Meditation for adults with non-specific low back pain: a systematic review and meta-analysis

Data sources and searches

This systematic review was prospectively registered on PROSPERO (CRD42017074693) and reported according to the PRISMA statement [32]. We searched PubMed, EMBASE, PEDro, Scopus, Web of Science, Cochrane Library and PsycINFO from their inception up to April 17, 2020. There were no restrictions with regards to year of publication or language. The detailed search strategy is described in Appendix. The study selection process involved screening the titles and reading the abstracts, after which potentially relevant articles were obtained in full text for further eligibility analysis. Two independent reviewers (LS, FR) performed the selection of the studies and, in the case of disagreement, a third reviewer obtained a consensus through discussion or arbitration (LN). Manual searches were performed by checking the reference lists of each eligible article by two authors (LS, FR).

Study selection

We only included randomized controlled trials published in peer-reviewed journals testing the effectiveness of CM and MM (including standardized mindfulness-based programs such as Mindfulness-Based Stress Reduction – MBSR, and Mindfulness-Based Cognitive Therapy – MBCT) practices in adults with non-specific low back pain aged 18 years or older [33]. We included trials that used meditation associated with some kind of exercises (mindful walking, stretching). However, interventions that were based mainly on physical practices such as Tai Chi, Yoga and Qigong were not included. Exclusion criteria consisted of (i) trials that enrolled participants with serious spinal pathologies, such as cancer, tumor, fractures, infection; (ii) participants with history of spinal surgery; (iii) trials involving pregnant woman; (iv) trials that included participants with evidence of nerve root compromise; and (v) trials that combined other types of interventions with meditation.

Outcome measure

The primary outcomes of this review were pain intensity and disability. Our secondary outcomes were symptoms of anxiety, symptoms of depression, kinesiophobia, pain catastrophizing, quality of life, and adverse events (nature and seriousness) reported in the trials included.

Data extraction and assessment of risk of bias

The data extraction from the eligible trials was performed using a standardized data extraction form containing: author (year), study site, sample size, duration of non-specific low back pain in months, mean age, details of the intervention group, details of the control group, outcomes, means and standard deviations of primary and secondary outcomes, duration of the intervention and time periods for outcome assessment and dropout rate. When the studies did not provide enough information for the analysis, we contacted the authors to request additional data. Two review authors (LS, FR) independently extracted data from all included studies. We resolved disagreements through discussion or arbitration by a third review author (LN).

Two reviewers (GF and FR) assessed risk of bias using the Physiotherapy Evidence Database - PEDro scale. Discrepancies between authors were resolved by discussion; in case of disagreement, a third reviewer was consulted (LN). This scale assesses the risk of bias and statistical reporting of randomized controlled trials considering 11 items: eight items relate to methodological quality (i.e., random allocation, concealed allocation, baseline comparability, blinded subjects, blinded therapists, blinded assessors, adequate follow-up and intention-to-treat analysis) and two items relate to the statistical reporting (between-group comparisons, and point estimates and variability). The first item (eligibility criteria) is not considered in the total score since it is related to external validity. The total PEDro score ranges from 0 to 10 points; higher scores mean greater methodological quality. This scale has good levels of validity and reliability [34], [35], [36]. There is recent evidence of convergent validity between the PEDro scale and the Cochrane risk of bias tool [37, 38].

Data analysis

Data were analyzed in Revman 5.3 (Review Manager, version 5.3 - The Nordic Cochrane Center, collaboration Cochrane, Copenhagen, Denmark). Data were pooled using a random-effects model. We used intention-to-treat analyses preferably over per-protocol (as-treated) analyses. For continuous outcomes, the standardized mean differences (SMDs) were calculated from extracted means and standard deviations (SD) collected post-intervention and at follow-up. The studies were grouped into two comparisons: (1) meditation vs. minimal intervention (waiting list, education); (2) meditation vs. usual care. The follow-up were grouped as follows: short-term (less than three months after randomization), intermediate-term (at least three months but less than 12 months after randomization), and long-term (12 months or more after randomization) follow-ups. The interpretation of the effect sizes (ES) was: small (SMD = 0.2); medium (SMD = 0.5); and large (SMD = 0.8) [39]. The heterogeneity (I²) was evaluated considering a value of I² close to 0% indicating no heterogeneity among the studies, close to 25% (low heterogeneity), close to 50% (moderate heterogeneity) and close to 75% (high heterogeneity) [40].

