HOBBIT WEEKLY · Original
Comparative Effects of Different Neuromuscular Training Modalities and Doses on Lower-Limb Injury Prevention in Athletes: A Systematic Review and Meta-analysis
For coachesIf you run injury prevention work, the study's most useful signal is that short bouts of 16 to 25 minutes, two or three times a week, were the ones with favourable estimates, though it does not prove a best dose.
Abstract
Background. Lower-limb injuries are common in athletes. Neuromuscular training (NMT) is widely promoted, but the comparative effectiveness of different modalities and the optimal dose remain uncertain.
Methods. We conducted a PRISMA-compliant, PROSPERO-registered systematic review and meta-analysis of randomised controlled trials. PubMed, Web of Science, Scopus, Cochrane Library and Embase were searched to 1 August 2025. Eligible trials compared NMT with a control in athletes and reported lower-limb injury outcomes. Random-effects models pooled risk ratios (RRs) with 95% confidence intervals (CIs). Two reviewers screened, extracted and assessed risk of bias. Prespecified subgroups examined modality and dose (frequency, session duration, intervention length), age, sex and sport. Certainty was rated with GRADE.
Results. Twenty-seven RCTs (27,663 participants) were included. NMT was associated with reduced lower-limb injury risk (RR = 0.72, 95% CI 0.64 to 0.82), although trim-and-fill adjustment attenuated the estimate. Structured (RR = 0.68, 95% CI 0.56 to 0.82) and multidimensional programmes (RR = 0.76, 95% CI 0.61 to 0.93) showed favourable pooled estimates, with no statistically significant between-modality difference. Favourable pooled estimates were also observed in adolescents (RR = 0.67, 95% CI 0.55 to 0.81) and young adults (RR = 0.70, 95% CI 0.50 to 0.99), and in football (RR = 0.71) and basketball (RR = 0.66); however, the between-age difference did not remain statistically significant after Benjamini to Hochberg false discovery rate correction, and no statistically significant differences were detected between sport subgroups. Low-certainty, study-level subgroup analyses yielded favourable pooled estimates for programmes delivered 2 to 3 times per week, for 16 to 25 min per session, and over intervention periods of 1.5 to 12 months; however, these associations did not establish a causal or optimal dose. Certainty was very low overall and low to very low across dosage comparisons.
Conclusions. NMT may reduce lower-limb injury risk, but the magnitude of benefit remains very uncertain. Structured and multidimensional programmes, together with programmes delivered at moderate frequency and duration, showed favourable pooled estimates in exploratory subgroup analyses; however, the dose-related findings represent study-level associations and should not be interpreted as evidence of a causal or optimal dose. These findings may inform the design of routine training programmes, while future high-quality RCTs should address adherence, long-term sustainability, and dose to response interactions across sports and populations.
Introduction
Lower-limb injuries are among the most prevalent health issues in both competitive and recreational sports, accounting for 55 to 70% of all sports-related injuries [1, 2]. These injuries not only impair athletic performance and result in prolonged absences from training, they are also closely associated with recurrent injuries and early-onset osteoarthritis, along with other long-term health consequences. As a result, they impose a significant personal and societal economic burden [3 to 5]. Therefore, the effective prevention of lower-limb injuries through evidence-based methods remains a crucial topic in both sports medicine and sports science.
Among various prevention strategies, neuromuscular training (NMT) has been widely adopted across different populations and sports due to its ability to simultaneously improve strength, balance, proprioception, and motor control. Since the 2000 s, well-established programmes such as FIFA 11+, Prevent Injury and Enhance Performance (PEP), and HarmoKnee have been introduced and have become representative approaches to injury prevention [6 to 8]. Numerous randomised controlled trials (RCTs) have demonstrated that NMT significantly reduces the overall risk of lower-limb injuries [9 to 11]. However, its effectiveness remains influenced by factors, such as population characteristics, training dosage, adherence, and intervention models, while the consistency and interpretability of the evidence remain subjects of ongoing debate [12 to 15].
