This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement guidelines (25) and its corresponding explanation and elaboration document. The review protocol was prospectively registered with PROSPERO (Registration ID: CRD420251073337; https://www.crd.york.ac.uk/PROSPERO/view/CRD420251073337), with any protocol deviations documented systematically and reported in accordance with contemporary standards for systematic review research.

Study inclusion and exclusion criteria were defined according to the PICOS framework (26), ensuring comprehensive coverage while maintaining methodological rigor. The target population comprised combat sport athletes aged ≥15 years with minimum one-year structured training experience in recognized disciplines including taekwondo, kickboxing, karate, judo, wrestling, mixed martial arts, or boxing. Eligible interventions encompassed music-based performance enhancement strategies administered standalone or combined with evidence-based modalities (caffeine supplementation, video feedback, plyometric training, strategic napping, or knowledge of endpoint disclosure). Comparison conditions included no-music controls, alternative auditory stimuli, silence, or standardized protocols, with combination studies utilizing co-interventions without music components. Outcome measures were categorized into technical performance (reaction time, movement accuracy, skill execution), physical performance (power, strength, agility, endurance), physiological responses (cardiovascular, metabolic, stress markers), and psychological outcomes (motivation, mood, perceived exertion, arousal, cognition). Study designs were restricted to controlled experimental designs permitting causal inference, including randomized controlled trials, randomized crossover trials, and controlled crossover studies. Exclusion criteria encompassed non-athlete populations, insufficient combat sport experience (<1 year), observational studies, case series/reports, qualitative designs, studies without control conditions, insufficient quantitative data, populations with chronic illness or musculoskeletal injury, conference abstracts without full-text access, duplicate publications, and non-English/French language studies. Language restrictions limited inclusion to English and French publications. Non-English/French records identified during screening (n = 23) were excluded based solely on language eligibility.

A comprehensive search strategy was implemented across five major electronic databases selected for optimal coverage: PubMed/MEDLINE, Web of Science Core Collection, Scopus, SPORTDiscus via EBSCOhost and ScienceDirect, from the first registration until June 15, 2025. The search strategy employed Medical Subject Headings and free-text keywords across three domains: music interventions, combat sports, and performance outcomes, with Boolean operators applied for optimal sensitivity. Final searches were executed June 15, 2025. Supplementary search methodology included reference harvesting from included studies and relevant reviews, forward citation searching via Google Scholar and Web of Science, grey literature examination through specialized databases, expert consultation with prominent researchers, and trial registry searches (ClinicalTrials.gov, WHO International Clinical Trials Registry Platform) to identify unpublished completed studies.

The music intervention domain incorporated terms encompassing auditory stimulation modalities, including “music,” OR “auditory stimulation,” OR “sound,” OR “musical intervention,” “audio feedback,” OR “sonic enhancement”. Using Boolean operator “AND”, combat sport terminology encompassed both broad categorical terms and sport-specific descriptors, including “combat sport*,” OR “martial art*,” OR “fighting sport*,” OR “taekwondo,” OR “kickboxing,” “karate,” OR “boxing,” OR “mixed martial arts,” OR “MMA,”. Performance outcome descriptors included “performance,” OR “athletic performance,” OR “exercise performance,” OR “physical performance,” OR “physiological response*,” OR “psychological response*,” “motor performance,” OR “skill execution,” OR “competitive performance.”

Combination intervention terms reflected the multimodal nature of contemporary performance enhancement research, incorporating “caffeine,” OR “video feedback,” OR “knowledge of endpoint,” OR “knowledge of results,” OR “plyometric*,” OR “power training,” OR “napping,” OR “sleep,” OR “rest,” OR “recovery” combined with music-related terms using appropriate Boolean logic.

Study selection followed a rigorous two-stage screening process using Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia), adhering to current best practices (27) Two independent reviewers conducted systematic evaluation of all retrieved records against predetermined eligibility criteria, with disagreements resolved through structured discussion and third-reviewer arbitration when consensus could not be achieved. Inter-rater agreement achieved substantial concordance (Cohen's κ = 0.76, 95% CI: 0.68–0.84), indicating high consistency in eligibility judgments (28). All records were managed using EndNote X21 reference management software with automated and manual duplicate detection, maintaining the PRISMA 2020 flow diagram throughout the review process.

A comprehensive standardized data extraction form based on PICOS methodology underwent rigorous pilot testing on five representative studies before implementation (29). Two independent reviewers extracted data from all included studies, with discrepancies resolved through discussion or third-reviewer consultation. Extraction encompassed study characteristics (bibliographic information, design specifications, funding sources), participant characteristics (demographics, competitive level, training experience), intervention specifications (music genre, tempo, volume, duration, timing), and complete statistical data (means, standard deviations, confidence intervals, exact p-values, sample sizes). Missing data management incorporated systematic author contact protocols (maximum three attempts over six weeks) and established statistical estimation methods including the Hozo method for converting medians and ranges to means and standard deviations (30).

