We identified and screened 16816 records from 10 academic databases. After removing duplicates, 10694 studies were assessed for title and abstract screening. Out of these, 10461 studies were excluded as they did not meet the eligibility criteria. Two reviewers independently screened 233 full-text articles and included 24 studies for meta-analysis. We manually searched and screened the references of the primary studies and included 41 more studies for meta-analysis. A total of 65 RCTs with 8009 participants, were included in the systematic review and meta-analysis. Figure 1 describes the study selection process.
Of the 65 studies, 56 were from High Income Countries, 6 studies were from Upper Middle-Income Countries and only 1 study was from a Lower- and Middle-Income country (i.e., India). In majority of the studies (96%, 62/65), the study participants were students. Mean age of adolescents was 19.09 (±2.92). In 80% (52/65) of the studies, participants were both male and female adolescents. Studies included adolescents with depression (21%, 14/65), anxiety (9%, 6/65), distress (20%, 13/65), combined symptoms of anxiety, distress and depression (17%, 11/65) and behavioural problems, academic concerns and eating problems (11%, 7/65). In 21% of the studies (14/65), the condition of the study participants was not specified (see Table 2 for the characteristics of the included studies).
Of the 65 studies, 12 (18%) studies reported ‘relaxation technique/s’ as a standalone intervention to reduce symptoms of anxiety, distress and depression in adolescents. The most commonly reported ‘relaxation techniques’ in the literature were Progressive Muscle Relaxation Techniques-PMR, breathing, exercise, walking meditation, stretching, relaxation imagery and meditation. In 82% (53/65) of the studies relaxation techniques were implemented as an integral component of other psychotherapies, multicomponent interventions such as Cognitive Behavioural Therapy (CBT), Mindfulness and Acceptance and Commitment Therapy (ACT). Other components of multicomponent interventions included identifying affect, psychoeducation and mindfulness exercises. A list of ‘relaxation techniques’ reported in the studies are mention in Table 3. The detail on characteristics of included interventions is mentioned in Table 6, supplementary material).
Interventions in 34/65 (52%) studies were delivered in an educational setting (school/colleges/universities). Relaxation techniques (either as a standalone or in combination with other elements) were delivered in group format in 57% (37/58) of the studies. Among the studies reporting relaxation techniques as a standalone intervention, the mean number of sessions was 22.27 (±12.48), over 9.55 (±2.92) weeks. Mean session duration was 66.5 (±32.32) minutes. Out of 12 studies, two studies did not report program duration; one study was of 2 hours and one study did not report session duration. The average program duration for multicomponent interventions was 7.39 (±6.35) weeks; average number of sessions were 9.04 (±8.40) and average session duration was 72.14 minutes (±37.61). Among studies using multicomponent interventions, nine studies did not report program duration; one study was of 2 hours, three studies did not report session duration and 11 studies did not report session duration. Booster sessions were delivered in 3 trials. In 37/65 (57%) studies, relaxation interventions (either as a standalone or in combination with other elements) were delivered by the specialists. Relaxation based interventions were self-administered in 18/65 (28%) studies and the intervention was delivered online in 4/65 (6%) studies. In 62/65 studies, mental health was reported as a primary outcome.
Different outcome measures were reported to measure anxiety, distress and depression in the included studies. In majority of the studies (n=12) Depression, Anxiety and Stress Scale (DASS) was used to measure all three outcomes (distress, anxiety and depression). The outcome measures used for anxiety were Spence Children’s Anxiety Scale (SCAS) (n=3), The Spielberger State-Trait Anxiety Inventory (STAI) (n=3), State Trait anxiety (STAI) (n=3), The Anxiety Sensitivity Index (n=3) and Beck Anxiety Inventory (BAI) (n=2). Many studies used several other tools for anxiety (n=20). The outcome measures used for depression were Beck Depression Inventory (BDI) (n=9), Centre for Epidemiologic Studies Depression Scale (CES-D) (n=7), Reynolds’ Adolescent Depression Scale (RADS) (n=4), Patient Health Questionnaire (PHQ-9) (n=3), Child Depression Rating Scale Revised (CDACR) (n=3). Moreover, various studies used different types of tools to measure depression (n=12). Outcome measures used to measure distress were Perceived Stress Scale (n=13), Penn State Worry Questionnaire (PSWQ) (n=3), Kessler Perceived Distress Scale (n=2), General Health Questionnaire (GHQ) (n=2). Other studies used different tools to measure distress (n=3) (list of outcome measures used in studies is given in Table 8, supplementary material).
In the included studies, relaxation techniques were compared with control groups, including wait-list controls (n=16), treatment-as-usual (n=7), no active intervention (n=23), placebo control (n=1) or active controls (n=18).
Effect of relaxation-based interventions to reduce symptoms of anxiety
Meta-analysis was conducted with 8009 participants. Effectiveness of relaxation-based interventions in the treatment of anxiety was explored in 46 studies, with a cumulative sample size of 5234 participants (2486 in intervention and 2748 in control arm). Meta-analysis with random effects model showed that relaxation techniques were effective to reduce the symptoms of anxiety in adolescents at post-intervention (SMD -0.386, 95% CI: -0.52 to -0.25) (Figure 2). There was evidence for substantial heterogeneity across the studies (I2= 79.16 %, t2= 0.15). No significant changes in pooled effect size were observed on sensitivity analysis. Egger’s regression statistic was significant (t= 2.32, p= 0.02, see supplementary material, Figure 5), demonstrating significant publication bias in reporting of anxiety outcome. The pooled effect size increased after adjusting for the publication bias (SMD-0.54, 95% CI: -0.69 to -0.40).