Previous systematic reviews and meta-analyses generally supported the potential preventive effects of NMT, though the robustness and generalisability of these findings were limited [12 to 23]. FIFA 11+ significantly reduced the risk of knee injuries (injury risk ratio ≈ 0.48), although the applicability of this finding was limited by the small number of cluster RCTs included [18]. The pooled effect estimate was risk ratio (RR) = 0.71, but the prediction interval crossed the null line, indicating uncertainty in the effect across settings [13]. Training frequency and duration may exhibit a dose to response relationship; however, inconsistent definitions and limited subgroup sample sizes reduce confidence in this conclusion [14]. High-frequency (4 to 5 times per week) and long-term (> 26 weeks) interventions were associated with an approximately 25% reduction in knee injury risk (RR = 0.75), but this proposed “optimal configuration” did not account for practical adherence, limiting its applicability [15].
Comparisons between different intervention models also lack a systematic framework. Integrated training has shown potential for both performance enhancement and injury prevention, but rigorous meta-analytic comparisons between different types of NMT were not performed [20]. NMT significantly reduced the risk of ACL injuries (RR = 0.36), but the evidence was limited to a single injury type and did not identify the most effective training model [21].
Furthermore, some previous studies had methodological limitations related to study inclusion and quality control, which further compromised the credibility of the evidence. For example, one previous review included 25 studies, but only eight were RCTs, with the remainder comprising non-randomised or quasi-randomised designs, which compromised internal validity [22]. A recent review reported a 27% reduction in lower-limb injury risk based on 24 studies involving approximately 22,950 participants; however, although the eligibility criteria specified RCTs, the review appears to have included non-randomised studies, raising concerns about consistency in study selection and reducing confidence in the conclusions [23].
While current evidence suggests that NMT can help prevent sports-related lower-limb injuries, several challenges persist: there is a lack of a consistent operational classification of NMT modalities based on programme structure and training content, which limits comparisons between programmes; insufficient stratification of dose-related parameters, making it challenging to explain variations in effectiveness; limited exploration of interaction effects, such as the interaction between intervention modality and duration, in the context of high heterogeneity; and methodological limitations in some previous studies that affect the interpretability and external validity of the evidence. Therefore, while these conclusions have academic value, they offer limited practical guidance for coaches, rehabilitation specialists, and athletes. This study aimed to address these methodological gaps and, through modality classification and dose-stratified analysis, to provide more practical and interpretable evidence to inform programme design.
Based on these gaps, this study undertook the following innovative work under a unified evidence framework: first, applying an operational classification of NMT modalities to compare structured, multidimensional, and sport-specific programmes; second, conducting systematic dose-stratified analyses focusing on training frequency, session duration, and intervention period; third, exploring the sources of effect differences through the interaction between intervention modality and duration in the context of high heterogeneity; and fourth, enhancing methodological transparency and evidence translation value through sensitivity analysis and Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) assessment. Through these efforts, this study aimed to provide more interpretable evidence to inform the design and implementation of programmes for the prevention of sports-related lower-limb injuries.
Methods
Methodology
The review protocol was registered with PROSPERO under registration number CRD420251140979, and the review was conducted in accordance with the PRISMA guidelines [24]. A systematic search was conducted in the following five electronic databases: PubMed, Web of Science, Scopus, the Cochrane Library, and Embase, from database inception to 1 August 2025. Only peer-reviewed articles published in English were considered eligible. The search strategy was structured according to the Population, Intervention, Comparison, Outcomes, Study design (PICOS) framework, and key terms related to NMT and sports injury prevention in athletes (both male and female) were combined using Boolean operators (AND/OR). A detailed search strategy is provided in Additional file 1: Table S1. In addition to database searching, reference lists of relevant articles were manually screened to identify additional eligible studies. Records were imported into EndNote (version 21, Clarivate, Philadelphia, PA, USA), and duplicate records were removed.