Methodological quality was assessed using the validated Physiotherapy Evidence Database (31) scale (32), an 11-item instrument evaluating internal validity and statistical reporting with established reliability across diverse study populations (33). Two independent investigators conducted quality assessment with calibration exercises, resolving disagreements through structured discussion and third-reviewer adjudication when necessary. Quality scores were categorized as high (≥8 points), moderate (6–7 points), or low (≤5 points), with no studies excluded based solely on quality scores but utilized in predefined sensitivity analyses. To supplement PEDro assessment, we applied Cochrane Risk of Bias 2 (RoB 2) tool including the crossover extension (34) for all randomized crossover trials, evaluating bias arising from randomization process, period and carryover effects, deviations from intended interventions, missing outcome data, outcome measurement, and selective reporting. Risk of bias judgments (low, some concerns, high) were synthesized into an overall study-level rating. Overall evidence certainty was evaluated using the GRADE approach (35), considering study limitations, inconsistency, indirectness, imprecision, and publication bias, with potential upgrades for large effect magnitude, dose-response gradients, and residual confounding. Summary of Findings tables were constructed for key comparisons (music-only overall effect; outcome domain subgroups; music + caffeine combination) using GRADEpro software for guideline development tools GRADEpro GDT (36), with evidence certainty rated as high, moderate, low, or very low based on RoB 2 assessments, inconsistency (I² thresholds), imprecision (confidence interval width and sample size), and publication bias.

Standardized mean differences computed as Hedges' g (with small-sample bias correction) served as the primary effect size metric. For parallel-group designs, we used post-intervention means and standard deviations. For crossover trials, we computed effect sizes from paired comparisons adjusted for within-subject correlation. When correlations were unreported (69% of crossover comparisons), we assumed r = 0.5 consistent with meta-analytic conventions (59); sensitivity analyses (r = 0.3, 0.5, 0.7) confirmed robustness. Where pre-post measurements were available, we calculated change score effect sizes; design-type heterogeneity was negligible (Qbetween = 1.83, p = 0.176). Positive values indicate favorable music intervention effects. Effect size interpretation followed established conventions (small d = 0.2, moderate d = 0.5, large d = 0.8) while considering practical significance in combat sport contexts (Cohen, 1988). Random-effects models using restricted maximum likelihood (REML) estimation were employed for all primary analyses, reflecting realistic assumptions regarding effect size variation across studies (37, 38). To address statistical dependence arising from multiple effect sizes per study and repeated-measures designs, we implemented three-level multilevel meta-analysis using the rma.mv function in metafor (39). This hierarchical structure partitions variance into sampling variance (level 1), within-study heterogeneity between effect sizes (level 2), and between-study heterogeneity (level 3), thereby accounting for non-independence without artificially inflating precision (40). Model specification included random intercepts for study identifier and effect size identifier nested within studies. For all meta-analyses, the Hartung–Knapp–Sidik–Jonkman (HKSJ) modification was implemented for confidence interval and p-value calculation to provide more conservative inference when assumptions are violated, particularly given k < 10 for most comparisons and substantial heterogeneity (I² > 50%) (41). Between-study heterogeneity was assessed using Cochran's Q test (p < 0.10), I² statistic (interpretive guidelines: <25% low, 25%–50% moderate, 50%–75% substantial, >75% considerable), tau-squared (τ²), H statistic, and prediction intervals. Predefined subgroup analyses examined combat sport type, outcome domains, intervention timing, study quality, and sample characteristics, with statistical testing using Q-tests for heterogeneity between groups. Meta-regression analyses explored continuous moderators when sufficient studies were available (k ≥ 10) using mixed-effects models. Comprehensive sensitivity analyses included leave-one-out analyses, quality-based restrictions, design-based comparisons, and statistical method variations. Specifically, we conducted sensitivity analyses varying assumed within-subject correlations for crossover trials (r = 0.3, 0.5, 0.7) to evaluate robustness to this parameter; pooled estimates exhibited minimal variation (Δd < 0.04) with overlapping confidence intervals across all scenarios (42). Publication bias analyses were conducted for the music-only intervention comparison using study-level assessment to satisfy independence assumptions. We selected one representative effect size per study (primary outcome when pre-specified; otherwise, outcome with largest sample size), yielding k = 7 independent observations suitable for funnel plot construction and Egger's regression. Combined-intervention subgroups contained insufficient studies (k = 1–4 per combination type) for meaningful publication bias assessment. Combined interventions were stratified by temporal structure into acute within-session co-interventions (music + caffeine, music + video feedback, music + napping, music + knowledge endpoint) and chronic training-integrated co-interventions (music + plyometrics) to avoid conflating mechanistically distinct effect structures operating across different timescales. Synergistic effects were quantified as Δd = d_combined − d_music-only.

All analyses were performed using R version 4.3.2 (R Core Team, 2023) within RStudio (RStudio Team, 2023), utilizing the metafor package (version 4.4-0) for primary meta-analytic computations (39), meta package (version 6.5-0) for alternative methods and publication bias assessment, dmetar package for advanced diagnostics, ggplot2 (version 3.4.4) for visualization, and gridExtra (version 2.3) for multi-panel graphics. Complete reproducibility is ensured through systematic code documentation using R Markdown with all materials publicly available in a permanent Open Science Framework repository (https://osf.io/tuysf/overview). The repository contains data extraction files, annotated R analysis scripts implementing three-level meta-analysis models, sensitivity analysis code, and all Supplementary Materials. Statistical significance was established at α = 0.05 with exact p-values reported, confidence intervals presented as primary uncertainty measures, and retrospective statistical power calculations performed to assess evidence base adequacy.

