The efficacy of music therapy in alleviating anxiety among college students: a systematic review and meta-analysis.

This is a systematic review and meta-analysis of 18 randomized controlled trials (RCTs) that tested music therapy (MT) against a control condition (no intervention, waitlist, or standard care) for reducing anxiety in college students.

The interventions fell into three categories:

**Passive music therapy:** Participants listened to pre-recorded or live music without active participation. This is sometimes called "music medicine" — the music is delivered to the person.

**Active music therapy:** Participants actively made music — singing, playing instruments, improvising, or songwriting — guided by a trained music therapist.

**Combined music therapy:** A mix of passive listening and active participation in the same session or across sessions.

The control groups received either no intervention, a waitlist (they got the intervention later), or standard care (e.g., usual campus mental health services). No study used an active placebo control (e.g., listening to white noise or audiobooks), which is a major limitation.

The primary outcome was anxiety measured by validated self-report scales. The meta-analysis combined results across studies using the standardized mean difference (SMD), which puts all scales on a common unit so they can be averaged.

**Total participants:** 1,679 college students across 18 RCTs (829 in the music therapy group, 850 in the control group).

**Population:** Undergraduate university students. Some studies specifically recruited students with elevated anxiety symptoms; others recruited general student populations.

**Setting:** 13 studies were conducted in China (domestic), 3 were international (other countries), and 1 study's location was not clearly reported. Most interventions took place on campus — in classrooms, music rooms, or dormitories.

**Age:** Not reported as a single mean across all studies, but typical college age (18–25).

**Gender:** Not reported in the aggregate — individual studies varied.

**Exclusion criteria:** Studies were excluded if participants had severe mental illness, were on psychiatric medication, or had hearing impairments, but this was not uniformly applied across all included RCTs.

Each included study used at least one validated self-report anxiety scale. The most common were:

**Self-Rating Anxiety Scale (SAS):** A 20-item questionnaire, scored 20–80 (raw) or 25–100 (index), with higher scores = more anxiety. A score ≥50 is considered clinically significant anxiety.

**Hamilton Anxiety Rating Scale (HAM-A):** A 14-item clinician-rated scale, scored 0–56, with higher = more anxiety. Scores >17 indicate mild anxiety; >25 indicate moderate-to-severe.

**State-Trait Anxiety Inventory (STAI):** Two 20-item subscales (State anxiety = how you feel right now; Trait anxiety = how you generally feel), each scored 20–80, higher = more anxiety.

**Beck Anxiety Inventory (BAI):** A 21-item self-report scale, scored 0–63, higher = more anxiety. Scores 0–7 = minimal; 8–15 = mild; 16–25 = moderate; 26–63 = severe.

**Generalized Anxiety Disorder-7 (GAD-7):** A 7-item scale, scored 0–21, with 5, 10, and 15 as cutoffs for mild, moderate, and severe anxiety.

Because different studies used different scales, the meta-analysis converted all results into a standardized mean difference (SMD), which expresses the effect in standard deviation units. An SMD of –1.54 means the music therapy group scored, on average, 1.54 standard deviations lower on anxiety than the control group.

**Design:** Systematic review and meta-analysis of randomized controlled trials (RCTs). The authors searched 8 databases (Chinese: CNKI, Wanfang, VIP; English: PubMed, Web of Science, Embase, Scopus, Ovid Medline) from database inception to December 1, 2025. Two reviewers independently screened studies, extracted data, and assessed risk of bias using the Cochrane Risk of Bias Tool (RoB 1). Disagreements were resolved by discussion or a third reviewer.

**Inclusion criteria:** Only RCTs. Participants had to be undergraduate students. The intervention had to be music therapy as a primary modality (not music as background). The control could be no intervention, waitlist, or standard care. Studies had to report anxiety as an outcome using a validated scale.

**Statistical methods:** Meta-analysis was performed using RevMan 5.2 and Stata 16.0. The standardized mean difference (SMD) with 95% confidence intervals was calculated using a random-effects model (appropriate when studies are expected to have different true effects). Heterogeneity was assessed using the I² statistic (0% = no heterogeneity, 100% = maximal heterogeneity). Subgroup analyses were pre-planned to explore sources of heterogeneity: type of MT (passive vs. combined), duration (<8 weeks vs. ≥8 weeks), and setting (domestic/China vs. international). Publication bias was assessed using funnel plots and Egger's test.