Certainty of the evidence

We used the GRADE approach including five aspects (study design and risk of bias, consistency of effect, imprecision, indirectness and publication bias) to assess the certainty of evidence [41]. The certainty of the evidence was determined using the approach by the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) working group [41]. The certainty of the evidence was graded as high, moderate, low, or very low based upon five main domains: (i) study design and risk of bias (downgraded by one level if >25% of participants were from studies with a PEDro score <7, and by two levels if >75% of participants were from studies with a PEDro score <7); (ii) inconsistency of findings (downgraded if I² > 50%); (iii) indirectness (generalizability of the findings; downgraded if >50% of the participants were from a different population); (iv) imprecision (downgraded if fewer than 400 participants were included in each comparison); and (v) publication bias. The certainty of the evidence was graded as: (i) high, which indicates that further research is unlikely to alter confidence in the effect estimate; (ii) moderate, which indicates that further research is likely to significantly alter confidence in the effect estimate and may change the estimate; (iii) low, which indicates that further research is likely to significantly alter confidence in the effect estimate and to change the estimate; and (iv) very low, which indicates that any effect estimate is uncertain.

Search results

The initial search of the databases identified a total of 3,575 studies. Duplicate studies were excluded resulting in a total of 2,424 articles. The screening of titles and abstracts identified 105 potentially eligible articles, and after a detailed analysis of the full text of the selected studies, eight randomized controlled trials fulfilled the inclusion/exclusion criteria (Figure 1). The main reasons for excluding articles were: patients with other chronic pain conditions (n = 1) and history of spinal surgery (n = 1), studies that were not a randomized controlled trial (n = 3). The median age of participants reported in the trials was 55 (IQR = 40 to 74; range = 40.3 to 78) years of age. The duration of the meditation programs ranged from 3 to 8 weeks and the frequency of practice varied from 60 to 90 min per week. The median time of follow-up of the studies was 10 (IQR = 8 to 24.5; range = 4 to 52) weeks. The median of dropout rate was 12% (IQR = 12 to 22.5; range = 5.7 to 44.3).

Figure 1:

Selection of studies for inclusion in the systematic review.

Characteristics of included studies

Eight randomized controlled trials were included with a total of 1,234 patients with chronic non-specific low back pain. Most trials provided MM (n=1,166) [42], [43], [44], [45], [46], [47], [48] in the experimental group with the exception of one which provided CM (n=68) (i.e., jyoti meditation – a practice for controlling and directing attention away from the physical body and sensations, from the emotions and thoughts to a place of relaxation or peace within the organism) [49].

Seven trials (n=892) assessed pain intensity at short-term [42], [43], [44], [45], [46, 48, 49], four trials (n=699) at intermediate-term [43], [44], [45, 48] and one trial (n=342) at long-term [43]. Disability levels were assessed in six trials (n=804) at short-term [43], [44], [45], [46, 48, 49], four trials (n=699) at intermediate-term [43], [44], [45, 48] and one trial (n=342) at long-term [43]. Regarding secondary outcomes, six trials (n=857) assessed quality of life [42], [43], [44], [45], [46, 49], four trials (n=732) assessed global impression of change [43], [44], [45, 49], two trials (n=410) assessed symptoms of anxiety and depression [43, 49], and two trials (n=624) assessed pain catastrophizing [45, 47]. None of the trials included assessed kinesiophobia. Table 1 presents the characteristics of the included studies. Data from one trial (n=40) [44] were excluded from quantitative analysis since data were not available in the publication and the authors did not provide information on request.

Risk of bias of the included studies

The scores of methodological quality and statistical reporting of the eligible trials are presented in Table 2. The total PEDro score ranged from three to 8. Five studies (62.5%) were considered as having low risk of bias (PEDro score ≥7). Most studies satisfied the items related to random allocation, baseline comparability, between-group comparisons, and point estimates and variability. The items least frequently satisfied were related to blinding of therapists, participants and outcome assessors.

Effect of meditation vs. minimal intervention

Primary outcomes

We included a total of three trials (n=387) in this comparison [45, 46, 49]. From those, two trials (n=350) presented low risk of bias [45, 49] and one (n=37) presented high risk of bias [46].

