1 Introduction

Although physical activity has multiple health benefits participation in sports also carries a risk of injury. Sports-related injuries are detrimental to an injured athlete’s health [1], may cause permanent disability, or even terminate the athlete’s sports career [2]. Injuries also burden the health care system as the treatment of sports injuries is often demanding and expensive [1, 3]. Injury prevention research is needed to promote safe participation in exercise [1, 4].

In the last 10 years, the number of published sports injury prevention studies has increased [4, 5]. The effects of several interventions aiming to prevent sports injuries have been studied in RCTs. Earlier RCTs have been more concerned with protective equipment, such as insoles and external joint supports, but recently an increased number of training programmes and multi-interventions have been studied [4, 5]. The effects of injury prevention methods have been studied in several systematic reviews earlier [6–21]. Some of the reviews have included only RCTs, but a few have also included other controlled trials. These earlier reviews have summarised the effects of specific injury prevention methods or prevention of specific types of injuries. As far as we know, the study by Aaltonen et al. [22] is the only systematic review of RCTs summarising the effects of all randomised controlled interventions intended to prevent sports injuries. Since then, many new trials have been published. Although the number of systematic reviews published is quite large, they have seldom included meta-analyses due to the small number of homogeneous studies.

Due to an increase in the number of sports injury prevention trials, what is known about preventing sports injuries needs updating. The aim of the study was to update and summarise the effects of preventive interventions.

2 Methods

2.1 Search Strategy

The literature search for this study was conducted by combining two independent, similarly conducted search processes. The first search was conducted by Aaltonen et al. [22] until December 31, 2005 and is described elsewhere. The second literature search was accomplished using the same search strategy as Aaltonen and co-workers, except that it began from 1 January 2006.

The systematic literature search was conducted in September 2013. Relevant trials were searched for in the following databases: PubMed, MEDLINE (Ovid), SPORTDiscus, the Cochrane Central Register of Controlled Trials, CINAHL (Cumulative Index to Nursing and Allied Health Literature), PEDro (the Physiotherapy Evidence Database), and Web of Science. The update search was conducted from 1 January 2006, to 24 September 2013. The following key words were used in the search: sports injury/ies, athletic injury/ies, prevention, preventive, randomiz/s/ed, controlled trial, and randomiz/s/ed controlled trial. Various combinations of the search terms were used. In addition, the reference lists of articles retrieved and relevant reviews were hand searched.

2.2 Trial Selection

The electronic search process yielded in total 5580 articles (4803 items in the first search and 777 in the second). The retrieved articles were first assessed on the basis of titles and abstracts. After the first screening, 5462 articles (4755 and 707) were excluded. The remaining 118 potentially eligible articles (48 and 70) were evaluated more thoroughly on the basis of the full article. Relevant reviews and reference lists of retrieved articles were searched hand searched, and two studies were included as a result of the hand search. Altogether 68 trials (32 and 36) were included in the present systematic review. Eight trials could not be pooled for meta-analysis due to lack of sufficient data. Thus, 60 trials provided adequate data and were included in the meta-analysis. The literature search is presented as a flow chart in Fig. 1.

Fig. 1
figure 1

Literature search flow chart

Full size image

2.3 Inclusion Criteria

Selected articles, published in English, had to examine the effects of any preventive intervention in sports injuries. The trials selected had to be randomised or quasi-randomised (the authors have acknowledged that the used randomisation procedure does not fulfill the current methodological recommendations of randomisation) controlled trials, and published in a peer-reviewed journal. In the second search, during the hand search of the reference lists of relevant reviews and retrieved articles, also articles published before 2006 were included if they met other inclusion criteria. The outcome of the trial had to be injury rate or the number of injured individuals. To maximize the information we can get from the literature we included studies on athletes from recreation to elite levels and from different sports disciplines as well as studies on military recruits although the target groups of the specific studies need to be taken into account when generalising the results.

