Training interventions to reduce the risk of injury to the lower extremity joints during landing movements in adult athletes: a systematic review and meta-analysis

Search strategy

The electronic databases (search engines) MEDLINE (PubMed), SPORTDiscus (EBSCO) and Scopus were searched from the database’s inception to 20 December 2021. The combination of the following keywords was used as search terms: ‘athletes OR sports’; ‘unilateral landing OR bilateral landing OR jump-landing OR jump task OR landing’; ‘intervention protocol OR intervention OR training protocol OR exercise OR training OR prevention’; ‘kinematics OR kinetics OR biomechanics OR knee flexion angle OR knee valgus OR injury risk OR injury rate OR landing error scoring system’ (online supplemental table 2).

Study selection

The eligibility criteria were created using the PICO(S) model for clinical questions and are summarised in table 1. Included interventions aimed to improve risk factors during jump landing performance and compared an intervention group to a control group with sham or without additional intervention ((non)-randomised controlled trials, RCTs) or to the intervention group’s baseline levels (eg, pre–post or cross-over trials).35 The outcome measurements should be assessed using reliable tools (eg, VICON Motion Systems, Landing Error Scoring System (LESS)) and analyse jump performance (eg, drop jumps, countermovement jumps).

English and German publications were considered, and geographical restrictions were not applied. Studies were excluded when participants were soldiers, because of their special military training programme. Also, simulation studies were excluded.

Two independent reviewers applied predetermined eligibility criteria to screen titles and abstracts in the Rayyan-Intelligent systematic review software (Rayyan System, Cambridge, Massachusetts, USA).36 An abstract was included when the two reviewers independently assessed it as satisfying the inclusion criteria. Afterwards, full texts were screened and were excluded if they did not fulfil inclusion criteria. Conflicting classifications, of both abstracts and full texts, were discussed and resolved.

Data extraction

Data extraction included: Study identifiers (authors name, year of publication), study design, study sample size and participants’ characteristics (age, gender, activity level). The data were organised according to the training intervention methods, and results were added for the main outcome variables (eg, biomechanical risk factors). Training interventions were classified as (A) technique training, that is, improving the landing technique by instructions, feedback or skill training, (B) dynamic strengthening, that is, plyometrics or a combination of strengthening and plyometrics, (C) static strengthening, that is, strengthening without plyometrics or (D) balance training (static or dynamic). An included training intervention was defined as a ‘prevention programme’ if an apparent sequence of different forms of exercises was used and the author of the study titled this form of training as a ‘prevention programme’. In addition, the different gender responses (ie, men and women) to the training interventions were identified. Furthermore, in terms of practicability of training interventions, the required training materials were sorted according to the training techniques (A–D) and the education of the coach and training time consumption were listed.

Methodological assessment

Two independent reviewers assessed the risk of bias. The Cochrane Risk of Bias tool for randomised trials (RoB2)37 was applied. The aimed target trial for the analysed studies, was a randomised clinical trial with a control and an intervention group consisting of female and male amateur athletes performing an intervention.33 34 In addition, the intention-to-treat effect was chosen to interpret study outcomes.38 The ‘Risk of bias in Non-randomised Studies-of Interventions’ (ROBINS-I) tool was used to analyse other study designs.38 Online supplemental table 3 summarises the adapted controlled questions for the corresponding seven bias domains.38

Data analysis

For RCTs with a biomechanical risk factor as a primary outcome in results, the standardised mean differences (SMDs), known as Cohen’s d effect sizes, with associated 95% CIs were calculated, and a meta-analysis was performed.39

Moreover, to avoid within-group effect size correlation, the SMDs were only calculated if two experimental groups or an additional control group were analysed (ie, RCT or CT study design).40 If a study contributed more than one effect size to the meta-analysis, a combining group approach (eg, putting means and SD of Group A and B together) was used as basis for effect size calculation37 to circumvent double-counting (eg, comparing group A with group C and group B with group C again). In addition, to capture the effectiveness of the intervention, only the effect size of the measurement directly after the intervention (ie, post-test) was considered if multiple measures were collected.

As the considerable between-study heterogeneity was anticipated, a random-effects model was used to pool effect sizes. The ‘DerSimonian-Larid’ estimator was used to calculate the heterogeneity variance   .41 Knapp-Hartung adjustment42 was applied to calculate the CIs around the pooled effect. Moreover, the heterogeneity index    with its 95% CI and additional prediction intervals were calculated.    was interpreted according to Higgins and Thompson 25%=low; 50%=moderate; 75%=substantial. In addition, if    >50% an outliers (ie, CI of studies did not overlap with CI of pooled effect) and influence analyses (ie, leave-one-out method) were performed to study the robustness of the true effect.43–45 Finally, asymmetries of the effect size distribution due to bias distribution were assessed using Egger’s test46 and were visually examined using a funnel plot of effect sizes relative to SE. All analyses were performed by the ‘meta’ V.5.2–0 of R language (R core team, Vienna, Austria). The thresholds for the interpretation of the effect sizes were as follows: 0.20=small; 0.50=moderate; 0.80=large.47 Statistical significance was set at p<0.05.