Effect of relaxation-based interventions to reduce symptoms of distress
Distress was reported as an outcome in a total of 23 studies, with a cumulative sample size of 2246 (1122 participants in intervention arm and 1124 participants in control arm). There was a substantial heterogeneity in reporting of this outcome (I2= 85.08%, t2= 0.26). Meta-analysis revealed a moderate effect size in favour of the intervention group (SMD= -0.48, 95% CI: -0.71 to -0.24) (Figure 3) to reduce the symptoms of distress in adolescents. Removal of outlier studies in sensitivity analyses did not reveal any significant changes in the pooled effect size for distress outcome. Egger’s regression statistic revealed a non-significant publication bias in reporting of distress outcome (P=0.30, see supplementary materials, Figure 6).
Effect of relaxation-based interventions to reduce symptoms of depression
Depression was reported as an outcome in a total of 50 studies, with a cumulative sample size of 5732 participants (2719 in intervention arm and 3013 in control arm). There was small evidence to support relaxation techniques being effective to reduce the symptoms of depression in adolescents (SMD= -0.28 (95% CI: -0.40% to -0.15) (Figure 4). There was a substantial heterogeneity in reporting of depression outcome (I2= 76.82%, t2= 0.13). The sensitivity analysis did not result in any change in the pooled effect size for depression outcome. There was no evidence of publication bias (Egger’s regression p=0.36, see supplementary material, Figure 7).
Sub-group and moderator analyses
We did a sub-group and meta-regression analyses to evaluate in which settings relaxation techniques work to reduce the symptoms of anxiety, distress and/or depression in adolescents. Anxiety- Subgroup analyses revealed that face-to-face, individually delivered, multicomponent interventions with a booster session yielded a higher effect size to reduce the symptoms of anxiety in adolescents. However, these results were statistically non-significant (Table 4). Data regarding intervention dosage (including number of sessions; duration of sessions and overall programme) was provided in 30 studies reporting anxiety outcome. To analyse association of dosage of intervention with improvement in the symptoms of anxiety (effect size) meta-regression analyses were run, after removing three of the outlier studies (34-36). The density of intervention dosage explained 44% of the variation in overall effect size (Table 5) (Figure 8, 9 & 10 in supplementary material). Age of participants explained 13% of variance in anxiety outcome, where age associated inversely with effect size (B= -0.07, SE= 0.03, p=0.04). Depression- Subgroup analyses did not reveal any significant subgroup differences based on the mode and format of intervention delivery, types of intervention and booster sessions for depression outcome (Table 3). A total of 33 studies reported statistics pertaining to density of dosage of interventions. The moderator analysis of the data revealed that characteristics of ‘relaxation techniques’ (format, dosage, components) did not have any effect to reduce the symptoms of depression in adolescents. Multivariate meta-regression analyses revealed that 5% of the variation in effect size was explained by density of dosage of interventions. None of the indicators of the dose of interventions reached statistical significance (Table 5, Figure 11, 12 & 13 supplementary material). Age did not yield any significant association with effect size, explained only 4% of variance in the depression outcome (B=-0.04, Se=0.04, p=0.33). Distress- Subgroup analyses revealed that studies with low risk of bias, multicomponent, individually delivered face-to-face interventions, with a booster dose were more effective in reducing the symptoms of distress. Only mode of delivery and booster dose were found to be statistically significant (Table 4). Multivariate regression model pertaining to density of intervention dosage of explained 47% of variation in effect size for distress outcome. None of the indicators of dose of interventions emerged as a significant predictor in this model (Table 5, Figure 14, 15 & 16 supplementary material). The results remained consistent even after removal of the outlier study (36). Age did not yield any significant association with effect size, explained only 1% of variance to reduce the symptoms of distress (B=-0.03, SE=0.06, p=0.64).
The results show that active components (i.e., individual effect of each intervention component) of interventions explained 9% to 25% of variance in heterogeneity across studies targeting distress, anxiety, and/or depression. Highest variance in heterogeneity amongst different studies was explained in the interventions targeting anxiety as an outcome (25%). Overall, multicomponent interventions were associated with a better improvement in anxiety than relaxation alone, still this effect was statistically non-significant.
Certainty of outcomes pertaining to anxiety, depression and distress using GRADE (Grading of Recommendations, Assessment, Development and Evaluations) framework
Certainty of outcomes pertaining to anxiety, distress and depression reported in the trials was assessed using the GRADE guidelines (37). The certainty for the outcome of anxiety was downgraded by three levels to very low for serious concerns pertaining to risk of bias in the studies, substantial heterogeneity and publication bias in reporting of this outcome. The outcomes of distress and depression were downgraded to low by two levels, due to high risk of bias in intervention design and presence of substantial heterogeneity, explained by clinically heterogeneous study samples and interventions (see Table 7, supplementary material).