2.4 Exclusion Criteria

Trials were excluded if they were not randomised, if there were no control group, or if the outcome was other than sports injuries. Also abstracts without the full text available were excluded. The study report had to contain adequate information about the trial protocol and the injury rate or the number of injured individuals as an outcome. One article was excluded because the article had been retracted afterwards on the basis of ethical reasons.

2.5 Data Extraction

Data from each study included was extracted from on the full text. In case of insufficient data, the authors were contacted via email. The study design, description of the intervention, characteristics of participants, and main outcomes from each article were extracted, and are presented in Electronic Supplementary Material Table S1.

Calculations for meta-analysis were made with Cochrane Collaboration Review Manager 5.1 software. All calculations were made according to the primary outcomes of the studies. The calculations were primarily based on the number of injured individuals in the intervention group and in the control group. If the number of injured individuals was not available, the number of injuries was used instead. Odds ratios (OR) with 95 % confidence interval were used as the effect measure, the statistical method was inverse variance, and the analysis model was based on random effects. Statistical heterogeneity (I 2) and test for overall effect was calculated and p values <0.05 were regarded as statistically significant.

2.6 Methodological Quality Assessment of the Selected Trials

Methodological quality assessment of the included trials (Electronic Supplementary Material Table S2) was made as recommended by Furlan et al. [23]. The quality assessment was made independently by two authors. In case of disagreement, consensus was found through discussion. The quality assessment list consists of 12 criteria: method of randomisation, concealed allocation, blinding of participants, blinding of care providers, blinding of outcome assessors, drop-out rate, analysis according to allocated group, reporting without selective outcome, baseline similarity of the groups, co-interventions, compliance, and timing of outcome assessment. Each criterion was scored as ‘yes’, ‘unclear’ or ‘no’, and ‘yes’ indicated one point. The studies were rated as having a low risk of bias when at least six out of twelve points were scored, and study had no other serious flaws (e.g. high drop-out rate in one group or compliance threshold less than 50 % of the criteria) [23]. Studies were rated as having a high risk of bias if fewer than six points were scored, or if a study had one or more serious flaws. Studies were not excluded due to low scores on methodological quality.

3 Results

Altogether, 68 randomised controlled trials examining the effects of preventive intervention on sports injuries were discovered through a systematic literature search [24–92]. The results of the methodological quality assessment are presented in Electronic Supplementary Material Table S2. The highest score a study received was 9/12 and the lowest 2/12. The average score received was 5/12.

Trials were divided into seven groups (insoles, external joint supports, training programmes, stretching, protective head equipment, modified shoes, and injury prevention videos) according to the type of the intervention. Some groups were further divided into subgroups according to intervention characteristics. This was done even though there might have been some heterogeneity in interventions and in methodological issues between the studies. Note that five trials [24–28] had two or more intervention groups tested, but only one control group. These interventions were pooled as individual trials. Therefore, in the forest plots, the number of participants in the control groups of these trials is actually multiple. Seven of the included trials [29–35] could not be pooled due to insufficient injury data. In addition, one included study [36] could not be pooled to any sub-group of studies because it was the only study to assess the effects of dietary supplements on sports injury risk.

3.1 Insoles

Nine of the trials [24, 30, 31, 37, 38, 62, 63, 73, 79] studied the effects of insoles to reduce the risk of lower limb injuries among military recruits (total of 4788 subjects). In the study by Smith et al. [24] two different types of insoles were used, and these interventions were treated as individual trials. Eight individual interventions were pooled and six of these yielded effective results. Although two interventions [37, 38] failed to show preventive effect, test for overall effect (Z = 2.83, p = 0.005) remained statistically significant, and insoles significantly reduced the risk of injuries (pooled OR 0.51, 95 % CI 0.32–0.81). Heterogeneity between the studies was strong (I 2 = 82 %, p < 0.001) (Fig. 2). Two studies could not be pooled; the result from the study by Larsen et al. [30] favoured intervention (RR 0.70, 95 % CI 0.50–1.10), whereas no group differences were observed in the study by Finestone et al. [31].