Search results

The initial database search yielded 2387 results. After removal of duplicates, 1717 records remained. Title and abstract screening resulted in 133 full-text articles that were assessed based on the inclusion criteria. The final full text screening resulted in 31 included studies, of which the flow of the PRISMA selection process is summarised in figure 1. The intervention studies included 15 RCT’s, 15 pre–post intervention studies without a control group, and one cross-over intervention study. The study characteristics are presented in table 2 and additional details are presented in online supplemental table 1.

Request permission

Study characteristics

Sample sizes of the RCTs ranged from 16 to 116 participants with an overall sample size of 682. The sample size of the pre–post intervention studies ranged from 8 to 37 participants with an overall sample size of 280. Eight participants participated in the single cross-over study. Eighteen of the 31 studies included only female participants, 5 studies had only male participants and 8 studies analysed both men and women. Accordingly, more than half of the results were derived from the analysis of female athletes (64%). Three-quarter of all participants were amateur athletes (730), followed by division 1 and 2 collegiate athletes (184) and high-performance athletes (60).

Intervention characteristics

Fifteen interventions were classified as technique training (A), 12 as dynamic strengthening (B), 2 as static strengthening (C) and 2 as balance training (D). The training intervention duration varied between studies with either 1 single training session in total (18 studies) or multiple training sessions performed 2–3 times per week for a total duration of three to 12 weeks (13 studies) (online supplemental table 1). The session duration of each training intervention lasted between 5 and 15 min (10 studies), 12–20 min (3 studies), 15–30 min (4 studies), 40–60 min (5 studies). In 10 studies, no exact training session duration was given, but they lasted 1 (8 studies), 2 (1 study) or a maximum of 3 (1 study) training sessions.

Different types of double limb (22 studies) and single limb (9 studies) jump landing tasks were observed to assess the outcome measurements and included variants of drop jumps (19 studies) and/or countermovement jumps (6 studies), volleyball, basketball or netball -specific jumps (3 studies), tuck jump (1 study), stop jump task (2 studies), lateral single leg jump (1 study) and maximal vertical jump (1 study).

Intervention outcomes

Fifteen studies reported ground reaction force (GRF) outcomes, 23 studies reported joint angle, 5 studies reported joint moment outcome and 3 studies reported LESS as outcome. All but two studies48 49 reported significant differences in outcomes before and after the intervention.

A significant reduction of biomechanical risk factors was reported for all four training types (A-D). In the following, the findings of the 15 RCTs organised by training type (A–D) and gender are reported to focus on cause-and-effect relationships instead of reporting correlations.50 In addition, the findings of the meta-analyses are reported.

Technique training focused on improving the athlete’s landing technique through three forms of instruction (three RCTs) and feedback (four RCTs). Forms of instruction consisted of external focus (directing the athlete’s attention away from the movement), internal focus (direct focus on the movement) and video instruction.

Postintervention, all three forms of instructions led to a significant improvement in knee flexion angle51 52 and a significant reduction in knee valgus moment in drop jump landing.51 A significant decrease in GRF was reported after all forms of technique training, except video instruction. The LESS scoring was significantly improved (fewer errors) when using external focus and video instruction or feedback.51–56

The long-term effect of technique training on biomechanical risk factors is not clear: 1-week postintervention, the vertical GRF during landing can increase again51 52 or further improve, that is, become less during landing.55 The positive effect of the intervention on the knee joint angle seems to reduce after 1–4 weeks52 57 or there is no clear trend.51 Benjaminse et al demonstrated an improved LESS and knee joint angle for external focus and video instruction groups 1 week after the intervention compared with baseline, however, postintervention data were not reported.54

Dynamic strengthening training interventions included plyometric exercises (two RCTs), a combination of plyometrics and lower extremity strengthening (one RCT) and plyometric prevalent prevention programmes consisting of exercises including plyometrics but where trainers are also instructed to improve technique (three RCTs).

One prevention programme with plyometric exercises of increasing intensity significantly improved the knee flexion angle landing pattern in males.58 A reduction in knee valgus and improvement of hip flexion and external rotation angle were reported after dynamic strengthening.58 59 Another programme used a combination of strengthening and plyometric exercises and was able to improve knee flexion angle at peak impact.60 In addition, one prevention programme consisted of a warm-up programme combining strengthening, stretching and balance exercises. The application of this programme significantly decreased LESS scores directly after an 8-week intervention.61 No clear effects of plyometrics were observed for the GRF, where both significant62 and non-significant findings were reported.48

Only Yang et al studied the long-term effect of a 4-week dynamic strengthening intervention. The positive effect on the knee flexion angle remained over time (up to 12 weeks), while the initial positive effect of the training on the vertical GRF was not consistent over time.60

Balance training included static balance training on a wobble board (ie, no external perturbation) (one RCT) and dynamic balance training consisting of increasing perturbation-enhanced neuromuscular training, such as sport-specific technique performance on a rocker board (one RCT).

