Effects of music therapy on depression: A meta-analysis of randomized controlled trials

Background We aimed to determine and compare the effects of music therapy and music medicine on depression, and explore the potential factors associated with the effect. Methods PubMed (MEDLINE), Ovid-Embase, the Cochrane Central Register of Controlled Trials, EMBASE, Web of Science, and Clinical Evidence were searched to identify studies evaluating the effectiveness of music-based intervention on depression from inception to May 2020. Standardized mean differences (SMDs) were estimated with random-effect model and fixed-effect model. Results A total of 55 RCTs were included in our meta-analysis. Music therapy exhibited a significant reduction in depressive symptom (SMD = −0.66; 95% CI = -0.86 to -0.46; P<0.001) compared with the control group; while, music medicine exhibited a stronger effect in reducing depressive symptom (SMD = −1.33; 95% CI = -1.96 to -0.70; P<0.001). Among the specific music therapy methods, recreative music therapy (SMD = -1.41; 95% CI = -2.63 to -0.20; P<0.001), guided imagery and music (SMD = -1.08; 95% CI = -1.72 to -0.43; P<0.001), music-assisted relaxation (SMD = -0.81; 95% CI = -1.24 to -0.38; P<0.001), music and imagery (SMD = -0.38; 95% CI = -0.81 to 0.06; P = 0.312), improvisational music therapy (SMD = -0.27; 95% CI = -0.49 to -0.05; P = 0.001), music and discuss (SMD = -0.26; 95% CI = -1.12 to 0.60; P = 0.225) exhibited a different effect respectively. Music therapy and music medicine both exhibited a stronger effects of short and medium length compared with long intervention periods. Conclusions A different effect of music therapy and music medicine on depression was observed in our present meta-analysis, and the effect might be affected by the therapy process.

Introduction music therapy at a hospital or community, whereas the control group not receiving any type of music therapy; (3) depression rating scale was used. The exclusive criteria were as follows: (1) non-human studies; (2) studies with a very small sample size (n<20); (3) studies not providing usable data (including sample size, mean, standard deviation, etc.); (4) reviews, letters, protocols, etc. Two authors independently (YPJ, HZH) searched and screened the relevant papers. EndNote X7 software was utilized to delete the duplicates. The titles and abstracts of all searched papers were checked for eligibility. The relevant papers were selected, and then the full-text papers were subsequently assessed by the same two authors. In the last, a panel meeting was convened for resolving the disagreements about the inclusion of the papers.

Data extraction
We developed a data abstraction form to extract the useful data: (1) the characteristics of papers (authors, publish year, country); (2) the characteristics of participators (sample size, mean age, sex ratio, pre-treatment diagnosis, study period); (3) study design (random allocation, allocation concealment, masking, selection process of participators, loss to follow-up); (4) music therapy process (music therapy method, music therapy period, music therapy frequency, minutes per session, and the treatment measures in the control group); (5) outcome measures (depression score). Two authors independently (TQS, ZH) abstracted the data, and disagreements were resolved by discussing with the third author (YPJ).

Assessment of risk of bias in included studies
Two authors independently (TQS, ZH) assessed the risk of bias of included studies using Cochrane Collaboration's risk of bias assessment tool, and disagreements were resolved by discussing with the third author (YPJ) [10].

Music therapy and music medicine
Music Therapy is defined as the clinical and evidence-based use of music interventions to accomplish individualized goals within a therapeutic relationship by a credentialed professional who has completed an approved music therapy program. Music medicine is defined as mainly listening to prerecorded music provided by medical personnel or rarely listening to live music. In other words, music medicine aims to use music like medicines.
Music therapy mainly divided into active music therapy and receptive music therapy. Active music therapy, including improvisational, re-creative, and compositional, is defined as playing musical instruments, singing, improvisation, and lyrics of adaptation. Receptive music therapy, including music-assisted relaxation, music and imagery, guided imagery and music, lyrics analysis, and so on, is defined as music listening, lyrics analysis, and drawing with musing. In other words, in active methods participants are making music, and in receptive music therapy participants are receiving music [6,7,9,[11][12][13].