Fig. 2
figure 2

Insoles vs. control. OR, pooled OR, 95 % confidence intervals, and test of heterogeneity

Full size image

3.2 External Joint Supports

The effects of external joint supports were studied in ten trials (total of 13808 subjects). All seven interventions [25, 27, 39–41, 43, 46] assessing various ankle supports (ankle stabilizers, and outside-the-boot braces) reduced ankle injuries compared to no ankle supports (pooled OR 0.40, 95 % CI 0.30–0.53, Fig. 3). The subjects in these trials (total of 6662 subjects) were young male and female athletes in basketball, male athletes in football and American football, and military paratroopers.

Fig. 3
figure 3

External joint supports vs. control. OR, pooled OR, 95 % confidence intervals, and test of heterogeneity

Full size image

In two trials [44, 45] assessing wrist supports, in a total of 5750 subjects, wrist supports were effective in protecting snowboarders against wrist injury (pooled OR 0.25, 95 % CI 0.12–0.51). Knee supports were studied in one trial [46] in which the use of prophylactic knee braces significantly reduced the number of knee injuries among 1396 military cadets while playing football (OR 0.43, 95 % CI 0.24–0.78) (Fig. 3).

According to ten trials studying the effects of external joint supports, the intervention group suffered significantly fewer injuries than the control group (pooled OR 0.39, 95 % CI 0.31–0.49). Statistical heterogeneity between the studies was low (I 2 = 13%, p = 0.32) (Fig. 3).

3.3 Training Programmes

The effects of training intervention on sports injury prevention were studied in 36 of the trials included. These interventions were divided into six subgroups: balance board training, multi-intervention with balance board training, other multi-interventions, warm-up programmes, strength training, and graded running programmes (Fig. 4).

Fig. 4
figure 4

Training programmes vs. control. OR, pooled OR, 95 % confidence intervals, and test of heterogeneity

Full size image

On the basis of seven trials [25, 27, 55, 67, 82, 85, 86] (1922 subjects), balance board training significantly reduced the number of sports injuries in the intervention group compared to the control group (pooled OR 0.45, 95 % CI 0.28–0.73). The results showed heterogeneity (I 2 = 61 %, p = 0.02). Examination of the studies investigating a multi-intervention with balance board training (3458 participants) yielded consistent results. Multi-interventions using balance board training were effective in reducing the risk of sports injuries (pooled OR 0.46, 95 % CI 0.31–0.64). Moreover, other multi-intervention studies (5429 subjects) also achieved preventive effects (pooled OR 0.63, 95 % CI 0.42–0.95), although the results were somewhat inconsistent, and heterogeneity was strong (I 2 = 84 %, p < 0.001).

Of the eight trials [64, 70, 71, 74, 80, 81, 87, 88] investigating the effects of a warm-up programme (13817 subjects), statistically significant results were found in half of them (ORs from 0.29 to 0.92). However, the pooled result reached statistical significance; warm-up programmes tended to reduce the risk of sports injuries (pooled OR 0.64, 95 % CI 0.49–0.83). The results showed statistically significant heterogeneity (I 2 = 66 %, p < 0.01).

The effects of strength training on the lower extremity injuries were assessed in four studies [27, 33, 47, 48] (1232 subjects). In the studies by Askling et al. [47] and Petersen et al. [48] eccentric strength training significantly reduced the risk of hamstring injuries among football players, whereas Mohammadi [27] found no significant effect of strength training on the recurrence of ankle sprain. The combined results confirmed that strength training achieved a significant reduction in the risk of injuries in the intervention group compared to the control group (pooled OR 0.27, 95 % CI 0.16–0.45). The results showed no heterogeneity (I 2 = 0 %, p = 0.59). The results of the study by Gabbe et al. [33] could not be pooled, but no differences were found in the rates of hamstring injuries between the intervention and control group (RR 1.2, 95 % CI 0.5–2.8).