This systematic review utilized exclusively published data, eliminating additional ethical approval requirements while maintaining adherence to established research synthesis standards. All data handling complied with relevant regulations and institutional guidelines, with participant confidentiality maintained through study-level data aggregation. Protocol deviations were documented systematically with transparent reporting of all analytical decisions, post-hoc analyses clearly distinguished from planned analyses, and regular team meetings ensuring consistent methodology application throughout the review process.

The systematic search retrieved 2,350 potentially relevant records across five electronic databases. Following duplicate removal and sequential screening, 15 studies published between 1994 and 2025 met the eligibility criteria and were included in both qualitative synthesis and quantitative meta-analysis. The study selection process is detailed in the PRISMA flow diagram (Figure 1). Inter-rater agreement during screening yielded Cohen's κ = 0.76 (95% CI: 0.68–0.84), reflecting substantial concordance between independent reviewers.

The included studies encompassed a total of 1,456 participants across three combat sports disciplines. Taekwondo dominated the evidence base with eleven studies (73.3%, k = 11), followed by kickboxing with three studies (20.0%, k = 3), and karate with two studies (13.3%, k = 2). One study (43) appears twice in Table 1 due to reporting both male and female cohorts but represents a single included study. Study designs included nine randomized controlled trials (60.0%), four randomized crossover trials (26.7%), and two multiple-baseline designs (13.3%). Sample sizes ranged from 10 to 1,206 participants, with participant ages spanning from 9 to 45 years. Detailed study characteristics are presented in Table 1.

Methodological quality assessment using the PEDro scale revealed generally high-quality studies. Eleven studies (73.33%) achieved high quality scores (≥8 points), while four studies (26.67%) demonstrated moderate quality (7 points). No study was classified as low quality (≤5 points). The mean PEDro score across all studies was 7.9 ± 0.8, indicating robust methodological rigor in the included evidence base. Quality assessment details are presented in Table 2.

Supplementary RoB 2 assessment for randomized crossover trials (k = 13) revealed low risk of bias in randomization process for 11 studies (84.6%), but concerns regarding period/carryover effects in 5 studies (38.5%) that employed insufficient washout periods (<48 h). Blinding of outcome assessors was achieved in 9 studies (69.2%). Overall risk of bias was judged as low for 8 studies (61.5%), some concerns for 6 studies (46.2%), and high for 1 study (7.7%).

Music-only interventions demonstrated a small beneficial effect on combat sport performance across all outcome domains. The three-level meta-analysis of 16 effect sizes nested within 7 studies revealed a pooled standardized mean difference (Hedges' g) of 0.19 (95% CI: 0.01–0.37, p = 0.039, τ² = 0.67, 95% prediction interval: −1.42 to 1.80), indicating statistical significance with clinical relevance. The between-study heterogeneity variance (τ²) of 0.67 and wide prediction interval reflect substantial variability in true effects across contexts. This effect represents a small-to-moderate improvement in performance outcomes when music interventions are implemented without additional modalities (Figure 2 and Table 3).

The music-only meta-analysis synthesized 16 distinct outcome measurements contributed by 7 studies; several studies assessed multiple outcome domains (physical, psychological, physiological), each treated as a separate effect size within the three-level framework. One study (Gacar 2021, n = 1,206) contributed substantially larger sample size via online survey methodology assessing psychological outcomes exclusively; however, influence diagnostics confirmed this study did not unduly drive pooled estimates (Cook's distance = 0.09, DFBETAS = 0.12). Leave-one-out sensitivity analysis excluding Gacar yielded d = 0.18 (95% CI: 0.01–0.35, p = 0.038), demonstrating robustness of the overall effect.

Substantial between-study heterogeneity was observed across the included studies (Q = 89.7, df = 14, p < 0.001; I² = 84.4%, τ² = 0.67). This considerable heterogeneity indicated that 84.4% of the total variance in effect sizes was attributable to true differences between studies rather than sampling error. The prediction interval of −1.42 to 1.80 suggested that future studies implementing similar music interventions could expect effects ranging from moderate detrimental to large beneficial impacts, underscoring the importance of contextual factors in determining intervention effectiveness. The three-level model partitioned total heterogeneity into within-study variance (σ² = 0.28, 42% of total) and between-study variance (σ² = 0.39, 58% of total), indicating that most heterogeneity stems from true differences between studies rather than multiple outcomes within studies. Despite subgroup analyses explaining portions of heterogeneity (outcome domain, timing, sport), residual unexplained variance remained substantial (I² = 71.2%), suggesting unmeasured moderators including individual differences in musical reward sensitivity, genetic polymorphisms affecting dopaminergic function (DRD2, DRD4, COMT), baseline arousal states, and nuanced sport-specific technical demands.

To explore sources of heterogeneity, we conducted pre-planned subgroup analyses by outcome domain. These analyses revealed meaningful differences in music's effectiveness across different types of performance measures, providing insight into the mechanisms underlying intervention effects (Figure 3 and Table 4).

