Original article

Intrinsic functional decits associated with increased risk of ankle injuries: a systematic review with meta-analysis
Jeremy Witchalls,1,2 Peter Blanch,1 Gordon Waddington,2 Roger Adams3
of Physical Therapies, Australian Institute of Sport, Canberra, Australia 2Department of Health, University of Canberra, Canberra, Australia 3Department of Health Sciences, University of Sydney, Sydney, Australia Correspondence to Jeremy Witchalls, Department of Physical Therapies, Australian Institute of Sport, Leverier Crescent, Australian Capital Territory 2617, Canberra, Australia; jeremy.witchalls@ausport. gov.au Received 22 April 2011 Accepted 20 October 2011 Published Online First 14 December 2011
1Department

ABSTRACT Background A history of ankle injury is known to be associated with an increased risk of future injuries. Prevention of a rst-time injury to an ankle will also prevent subsequent re-injury; yet these participants are often overlooked in reports of preventive testing. Determining the functional decits which promote injury risk in all ankles, through studies inclusive of previously injured and never injured ankles, will enable training to be directed at improving known decits in all sports participants. Objective To review studies investigating the measurement of intrinsic functions in healthy ankles and assess their predictive value for injury. Method Systematic review and meta-analysis of journal articles from selected electronic databases. Using all papers that included sufcient data for extraction in any paradigm, the authors pooled results for measures of strength, postural control, proprioception, muscle reaction time in response to perturbation, range of movement and ligament stability. Results Thirteen papers were found with adequate data reporting to allow calculation of pooled standardised mean difference (SMD) or pooled RR. The following are all associated with an increased risk of ankle injury: higher postural sway (SMD=0.693, 95% CI=0.151 to 1.235, p=0.012), being in the lower postural stability group (RR=2.06, 95% CI=1.364 to 3.111, p=0.001), lower inversion proprioception (0.573, 0.244 to 0.902, <0.001), higher concentric plantar exion strength at faster speeds (0.372, 0.092 to 0.652, 0.009) and lower eccentric eversion strength at slower speeds (0.337, 0.117 to 0.557, 0.003). Conclusion There is a set of intrinsic functional and structural ankle decits associated with signicantly increased risk of ankle injury. These ndings will enable clinicians and sports trainers to measure and train specic decits in sports people for the prevention of ankle injury.

has recently been shown to be associated with a mixture of mechanical instability, self-reported instability and a history of recurrent sprains.6 Identication of intrinsic ankle performance characteristics that can be improved to reduce the risk of rst-time ankle injury and subsequent re-injury has clinical value, and the potential to raise sports participation rates, prolong sporting careers and improve quality of life.4

Sensorimotor terminology
The terminology and theory underlying the functional control of the ankle has been summarised elsewhere. 2 7 Control is provided by an interaction between active restraint from the muscles and passive stability from the non-contractile soft tissues and bony geometry. Active stability of the ankle joint and control of movement is achieved through the interaction of sensory reception, central nervous system interpretation and movement planning, and peripheral motor output.8 This system of sensory and motor (sensorimotor) control is summarised in gure 1. The quality of sensorimotor control can be measured at each stage of the reception/processing/ motor output process or to assess as an entirety (gure 2).913

Reviews of intrinsic decits associated with ankle injury risk

The risk of re-injuring an ankle is greater than the risk of rst-time injury. When viewed in isolation, the prevention of injury to previously healthy ankles shows lower clinical power14 and cost-effectiveness.15 16 However, this preventive value is much greater when it is remembered that every rst-time injury prevented is also removing a predisposition to re-injury. In this context, the prevention of rst-time injury assumes much greater importance despite its primary benet being relatively smaller, because the risk of rsttime and secondary re-injury is compounded. Hence, in terms of risk: Risk of future ankle injury=risk of rst-time injuryrisk of re-injury. An individuals history of ankle injury is not something which a clinician or a trainer can change. All sports participants, whether previously injured or not, will benet from the prevention of ankle injuries. It is more useful to determine the functional traits which promote injury risk, to enable training to be directed at improving known decits, whether in a previously injured or in a never-injured ankle.
515

INTRODUCTION
In court and eld sports, the ankle is one of the most frequently injured body regions.1 Injury to an ankle puts an individual at risk of chronic ankle dysfunction. 2 In a general clinical population, 73 % of patients reported ongoing symptoms 18 months after an ankle sprain, 3 and of these 40 % have moderate to severe symptoms. Similarly, 74 % of patients with ankle injury attending a sports clinic reported ongoing symptoms.4 Chronic ankle symptoms of pain, recurrent sprains and subjective feelings of instability persist beyond 3 years after injury. 5 Functional ankle instability (FAI)
Br J Sports Med 2012;46:515523. doi:10.1136/bjsports-2011-090137

Original article

Figure 1 The sensorimotor process in perceiving activity and producing joint stability.