In the studies by Bredeweg et al. [49] and Buist et al. [50], a graded training programme in the prevention of running related injuries among novice runners (848 subjects) failed to show any preventive effects (pooled OR 0.97, 95 % CI 0.69–1.38). No heterogeneity was observed (I 2 = 0 %, p = 0.69).

According to the 35 trials, training programmes were effective in reducing the risk of sports injuries (pooled OR 0.55, 95 % CI 0.46–0.66). Heterogeneity between the studies was marked (I 2 = 75 %, p < 0.001) (Fig. 4).

3.4 Stretching

Four trials [34, 69, 77, 78] investigated the effects of stretching on lower extremity injuries (total of 4812 participants). Stretching appeared to have no effect on the rate of injuries (pooled OR 0.92, 95 % CI 0.80–1.06) (Fig. 5). There was no statistical heterogeneity between the studies. The study by Bello et al. [34] could not be pooled, but the results showed no differences in the risk of injury in a rhythmic stabilisation method from the normal stretching.

Fig. 5
figure 5

Stretching vs. control. OR, pooled OR, 95 % confidence intervals, and test of heterogeneity

Full size image

3.5 Protective Head Equipment

Three trials [28, 32, 51] (total of 5010 subjects) with four different comparisons studied the effects of protective head equipment on head injuries or concussions. In the study by McIntosh et al. [28], two different types of headgear were tested among 4095 rugby players, with no preventive effect (pooled OR 1.06, 95 % CI 0.91–1.24). Similar results were found in the study by Barbic et al. [51], in which the use of mouth guards was not effective in reducing the rate of concussions among 614 university American football and rugby players (OR 1.04, 95 % CI 0.56–1.94). Consequently, the pooled OR of protective head equipment was 1.06 (95 % CI 0.91–1.24), with no statistical heterogeneity between the studies (I 2 = 0 %, p = 0.66) (Fig. 6). The results of the study by Finch et al. [32] with 301 Australian football players could not be pooled, but showed opposite results: custom-made mouth guards had a significant effect on the rates of head and orofacial injuries in the intervention group compared to the control group (RR 0.56, 95 % CI 0.32–0.97).

Fig. 6
figure 6

Protective head equipment vs. control. OR, pooled OR, 95 % confidence intervals, and test of heterogeneity

Full size image

3.6 Modified Shoes

The effects of modified shoes on lower limb injuries were studied in four trials [26, 29, 35, 52] (total of 1408 subjects) with five different interventions. The results of the three individual comparisons were pooled, and are illustrated in Fig. 7. Barrett et al. [26] compared two different types of high-top basketball shoes to low-top shoes among basketball players. Milgrom et al. [52] studied modified basketball shoes among military recruits. None of these three interventions were able to show any reduction in the risk of injury (pooled OR 1.23, 95 %CI 0.81–1.87). No heterogeneity was observed (I 2 = 0 %, p = 0.79). In addition, two studies could not be pooled. In the study by Finestone et al. [29], no group differences were observed, but the results of the study by Kinchington et al. [35] investigating the effects of a tailored footwear programme, favoured the intervention group.

Fig. 7
figure 7

Modified shoes vs. control. OR, pooled OR, 95 % confidence intervals, and test of heterogeneity

Full size image

3.7 Injury Prevention Videos

Three interventions [53, 54, 92] with injury prevention videos (total of 1103 subjects) yielded contradictory results. An instructional ski video reduced the injury risk in downhill skiers [53], whereas another ski and snowboard injury prevention program including video and brochure was not effective to reduce injuries in school-aged children [92]. Similarly, a video-based awareness programme had no effect on the rate of injuries among football players [54].There was considerable heterogeneity between the studies (I 2 = 67 %, p = 0.05) and the combined effects of the injury prevention videos were not significant (pooled OR 0.86, 95 % CI 0.44–1.68) (Fig. 8).