The subgroup analysis revealed that psychological outcomes demonstrated the strongest and most consistent responses to music interventions (d = 0.52, p = 0.011), while physical performance showed variable effects (d = 0.18, p = 0.583), and physiological measures exhibited minimal impact (d = 0.05, p = 0.921). The significant test for subgroup differences (p = 0.015) confirmed that effect sizes varied meaningfully across outcome domains, explaining a portion of the observed heterogeneity.

Studies examining music combined with other modalities demonstrated substantially larger effects than music alone. We stratified eight combination studies by temporal structure: seven acute within-session protocols and one chronic training protocol. Acute combinations (music + caffeine, music + video feedback, music + napping, music + knowledge endpoint) yielded d = 0.89 (95% CI: 0.67–1.11, p < 0.001, k = 7, N = 122), representing immediate neurochemical and cognitive synergies. The single chronic protocol (music + plyometric training over four weeks) produced d = 1.18 (95% CI: 0.72–1.64, p < 0.001, k = 1, N = 16), reflecting cumulative neuromuscular adaptations. Among acute combinations, music + caffeine showed greatest efficacy (d = 1.24, 95% CI: 0.85–1.63, p < 0.001, k = 4, N = 64). Overall combined-intervention effect (d = 0.93) exceeded music-only effect (d = 0.19) by Δd = 0.74 (Figure 4, Table 5).

Publication bias assessment for music-only interventions was conducted at the study level (k = 7 studies) to satisfy independence assumptions, with one representative effect size selected per study. Visual inspection of the funnel plot revealed modest asymmetry, with smaller studies showing greater variability (Figure 5). Study-level Egger's regression intercept was −0.89 (95% CI: −2.31 to 0.53, p = 0.184), not statistically significant. The trim-and-fill procedure estimated one potentially missing study; imputation yielded adjusted d = 0.16 (95% CI: −0.04 to 0.36, p = 0.102). However, statistical power for publication bias detection remains severely limited with k = 7 studies, and these tests should be interpreted cautiously. The fail-safe N calculation estimated 18 additional null studies would be required to nullify the observed effect, providing modest evidence of robustness against unreported negative findings (Table 6).

Comprehensive sensitivity analyses were conducted to examine the robustness of findings across different analytical decisions and study characteristics. Sensitivity analyses varying assumed within-subject correlations for crossover trials demonstrated robustness across plausible values: pooled music-only effect was d = 0.17 (95% CI: −0.01 to 0.35, p = 0.064) assuming r = 0.3; d = 0.19 (95% CI: 0.01 to 0.37, p = 0.039) assuming r = 0.5; and d = 0.21 (95% CI: 0.03–0.39, p = 0.022) assuming r = 0.7. Maximum variation across correlation assumptions was Δd = 0.04, with all confidence intervals overlapping substantially, confirming that conclusions are insensitive to this parameter within plausible ranges. These analyses demonstrated that conclusions remained stable across multiple scenarios, enhancing confidence in the reliability of the overall findings (Figure 6 and Table 7).

To identify potential sources of heterogeneity, we conducted exploratory subgroup analyses for moderators with adequate between-study variation. Music selection type moderated effects: self-selected music yielded d = 0.28 (95% CI: 0.06–0.50, k = 9) vs. researcher-selected music d = 0.03 (95% CI: −0.31 to 0.37, k = 4; Qbetween = 5.12, p = 0.024), suggesting personalized selection enhances efficacy. Intervention timing showed marginal moderation: music during exercise produced d = 0.31 (95% CI: 0.05–0.57, k = 6) vs. pre-exercise d = 0.09 (95% CI: −0.15 to 0.33, k = 7; Qbetween = 3.94, p = 0.047), consistent with real-time neurobiological effects. Study quality did not moderate effects (high PEDro ≥8: d = 0.22 vs. moderate PEDro 6–7: d = 0.14; Qbetween = 0.89, p = 0.345). Sex-specific analysis was unfeasible given 87% of studies included male-only samples. Tempo (range: 120–145 BPM), volume (range: 70–80 dB), and delivery mode (93% headphones) exhibited insufficient variability for subgroup comparison. These exploratory findings require cautious interpretation given multiple testing without correction and small subgroup sizes (Table 8).

This systematic review and meta-analysis represent the first comprehensive synthesis examining music interventions in combat sports, revealing a complex landscape of neurobiological effects that transcend simplistic ergogenic interpretations. The analysis of 15 studies encompassing 1,456 participants demonstrated that music-only interventions produce a small but statistically significant beneficial effect on combat sport performance.

However, the substantial between-study heterogeneity indicates considerable variation in intervention effectiveness, suggesting that music's ergogenic potential operates through multiple, context-dependent pathways rather than a singular mechanism. The most striking finding emerged from the analysis of combined interventions, which demonstrated markedly superior efficacy compared to music alone (d = 0.93 vs. d = 0.19), representing a 389% relative improvement. Music plus caffeine supplementation yielded the largest synergistic benefits (d = 1.24), providing compelling evidence for additive neurochemical mechanisms.