Inclusion and Exclusion Criteria
Inclusion Criteria
The studies included in this meta-analysis adhered to the PICOS framework: (1) population: competitive athletes of all age groups, including youth, collegiate, professional, and recreational veteran players. Both male and female athletes were eligible to participate. (2) Intervention: NMT was designed to enhance balance, agility, strength, jump-landing mechanics, or neuromuscular control, including standardised protocols, such as FIFA 11+, PEP, and Knee Injury Prevention Programme (KIPP). Eligible studies were required to provide sufficient information to characterise the NMT intervention and assign it to a training-modality category. (3) Comparator: control groups that received no intervention, standard warm-up routines, or non-NMT protocols, such as general fitness programmes or static stretching routines. (4) Outcome: the primary outcome was the incidence of lower-limb injuries, defined as any injury to the foot, ankle, lower leg, knee, thigh, groin, or hip sustained during training or match play. Eligible studies were required to report sufficient lower-limb injury data to calculate RRs between the intervention and control groups. (5) Study design: RCTs, including cluster RCTs, published in peer-reviewed journals.
Exclusion Criteria
Studies were excluded from this meta-analysis if they were (1) non-RCTs, observational studies, or case reports; (2) duplicate publications or review articles; (3) involving animals or participants with pre-existing diseases; (4) non-English publications or with inaccessible full texts; (5) lacking data on injury incidence or reporting outcomes unrelated to injury prevention; and (6) studies in which protective equipment constituted the primary injury-prevention intervention. Exercise-training aids used as components of NMT, such as balance boards, were permitted.
Study Selection and Data Extraction
Two reviewers (GJ and HZ) independently screened titles and abstracts against the eligibility criteria. The full texts of potentially eligible reports were subsequently retrieved and independently assessed by the same reviewers. Two reviewers (GJ and HZ) independently extracted data from the included studies using an author-developed Microsoft Excel data-extraction form (Microsoft Corporation, USA). The extracted data included study characteristics, participant demographics, intervention details, and injury outcomes. Disagreements during study selection or data extraction were resolved through discussion or consultation with a third reviewer (ZR).
Risk-of-Bias Assessment
Risk of bias was assessed for each trial result using the Cochrane RoB 2 tool [25]. Cluster-randomised trials were assessed using the RoB 2 extension for cluster-randomised trials, including the additional domain concerning bias arising from the identification or recruitment of individual participants within clusters (Domain 1b) [25]. Individually randomised trials were assessed using the standard RoB 2 tool. The target effect was the effect of assignment to intervention, and the outcome assessed was lower-limb injury incidence at the principal follow-up. Trial protocols, registration records, statistical analysis plans, and supplementary materials were sought where available to inform judgements concerning selection of the reported result. Two reviewers (HZ and GJ) independently performed the assessments. Disagreements were resolved through discussion or consultation with a third reviewer (ZR).
Certainty of the Evidence
The certainty of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach and was classified as high, moderate, low, or very low level [26]. Evidence from randomised trials initially started at high certainty and could be downgraded for risk of bias, inconsistency, indirectness, imprecision, or publication bias. Certainty was assessed separately for the overall effect and for the prespecified dosage comparisons of training frequency, session duration, and intervention period.
Data Analysis
A meta-analysis was conducted using R 4.4.3 and RStudio to estimate pooled effect sizes. Injury data were treated as dichotomous outcomes, with RR and 95% confidence intervals (CIs) used to calculate overall effects. Heterogeneity was quantified using the I2 statistic. Random-effects models were used for the primary meta-analyses, because clinical and methodological heterogeneity was anticipated across the included studies. A 95% prediction interval for the overall meta-analysis was calculated from the pooled effect estimate and between-study variance.