Statistical analysis
The pooled effect were estimated by using the standardized mean differences (SMDs) and its 95% confidence interval (95% CI) due to the different depression rate scales were used in the included papers. Heterogeneity between studies was assessed by I-square (I 2 ) and Q-statistic (P<0.10), and a high I 2 (>50%) was recognized as heterogeneity and a random-effect model was used [14][15][16]. We performed subgroup analyses and meta-regression analyses to study the potential heterogeneity between studies. The subgroup variables included music intervention categories (music therapy and music medicine), music therapy methods (active music therapy, receptive music therapy), specific receptive music therapy methods (music-assisted relaxation, music and imagery, and guided imagery and music (Bonny Method), specific active music therapy methods (recreative music therapy and improvisational music therapy), music therapy mode (group therapy, individual therapy), music therapy period (weeks) (2-4, 5-12, �13), music therapy frequency (once weekly, twice weekly, �3 times weekly), total music therapy sessions (1-4, 5-8, 9-12, 13-16, >16), time per session (minutes) (15-40, 41-60, >60), inpatient settings (secure [locked] unit at a mental health facility versus outpatient settings), sample size (20-50, �50 and <100, �100), female predominance(>80%) (no, yes), mean age (years) (<50, 50-65, >65), country having music therapy profession (no, yes), pre-treatment diagnosis (mental health, depression, severe mental disease/psychiatric disorder). We also performed sensitivity analyses to test the robustness of the results by re-estimating the pooled effects using fixed effect model, using trim and fill analysis, excluding the paper without information on music therapy, excluding the papers with more high biases, excluding the papers with small sample size (20< n<30), excluding the papers using an infrequently used scale, excluding the studies focused on the people with a severe mental disease. We investigated the publication biases by a funnel plot as well as Egger's linear regression test [17]. The analyses were performed using Stata, version 11.0. All P-values were two-sided. A P-value of less than 0.05 was considered to be statistically significant.

Characteristics of the eligible studies
Fig 1 depicts the study profile, and a total of 55 RCTs were included in our meta-analysis . Of the 55 studies, 10 studies from America, 22 studies from Europe, 22 studies from Asia, and 1 study from Australia. The mean age of the participators ranged from 12 to 86; the sample size ranged from 20 to 242. A total of 16 different scales were used to evaluate the depression level of the participators. A total of 25 studies were conducted in impatient setting and 28 studies were in outpatients setting; 32 used a certified music therapist, 15 not used a certified music therapist (for example researcher, nurse), and 10 not reported relevent information. A total of 16 different depression rating scales were used in the included studies, and HADS, GDS, and BDI were the most frequently used scales (Table 1).
Of the 55 studies, only 2 studies had high risks of selection bias, and almost all of the included studies had high risks of performance bias (Fig 2).

The overall effects of music therapy
Of the included 55 studies, 39 studies evaluated the music therapy, 17 evaluated the music medicine. Using a random-effects model, music therapy was associated with a significant   (Fig 3). Twenty studies evaluated the active music therapy using a random-effects model, and a moderate-sized mean effect (SMD = −0.57; 95% CI = -0.90 to -0.25; P<0.001) was observed with a high heterogeneity across studies (I 2 = 86.3%, P<0.001). Fourteen studies evaluated the receptive music therapy using a random-effects model, and a moderate-sized mean effect (SMD = −0.73; 95% CI = -1.01 to -0.44; P<0.001) was observed with a high heterogeneity across studies (I 2 = 76.3%, P<0.001). Five studies evaluated the combined effect of active and receptive music therapy using a random-effects model, and a moderate-sized mean effect (SMD = −0.88; 95% CI = -1.32 to -0.44; P<0.001) was observed with a high heterogeneity across studies (I 2 = 70.5%, P<0.001) (Fig 4).

Sub-group analyses and meta-regression analyses
We performed sub-group analyses and meta-regression analyses to study the homogeneity. We found that music therapy yielded a superior effect on reducing depression in the studies with a small sample size , with a mean age of 50-65 years old, with medium intervention frequency (<3 times weekly), with more minutes per session (>60 minutes). We also found that music therapy exhibited a superior effect on reducing depression among people with severe mental disease /psychiatric disorder and depression compared with mental health people. While, whether the country have the music therapy profession, whether the study used group therapy or individual therapy, whether the study was in the outpatients setting or the inpatient setting, and whether the study used a certified music therapist all did not exhibit a    remarkable different effect ( Table 2). Table 2 also presents the subgroup analysis of music medicine on reducing depression. In the subgroup analysis by total session, music therapy and music medicine both exhibited a stronger effects of short (1-4 sessions) and medium length (5-12 sessions) compared with long intervention periods (>13sessions) (Fig 6). Meta-regression demonstrated that total  music intervention session was significantly associated with the homogeneity between studies (P = 0.004) ( Table 3).