The domain-specific analysis revealed that music's primary sphere of influence operates through psychological pathways (d = 0.52), with moderate effects on physical performance (d = 0.18) and minimal impact on physiological measures (d = 0.05, p = 0.921). These differential effects across outcome domains provide crucial insights into the underlying neurobiological mechanisms governing music's influence on human performance.

To understand how music influences combat sport performance, we must first examine the basic neurobiological principles underlying music processing and reward. Music listening activates a sophisticated network of brain regions that extends far beyond simple auditory processing, engaging what neuroscientists term the “music-reward circuit” (8, 44). This circuit operates through two fundamental neurochemical pathways that work in concert to produce the performance effects observed in our analysis.

The neuro affective pathways outlined above offer a plausible explanation regarding the notable changes seen psychologically (d = 0.52), especially considering modulation of emotions, arousal levels, and attention during the execution of particular tasks. The primary pathway involves the mesolimbic dopaminergic system, particularly the ventral tegmental area (VTA) projecting to the nucleus accumbens and caudate nucleus (45). When individuals listen to preferred music, this system releases dopamine in a temporally distinct pattern: the caudate shows greater activation during anticipation of musical rewards, while the nucleus accumbens becomes more active during the actual experience of musical pleasure (8). This anticipatory-experiential dissociation explains why music's psychological effects (d = 0.52) were substantially stronger than physical effects (d = 0.18) in our analysis. This is aligned with central top-down control patterns rather than bottom-up neuromuscular enhancement. Statistically, this divergence aligns with the domain-specific heterogeneity observed in subgroup analyses (p = 0.015), reinforcing outcome classification relevance for interpreting effect size suffices.

The secondary pathway involves the opioidergic system, which mediates the hedonic aspects of musical experience. Research using naltrexone, a μ-opioid receptor antagonist, demonstrates that opioid signaling contributes significantly to music-evoked pleasure and emotional responses (46). The relatively high internal consistency of psychological outcome measures may stem from employing integrated dopaminergic and opioidergic systems (low variance within d = 0.52 subgroup). Both systems converge on shared limbic structures like the amygdala and orbitofrontal cortex, enhancing transcendental affective homeostasis while allowing stress response modulation. Effects are critical during combat sports considering emotionally driven regulation impacts reflex response latency, decision-making speed, and endurance to fatigue. This dual neurotransmitter system, dopaminergic and opioidergic, creates a powerful neurochemical foundation that explains why music interventions in our review consistently improved psychological outcomes across diverse combat sport contexts.

The prediction error theory provides additional mechanistic insight into music's ergogenic effects. Music listening engages reward-related predictive processes through continuous prediction and violation of musical expectations, creating what Sayal et al. (45) describe as “musical surprises” that trigger pleasure-related neural networks. These prediction errors activate the same mesolimbic circuits involved in processing traditional rewards like food or monetary gains, explaining why an abstract stimulus like music can produce measurable performance benefits in combat sports. This supports more recent meta-analytic anticipatory reward frameworks aimed at explaining psychologically-based interventions' increased adherence, participation, and effort responsiveness. Importantly, these modulations are made via reward pathways without reliance on cardiorespiratory or metabolic workload markers providing physiologic rationale for weak or null physiological results impact findings in our review (d = 0.05).

The superior effects of combined interventions, particularly music and caffeine (d = 1.24), reflect sophisticated neuropharmacological interactions that operate through complementary rather than redundant pathways. Understanding these interactions requires examining caffeine's mechanism of action and how it synergizes with music-induced neurochemical changes to optimize performance outcomes. Caffeine functions primarily as a competitive antagonist at adenosine A1 and A2A receptors in the central nervous system (47, 48). Under normal physiological conditions, adenosine accumulates during periods of high neural activity and binds to these receptors, producing effects that promote rest and recovery: reduced neural excitability, decreased release of excitatory neurotransmitters like dopamine and noradrenaline, and increased perception of fatigue (49). When caffeine blocks these adenosine receptors, it effectively removes this “neural brake,” allowing for enhanced neurotransmission and reduced perception of effort effects that are particularly beneficial during high-intensity combat sport activities.

The synergistic interaction between music and caffeine operates through convergent effects on the dopaminergic system. While music directly stimulates dopamine release in the nucleus accumbens and caudate (8), caffeine enhances this effect by blocking adenosine's inhibitory influence on dopaminergic neurons (50). This dual activation leads to more sustained and temporally congruent activation of mesolimbic reward circuits, which is especially helpful in contexts of high-arousal competition. To a statistical modeler, these synergistic effects are in line with multiplicative instead of additive interaction terms, and can partly account for the increased effect size heterogeneity in the combined interventions subgroup (I² = 78.2%). Additionally, caffeine's antagonism of adenosine A2A receptors in the striatum facilitates glutamate and dopamine release, further amplifying the motivational and arousal effects initiated by music listening (47, 51). These neurochemical cascades would also promote subcortical and cortical adaptations that improve attentional allocation, motor planning, and psychomotor vigilance all primary determinants of combat sports. This is the mechanistic integration that contextualizes the strong physical and technical improvements seen in multi-arm trials with music-caffeine combinations.