Fig. 8
figure 8

Injury prevention videos vs. control. OR, pooled OR, 95 % confidence intervals, and test of heterogeneity

Full size image

3.8 Other Interventions

Lappe et al. [36] investigated the effects of calcium and vitamin D supplements on the incidence of stress fractures in female military recruits (total of 5201 subjects). The results showed that calcium and vitamin D supplements were effective in reducing the risk of stress fractures (OR 0.80, 95 % CI 0.67–0.97) [36].

4 Discussion

4.1 Principal Findings

In our systematic review we included 68 RCTs examining the effects of various preventive interventions on the risk of sports injuries. Meta-analyses were conducted using 60 RCTs with 66 comparisons. According to the data available, insoles, external joint supports, and training programmes with different components appear to be effective in reducing the risk of sports injuries, whereas stretching, modified shoes, and injury prevention videos failed to show preventive effects.

4.2 Comparison to Previous Reviews

Orthotic insoles are widely used to prevent overuse injuries [38]; nevertheless, the evidence of their effectiveness has been inconsistent [38, 93]. Evidence from earlier reviews has shown that the use of orthotic insoles may prevent first-time lower limb injuries [9] and tibial stress fractures [93], and that shock-absorbing insoles may be effective in reducing the incidence of stress fractures [14]. However, insoles appear not to be effective in reducing lower limb soft-tissue running injuries [19]. In our meta-analysis, six of the eight pooled studies supported the use of insoles to prevent lower limb injuries, whereas the studies by Withnall et al. [37] and Mattila et al. [38] reported no preventive effects. Interestingly, the quality assessment of these trials revealed that only the trials by Withnall et al. and Mattila et al. were rated as having a low risk of bias, whereas the other trials had a high risk of bias. Although the pooled results showed a significant preventive effect, there is a potential risk of bias. These results therefore need to be interpreted with caution. The applicability of these results is moreover limited primarily to young men undergoing military training and cannot be directly generalized to athletes even though military training includes high-intensity physical training [22].

Trials assessing the effectiveness of external joint supports have mostly been conducted among subjects in high-risk sporting activities, such as football, basketball, American football, parachute jumping, and snowboarding. In all the trials included in the present meta-analysis except one, the use of external joint supports provided beneficial protection against ankle, knee, or wrist injuries. Interestingly, the statistical heterogeneity between the trials of external joint supports was very low, even though there were differences in the study designs, whereas in most of the other groups with multiple trials, heterogeneity was high. Ankle sprains have often been reported as the most common type of injury encountered in sports [15, 94]. The present finding of a preventive role of external ankle supports is consistent with earlier research [15, 94, 95]. However, in the present analysis, the external ankle supports used were different kinds of stabilizing devices, such as orthoses and braces, but for instance taping, also reportedly effective for reducing ankle sprains [15, 94], was not studied in these RCTs. Although anterior cruciate ligament (ACL) injuries of the knee are a well-known problem, especially among female athletes participating in pivoting and cutting sports [96], the effects of prophylactic braces to reduce the incidence of knee injuries has not been commonly studied in RCTs. This is probably because earlier studies have reported a lack of evidence of their effectiveness in preventing ACL injuries [96], but also due to a growing interest in the effects of specific preventive training interventions to reduce the risk of lower extremity injuries. In this meta-analysis only one trial [46] demonstrated the effect of prophylactic knee braces and reported a reduced risk of knee injuries. Note that this trial was conducted among military recruits while playing football, and the reduction of knee injuries was dependent on a player’s position. Based on one trial only, the implications of the effectiveness of knee bracing cannot be drawn.

The number of interventions assessing training programmes has increased nearly three-fold since the previous systematic review [22]. This reflects the current trend in sports injury research. Exposure to extrinsic risk factors, such as environmental conditions or other players, can seldom be influenced. Even though not all intrinsic risk factors can be changed, factors such as physical fitness, muscle strength, motor abilities and sports specific skills are highly trainable. Most of the training programmes designed for the prevention of injuries aim to influence these risk factors by enhancing athletes’ intrinsic abilities. What makes the interpretation of these results complex is the variety of components used in training interventions.