Figure 2 Measuring sensorimotor function.

Prediction and prevention of ankle injuries interests sports trainers and musculoskeletal health professionals and a number of authors have reviewed the literature in this area of study.15 1720 The relationship between previous injury and ongoing injury risk is clear and well proven. 2124 However, to date, no meta-analysis has been done to investigate the association between pre-existing decits in ankle structure and the function and risk of future ankle injury in all ankles, whether previously injured or never injured. The aim of this paper was to identify studies that have found pre-existing functional and structural decits associated with an increased risk of injury in previously healthy ankles and to combine their results through meta-analysis. Identication of decits that can be improved should enable further development of training and treatment interventions to reduce the risk of ankle injury.

SPORTDiscus (via EBSCO); MEDLINE (via EBSCO); as subscribed by the Australian Institute of Sport, National Sport Information Centre (AIS, NSIC). CINAHLPlus (via EBSCO); PEDro; ScienceDirect; Cochrane Library; as subscribed by the University of Canberra (UC). The search term Ankle AND Injury AND Prospective AND (Risk OR Prediction OR Incidence OR Prevention) was entered for each database, and the resulting references lists combined. The search was not historically date limited and nished in April 2011. The sequence for selection of papers suitable for metaanalysis is shown in gure 3.

Inclusion and exclusion criteria

We selected papers reporting studies that were prospective in design, and investigated ankle injury occurrence in association with measurements of biometric traits and functional abilities of the participants. The studies had to state the method used to de ne the participants as healthy, prior to baseline measurements. To facilitate data extraction, we selected Englishlanguage papers that presented injured and un-injured group
Br J Sports Med 2012;46:515523. doi:10.1136/bjsports-2011-090137

METHOD Literature search

In order to identify suitable literature, we searched the following databases:
516

Original article
scores, SD (or an alternative from which SD can be calculated) and the number of participants in injured/un-injured groups. For categorical variables, the number of participants in each category and outcome had to be available, rather than a single OR or RR. Papers that did not meet these inclusion criteria were excluded from meta-analysis, but considered in our discussion of the body of research in this area. Papers were also excluded if the study was of a population with medically compromised ankle function or with general mobility. the same score as a no answer, since this implied poor quality of design or reporting. If a question was not relevant to a particular paper, it was excluded from the percentage calculation.

Data analysis
Each paper was searched for raw scores and injury incidences and performed a meta-analysis where two or more papers presented data on the same type of decit. StatsDirect statistical software (http://www.statsdirect.com StatsDirect Ltd. England. Version 2.7.8, released 15/03/10) random effects model was used 28 to construct Forest plots of multiple study outcomes on the same feature of ankle function, to calculate pooled mean effect sizes (d+) of standardised mean differences (SMD) or pooled RR for all group comparisons and the 95% CI. 29 If a study provided sufcient information for SMD and RR, it was included in each meta-analysis. The likelihood that the pooled effect size or RR differed from zero was calculated, 30 with a p value of 0.0531 using the test. Effect size thresholds were de ned as negligible (<0.2), small (0.20.6), moderate (0.61.2) and large (>1.2). 32 Funnel plots were created when results from four or more studies were pooled, to assess the inuence of publication bias in results where the p value was 0.05. 33 To evaluate the inuence of study quality on effect size outcomes, the mean quality score of the papers was calculated for each of the pooled groups and correlated with the pooled effect size produced by each meta-analysis. The difference in the mean QI score between the meta-analyses which produced a p value of 0.05 and those with a p value > 0.05 was also determined.

Quality assessment
To assess the quality of design and reporting of the papers included for meta-analysis, we used the quality index (QI). 25 An adapted version of this index 26 has been used in a review of studies of ankle sensorimotor deficits in participants with FAI. 27 The earlier version was used, 26 because our review involved studies that do not need to blind researchers for a prior injury outcome and that only measure prospectively. The modied QI contains 15 questions. We calculated a percentage score, giving a point for each question that was answered yes. If the response was unclear, it was given

RESULTS Papers
An initial 2397 papers matched the electronic database search parameters. After the systematic review, 13 papers remained and these provided results which were able to be used in the meta-analysis (gure 3). The participant numbers, injury numbers and sports involved are detailed in table 1.