From a clinical perspective, the combination of preferred music and moderate caffeine consumption (∼3 mg/kg) produces an optimal synergistic effect. Caffeine reaches its peak mode of action 30–45 min before exercise, when the music induces the peak of dopaminergic activity (52, 53). This is the ideal time to take caffeine to achieve optimal performance and minimize side effects at high doses. The practical implementation of these combined protocols requires careful consideration of individual response variability. Genetic polymorphisms in the CYP1A2 gene, which encodes the primary enzyme responsible for caffeine metabolism, significantly modulate caffeine's ergogenic effects (54). Athletes with the AA genotype (fast metabolizers) show optimal responses to standard caffeine doses, while those with the CC genotype (slow metabolizers) may require dose adjustments or extended timing protocols to achieve similar benefits.

This interindividual variability can account for a portion of the remaining unexplained heterogeneity (residual τ² = 0.67) and would have to be controlled for in subsequent analysis or Bayesian modeling analysis to enhance precision. Similarly, polymorphisms in the ADORA2A gene, which encodes the adenosine A2A receptor, influence both caffeine sensitivity and music reward responses (47). Athletes with certain ADORA2A variants show enhanced dopaminergic responses to both music and caffeine, potentially requiring lower caffeine doses when combined with music interventions. These pharmacogenetic considerations highlight the future potential for precision medicine approaches in combat sport performance enhancement.

Exploratory subgroup analysis suggested differential effects across combat sports (taekwondo: d = 0.12, k = 11; kickboxing: d = −0.10, k = 3; karate: d = 0.80, k = 2; Qbetween = 9.67, p = 0.008). However, these findings must be interpreted cautiously given small study numbers per sport, particularly for kickboxing and karate. The apparent karate superiority (d = 0.80) derives from only two studies and may reflect publication bias, sampling variation, or true sport-specific mechanisms. Adequately powered comparisons require k ≥ 10 studies per sport. These exploratory patterns warrant hypothesis-driven replication rather than definitive conclusions regarding sport-specific effectiveness.

Karate's large observed effect (d = 0.80, k = 2) may reflect movement structure compatibility with musical rhythm, as kata performance involves predetermined technique sequences with specific timing patterns potentially enhanced through auditory-motor coupling (55). However, this interpretation remains speculative given the limited evidence base. The small kickboxing effect (d = −0.10, k = 3) could reflect attentional interference during reactive intermittent efforts requiring rapid adaptive responses, though insufficient data preclude mechanistic conclusions. Sport-specific hypotheses should be tested in prospective trials with adequate statistical power rather than post-hoc subgroup interpretation.

As previously reported, the effect of music on kickboxing performance was minimal, with a d value of −0.10. This suggests that music did not contribute to kickboxing performance. This result could be due to a highly reactive test, a type of intermittent test that requires adaptation to cope with the movements of the punching bag, and a strong need for flexible cognitive control for sustained concentration. In this context, music can distract attention, which is valuable for making quick decisions, but the data collected do not allow for definitive conclusions. Further research is needed to determine the role of music's effectiveness in simulated kickboxing combat. The timing of music exposure emerged as a critical moderator, with concurrent administration (during exercise) showing superior effects (d = 0.31) compared to pre-exercise protocols (d = 0.09). This temporal specificity reflects the real-time nature of music's neurobiological effects. Dopaminergic activation and prediction error signaling occur during active music listening, not as residual effects following exposure (45). This variation is in accordance with pharmacokinetic principles of performance science whereby timing relative to the task onset determines neuromodulatory effectiveness. Meta-analytically, this effect was significant statistically (Δd = 0.22, p = 0.047) and indicates that simultaneous music use is an optimal protocol when sport regulation supports the reception of auditory stimuli during warm-up or exercise. For combat sports, this suggests that music should be integrated into training sessions rather than limited to pre-training preparation, when regulations permit.

The substantial individual variability in responses (prediction interval: −1.42 to 1.80) emphasizes the importance of personalized approaches based on musical preferences, personality traits, and sport-specific contexts. Research demonstrates that musical reward responses are highly individual, influenced by cultural background, previous musical experiences, and genetic factors affecting dopamine sensitivity (56). Combat sport practitioners must therefore adopt flexible, athlete-centered approaches rather than standardized protocols. Individual response heterogeneity mandates personalized protocols. Baseline phenotyping should assess: (1) musical reward sensitivity via Brunel Music Rating Inventory-3 to identify responders likely to benefit; (2) optimal arousal profiles through pre-training feeling scale and arousal monitoring to match music tempo and intensity to individual inverted-U zones; (3) caffeine sensitivity via CYP1A2 genotype or empirical dose-response trials to optimize combined protocols; (4) sport-specific technical demands to align intervention timing with performance windows. Female athletes may exhibit distinct response profiles given documented differences in caffeine metabolism (slower CYP1A2 activity, heightened anxiety sensitivity) and enhanced mood responsiveness to musical stimuli, though limited evidence (k = 3 studies with female samples) precludes definitive sex-specific recommendations pending adequately powered comparative trials.

Several methodological limitations must be acknowledged when interpreting these findings. The substantial heterogeneity between studies (I² = 84.4%) limits generalizability and suggests the presence of important moderating variables not captured in our analysis. This heterogeneity likely reflects genuine differences in intervention protocols, participant characteristics, and contextual factors rather than methodological inconsistencies, given the generally high study quality (mean PEDro score: 7.8 ± 1.2).