Sensitivity analyses
We performed sensitivity analyses and found that re-estimating the pooled effects using fixed effect model, using trim and fill analysis, excluding the paper without information regarding music therapy, excluding the papers with more high biases, excluding the papers with small sample size (20< n<30), excluding the studies focused on the people with a severe mental disease, and excluding the papers using an infrequently used scale yielded the similar results, which indicated that the primary results was robust ( Table 4).

Evaluation of publication bias
We assessed publication bias using Egger's linear regression test and funnel plot, and the results are presented in Fig 7. For the main result, the observed asymmetry indicated that either the absence of papers with negative results or publication bias.

Discussion
Our present meta-analysis exhibited a different effect of music therapy and music medicine on reducing depression. Different music therapy methods also exhibited a different effect, and the recreative music therapy and guided imagery and music yielded a superior effect on reducing depression compared with other music therapy methods. Furthermore, music therapy and music medicine both exhibited a stronger effects of short and medium length compared with long intervention periods. The strength of this meta-analysis was the stable and high-quality result. Firstly, the sensitivity analyses performed in this meta-analysis yielded similar results, which indicated that the primary results were robust. Secondly, considering the insufficient statistical power of small sample size, we excluded studies with a very small sample size (n<20). Some prior reviews have evaluated the effects of music therapy for reducing depression. These reviews found a significant effectiveness of music therapy on reducing depression among older adults with depressive symptoms, people with dementia, puerpera, and people with cancers [4,5,[73][74][75][76]. However, these reviews did not differentiate music therapy from music medicine. Another paper reviewed the effectiveness of music interventions in treating depression. The authors included 26 studies and found a signifiant reduction in depression in the music intervention group compared with the control group. The authors made a clear distinction on the definition of music therapy and music medicine; however, they did not include all relevant data from the most recent trials and did not conduct a meta-analysis [77]. A recent meta-analysis compared the effects of music therapy and music medicine for reducing depression in people with cancer with seven RCTs; the authors found a moderately strong, positive impact of music intervention on depression, but found no difference between music therapy and music medicine [78]. However, our present meta-analysis exhibited a different effect of music therapy and music medicine on reducing depression, and the music medicine yielded a superior effect on reducing depression compared with music therapy. The different effect of music therapy and music medicine might be explained by the different participators, and nine studies used music therapy to reduce the depression among people with severe mental disease /psychiatric disorder, while no study used music medicine. Furthermore, the studies evaluating music therapy used more clinical diagnostic scale for depressive symptoms.
A meta-analysis by Li et al. [74] suggested that medium-term music therapy (6-12 weeks) was significantly associated with improved depression in people with dementia, but not short-  On the contrary, our present meta-analysis found a stronger effect of short-term (1-4 weeks) and medium-term (5-12 weeks) music therapy on reducing depression compared with long-term (�13 weeks) music therapy. Consistent with the prior meta-analysis by Li et al., no significant effect on depression was observed for the follow-up of one or three months after music therapy was completed in our present meta-analysis. Only five studies analyzed the therapeutic effect for the follow-up periods after music therapy intervention therapy was completed, and the rather limited sample size may have resulted in this insignificant difference. Therefore, whether the therapeutic effect was maintained in reducing depression when music therapy was discontinued should be explored in further studies. In our present meta-analysis, meta-regression results demonstrated that no variables (including period, frequency, method, populations, and so on) were significantly associated with the effect of music therapy. Because meta-regression does not provide sufficient statistical power to detect small associations, the non-significant results do not completely exclude the potential effects of the analyzed variables. Therefore, meta-regression results should be interpreted with caution.
Our meta-analysis has limitations. First, the included studies rarely used masked methodology due to the nature of music therapy, therefore the performance bias and the detection bias was common in music intervention study. Second, a total of 13 different scales were used to evaluate the depression level of the participators, which may account for the high heterogeneity among the trials. Third, more than half of those included studies had small sample sizes (<50), therefore the result should be explicated with caution.

Conclusion
Our present meta-analysis of 55 RCTs revealed a different effect of music therapy and music medicine, and different music therapy methods also exhibited a different effect. The results of subgroup analyses revealed that the characters of music therapy were associated with the therapeutic effect, for example specific music therapy methods, short and medium-term therapy, and therapy with more time per session may yield stronger therapeutic effect. Therefore, our present meta-analysis could provide suggestion for clinicians and policymakers to design therapeutic schedule of appropriate lengths to reduce depression.