An earlier systematic review of research on the prevention of sports injuries yielded preliminary findings of balance board training as a preventive strategy [22]. Since then seven new RCTs on balance board training alone or as a part of multi-intervention have been published. Balance training can improve both static and dynamic balance and enhance postural control during sports which may reduce the risk of injury [86], and moreover likely improve neuromuscular control. The present findings further support the benefits of balance board training. Balance board training seems to be effective especially in reducing the risk of ankle injuries, but also to some extent, when part of neuromuscular training, of other injuries of the lower extremities. However, when balance board training is combined with the other components of a multi-interventional training programme, the actual effect of the balance board remains unclear. Interestingly, data from Electronic Supplementary Material Table S3 indicates that 80 % of all effective training interventions included some sort of balance or coordination component. Five trials used balance board training as home-based training and four of these home-based interventions were demonstrably effective. At least among athletes frequently accustomed to also doing exercises in their leisure time, a home-based training intervention can be as effective as a supervised intervention.

According to the review by Fradkin et al. [7], the evidence of warming up has been insufficient. The present finding on the effectiveness of warm-up programmes is more promising, but still, to some extent contradictory. Although the pooled result showed a preventive effect, half of the trials failed to achieve significant results. The conflicting results may be attributable to the variety of training components used in these interventions. While using a wide spectrum of different types of exercises (e.g. balance, stretching and strength), it remains unclear which are the most beneficial components of a preventive intervention. Independent trials had some major limitations, such as lack of compliance in the study by Steffen et al. [87] and small number of events in the study by Waldén et al. [88], which could have influenced the pooled result. Most of the warm-up programmes were conducted among soccer players and included exercises intended to enhance neuromuscular control.

When dividing training interventions into subgroups, some overlap in the contents of the training programs used in each group was inevitable. Therefore a neuromuscular training method was included in many training interventions. It has been hypothesised that neuromuscular training can induce such changes within the neuromuscular system that may affect the risk of injuries [20, 89]. Neuromuscular training is thought to have beneficial effects on sense of joint position, stability, and reflexes [16, 20]. Also, according to recent review [20], neuromuscular training can be implemented effectively with no additional equipment, thereby offering a practical, cost-effective way to reduce the risk of injury.

Eccentric strength training, according to two high quality trials, seemed to reduce the risk of hamstring injuries among soccer players [47, 48]. This finding is consistent with the other systematic review [10]. However, strength training was not effective among Australian football players [33], and among football players practicing strengthening the evertor muscles to reduce the recurrence of ankle sprains [27]. Interventions intended to increase strength and power have not yet been extensively studied in RCTs. Instead, strength and power training has successfully been used as a part of multicomponent interventions. Approximately half of the effective training multicomponent interventions included strength or power training components, and almost all of them combined elements from both strength and power, and balance and coordination training (Electronic Supplementary Material Table S3).

It is likely that the preventive effect of the training programmes using several components is the sum of individual effective methods. It is almost impossible to identify which part of the training intervention is the actual effective component and which part has no influence on the risk of injury [22]. The preventive effect may also be a result of the interaction of the different components, when two single elements may not be effective in isolation from each other, but have a desirable combined effect. Nevertheless, effective training programs require carefully planned injury specific exercises and the programmes need to be adjusted to the injury problem within the target population at hand. In addition, one major issue concerning the effectiveness of any intervention is compliance to the intervention. Knowing that high compliance can further reduce injury risk [97], the challenge is how to motivate athletes and their coaches to follow injury prevention program.