The predominance of taekwondo studies (73.3% of included research) may bias findings toward this specific combat sport, limiting applicability to other martial arts disciplines with different technical, tactical, and physiological demands. Additionally, most studies included relatively small sample sizes (range: 10–1,206 participants), potentially limiting statistical power for detecting true effects and increasing susceptibility to Type II errors. The sport-specific subgroup analysis represents exploratory hypothesis generation constrained by small study numbers per discipline. With k = 2 for karate and k = 3 for kickboxing, these comparisons possess insufficient statistical power to distinguish true moderator effects from sampling variation. The observed heterogeneity test (p = 0.008) may reflect Type I error given multiple testing without adjustment. Future evidence synthesis should prioritize within-sport accumulation before cross-sport comparison.

Gender-related considerations represent another important limitation. Most included studies either focused on male participants or failed to conduct sex-specific analyses, despite established differences in caffeine metabolism, music preferences, and reward sensitivity between sexes. Women typically show enhanced sensitivity to caffeine's anxiogenic effects and different temporal patterns of dopaminergic responses to rewarding stimuli, suggesting that optimal protocols may require sex-specific modifications (48). The substantial unexplained heterogeneity (residual I² = 71.2%) despite comprehensive subgroup and meta-regression analyses represents a fundamental limitation constraining generalizability. This heterogeneity likely reflects complex interactions between measured and unmeasured moderators operating at participant, intervention, and measurement levels. Future research should prioritize individual participant data meta-analysis enabling precise modeling of person-level moderators (musical preferences, personality traits, genetic factors), intervention characteristics (tempo-athlete arousal matching, volume titration), and contextual factors (competitive pressure, environmental conditions).

The detection of publication bias (Egger's test: p = 0.032) suggests potential overestimation of positive effects, particularly among smaller studies. However, the trim-and-fill analysis indicated minimal impact on overall conclusions, and the fail-safe N of 23 studies suggests reasonable robustness to unpublished null findings. The intervention heterogeneity across studies, including variations in music characteristics (tempo, volume, genre), delivery methods (headphones vs. environmental sound), and timing protocols, complicates direct comparisons and mechanistic interpretations.

This review represents several methodological advances over previous syntheses of music and exercise performance. The specific focus on combat sports provides unprecedented insight into a distinct athletic population with unique performance demands and cultural contexts. The inclusion of combined intervention analyses addresses a critical gap in the literature, moving beyond simple music-only effects to examine clinically relevant multimodal protocols. The comprehensive search strategy across five major databases, combined with prospective protocol registration and adherence to PRISMA 2020 guidelines, ensures systematic and unbiased evidence synthesis. The application of robust statistical methods, including random-effects meta-analysis with restricted maximum likelihood estimation and comprehensive sensitivity analyses, enhances confidence in the reliability of findings. The high methodological quality of included studies (66.7% achieving high-quality PEDro ratings) strengthens the evidence base, while the systematic application of GRADE criteria provides transparent quality assessment for clinical implementation. The detailed examination of domain-specific effects (psychological, physical, physiological) offers unprecedented granularity in understanding music's mechanisms of action in combat sports.

Music intervention feasibility differs across competitive contexts and combat sport regulations. World Taekwondo, World Karate Federation, and major kickboxing federations prohibit electronic devices and auditory aids during competition bouts but permit music in designated warm-up areas. Consequently, practical protocols target pre-bout preparation (20–45 min before competition) when dopaminergic activation and arousal optimization can influence subsequent performance through residual neurobiological effects lasting 30–60 min post-exposure. Training implementation faces fewer restrictions. Music integration during moderate-intensity technical sessions, conditioning work, and skill refinement capitalizes on concurrent neurobiological effects (arousal modulation, perceived exertion reduction, enjoyment enhancement) without interfering with coach instruction or partner communication. High-intensity sparring sessions may benefit from music during rest intervals rather than active rounds to preserve attentional resources for tactical decision-making. Periodization considerations include increasing music exposure frequency during competition mesocycles (final 3–4 weeks pre-competition) to stabilize protocol familiarity and minimize novelty-related arousal perturbations. Combined interventions (particularly music + caffeine) require competition-specific trialing during training to identify optimal timing: caffeine 45–60 min pre-competition coinciding with music-enhanced warm-up 20–30 min pre-bout. Athlete autonomy in music selection is critical; coaches should facilitate playlist development reflecting individual preferences and arousal regulation needs rather than imposing standardized selections. Practical barriers include venue acoustic constraints, equipment availability (wireless headphones, portable speakers), and individual athlete preferences for silence or self-generated psychological preparation.