Figure 3 The search process and papers produced at each stage of the literature search.

Table 1
Paper

Details of participants and quality ratings of studies suitable for meta-analysis

Participants (number and gender (total)) 72 F, 73 M (145) 68 F, 50 M (118) 508 M 94 F, 21 M (115) 210 M 91 F, 119 M (210) 68 F, 101 M (169) 80 F, 130 M (210) 127 M 42 M 80 M 152 M 76 F 2162 (total) 549 F, 1613 M Injuries (number and gender (total)) 7 F, 8 M (15) 13 F, 7 M (20) 43 33 (total) 21 23 (total) 20 (total) 12 F, 16 M (28) 23 18 52 44 32 372 (total) Sports and organisations Collegiate football, hockey and lacrosse Collegiate football, hockey and lacrosse (separate M/F cohorts) Amateur football Adolescent dancers Professional Australian football High school basketball University American football (M only), M and F basketball, F gymnastics, M and F football One high school, 3 colleges. F volleyball, M and F football, M American football Division 4 football team High school basketball Gaelic football, hurling Physical education, university students Physical education, university students Quality index (%) 14 13 14 12 12 13 10 15 12 15 11 15 15 Mean 13.15 (SD=1.68)

Baumhauer et al34 Beynnon et al35 Engebretsen et al36 Hiller et al37 Hrysomallis et al38 McGuine et al39 McHugh et al40 Trojian et al41 Tropp et al42 Wang et al43 Watson44 Willems et al45 Willems et al46 Combined total

M/F, male/female; football=soccer.

Br J Sports Med 2012;46:515523. doi:10.1136/bjsports-2011-090137

517

Original article
Table 2 Meta-analyses results for strength at slower speeds
Pooled SMD 0.055 0.119 0.103 0.224 0.062 0.337 0.187 +0.020 0.101 0.353 0.104 0.188 95% CI 0.18 to 0.289 0.322 to 0.084 0.356 to 0.15 0.623 to 0.175 0.425 to 0.241 0.557 to 0.117 0.669 to 0.296 0.462 to 0.503 0.688 to 0.487 0.708 to 0.003 0.462 to 0.254 0.671 to 0.294 Z value 0.457 1.151 0.798 1.101 0.539 3.003 0.757 0.083 0.335 1.945 0.568 0.765 p 0.648 0.250 0.425 0.271 0.590 0.003* 0.449 0.934 0.737 0.052 0.570 0.444 Mean QI% score 14.167 14.167 14 14 14.2 14.2 13 13 13.75 13.67 13.33 13

Muscle test (<110 /s) Concentric eversion Concentric inversion Concentric PF Concentric DF Eccentric Ev Eccentric Inv Eccentric PF Eccentric DF Concentric Ev/Inv ratio Eccentric Ev/Inv ratio Concentric DF/PF ratio Eccentric DF/PF ratio

DF, dorsiexion; Ev, eversion; Inv, inversion; PF, plantar exion; QI, quality index; SMD, standardised mean difference. * Reaches statistical signicance at p<0.05.

Table 3

Meta-analysis results for strength at faster speeds

Pooled SMD +0.108 0.215 +0.372 +0.149 +0.027 0.148 95% CI 0.145 to 0.362 0.468 to 0.039 0.092 to 0.652 0.129 to 0.428 0.401 to 0.347 0.438 to 0.141 Z value 0.84 1.659 2.605 1.052 0.143 1.003 p 0.401 0.097 0.009* 0.293 0.886 0.316 Mean QI% score 15 15 15 15 15 15

Muscle test (>110 /s) Concentric eversion Concentric inversion Concentric plantarexion Concentric dorsiexion Eccentric eversion Eccentric inversion

QI, quality index; SMD, standardised mean difference. * Reaches statistical signicance at p<0.05.

Figure 4 Eccentric inversion strength (slow).