For pain intensity, three trials (n=387) [45, 46, 49] provide moderate-certainty evidence (downgraded due to imprecision) that there is no significant difference between meditation and minimal intervention at short-term (SMD = −0.14; 95% CI = −0.36 to 0.08). One trial (n=282) [45] provides low-certainty evidence (downgraded by two levels for imprecision) that there is no significant difference between meditation and minimal intervention at intermediate-term (SMD = −0.22; 95% CI = −0.46 to 0.01) (Figure 2A).

Figure 2:

Forest plot of comparison: Meditation vs. minimal intervention, outcome: Pain intensity (A); outcome: Disability (B).

With regards to disability, based upon three trials (n=387) [45, 46, 49] and moderate-certainty evidence (downgraded due to imprecision), there is a small effect in favor of meditation at short-term follow-up (SMD = −0.22; 95% CI = −0.42 to −0.02). At intermediate-term follow-up, one trial (n=282) [45] provides low-certainty evidence (downgraded by two levels for imprecision) that there is no difference in disability levels when compared to minimal intervention (SMD = −0.08; 95% CI = −0.31 to 0.15) (Figure 2B).

Secondary outcomes

For mental health-related of quality of life, based on two trials (n=319) [45, 46] and low-certainty evidence (downgraded due to risk of bias and imprecision), no difference between meditation and minimal intervention was observed at short-term (SMD = −0.10; 95% CI = −0.30 to 0.49) (Figure 4A). For physical health-related of quality of life, based on three trials (n=387) [45, 46, 49] and low-certainty evidence (downgraded due to risk of bias and imprecision) no difference was observed at short-term (SMD = −0.11; 95% CI = −0.10 to 0.31) and based on a single trial (n=282) [45] and low-certainty evidence (downgraded by two levels for imprecision) no difference was observed at intermediate-term (SMD = −0.00; 95% CI = −0.23 to 0.23) (Figure 4B). Based on a single trial (n=68) [49] and low-certainty evidence (downgraded by two levels for imprecision), no differences for symptoms of anxiety (SMD = −0.12; 95% CI = −0.36 to 0.59) and for depression (SMD = −0.12; 95% CI = −0.36 to 0.59) were found at short-term follow-up. None of the evaluated trials measured these outcomes at intermediate and long-term. For pain catastrophizing, based on a single trial (n=282) [45] and low-certainty evidence (downgraded by two levels for imprecision) no difference between meditation and minimal intervention was observed at short-term (SMD = 0.12; 95% CI = −0.14 to 0.38), at intermediate-term (SMD = −0.07; 95% CI = −0.33 to 0.19) and at long-term (SMD = −0.04; 95% CI = −0.22 to 0.30). Based on a single trial (n=282) [45] and low-certainty evidence (downgraded by two levels for imprecision), 106 participants (80.3%) in the intervention group and 51 (37.0%) in the control group reported perceived improvement in back pain symptoms as a result of the intervention at short-term (RR = 2.17; 95% CI = 1.72 to 2.74). At intermediate-term 89 participants (76.1%) stated at least minimal improvement compared with 57 (42.2%) in the control group (RR = 1.81; 95% CI = 1.45 to 2.26).

Qualitative analysis

The trial included in the qualitative analysis (n=40) [44], compared meditation (MM 8-week program) to minimal intervention (health education) in a sample of older adults (≥65 years). The MM group and the control group consisted of 16 and 19 people, respectively. The authors did not find differences between groups for the outcomes pain intensity, disability and quality of life at both short and intermediate-term follow-ups.

Effect of meditation vs. usual care

Primary outcomes

Four trials (n=807) were included in this comparison [42, 43, 47, 48]. For pain intensity, based on low-certainty evidence (downgraded due to risk of bias and inconsistency), there was no difference between meditation and usual care (SMD = −0.68; 95% CI = −1.87 to 0.52) [42, 43, 48] at short-term follow-up. At intermediate-term follow-up (n=377), based on low-certainty evidence (downgraded due to imprecision and inconsistency), there is no difference between meditation and usual care (SMD = 0.08; 95% CI = −0.49 to 0.65) [43, 48]. At long-term follow up, based on one trial (n=342) [43] and low-certainty evidence (downgraded by two levels for imprecision), there is a statistically significant difference in pain intensity (SMD = −0.28; 95% CI = −0.54 to −0.02) (Figure 3A).