Stretching has not generally been studied in RCTs. In this review four RCTs studied the effectiveness of stretching and only three of these were eligible for inclusion in the meta-analysis. This limited number of trials is probably due to the existence of evidence from earlier studies implying that stretching has no effect on overall injury risk [12, 98, 99]. The present finding does nothing to change to this. However, some reviews have stated that there is preliminary evidence that stretching may reduce the risk of musculotendinous strains [12, 100], and ligament sprains [12], but this needs further investigation. Evidently, the lack of well-conducted controlled trials is a problem, and therefore the definitive role of stretching cannot be confirmed [99, 100].

Some of other methods, such as protective head equipment, modified shoes and injury prevention videos, have been studied even less in RCTs, and therefore drawing firm conclusions on their effectiveness is not reasonable. In addition, there were other limitations, such as poor compliance and co-interventions, affecting the results of individual trials.

4.3 Limitations and Future Studies

This study had some limitations. The quality assessment of the trials revealed various methodological weaknesses in the trials. The quality score varied between two and nine points out of 12 the average being five points. According to the recommendations of Furlan et al. [23], studies should be rated as having a low risk of bias when at least six of the 12 criteria have been met and the study has no serious flaws. Studies should be rated as having a high risk of bias if the score given is less than six or if the study has serious flaws [23]. Based on these recommendations, in this review, only 21 studies were rated as having a low risk of bias, whereas 47 studies were considered as having a high risk of bias. Under these circumstances, issues of the internal validity of the studies ought to be acknowledged. There is a possibility of selection bias because of inadequate concealment of allocation or baseline differences between groups in 64 trials (including all three quasirandomised trials); performance bias due to the lack of blinding of participants or care providers, the effect of co-interventions and inadequate compliance in all 68 trials; detection bias due to the lack of blinding of outcome assessors or different timing of the outcome assessment in 48 trials; and attrition bias due to lack of intention-to-treat analysis or unacceptable drop-out rate in 44 trials. Given the nature of interventions to prevent sports injuries, the blinding of the patients, care providers, and outcome assessors simultaneously, and avoiding co-interventions is in most cases extremely problematic. Therefore it is almost impossible for a study to attain the highest quality assessment score [22]. Of the 21 studies considered that have a low risk of bias, all except one were conducted in the 2000s, suggesting that studies are nowadays conducted and reported more correctly than before.

Furthermore, there were limitations in external validity. In general, most of the participants in the studies were young adult athletes, both male and female. Only a few of the studies included senior athletes or children. The studies involving the use of insoles in particular used exclusively military recruits, most of them male and aged under 30. Interventions involving strength training were also all conducted among male football players and with rather small sample sizes. Consequently the applicability of these findings to other age and sports groups is limited and should be tested in future studies.

The trials were divided into subgroups according to the similarity of the preventive measures. This was done even though the interventions were not identical in all respects, and the study designs and participants were decidedly heterogeneous.

Our goal was to gather information as reliable as possible by using data from randomised studies only. Inclusion of controlled trials or cohort studies might have weakened the quality of evidence, although in some reviews (e.g. reviews assessing effectiveness of neuromuscular training [16, 17, 20]) inclusion of non-randomised trials has still given results similar to ours.

Future studies should focus on time and resource-efficient preventive methods. The mechanisms behind effective methods and the most beneficial elements of preventive interventions need to be ascertained. Sports specific preventive training programmes should be developed and tested properly, not only in the prevention of acute but also of stress injuries, because specified training intended to increase neuromuscular control and sports specific skills probably has multiple advantages. Finally, wide-scale implementation studies are needed to find out how interventions proven to be effective in smaller controlled intervention studies work in real life.

5 Conclusions

This meta-analysis takes into account all randomised controlled intervention trials intended to prevent sports injuries. It provides information on what has been studied and what needs to be studied more, and also on which methods are truly preventive versus those which have no preventive effect in light of the evidence from RCTs.

The significance of the current findings is that sports injuries can be prevented at least to a certain extent by using injury and sports specific methods, and by taking such preventive actions in practice may yield major benefits. Because sports injuries are detrimental to an athlete’s career and health, and incur major costs for society, it is essential to promote evidence-based preventive methods.