Figure 5 Concentric plantar exion strength (fast). belief that slower speeds test muscular strength, while faster speeds test power. 30 Based on this theoretical premise, study results were pooled into slow (<110 /s) and fast (110 /s) testing30 and then pooled effect sizes for the SMD by movement direction (tables 2 and 3). A negative effect size indicates that the injured group was weaker in that test. In comparing strength ratios, a negative sign indicates that the injured group had a lower ratio than the un-injured, showing a relative weakness in the rst direction of the parameter (eversion and dorsiexion, DF). Forest plots are shown for the comparisons with a p value 0.05 (gures 4 and 5) and a funnel plot for the slow eccentric inversion strength meta-analysis that included more than four papers (gure 6).
Br J Sports Med 2012;46:515523. doi:10.1136/bjsports-2011-090137

Features of ankle function Strength

Five papers presented the results of the tests of ankle strength between injured and un-injured groups. 34 35 43 45 46 However, Beynnon et al 35 reported male and female groups separately, so these were entered separately into the meta-analyses. The methodology of testing was similar in all studies. A pooled effect size was calculated for each strength direction and each agonistantagonist ratio reported for two or more participant groups. Some studies also tested their participants at a range of different isokinetic speeds: 30 /s, 34 35 45 46 60 /s,43 120 /s 45 46 and 180 /s.43 It has been suggested that researchers have historically tested the strength at a mixture of speeds, from a
518

Original article
Postural stability
Eleven papers presented results from various tests of balance, in 12 cohorts of participants. Four studies used similar instrumented measures of postural sway (in ve cohorts). 35 39 43 45 Five gave sufcient information to perform the meta-analysis of RR for their selected categorical designation of poor balance between their cohorts. 37 38 41 42 44 Three studies used different scoring systems for counting balance errors, allowing some subjectivity on the part of the scorer. 36 40 45 We pooled the SMD and RRs for each paradigm, with the results as shown in table 4 and gures 710. A positive effect size in error counting or the amount of sway indicates a greater error rate and the range of sway instability in the injured cohorts. The RR is the risk of injury relative to the un-injured control cohort. size indicates that the injured group had a greater mean error in replicating the test angle.

Muscle reaction time

Three studies measured muscle reaction times to a perturbation or to a movement and related these to future injury occurrence, 35 45 46 but one paper did not report these results.46 So two papers, providing three cohorts, were entered into a meta-analysis for each of the muscle groups tested. Both papers included an inversion component in their perturbation movement. In table 6, a negative effect size indicates that the injured group had a shorter muscle reaction time.

Range of movement
Six studies measured range of movement (ROM) in various planes of movement 3436 43 45 46 in seven cohorts, with ve testing inversion and eversion ROM, 3436 45 46 a different grouping of ve testing DF with two associated knee positions34 35 43 45 46 and two studies including plantar exion (PF).45 46 When meta-analysis was rst conducted for inversion and eversion ROM, the results of one paper36 were more than two SD outside the effect sizes for the other four studies (ve cohorts) included in the meta-analysis, indicating a heterogeneous bias in this studys ndings relative to the other four. Subsequently, this paper was excluded, resulting in a reduced pooled effect size for both analyses, but no change in the statistical signicance of the outcome (table 7).

Proprioception (JPS)
Our search found three papers that reported measures of joint position sense (JPS) in a prospective study of ankle injury risk,4547 but none using the proprioceptive paradigms of kinaesthesia or force perception. However, only one paper presented full data on the different JPS tests used, including scores with a p value of >0.05.46 This study reported the results of multiple test formats separately: active and passive position replication tests, with two inversion positions and an eversion position for each paradigm. The results of the two passive and the two active inversion JPS tests could be pooled separately, to provide an overall effect size for each testing paradigm. Since the pattern of results was similar for absolute and exact methods of scoring the same test, we analysed the absolute results. The passive tests show a statistically signicant difference between injured and un-injured groups (gure 11), while the active tests do not reach signicance (table 5). A positive effect

Ligament stability
Three studies tested the ligament integrity, giving four cohorts. 3537 Two different tests of ligament stability were used: (A) an anterior draw mechanism, intended to stress the anterior talo-bular ligament (ATFL) and (B) the talar tilt into inversion, believed to stress the calcaneo-bular ligament and the ATFL. These papers presented results from four cohorts

Figure 6 Funnel plot of eccentric inversion papers.

Figure 7 Instrumented postural sway.

Table 4
Method

Meta-analysis results for the measures of postural stability

Pooled results +0.091 (SMD) +0.693 (SMD) 2.06 (RR) 95% CI 0.284 to 0.466 0.151 to 1.235 1.364 to 3.111 Z value 0.475 2.505 2 11.808 (df=1) p 0.635 0.012* 0.001* Mean QI% score 13 13.8 12.4

Count errors Instrumented sway Risk ratio

df, Degrees of freedom; QI, quality index; SMD, standardised mean difference. * Reaches statistical signicance at p<0.05.

