but did not publish the results of all outcomes they had prespecified on the registration Web site. We also found 9 trials that appeared to have been completed before January 1, 2010 but that we could not find any publication for, and had prespecified at least one of our outcomes. 10 registered trials had prespecified one or more KQ1 outcomes but did not publish them, 2 registered trials had prespecified attention as an outcome but did not publish, 5 registered trials prespecified one or more KQ3 outcomes but did not publish, and 5 registered trials prespecified one or more KQ4 outcomes but did not publish. For 8 of the 9 registered trials for which we could not find a publication, it was not possible to tell if those trials had actually been conducted or completed While examining for selective outcome reporting, we found one trial that selectively reported on positive outcomes. Among 109 outcomes in 41 trials, trials did not give enough information to calculate a relative difference-in-the-change score (our primary analysis) for six outcomes due to statistically insignificant findings. These are represented as solid grey boxes in the figures. Trials did not give enough information to conduct a meta-analysis on 16 outcomes. Our findings from the primary analysis are therefore less likely to be affected by publication bias than the meta-analysis.

Applicability

Eleven of 14 trials took place in the United States, the remainders took place in Canada, Germany, and Hong Kong, so that these findings might be inapplicable to patients or settings in other regions. All of the trials took place in outpatient settings, so these findings would not be applicable to the inpatient setting.

Regarding the population characteristics of the trials for this KQ, only one trial reported the racial or ethnic characteristics of its study population. In addition, these trials under represent younger patients (less than 45) and older patients (age greater than 80), making these findings less applicable to those groups. The most important proviso regarding the population characteristics is that the trials for this KQ were of two different kinds: those in populations with chronic pain, and those predominantly with another condition. Thus, the populations in these trials were heterogeneous as a group, making it difficult to identify the clinical populations for which these findings would be most applicable.

Characteristics of the interventions represented in this KQ were diverse, making it difficult to foresee how these findings would be applicable to a similarly wide array of mindfulness practices under everyday clinical situations. For example, only a few trials specified the certification and training of instructors or the time spent in home training.