The intervention programme focusing on static balance49 did not result in significant changes in biomechanical risk factors, whereas the dynamic balance training significantly increased the initial knee flexion angle during landing.63

Regarding gender differences, both genders significantly improved risk factors in terms of knee flexion angle significantly by technique training such as teaching plantarflexed landing, feedback, external instruction,51 54–56 and dynamic balance training with increasing perturbation-enhanced neuromuscular tasks63 (five RCTs).

For meta-analyses, 15 RCTs were available. Due to inconsistent evaluation and reporting of outcome variables, only two variables, the knee flexion angle and vertical GRF, could be examined. The meta-analyses used 11 RCTs in total.

Four studies showed a large (>0.80),52 57 58 63 one study showed a moderate (>0.50)57 and one study a small (>0.20)51 effect size for knee flexion angle. The between-study heterogeneity variance was estimated at    =0.53 (95% CI 0.25 to 5.72) with a high heterogeneity (   =83%; 95% CI 64% to 92%). Based on substantial heterogeneity in the main analysis (   >50%), an outliers and influence analysis was performed and effect sizes of Letefatkar et al and Makaruk et al were removed from the random-effect model.62 63 Finally, a significantly medium pooled effect size (g=0.77; 95% CI 0.33 to 1.21) was found for the three technique training interventions and one prevention programme (dynamic strengthening) on knee flexion angle (p=0.01).

A large effect size (>0.80) for GRF was shown in one study.56 Four studies showed a small effect size (>0.20)51 52 55 62 and three studies had no effect.48 49 60 The between-study heterogeneity was estimated at    =0.11 (95% CI 0.00 to 0.80) with a moderate heterogeneity (   =48%; 95% CI 0 to 77%) and a non-significant pooled small effect (g=−0.24; 95% CI −0.63 to 0.16).

Publication bias analysis of both meta-analyses showed no asymmetry in a visual inspection of funnel plots, and application of Egger’s test was not possible due to the limited number of studies analysed (<10 studies each) (figure 2).

Request permission

Required material and coach education characteristics

Table 3 summarises the practicability of the interventions in terms of training material and coach education required to perform the intervention. Most interventions required materials such as boxes (height ranged from 20 to 50 cm), a BOSU, ball or balance board. In addition, cones, hurdles, walls and mattresses were the second most common training materials. For feedback interventions, a video camera and monitor were needed in two out of four interventions, and in one intervention, a video of an expert performing a jump landing model must be provided. For instruction interventions, 3 out of 10 studies required an expert performing or instructing the landing manoeuvre on video.

Fourteen studies required coach education, with three training interventions including a training brochure or poster with detailed information on the exercises.58 59 61 No information was provided on how long it takes to master these programmes. One training intervention introduced instructions via DVD for coaches,58 three interventions required knowledge about LESS51 54 57 and one provided a checklist of jump landing performance for coaches.55 Five exercises required knowledge and performance of correct jump landing.51 54 55 57 64

Risk of bias in studies

The risk of bias of the included studies, sorted by study design, are shown in figure 3. Twenty-six studies showed an overall moderate risk of bias. Across the RCTs, higher risk of bias was due to lack of blinding of participants and researchers (11 studies), lack of a randomisation tool (3 studies), or due to missing information regarding a prespecified analysis plan that was finalised before unblinded outcome data were available for the analysis (4 studies). All pre–post and cross-over studies had a moderate risk of bias mostly due to missing follow-up outcome measurement on a second date (16 studies).

Request permission

Relevant intervention types

Technique training and dynamic strengthening improved the knee joint flexion angle in the meta-analysis and thus can be recommended interventions in amateur setting. These intervention types were accompanied by other improvements in biomechanical risk factors and were also successfully integrated into other study programmes in this review. First, technique training as external focus instruction resulted in a significant improvement in knee flexion angle65 and LESS score54 compared with internal focus instruction. The positive effect of external focus is based on the enhanced motor system to self-organise the movement, which will result in an improved movement pattern (ie, optimal kinetic solution), compared with internal focus.65 These results are in line with the general positive effects of external focus instruction on motor learning and movement technique when compared with internal focus instruction.66 67 Furthermore, the review showed that one to three technique training session based on feedback or instruction improved knee flexion angle,57 LESS score53 54 and reduced GRF.52 55 This rapid performance improvement after a maximum of three sessions seems to be useful for integrating this intervention into training routine.68