Br J Sports Med 2012;46:515523. doi:10.1136/bjsports-2011-090137

519

Original article

Figure 8 Funnel plot of instrumented sway papers.

Figure 10 Funnel plot of postural stability risk ratio.

Figure 9 Postural stability risk ratio.

Figure 11 Passive inversion JPS (deg=degrees, max=maximum, inv=inversion).

Table 5

Inversion JPS at multiple joint angles

Pooled SMD +0.573 0.203 95% CI 0.244 to 0.902 0.526 to 0.12 Z value 3.417 1.233 p <0.001* 0.218 Mean QI% score 15 15

Inversion JPS Passive Active

JPS, joint position sense; QI, quality index; SMD, standardised mean difference. * reaches statistical signicance at p<0.05.

Table 6

Meta-analysis results for muscle reaction times

Pooled SMD 0.346 0.252 0.278 95% CI 0.699 to 0.008 0.679 to 0.175 1.179 to 0.623 Z value 1.917 1.156 0.605 p 0.055 0.248 0.545 Mean QI% score 13.67 13.67 13.67

Muscle reaction time Peroneus brevis Peroneus longus Tibialis anterior

QI, quality index; SMD, standardised mean difference.

tested by the anterior draw, 3537 and two for the talar tilt 35 (table 8).

The inuence of study quality on effect size of results

The effect sizes of all pooled comparisons shown in tables 28 were correlated against the mean QI percentage score for their respective groups, with the scores for ligament testing and RR associated with postural stability excluded.
520

Mean QI score and pooled effect size for each meta-analysis were not signicantly correlated (r=0.001, p=0.497 onetailed, n=31). An independent-samples t test on the QI scores between those meta-analyses with p value greater or less than 0.05 (mean QI score 13.987 and 14.08 for p values >0.05 and 0.05, respectively) showed these not to be signicantly different (p=0.402 one-tailed, n=34).
Br J Sports Med 2012;46:515523. doi:10.1136/bjsports-2011-090137

Original article
Table 7 Results of meta-analysis of ankle range of movement
Pooled SMD +1.040 +0.173 +0.032 0.064 <+0.001 95% CI 1.049 to 3.13 0.199 to 0.545 0.177 to 0.242 0.4 to 0.272 0.277 to 0.279 Z value 0.976 0.91 0.30 0.374 0.006 p 0.329 0.363 0.764 0.708 0.995 Mean QI% score 14 14 14.167 14.167 15

Ankle movement Inversion Eversion Dorsiexion (with knee exion) Dorsiexion (with knee extension) Plantarexion

QI, quality index; SMD, standardised mean difference.

Table 8

Meta-analysis results for ligament testing

Pooled RR +3.144 +1.138 95% CI 0.755 to 13.083 0.564 to 2.299 2 value (df=1) 2.478 0.131 p 0.115 0.717 Mean QI % score 13 13

Ligament test Talar tilt Anterior draw

df, Degrees of freedom; QI, quality index.

DISCUSSION
This review has enabled the meta-analysis of the features of ankle function that have been examined in research on the risk of ankle injury.

the studies found by our search included these ratios at the higher speeds.

Postural stability
Reduced postural control has been shown repeatedly to indicate heightened injury risk in healthy ankles, whether tested by functional-specic tasks, 37 single leg balance,41 more complex balance postures45 or postural sway. 39 42 43 One study of balance on demi-pointe in the adolescent dancers revealed a correlation in the rst test cohort, which did not then replicate in a second test group of ankles. 37 Meta-analysis revealed that participants who subsequently injured their ankle had a larger mean area of sway when tested using an instrumented measurement. They also have a higher risk ratio if they were determined to be in the lower category of postural control when scored by observation or by sway. However, methods that test postural control by scoring the number of errors during a test have not shown this pattern, suggesting that the increased subjectivity of these methods may increase the variability in scores and render them less useful for this particular purpose.