Figure 3:

Forest plot of comparison: Meditation vs. usual care, outcome: Pain intensity (A); outcome: Disability (B).

*Quality of life mean values were converted to positive into negative to keep the consistency of the forest plot.

Regarding disability, two trials (n=377) [43, 48] were included in the meta-analysis. There is moderate-certainty evidence (downgraded due to imprecision) that there is no difference between meditation and usual care at short-term (SMD = −0.08, 95% CI = −0.32 to 0.17), intermediate-term follow-up (SMD = −0.07, 95% CI = −0.32 to 0.17) and at long-term follow-up (SMD = −0.20, 95% CI = −0.46 to 0.06) (Figure 3B).

Secondary outcomes

For mental health-related of quality of life, based on two trials (n=430) [42, 43] and low-certainty evidence (downgraded due to risk of bias and inconsistency), meditation was better to usual care (SMD = −0.74; 95% CI = −1.42 to −0.06) at short-term. This difference was not observed at intermediate-term (SMD = −0.24; 95% CI = −0.50 to 0.02). Based on a single trial (n=342) [43] and low-certainty evidence (downgraded by two levels for imprecision), there is a significant difference in favor of meditation at long-term follow-up (SMD = −0.26; 95% CI = −0.52 to −0.00) (Figure 4C). For physical health-related of quality of life, based on two trials (n=430) [42, 43] and very low-certainty evidence (downgraded due to risk of bias, inconsistency and imprecision), meditation was better to usual care at short-term with a large effect size (SMD = −3.07; 95% CI = −4.05 to −2.10). On the other hand, based on a single trial (n=342) [43] and low-certainty evidence (downgraded by two levels for imprecision) there was no difference between meditation and usual care at intermediate-term (SMD = 0.60; 95% CI = −1.64 to 2.84 to) and at long-term (SMD = −0.30; 95% CI = −2.39 to 1.79) (Figure 4D). Based on a single trial (n=342) [43], and low-certainty evidence (downgraded by two levels for imprecision), there are no differences for symptoms of anxiety at short-term (SMD = −0.22; 95% CI = −0.48 to 0.04), intermediate-term (SMD = −0.05; 95% CI = −0.31 to 0.21), and long-term (SMD = −0.05; 95% CI = −0.31 to 0.21).

Figure 4:

Forest plot of comparisons: Meditation vs. minimal intervention (A and B), outcome: Quality of life mental (A); Quality of life – physical (B). Meditation vs. usual care (C and D), outcome: Quality of life mental (C); Quality of life – physical (D).

For depression, based on a single trial (n=342) [43], there is low-certainty evidence (downgraded by two levels for imprecision) that meditation reduces symptoms of depression compared with usual care at short-term follow-up, with a medium effect size (SMD = −0.39; 95% CI = −0.65 to −0.13). The same trial reported a non-significant difference in symptoms of depression at intermediate-term (SMD = −0.18; 95% CI = −0.44 to 0.08) and long-term follow-up (SMD = −0.18; 95% CI = −0.44 to 0.08).

For pain catastrophizing, based on a single trial (n=342) [47] and low-certainty evidence (downgraded by two levels for imprecision), no differences were reported at short-term (SMD = 0.12; 95% CI = −0.14 to 0.38), intermediate-term (SMD = −0.07; 95% CI = −0.33 to 0.19), and long-term (SMD = 0.04; 95% CI = −0.22 to 0.30).

Global impression of change was presented in one study (n=342) [43] with a significant difference at intermediate- and long-term. At intermediate-term, 30 (26.2%) patients in the meditation group reported that their pain was much better or had completely gone in comparison to 15 (13.6%) patients in the usual care (RR = 1.94; 95% CI = 1.10 to 3.42). At long-term, a 34 (30%) patients in the meditation group reported that their pain was much better or had completely gone in comparison to 20 (18%) patients in the usual care (RR = 1.65; 95% CI = 1.02 to 2.70). None of the included trials investigated kinesiophobia. Supplementary material 1 provides the summary of findings.