**What this design can and cannot prove:**

**Can prove:** That, across the existing literature, music therapy is associated with larger reductions in anxiety scores compared to control conditions. The meta-analytic approach increases statistical power and generalizability by pooling data.

**Cannot prove:** Causality in any individual study — that's what the original RCTs are for, but many had weak designs. Cannot prove that music therapy works better than an active placebo (e.g., a relaxing audiobook or guided meditation), because no study included such a control. Cannot prove long-term effects, because most studies measured anxiety only immediately after the intervention (no follow-up). Cannot prove that the effect is clinically meaningful for any specific individual, because the SMD is an average across very different studies.

**Major methodological weaknesses:**

**Very high heterogeneity (I² = 95%):** This means 95% of the variation in results across studies is due to real differences between studies, not random chance. This is extremely high and suggests the studies are measuring different things or have very different populations, interventions, or designs. The average effect size may be misleading.

**Poor blinding:** In music therapy RCTs, it is nearly impossible to blind participants or therapists. Only 1 of the 18 studies attempted blinding of outcome assessors. This introduces high risk of performance bias (participants know they're getting music and expect to feel better) and detection bias (assessors who know the group assignment may rate anxiety differently).

**Small sample sizes:** Many individual studies had fewer than 50 participants per group, making them underpowered to detect small-to-moderate effects and prone to inflated effect sizes (the "winner's curse").

**Publication bias:** The funnel plot and Egger's test were not reported in the abstract, but given that 17 of 18 studies were from China and all showed positive effects, there is a strong possibility of publication bias (negative or null studies not getting published).

**No active control:** Without an active placebo, the observed effect could be due to expectation, attention from a therapist, or simply taking a break from studying — not the music itself.

**Primary outcome — anxiety reduction:** Music therapy significantly reduced anxiety compared to control conditions. The overall standardized mean difference was SMD = –1.54 (95% CI: –2.08 to –1.01, P < 0.00001). This is considered a "large" effect size by conventional benchmarks (0.2 = small, 0.5 = medium, 0.8 = large).

**Heterogeneity:** I² = 95%, indicating extremely high variability between studies. This means the true effect in any given study could be very different from the average.

**Subgroup analysis — type of MT:**

- Passive MT (listening only): SMD = –1.10 (95% CI not reported in abstract, but significant)

- Combined MT (active + passive): SMD = –1.89 (95% CI not reported, but significant)

- The difference between passive and combined was statistically significant, suggesting combined approaches may be more effective.

**Subgroup analysis — duration:**

- Short-term (<8 weeks): Significant anxiety reduction

- Long-term (≥8 weeks): Significant anxiety reduction

- The abstract does not report whether the effect sizes differed significantly between these subgroups.

**Subgroup analysis — setting:**

- Domestic (China, 13 studies): Larger effect than international (3 studies)

- The abstract does not report the specific SMDs for this comparison.

**Secondary outcomes:** Not reported in the abstract. The review focused exclusively on anxiety as the outcome.

An SMD of –1.54 is a very large effect by statistical standards. To translate this into something more intuitive:

If you imagine two college students — one who receives music therapy and one who does not — the average student in the music therapy group would be less anxious than approximately 94% of students in the control group (assuming a normal distribution).

On the Self-Rating Anxiety Scale (SAS, raw score 20–80), a typical study might show a reduction of about 8–12 points in the music therapy group versus 2–4 points in the control group. For context, a 10-point drop on the SAS can move someone from "moderate anxiety" to "normal" range.

On the GAD-7 (0–21), a reduction of roughly 4–6 points in the music therapy group versus 1–2 points in controls would be consistent with this SMD. A 5-point drop on the GAD-7 is considered a clinically meaningful improvement.

**However**, the 95% confidence interval is very wide (–2.08 to –1.01). This means the true average effect could be as small as SMD = –1.01 (still large) or as large as SMD = –2.08 (extremely large). And with I² = 95%, individual studies ranged from showing almost no effect to showing enormous effects. The average is not necessarily representative of what any one person would experience.

**Acknowledged by authors:**

High heterogeneity across studies, limiting the precision of the pooled estimate.

Most studies were conducted in China, limiting generalizability to other cultural contexts.

Lack of standardized MT protocols — interventions varied widely in type, duration, frequency, and music selection.

Short intervention durations and lack of long-term follow-up — most studies measured anxiety only immediately post-intervention.

Small sample sizes in many individual studies.