Subgroup analyses were conducted to explore the influence of intervention characteristics (session duration, session frequency, and intervention period) and participant factors (age group, sex, and sport type). For subgroup analyses, interventions were classified according to programme structure and training content as structured NMT (predefined programmes with a standardised exercise sequence), multidimensional NMT (programmes combining multiple neuromuscular components), or sport-specific NMT (programmes designed for a particular sport and/or focused on a limited set of sport-relevant neuromuscular components). For each dose-specific subgroup analysis, studies were included only when the relevant variable, training frequency, session duration, or intervention period, was reported or could be derived from the available information. Studies with missing, unclear, or non-classifiable values were excluded only from the corresponding dose-specific subgroup analysis. Stratified analyses further compared intervention modalities (structured, multidimensional, and sport-specific programmes). Additional subgroup analyses were conducted specifically for structured NMT interventions across participant and dosage characteristics. Finally, effect modification by intervention duration was assessed using random-effects meta-regression based on study-level log RRs and corresponding standard errors, with NMT modality, intervention duration, and their interaction term included as moderators. Between-study variance was estimated using restricted maximum likelihood (REML), and the overall modality × duration interaction was evaluated using an omnibus test of the interaction coefficients. To account for multiplicity, p values for the formal subgroup-difference and moderator tests were adjusted using the Benjamini to Hochberg false discovery rate (BH-FDR) procedure. Because training frequency, session duration, and intervention duration were study-level characteristics rather than randomised dose contrasts, dose-related subgroup analyses were interpreted as between-study associations rather than evidence of causal dose effects. All subgroup and interaction analyses were considered exploratory.
Leave-one-out sensitivity analyses were conducted by sequentially excluding each individual study to assess whether the pooled effect was influenced by any single study. As an additional model-specification sensitivity analysis, the overall pooled estimate was recalculated using a fixed-effect model and compared with the primary random-effects estimate. Publication bias was evaluated visually using funnel plots and statistically using Egger's regression and Begg's rank correlation tests (when 10 or more studies were available). The trim-and-fill method was used to adjust for potential missing studies and to recalculate the overall effect.
Results
Literature Search Results
A total of 1607 records were retrieved through database searches. After removing duplicates and screening titles and abstracts, 92 reports were sought for retrieval. Eight reports could not be retrieved, and 84 were assessed for eligibility. Following full-text review, 60 reports were excluded because of different outcome measures (n = 21), ineligible study type (n = 21), duplicate or secondary reports (n = 3), ineligible populations (n = 12), or ineligible interventions (n = 3), leaving 24 eligible studies. An additional nine reports were identified through other sources; three could not be retrieved, and three of the six reports assessed for eligibility were excluded because of ineligible study type (n = 2) or insufficient data (n = 1). Consequently, a total of 27 RCTs were included in the systematic review and meta-analysis [6, 9 to 11, 27 to 49], as shown in Fig. 1.
Characteristics of Included Studies
Additional file 4: Table S2 summarises the characteristics of the 27 included trials (randomised or cluster-randomised) [6, 9 to 11, 27 to 49]. Most studies employed a cluster-randomised design, while a smaller number used individual randomisation. Participants ranged from children (7 years) to veteran athletes over 30 years, with both sexes represented. The majority of trials were conducted in football, with additional studies in basketball, volleyball, handball, futsal (a five-a-side form of indoor football), Australian Rules football, and endurance running.
Intervention durations varied between 7 and 43 weeks, with sessions delivered at frequencies of one to six sessions per week, and session lengths typically ranging from 10 to 25 min. Many trials investigated established warm-up programmes, including FIFA 11, FIFA 11+, FIFA 11+ Kids, PEP, and bounding exercise programme (BEP), whereas others evaluated jump-landing, balance, proprioceptive, strength, or other multicomponent NMT approaches. Sample sizes varied substantially across trials. Outcomes predominantly focused on lower-limb injuries, with specific attention given to knee, hamstring, and patellar injuries in selected trials. Control groups generally continued their usual warm-up or training practices, as presented in Additional file 4: Table S2.