Adverse events

Adverse effects of meditation were assessed in six trials (n=797), being five related to MM (n=736) [43], [44], [45], [46, 48] and one to jyoti meditation (n=68) [49]. Three trials (n=359) [44], [45], [46] reported no serious adverse events. In the study with jyoti meditation (n=68) [49], the authors reported an increase of an already existing tinnitus during the first weeks of in two participants in the meditation group and one participant reported slightly increased dizziness and headache during the meditation. Cherkin et al. [43] reported a temporary increase in pain intensity in at least one meditation session in 29% (n = 103) of patients. Zgierska et al. [48] also observed mild and self-limited adverse effects during the intervention, such as an increase in pain intensity of 0.7% (n = 2); increased anxiety during practice 1.5% (n = 3); increase in smoking 0.3% (n = 1) and increase in weight 0.3% (n = 1).

Main findings

This study aimed to summarize the evidence from trials evaluating the effects of meditation (MM or CM) for adults with chronic non-specific low back pain. In the current review, meditation group was compared to minimal intervention or usual care. The reason to group waitlist and booklet into minimal intervention was based on the following criteria: the intervention was not specific to treat patients with chronic low back pain (education program about healthy aging) [45]; a wait-list for meditation plus a self-care manual [49]. Usual care consisted mainly in physiotherapy, medical care and/or mental health. Moderate-certainty evidence suggested that meditation may reduce disability levels with small effect size when compared with minimal intervention at short-term follow-up period. Low-certainty evidence suggested a small effect size for pain relief with meditation in the long-term but this finding was based on only one trial. Low-certainty evidence showed that meditation provided medium effect size on mental health-related quality of life in the short-term when compared to usual care. There is very low-certainty evidence that meditation has large effect size on physical health-related quality of life. Moderate-certainty evidence suggests that meditation improved symptoms of depression with a medium effect size in the short-term in comparison to usual care. In the few comparisons where we identified a statistically significant difference, the magnitude of the effects were small, which are in line with most interventions for patients with chronic low back pain.

Comparison with literature

Our findings are partly in line with previous systematic reviews that investigated the effect of MM including people with specific and non-specific low back pain [18, 19]. Based on one trial composed of people with failed back surgery syndrome, Cramer et al. [19] reported significant and clinically important short-term improvements in pain intensity and disability for MM compared to no treatment. In the most recent review, Anheyer et al. [18] found statistically, but not clinically important, improvement in pain intensity when compared with usual care at short-term follow-up. The authors also reported improvements on physical health–related quality of life at short-term, but not for mental health–related quality of life nor disability at short or long-term. Divergences may be explained by three reasons: (i) both reviews included one trial composed of people submitted for lumbosacral spinal surgery; (ii) different methods to group treatment comparisons; and (iii) the inclusion of one trial with a different meditation practice in this review.

Strengths and limitations

The strengths of the current review remains on the inclusion of several outcomes such as symptoms of anxiety, depression, pain catastrophizing, and global perception of change that were not reported previously. Besides, we included the overall certainty of the evidence assessed using the GRADE approach, which is the most widely used approach for summarizing confidence in effects of interventions by outcome across studies [50]. On the other hand, some limitations should be addressed. First, the small number of studies included with some results provided by single trials. Second, interpretation of the results should consider the sample characteristics (mean age higher than 50 years), different meditation programs, instructor experience, and outcome measures. Third, although we tried to include other meditation practices, seven out of eight trials in our review had MM as the intervention. Last, we had a deviation from protocol in respect to the outcome return to work. None of the trials included in the current review assessed this outcome. We recognize that return to work is a complex and difficult outcome to assess in the general population. This is more complex when comparing across several studies where it is possible to have different proportions of working/non-working participants in different studies.

Clinical and research implications

In the included studies, the duration of the sessions ranged from 30 min to 2 h, with a weekly meeting for 8–12 weeks and encouraging daily home practice, which may be an important barrier for adherence [33]. Professionals who conducted the meditation programs were all instructors with experience ranging from 5 to 30 years. This finding highlights that meditation intervention probably needs long periods of training. Therefore, in clinical practice, the decision to use meditation as an intervention in people with non-specific chronic low back pain may depend on treatment availability, instructor expertise and, importantly, participant or provider preference. Future research can elucidate the effects of different types of meditation, the acute effects of meditative practice, the dose of meditation programs, the effects when conducted with different degrees of supervision (individual, group or remotely) and instructor experience, as well as identifying whether there is a subgroup of patients who benefit most from meditative practices. In addition, it is important to measure the feasibility and economic evaluation alongside a clinical trial of the meditation.