Risk of bias due to lack of blinding.

**Additional limitations a critical reader would note:**

**No active control condition:** Without comparing music therapy to another engaging activity (e.g., listening to an audiobook, doing a puzzle, guided relaxation), we cannot attribute the effect specifically to music. The effect could be due to taking a break, receiving attention, or simply having a structured activity.

**Self-report bias:** All anxiety measures were self-report. Participants who know they received music therapy may report lower anxiety due to demand characteristics (they think the researchers want to see improvement).

**Publication bias is highly likely:** 17 of 18 studies were from China, and all showed positive effects. This is statistically implausible and suggests that negative or null studies were not published or not included. The authors did not report a formal test for publication bias (Egger's test) in the abstract.

**No intention-to-treat analysis reported:** Many small RCTs exclude dropouts from analysis, which can inflate the apparent effect.

**No assessment of adverse effects:** Music therapy is generally safe, but some people may find certain music irritating or anxiety-provoking. No study reported on negative reactions.

**Population homogeneity:** College students are a specific population — young, generally healthy, with high access to music. Results may not generalize to older adults, clinical populations, or people with hearing impairments.

**The "combined" therapy effect may be confounded by dose:** Combined therapy typically involves more total time with a therapist, so the larger effect may be due to more attention, not the specific combination of active + passive music.

For someone running their own n=1 experiment:

**What to test:**

**Intervention:** Listen to calming instrumental music (e.g., classical, ambient, or nature sounds with slow tempo, ~60–80 beats per minute) for 20–30 minutes per day. For a stronger test, add a brief active component: after listening for 10 minutes, spend 5 minutes humming or singing along softly, or tapping a simple rhythm. This mirrors the "combined" approach that showed larger effects in the meta-analysis.

**Dose:** 20–30 minutes per session, once daily, for at least 4 weeks (the minimum duration in the included studies). The meta-analysis found effects at both <8 weeks and ≥8 weeks, so 4 weeks is a reasonable starting point.

**Music selection:** Use the same playlist each session to control for variability. Choose music without lyrics (lyrics can engage language processing and distract from relaxation). Instrumental piano, ambient electronic, or nature sounds with water/birds are common choices.

**Minimum meaningful duration:**

Run the experiment for at least 4 weeks. The meta-analysis found significant effects even in short-term interventions (<8 weeks), but 2 weeks is likely too short to see a reliable change, especially if your anxiety fluctuates day-to-day.

Measure anxiety daily for 1 week before starting (baseline), then daily throughout the 4-week intervention, and ideally for 1 week after stopping (washout) to see if effects persist.

**What to measure (specific metrics):**

**Primary metric:** A validated anxiety scale. The **GAD-7** is free, takes 2 minutes, and is widely used. Score it daily at the same time (e.g., 30 minutes after your music session). Track the total score (0–21).

**Secondary metric:** A single-item rating: "On a scale of 0 (completely relaxed) to 10 (most anxious I've ever felt), how anxious do you feel right now?" Rate this immediately before and after each music session. This gives you a within-session effect.

**Physiological metric (optional but useful):** Heart rate variability (HRV) measured with a chest strap or smartwatch. Higher HRV is associated with lower stress. Measure for 5 minutes before and after each session, at the same time of day.

**Confound tracker:** Keep a daily log of sleep hours, caffeine intake (cups), alcohol (drinks), exercise (minutes), and any major stressors (e.g., exam, argument). These can all affect anxiety independently.

**Key confounds to control for:**

**Time of day:** Always do your music session at the same time. Anxiety naturally fluctuates throughout the day. If you do music in the evening (when anxiety is often lower) and compare to morning baselines, you'll see a fake effect.

**Expectation:** You know you're getting music therapy, so you may expect to feel better. To partially control for this, add a "control week" where you sit quietly for 20 minutes without music (or listen to an audiobook). Compare the music week to the quiet week.

**Regression to the mean:** If you start the experiment during a particularly anxious period (e.g., exam week), your anxiety will naturally drop afterward regardless of the intervention. Run the experiment during a stable period, or use a longer baseline (2 weeks) to establish your typical range.

**Novelty effect:** The first few days of any new routine can feel good just because it's new. Effects may fade after 1–2 weeks. Look at the trend across all 4 weeks, not just the first week.

**Music preference:** If you hate classical music, don't use it. Choose music you find genuinely pleasant and relaxing. The meta-analysis included studies using participant-selected music