Risk-of-Bias Assessment
Of the 27 included trials, 25 cluster-randomised trials were assessed using the Cluster RoB 2 extension, and two individually randomised trials were assessed using the standard RoB 2 tool. No trial was judged to be at low risk of bias overall; 26 trials were rated as having some concerns and one trial was rated as having a high risk of bias. Concerns frequently arose in relation to the identification or recruitment of participants within clusters (Domain 1b), measurement of injury outcomes (Domain 4), and selection of the reported result (Domain 5). Protocol, registration, analysis plan, or other sufficiently prespecified information was identified for nine trials, whereas such information was not identified in the materials available to the review team for 18 trials. The study-level assessments are presented in Additional file 6: Table S4 and Fig. 2.
Evaluation of Meta-analysis Results
Overall Effect Test
This meta-analysis included 27 RCTs with a total of 27,663 participants. The overall analysis of all NMT interventions demonstrated a significant reduction in the risk of lower-limb injuries, with a pooled RR of 0.72 (95% CI 0.64 to 0.82; Z = 5.23, p< 0.00001). Subgroup analyses indicated consistent protective effects in studies reporting isolated lower-limb injuries (RR = 0.75, 95% CI 0.66 to 0.85) and in those reporting mixed injuries predominantly involving the knee, ankle, or hamstrings RR = 0.61, 95% CI 0.43 to 0.86). Considerable heterogeneity was observed (I2 = 74%). The 95% prediction interval ranged from 0.40 to 1.30, suggesting that the true effect in a future comparable study may range from a beneficial effect to little or no effect, reflecting uncertainty across settings. A summary of the individual study RRs and the pooled effect estimate is provided in Fig. 3.
Subgroup Analysis Results
To explore potential sources of heterogeneity, subgroup analyses were conducted by intervention type, sex, age group, sport, and characteristics of intervention dosage. The main results are presented in Fig. 4 (type, sex, age, and sport) and Fig. 5 (dosage characteristics). Conclusions regarding differences between subgroups were based on formal between-subgroup tests rather than comparisons of statistical significance within individual subgroups.
By intervention type, NMT significantly reduced injury risk with multidimensional programmes (RR = 0.76, 95% CI 0.61 to 0.93) and structured programmes (RR = 0.68, 95% CI 0.56 to 0.82), while sport-specific programmes showed no significant effect. No significant subgroup differences were detected across intervention types.
By sex, injury risk was significantly reduced in both male (RR = 0.65, 95% CI 0.51 to 0.83) and female athletes (RR = 0.75, 95% CI 0.61 to 0.91), with no statistically significant between-sex subgroup difference.
By age group, significant effects were observed in adolescents (RR = 0.67, 95% CI 0.55 to 0.81) and young adults (RR = 0.70, 95% CI 0.50 to 0.99), whereas no significant effect was observed in children (RR = 0.64, 95% CI 0.37 to 1.12) or adults >32 years (RR = 1.10, 95% CI 0.81 to 1.50). The between-age subgroup difference was statistically significant before multiplicity correction (p = 0.047), but it did not remain statistically significant after BH-FDR correction.
By sport, significant reductions in injury risk were observed in football (RR = 0.71, 95% CI 0.60 to 0.84) and basketball (RR = 0.66, 95% CI 0.46 to 0.94), whereas no significant effect was observed in other sports (RR = 0.76, 95% CI 0.52 to 1.10) or mixed sports (RR = 0.78, 95% CI 0.51 to 1.18). No statistically significant subgroup differences were detected across sport types.
By training frequency, significant effects were observed for 2 to 3 sessions per week (RR = 0.71, 95% CI 0.61 to 0.82) and mixed frequencies (RR = 0.61, 95% CI 0.52 to 0.71), whereas one session per week and 4 to 5 sessions per week did not show significant effects. Subgroup comparisons indicated significant differences (p < 0.00001), and the between-subgroup difference remained statistically significant after BH-FDR correction.
By intervention duration, programmes lasting 1.5 to 6 months (RR = 0.73, 95% CI 0.62 to 0.88) and 7 to 12 months (RR = 0.71, 95% CI 0.60 to 0.84) significantly reduced the risk of injury. No significant difference was observed between intervention-duration subgroups, and this conclusion was unchanged after BH-FDR correction.
