TDMUSIC ยท progress briefing
โ† Back to briefing
EMPATHIZE ยท DISCOVER ยท LITERATURE & MARKET

Literature & market: what the evidence actually supports

Six independent research streams โ€” clinical evidence, competitor landscape, market size & willingness-to-pay, regulation, wearables/biofeedback, and user sentiment โ€” were pulled into one synthesis, then adversarially fact-checked claim by claim against primary sources. This page carries only what survived that check, in the form it was corrected to.

Source: research/SYNTHESIS.md, compiled from a 6-angle fact-checked research sweep ยท every numeric claim below was re-opened against its cited primary source and marked CONFIRMED, CORRECTED or UNSUPPORTED before being used here.

Included: only CONFIRMED / CORRECTED figures from the verified findings files Excluded: Calm's disputed "$596.4M 2024 revenue" figure โ€” internally inconsistent with its own 2025 numbers Excluded: unverified "180+ BBB complaints" and "100M brown-noise views" claims โ€” untraceable to source
/ AClinical evidence

Music reliably reduces anxiety โ€” but the effect size depends on which study you read

Four independent meta-analyses converge on a genuine anti-anxiety effect of music listening, but the pooled effect size ranges from dโ‰ˆโˆ’0.30 to dโ‰ˆโˆ’0.97 depending on population, outcome type (self-report vs. physiological) and how strictly the review screens for trial quality.

Cochrane 2013 ยท perioperative
โˆ’5.72

STAI-S anxiety units [95% CI โˆ’7.27, โˆ’4.17] across 26 trials, N=2,051; one N=327 trial matched the sedative midazolam. GRADE: low quality (no blinding).

Bradt, Dileo & Shim (2013), Cochrane CD006908 โ€” source

de Witte 2020 ยท stress
.545 / .380

Psychological vs. physiological stress reduction across 104 RCTs (N=9,617). Heart rate is the strongest physiological marker (d=.456); cortisol the weakest (d=.349).

de Witte et al. (2020), Health Psychology Review โ€” source

Harney 2023 ยท anxiety
โˆ’0.77

[95% CI โˆ’1.26, โˆ’0.28] across 21 controlled studies; rises to d=โˆ’0.97 in a low-bias-only sensitivity subset (k=10).

Harney et al. (2023), Musicae Scientiae โ€” source

Lassner 2024/25 ยท delivery mode
0.47

Music therapy, music medicine and receptive delivery achieve a similar SMD on anxiety (k=23, n=1,084) โ€” but GRADE "very low" and not maintained at follow-up.

Lassner et al. (2024/2025), BJPsych Open โ€” source

ClaimSourceEvidence type
Perioperative music reduces state anxiety by โˆ’5.72 STAI-S units; comparable to midazolam in one trialBradt, Dileo & Shim, 2013 (Cochrane CD006908)Cochrane review ยท 26 RCTs, N=2,051 ยท GRADE low
Music reduces physiological stress (d=.380) and psychological stress (d=.545); cortisol weakest at d=.349de Witte et al., 2020, Health Psychology ReviewMeta-analysis ยท 104 RCTs, N=9,617
Music reduces anxiety with d=โˆ’0.77, rising to โˆ’0.97 in low-bias studiesHarney et al., 2023, Musicae ScientiaeMeta-analysis ยท 21 controlled studies
Therapist-led, passive-listening and receptive delivery modes achieve similar anxiety effects (SMD=0.47); not sustained at follow-upLassner et al., 2024/2025, BJPsych OpenMeta-review ยท GRADE very low
Slow-tempo (60โ€“80 BPM), relaxing music raises parasympathetic/vagal tone (HRVโ†‘); fast tempo shifts toward sympathetic dominance (HRVโ†“)Bernardi, Porta & Sleight, 2006, Circulation; Mitrovic & Paladin, 2026, Front. Cardiovasc. Med.Controlled physiology study + 2026 review
Self-reported anxiety drops significantly (d=โˆ’0.30); the psychophysiological effect is not statistically significantPanteleeva et al., 2018, Psychology of MusicMeta-analysis ยท caution / counterweight
Read with caution

Effect sizes span d=โˆ’0.30 to d=โˆ’0.97 โ€” and self-report doesn't always track physiology

Every review agrees music reduces anxiety; none agree by how much. The spread reflects differing inclusion criteria (clinical vs. non-clinical samples, control conditions, outcome type). Cortisol โ€” the marker closest to a "stress hormone" claim โ€” is consistently the weakest of the three physiological markers [de Witte et al., 2020], and self-report and physiological anxiety measures do not always move together [Panteleeva et al., 2018]. Any claim that leans on "music reduces anxiety" should name which review, which outcome, and which population it means.