Second, changing the unilateral landing technique to landing on the toes with the body leaning forward also resulted in a reduction of risk factors in three pre–post studies.64 69 70 Particularly, a toe landing pattern increased the knee flexion angle and reduced GRFs.69 Whereas heel contact (dorsiflexion) at initial ground contact can be the cause of non-contact ACL injuries.71 Thus, analysing the athletes landing technique should be considered and landing on the toe should be preferred when performing single or double leg landings.69

Lastly, a warm-up programme including a combination of dynamic strengthening, technique and balance training was useful as injury prevention strategy.61 These findings are in line with the study of Herman et al, who showed the reduction of lower extremity injuries in amateur athletes using neuromuscular warm-up strategies combining, among others, dynamic strengthening and perturbation training.72

Large variation of outcome measures

General consideration of biomechanical risk factors within the included studies showed large variation in outcome measures, which affected direct comparison of risk factors (eg, maximum knee flexion angle and knee flexion angle over time). Future studies should consider analysing jump landing movements using, for example, the LESS, which identifies movement deficiencies and poor landing technique, as it serves as a validated method and simplifies assessment in the field and direct comparison between studies.73 Using the LESS score enables extraction of individual biomechanical risk factors from the score (eg, knee valgus) and comparison of these partial results between studies.74 In addition, caution should be taken when using two-dimensional motion capture analysis while focusing on the dynamic knee valgus during landing movements, as it consists of proximal (hip adduction/internal rotation) and distal (tibial abduction) movement to the knee, which take place in two different planes of motion. Therefore, three-dimensional analysis represents the gold standard for kinematic analyses.75

Gender differences

Regarding gender differences, 65% of the results were obtained for female samples. As women have been reported to be at higher risk for knee injuries, injury prevention is crucial for them. Female athletes reduced biomechanical risk through dynamic and static strengthening, technique training and dynamic balance training, like male athletes. A study by Crossley et al has shown that it is beneficial to use a multifaceted intervention programme (eg, strength, plyometric exercises, dynamic balance training) to reduce the rate of lower extremity injury in women.76 The included studies support these findings and show significant reduction of biomechanical risk factors by applying a combination of plyometrics and strengthening exercise (ie, dynamic strengthening).59

Practicability of intervention implementation

When referring to the practicability of the training intervention the results suggested a moderate time commitment (around 30 min), minimal training material and coach education, which make the interventions integrable into training routine (table 3 and online supplemental table 1). The examined interventions mainly used materials which are often part of the existing training equipment, for example, hurdles and boxes or no additional material, and thus facilitate the application of the training programme.52 67 77 A barrier to implement the intervention in amateur sports could be the incurred costs for monitors or cameras that were used in some studies to provide feedback regarding jump landing performance. However, significant improvement in LESS score were also obtained when visual feedback was provided with mirrors, which drastically reduced costs if available.53

Although good coach education reduces injury risk,77 the present review demonstrated that targeted simple training interventions can significantly reduce biomechanical risk factors, thus circumventing the time and cost hurdle of continuing education in amateur sports. However, successful training intervention depend on programme adherence. Steffen et al reported a larger positive effect of FIFA 11+ when participants showed high adherence compared with low adherence athletes.78

Limitations and strength of the review

As all research, this review was subject to limitations of the included studies. First, the unaccounted-for influence on outcomes, such as sample frequency, footwear, age and activity level of participants was acknowledged as limitations. Second, more than half of the results of this review came from the analysis of female athletes (65%). Therefore, caution is advised in the general transfer of results for male amateur sport athletes. Third, direct comparison of biomechanical risk factors between studies was difficult because the collection of these parameters in the studies was not uniform. Fourth, the meta-analysis, which showed a significant result of increasing knee flexion angle, only included four RCTs (figure 2). Furthermore, baseline values differed between groups and influenced the intervention effects.49 55 Nevertheless, risk of bias analyses of the 31 included studies showed a mostly moderate risk. Main reason for this was the missing follow-up outcome measurement of the pre–post and cross-over studies on a second date. Therefore, future studies should aim to perform a RCT to increase the quality of the results and add a follow-up measurement to reduce the risk of bias. So far, the results suggest that the interventions have a positive effect on the biomechanical risk factors directly post intervention, but it is not clear if these effects hold long-term if training is not performed or neglected.51 52 57 60

The key strength of this review is the comprehensive literature search conducted in three databases, rigorous checking of the extracted data performed by two independent reviewers, and the additional risk of bias analyses carried out with the most reliable Cochrane Risk of Bias Tools (RoB2, ROBINS-I) performed independently by the main author and a third reviewer. In addition, 70% of the included studies analysed amateur athletes, so the transfer of the results to amateur sports is given.