Two studies with unclear or phase-dependent session durations were excluded from this analysis. Significant effects were found for 1 to 15-min sessions (RR = 0.88, 95% CI 0.78 to 0.98) and 16 to 25-min sessions (RR = 0.66, 95% CI 0.56 to 0.79); however, 26 to 30-min sessions did not reach statistical significance. Subgroup differences were significant (p = 0.03), but this difference did not remain statistically significant after BH-FDR correction.
Subgroup Analysis of Structured NMT
Structured NMT interventions demonstrated a significant protective effect in the overall analysis; however, substantial heterogeneity was observed (I2 = 80%). To further explore potential sources of variability, additional subgroup analyses were conducted by age, intervention duration, session duration, and training frequency. The results are shown in Fig. 6.
By age, significant reductions were observed in school-aged children (RR = 0.50, 95% CI 0.36 to 0.69), adolescents (RR = 0.67, 95% CI 0.49 to 0.90), and young adults (RR = 0.69, 95% CI 0.49 to 0.97), whereas no significant effect was found in adults > 32 years (RR = 1.17, 95% CI 0.78 to 1.74). Subgroup differences were significant (p = 0.01), but this difference did not remain statistically significant after BH-FDR correction.
By intervention duration, the pooled estimates were RR = 0.64 (95% CI 0.41 to 1.00) for programmes lasting 1.5 to 6 months and RR = 0.71 (95% CI 0.59 to 0.86) for programmes lasting 7 to 12 months. No significant difference was observed between intervention-duration subgroups, and this conclusion was unchanged after BH-FDR correction.
By session duration, interventions with 20-min sessions showed significant effects (RR = 0.59, 95% CI 0.47 to 0.74), whereas other durations did not. Subgroup comparisons indicated significant differences (p = 0.02), but this difference did not remain statistically significant after BH-FDR correction.
By training frequency, significant effects were observed for 2 to 3 sessions per week (RR = 0.54, 95% CI 0.40 to 0.72) and >3 sessions per week (RR = 0.73, 95% CI 0.58 to 0.90), whereas one session per week did not show asignificant effect (RR = 1.03, 95% CI: 0.89 to 1.19). Subgroup comparisons were significant (p < 0.0001), and this differenceremained statistically significant after BH-FDR correction.
Interaction Between NMT Modality and Duration
The formal random-effects meta-regression showed no statistically significant interaction between NMT modality and intervention duration (QM = 1.07, df = 2, p = 0.587). Stratified pooled estimates are presented in Additional file 1: Figure B1. For interventions lasting 1.5 to 6 months, the pooled estimates were RR = 0.83 (95% CI 0.697 to 0.996) for multidimensional NMT, RR = 0.64 (95% CI 0.41 to 1.00) for structured NMT, and RR = 0.84 (95% CI 0.69 to 1.02) for sport-specific NMT. For interventions lasting 7 to 12 months, the corresponding estimates were RR = 0.37 (95% CI 0.08 to 1.73), RR = 0.71 (95% CI 0.59 to 0.86), and RR = 0.79 (95% CI 0.48 to 1.31), respectively. These stratified estimates should be interpreted descriptively, because the formal modality × duration interaction was not statistically significant.
Sensitivity Analysis of Pooled Effect
Leave-one-out sensitivity analyses were performed by sequentially excluding each individual study. The pooled RRs ranged from 0.71 (95% CI 0.62 to 0.80) to 0.75 (95% CI 0.67 to 0.84), and all estimates remained statistically significant and directionally consistent, suggesting that the overall finding was not materially altered by the removal of any single study. In an additional model-specification sensitivity analysis, the fixed-effect model yielded a directionally consistent and statistically significant estimate (RR = 0.79, 95% CI 0.75 to 0.83; p < 0.00001), compared with the primary random-effects estimate (RR = 0.72, 95% CI 0.64 to 0.82; p < 0.00001). Given the substantial heterogeneity (I2 = 74%), the random-effects model was retained as the primary analysis.