/ BMarket & prevalence

A growing category, built on top of very different anxiety rates

Market-research firms diverge sharply by category definition; anxiety-prevalence estimates diverge even more sharply by country and diagnostic instrument.

Meditation apps
$2.2B โ†’ $7.0B

2025 โ†’ 2033, 14.7% CAGR; North America ~43% share.

Grand View Research, 2025 โ€” source

Mental-health apps
$7.48B โ†’ $35.29B

2025 โ†’ 2034, 19.2% CAGR โ€” a broader category than meditation-only; not directly comparable.

Fortune Business Insights, 2025 โ€” source

China emotion economy
ยฅ2.31T

2024, forecast >ยฅ4.5T by 2029 โ€” bundles blind-box toys, pets, escape rooms; not an app-specific figure.

iiMedia, 2025 โ€” source

Global anxiety burden
359M

People with an anxiety disorder, 2021 (~4.4% of world population); only 27.6% receive treatment.

WHO, 2023 โ€” source

Region12-month prevalenceLifetime prevalenceSource
China5.0%7.6%Huang et al., 2019, The Lancet Psychiatry
United States19.1%31.1%NIMH, NCS-R survey
Global~4.4% (359M people)โ€”WHO, 2023

The roughly 4ร— gap between the China and US figures is best read as a measurement artifact โ€” underdiagnosis, stigma and differing diagnostic instruments โ€” rather than a true 4ร— difference in psychological distress [Huang et al., 2019]; [NIMH].

Willingness to pay differs by market as much as prevalence estimates do: Calm prices its annual plan at $69.99 [Business of Apps, 2026]; China's leading meditation app, Tide (ๆฝฎๆฑ), prices its annual plan at roughly ยฅ218 (~$30) โ€” about 40% of Calm's price [SSPai, 2026].

Tension to note

Forecasts say growth; incumbents say maturity

Grand View Research and Fortune Business Insights both forecast double-digit CAGR through the early 2030s โ€” yet Calm's own 2025 numbers show revenue down 24% year-on-year to about $210M and subscribers down roughly 500,000 [Business of Apps, 2026]. Headspace shows a similar softening. Category-level growth forecasts and incumbent-level performance are telling two different stories, and both should be held at once, not resolved in favour of one.

/ CRegulation

Wellness, not medicine โ€” a line every regulator draws, but draws differently

US, EU/Germany and China all separate low-risk "wellness" claims from regulated "treatment" claims โ€” but where that line falls, and what it costs to cross it, differs by market.

MarketWellness-safe languageWhat triggers medical/regulatory status
US (FDA)"Helps you relax," "supports stress management" โ€” general-wellness two-factor test (low risk + lifestyle-only intent)"Treats anxiety disorder," diagnosing a named condition โ†’ Software-as-a-Medical-Device (SaMD), requiring 510(k)/De Novo clearance
EU / Germany (DiGA)Unlimited wellness/lifestyle claims outside the DiGA reimbursement systemClaiming a therapeutic benefit for a named ICD-10 condition + seeking reimbursement โ†’ CE-marked medical device + BfArM DiGA Fast-Track clinical evidence
China"ๆƒ…็ปช็ฎก็†" (mood management), "ๅ‡ๅŽ‹" (stress relief), "ๅฅๅบท็ฎก็†" (health management) โ€” no diagnostic or treatment claim"ๆฒป็–—็„ฆ่™‘" (treat anxiety) โ€” barred by Advertising Law Art. 17 (fines of 1โ€“3ร— ad spend or ยฅ100,000โ€“200,000); AI-driven diagnosis/treatment software risks NMPA Class III classification

Sources: FDA General Wellness guidance ยท BfArM DiGA-Verzeichnis ยท Advertising Law of the PRC, via Zhonglun, 2026

Germany's DiGA registry lists roughly 60 apps, and mental health (anxiety, depression, sleep) is the single largest category (31 of 60) [BfArM DiGA-Verzeichnis, 2026] โ€” but clinician skepticism is real: in a 2022 survey of ~2,600 German physicians, 27.7% considered DiGAs prescribed for depression ineffective or contraindicated [Schmitz et al., 2025].