Future research must address methodological limitations constraining current evidence. Adequately powered trials require prospective sample size calculation; detecting small effects (d = 0.2) with 80% power (α = 0.05, two-tailed) necessitates n = 394 per group for parallel designs or n = 199 for crossover designs assuming r = 0.5 within-subject correlation. Prospective registration (ClinicalTrials.gov, PROSPERO) with pre-specified primary outcomes, analysis plans, and stopping rules is essential to minimize reporting bias and analytical flexibility. Grappling combat sports (judo, wrestling, Brazilian jiu-jitsu) remain underrepresented (k = 0 in current review), limiting generalizability across combat sport modalities with distinct energetic and technical demands. Standardized outcome batteries should employ sport-specific validated tests: taekwondo-specific agility test, frequency-speed kick test, and simulated competition protocols with temporal-technical analysis. Psychological assessment should prioritize validated instruments (PACES, feeling scale, STAI) administered at standardized time points. Mechanistic neuroimaging substudies employing functional MRI, PET dopamine imaging, or high-density EEG during music + caffeine protocols would elucidate neurobiological pathways and identify predictive biomarkers. Pharmacogenetic research investigating CYP1A2-ADORA2A-DRD2 haplotype interactions with combined intervention responses enables precision medicine implementation. Individual participant data meta-analysis consortia pooling raw data across studies permit person-level moderator modeling (baseline arousal, personality traits, genetic variants) unachievable with aggregate study-level data. All future trials should mandate open data sharing via persistent repositories (Open Science Framework, Figshare) with analysis code (R, Python) and materials (intervention protocols, assessment instruments) deposited contemporaneous with publication. Long-term investigations examining chronic adaptation patterns, tolerance development across competitive seasons, and optimal periodization strategies (loading/tapering phases) address applied questions unresolved by acute effect studies dominating current evidence.

Personalized implementation requires athlete-specific assessment rather than standardized protocols. Practitioners should: (1) conduct individual music preference profiling using validated instruments (Brunel Music Rating Inventory-3) to identify motivational and arousal-regulating selections; (2) match tempo to athlete's optimal arousal zone via systematic feeling scale monitoring across training sessions; (3) phenotype caffeine response through graduated dose trials (1.5, 3.0, 4.5 mg/kg) assessing performance benefits against anxiety/jitteriness side effects; (4) for female athletes, initiate with lower caffeine doses (2.0 mg/kg) given slower metabolism and adjust based on individual tolerance; (5) implement 4-week familiarization during training before competition deployment to minimize novelty-related perturbations. Population-level effect sizes (d = 0.19 music-only; d = 1.24 music + caffeine) represent central tendencies; individual responses span wide ranges (prediction interval: −1.42 to 1.80), with some athletes experiencing negligible or negative effects necessitating alternative arousal regulation strategies. The optimal evidence-based approach combines preferred music (120 BPM, 70–80 dB intensity) during warm-up with low-dose caffeine supplementation (3 mg/kg body weight) administered 30–45 min before training or competition. Individual responsiveness varies significantly based on genetic factors (CYP1A2, ADORA2A polymorphisms), musical preferences, and sport-specific demands, necessitating personalized protocols with systematic monitoring of psychological responses using validated measures (feeling scale, PACES, RPE). Athletes seeking primary psychological enhancement may benefit from music-only interventions, particularly for mood regulation and perceived exertion reduction. Practitioners should recognize that music's effects operate primarily through dopaminergic and opioidergic pathways affecting psychological rather than physiological outcomes, making these interventions most appropriate for sports emphasizing mental preparation and motivation. Regular assessment and protocol adjustment ensure sustained effectiveness while avoiding potential habituation or adverse responses.

Music + caffeine protocols require safety assessment given inter-individual variability in caffeine response and potential anxiogenic effects under competitive stress. The 3 mg/kg dose employed across included combination studies (∼200–210 mg absolute dose for 70 kg athlete) remains below thresholds associated with performance decrements (∼9 mg/kg) or World Anti-Doping Agency prohibitions (urinary caffeine >15 μg/mL, achievable only at ∼9–13 mg/kg) (42). However, CYP1A2 polymorphisms create 3–40 fold inter-individual variation in caffeine metabolism (49). Athletes with AA genotype (fast metabolizers, ∼45% prevalence) exhibit rapid caffeine clearance (half-life ∼3 h) and optimal ergogenic responses at standard doses. AC and CC genotypes (slow metabolizers, ∼55% combined prevalence) demonstrate prolonged half-lives (5–7 h) with elevated adverse effect risk (anxiety, jitteriness, sleep disruption) at equivalent doses, necessitating dose reduction (1.5–2.0 mg/kg) or extended pre-exercise timing (75–90 min) (54). ADORA2A polymorphisms further modulate anxiety responses, with T/T genotype exhibiting heightened anxiogenic sensitivity under combined caffeine-stress exposure (47). Practical safety protocols include: (1) pre-competition genotyping (where accessible) or empirical dose-response phenotyping during low-stakes training; (2) systematic anxiety monitoring via State-Trait Anxiety Inventory during training trials; (3) competition-day dose individualization based on pre-competition arousal states (reduce dose if baseline arousal exceeds optimal zone); (4) 4-week minimum familiarization with consistent dose-timing protocols; (5) avoidance of novel dosing on competition day. Athletes experiencing tachycardia, tremor, or excessive anxiety during training trials should discontinue caffeine co-intervention and utilize music-only protocols. Pre-existing anxiety disorders, cardiovascular conditions, or concurrent stimulant medications represent relative contraindications requiring medical consultation before implementation.