Assessment of Publication Bias
Funnel plot inspection suggested a certain degree of asymmetry, indicating potential publication bias. Egger's regression test indicated significant asymmetry (t = −3.38, p = 0.0024), whereas Begg's rank correlation test was not statistically significant (z = −1.77, p = 0.076). Using the trim-and-fill method, six potentially missing studies were imputed on the right side of the funnel plot. After adjustment, the pooled effect estimate shifted from an RR of 0.72 (95% CI 0.64 to 0.82) to an RR of 0.80 (95% CI 0.71 to 0.91), indicating attenuation in the estimated magnitude of benefit despite remaining statistically significant (Additional file 2: Figure B2).
Certainty of the Evidence
The certainty of evidence for the overall effect of NMT on lower-limb injury incidence was rated as very low. The evidence was downgraded one level for risk of bias, because all contributing trials had at least some concerns, including one trial judged to be at high risk, one level for substantial inconsistency (I2 = 74%), and one level for suspected publication bias. The publication-bias downgrade was supported by funnel-plot asymmetry, a significant Egger’s regression test (p = 0.0024), and attenuation of the pooled estimate after trim-and-fill adjustment, despite the adjusted estimate remaining statistically significant. Funnel-plot asymmetry may also arise from causes other than publication bias.
For training frequency, the certainty of evidence was low for one session per week, 2 to 3 sessions per week, and mixed-frequency programmes, and very low for 4 to 5 sessions per week. Certainty was low for interventions lasting 1.5 to 6 months and 7 to 12 months. For session duration, certainty was low for 1 to 15-min and 16 to 25-min sessions and very low for 26 to 30-min sessions. Detailed GRADE assessments are presented in Additional file 5: Table S3.
Discussion
Major Findings
The core significance of this study lies in comparing different training modes and dosages and providing evidence that may inform practice. The added value of this study lies in three aspects: first, the application of an operational classification of NMT modalities (structured, multidimensional, and sport-specific) to support subgroup comparisons; second, the implementation of dose-stratified analyses, including analyses within structured NMT; and third, the formal evaluation of the interaction between intervention modality and duration. These subgroup and interaction findings were exploratory and should be interpreted cautiously.
Comparison with Previous Studies
Previous reviews and meta-analyses have supported the preventive effects of NMT, particularly FIFA 11+ in football [16 to 18, 50]. An umbrella review similarly supported the effectiveness of multicomponent lower-extremity injury-prevention programmes in team-sport athletes, indicating that the more important remaining questions concern programme composition, modality, and dosage [51]. Previous dose-focused meta-analyses reported differing patterns, including a possible dose to response relationship in youth athletes and potential benefits of higher-frequency or longer interventions for knee injury prevention [14, 15]. In contrast, the present subgroup analyses did not establish a monotonic dose to response relationship or an optimal dose. Earlier reviews also did not directly compare the three modality categories used in this study [20, 21]; although structured and multidimensional programmes had significant pooled estimates, no statistically significant differences were detected between modalities. These findings complement previous work but should be interpreted as exploratory. Differences between previous findings and the present analysis may partly reflect differences in eligible populations, outcome definitions, intervention classification, and inclusion of non-randomised evidence.
Sources of Heterogeneity
Age-related variation has been proposed as a potential source of heterogeneity in previous research. Previous studies have reported favourable effects in adolescent and young adult populations, whereas findings in children and individuals over 32 years of age were less consistent [13, 34, 52]. However, after correction for multiple subgroup comparisons, the between-age difference did not remain statistically significant, and age was, therefore, not established as a definitive source of heterogeneity. Effect estimates also varied across training-frequency subgroups [44, 46] and session-duration subgroups. After multiplicity correction, only the training-frequency subgroup difference remained statistically significant, whereas the session-duration difference did not. The formal interaction between intervention modality and duration was not statistically significant. Overall, age and dosage characteristics may contribute to heterogeneity, although the specific sources could not be determined from the current analyses.