Pear Therapeutics
Ch. 11 ยท Apr 2023

Maker of the first FDA-cleared prescription digital therapeutic filed for bankruptcy despite clinical and regulatory success.

Fierce Biotech, 2023 โ€” source

Akili Interactive
$0.434/share ยท Jul 2024

EndeavorRx's maker was delisted from Nasdaq and taken private after a failed prescription-to-OTC pivot.

Nasdaq/BusinessWire, 2024 โ€” source

Regulatory clearance โ‰  commercial viability

Both major US digital-therapeutic precedents reached clearance and still failed commercially

Pear and Akili both achieved genuine FDA clearance โ€” and both still failed commercially. Reimbursement, not clinical efficacy, was the repeatedly cited constraint. Medicare only began reimbursing FDA-cleared digital mental-health devices from January 2025 (HCPCS G0552โ€“G0554), and clinician cost-fronting has limited uptake so far [Healthcare Brew, 2025].

/ DWearables & closed loop

The "real-time biometric loop" is mostly heart rate, not HRV

A closed-loop product concept rests on one assumption: that a wearable can feed live physiological state back into the audio. The data says that's true for heart rate โ€” and not yet true for HRV.

Feasibility flag

Real-time HRV is not continuously available on today's consumer platforms

Apple HealthKit exposes HRV (SDNN) to third-party apps only with explicit permission, and only as opportunistically-sampled values around Mindfulness/Breathe sessions and background rest โ€” not a continuous stream. Only plain heart rate can update every few seconds; HRV itself can lag up to ~30 minutes [Apple Developer Forums, cited in Wearables-verified, 2026]. Oura's API returns HRV/heart-rate at 5-minute granularity, but only captured during sleep and released the following morning โ€” there is no live endpoint, and since 2024 Oura has paywalled this data behind an active Membership subscription for newer rings [Oura developer documentation, 2024]. Any "real-time HRV-adaptive" product claim needs to be scoped to what's actually available โ€” live heart rate, next-morning HRV โ€” not live HRV.

Data sourceWhat's actually availableReal-time?
Apple HealthKit โ€” heart rateUpdates every few seconds during an active sessionYes
Apple HealthKit โ€” HRV (SDNN)Opportunistic samples around Mindfulness/rest; can lag up to ~30 minutesNo
Oura Ring API5-minute HRV/HR granularity, sleep-only, released the next morning; paywalled behind Membership for Gen3+/Ring 4 (2024)Next-day only
Endel's closed loop (current market precedent)Real-time Apple Watch heart rate (not HRV) + motion, time of day, lightHR only
Apple Watch HRV accuracy
โˆ’8.31 ms

Series 9/Ultra 2 underestimated HRV against a Polar H10 chest-strap reference (MAPE 28.9%), failing pre-set equivalence bounds โ€” though resting heart rate was accurate (5.9% error).

Sensors, 2024 โ€” source

HRV-biofeedback evidence
g=0.81 / 0.83

Pre-post and vs.-control effect of HRV-biofeedback training on stress/anxiety, across 24 studies (n=484) โ€” the strongest indirect evidence for a closed loop, but for breathing-paced biofeedback, not music-driven adaptive audio.

Goessl, Curtiss & Hofmann, 2017 โ€” source

Endel remains the clearest commercial precedent for closed-loop biometric-adaptive audio โ€” using real-time Apple Watch heart rate (not HRV) plus motion, time of day and light, and winning Apple's "Watch App of the Year" in 2020 โ€” but no independent, peer-reviewed outcome study of Endel or any comparable commercial biometric-adaptive audio app was found [Wearables-verified, 2026]. This is a genuine generalization gap worth naming explicitly: Goessl et al. (2017) validates HRV-biofeedback protocols generally โ€” it is not direct evidence that music-driven adaptive audio produces the same effect [Goessl, Curtiss & Hofmann, 2017].