Based on the updated GRADE assessment, the certainty of evidence for the overall effect was very low because of risk of bias, serious inconsistency, and suspected publication bias. Therefore, the findings, particularly those from dosage subgroup analyses, should be interpreted cautiously. This suggests that higher-quality trials are needed to enhance the certainty of the evidence.
Clinical and Practical Implications
The findings of this study may inform sports medicine and coaching practice. First, structured or multidimensional NMT may be considered as part of regular warm-up routines, such as during pre-competition warm-ups or sport-specific training sessions, when feasible. Second, favourable pooled estimates were observed in adolescent and young adult athletes, although between-age differences did not remain statistically significant after multiplicity correction. In the overall dose subgroup analyses, programmes delivered 2 to 3 times per week, for 16 to 25 min per session, and over intervention periods of 1.5 to 6 or 7 to 12 months showed favourable pooled estimates. These ranges should be interpreted as exploratory study-level associations rather than evidence of a causal or optimal dose, because the comparisons may be confounded by programme type, population, sport, adherence, and other trial-level characteristics. Third, favourable pooled estimates were observed in football and basketball, although no statistically significant differences were detected between sport subgroups; evidence for other sports remains limited. Finally, at the public health and policy levels, NMT may inform injury-prevention programmes in schools and sports clubs, with an emphasis on training coaches and medical personnel in programme delivery and adherence monitoring. These practical considerations should be interpreted cautiously, because the certainty was very low for the overall effect and ranged from low to very low across the dosage comparisons; they should not be regarded as definitive recommendations.
Limitations
This study has several limitations. First, considerable heterogeneity existed across studies, and its specific sources could not be determined. Second, funnel-plot asymmetry and Egger’s regression suggested possible publication bias, which may have affected the estimated magnitude of benefit. Third, the sample sizes in some subgroups (such as older athletes and non-mainstream sports) were small, and effect estimates should be interpreted with caution. Although multiplicity was addressed using BH-FDR correction, the subgroup analyses remained exploratory, and several categories contained relatively few studies. In addition, training frequency, session duration, and intervention duration were study-level characteristics rather than randomised dose contrasts; therefore, causal dose to response inferences cannot be made. Moreover, most of the studies included in this analysis had intervention and follow-up periods of less than 1 year, which limited our understanding of the long-term protective effects and sustainability. Evidence concerning high-frequency and longer-duration interventions was limited and uncertain, particularly because several subgroups contained few studies and had low or very low certainty ratings.
Conclusion
In summary, NMT was associated with reduced lower-limb injury risk in athletes (RR = 0.72, 95% CI 0.64 to 0.82), but the very-low-certainty evidence limits confidence in the magnitude of benefit. Structured and multidimensional programmes showed favourable pooled estimates, while exploratory study-level dose subgroup analyses showed favourable estimates for programmes delivered 2 to 3 times per week, for 16 to 25 min per session, and over intervention periods of 1.5 to 12 months; however, these associations do not establish a causal or optimal dose, and neither between-modality differences nor the modality × duration interaction were statistically significant. Future trials should improve methodological quality and examine adherence, long-term sustainability, dose to response relationships, and modality-by-duration interactions across sports and populations.
Gu et al., Sports Medicine - Open, 2026, doi:10.1186/s40798-026-01142-6. Published under CC BY 4.0: https://creativecommons.org/licenses/by/4.0/.
pubmed.ncbi.nlm.nih.gov
Reprinted in full from the original. Copyright stays with the author and the original publisher.
More in Body and injuries
- Effects of Combined Sprint and Plyometric Training vs Plyometric Training only on Physical Fitness and Sport-Specific Performance in Male Pubertal Soccer PlayersSports Medicine - Open · 18 Nov
- ParticipACTION’s Early Years Report Card: Are Young Kids Developing Physical Literacy?Sport for Life · 12 Nov
- A Parent’s Guide to Strength and ConditioningWorking with Parents in Sport · 11 Nov
- Heading in youth soccer: The scientific and practical guide for coachesAprende Entrenando · 5 Nov