Strength
Weaker eccentric inversion strength is associated with an increased occurrence of ankle injury at testing speeds slower than 110 /s. The effect size (0.337) is small, suggesting that the injured participants were only slightly weaker than those who remained uninjured. It is important to note that although the effect size was small, it was still able to differentiate between those that subsequently were injured and those that were not. Although none of the other strength comparisons at slow speeds reached signicance, the effect size for the eversion/ inversion strength ratio is actually greater than that for the eccentric inversion (0.353), suggesting that the relative weakness of the evertor muscles (peroneii) is also worthy of further study. It might be expected that the evertors, which assist in reducing or controlling inversion movements, would show a decit in those who injure more readily. This is borne out in meta-analysis of individuals suffering FAI or recurrent sprains postinjury.30 However, the nding here regarding inversion strength is consistent with another meta-analysis of ankle strength in FAI.48 In this type of postinjury dysfunction, the inversion strength was found to be signicantly less than in healthy control groups, and as a consistent nding in injury prone participants before injury and postinjury. Although Arnold et al 30 found no meaningful difference between fast and slow strength testing speeds in their metaanalysis investigating decits in FAI, our meta-analyses was prospective in nature and produced different results. At speeds faster than 110 /s, the injured participants were stronger in concentric PF than those who remained un-injured. The effect size (0.372) is again small. However, greater PF strength may not be a direct cause of heightened injury risk; rather greater strength may be associated with other variables which increase the ankle injury risk, for example, higher grade of sports participation or higher intensity of workload during their chosen sport. Studies which evaluate strength in association with other covariants are required to address this question. Higher PF strength may also alter the DF/PF ratio, which may be more meaningful for injury risk. However, none of
Br J Sports Med 2012;46:515523. doi:10.1136/bjsports-2011-090137

JPS
The two different joint positions used to test inversion JPS produced similar ndings.46 These indicate that relatively poor active position-replication JPS performance in inversion is associated with ankle injury risk, in the female population studied. The effect size for these two tests combined is 0.573, with the group which subsequently injured their ankles having a greater error in angle when attempting to replicate the test position. Eversion JPS was tested at 10 45 46 and 20 eversion.47 As only one paper presented the scores in sufcient detail,46 metaanalysis was not possible. Both studies at 10 found no difference between injured/un-injured groups, while at 20 right ankle inversion correlated signicantly with right ankle sprain, in women only. Three of four ankle groups tested in this paper (men/women, right/left ankle) did not show a signicant difference between injured/un-injured groups. These results generally do not suggest that reduced eversion JPS is associated with increased risk of ankle injury. Although a number of papers have found that there is evidence of a decit in joint position sense associated with CAI,19 because of the retrospective nature of this study it is
521

Original article
possible that this decit may not exist prior to injury in healthy ankles. The method of testing ankle JPS in two of these three studies involved supine lying, with the knee at 90 exion and the limb non-weight-bearing.45 46 The other study used a sitting position with knee-exion for the test.47 Since these studies were conducted, it has been found that higher percentages of weight-bearing are positively correlated with ankle JPS acuity, and this testing position may have reduced the sensitivity of the test.49 Gross limb positioning also has a signicant effect on ankle JPS, giving much greater acuity when the knee is extended, whether in sitting or in standing. 50 Further, psychophysical testing requires a large number of trials for each movement direction, as thresholds vary with time and repetition. 51 This aspect was missing in the reported studies, with each subject tested only twice45 46 or three times47 for each of the joint positions. An additional factor of the JPS test used in all three studies is that they tested isolated ankle joint mechanisms in relatively mid-range joint positions. The role of joint capsule receptors and their afferent neural pathways has been described as being particularly important in determining the extreme limits of movements. 52 This suggests testing for injury risk, using midrange positioning may exclude some of the receptors which are most active at the time when the joint is at riskclose to the limit of safe joint range. In assessing functional decits, ethical considerations (necessarily) restrict putting the participants being placed in positions where damage is imminent. However, it is most likely to be at this range where the necessary sensorimotor skills to stabilise real-life critical loads and positions are demonstrated. Some authors consider mid-range testing incompatible with the functional severity of injuring movements, and with the joint excursion present in healthy sporting activities.9 Studies that move closer to the limits of ankle control may reveal sensorimotor decits in healthy sportspeople, and reveal a stronger correlation between ankle function and risk of injury. The lack of studies and the limited publication of detail in results published provide limited information about the utility of JPS in predicting ankle injury. Are current testing methods sufciently specic and sensitive, and is JPS more signicant to injury risk for women as opposed to men? The results from our meta-analysis of two different inversion positions suggest that poor inversion JPS is associated with increased ankle injury risk, and may be useful as part of a preventive screening protocol. Other methods of proprioceptive testing, such as kinaesthesia and force sense, have not been used in the prospective studies of injury prediction. They are also potential avenues for further study in this area.

ROM
Analysis of ROM for inversion, eversion, PF and dorsiexion did not show a link to ankle injury risk. This nding differs from an earlier systematic review,18 but the criteria for metaanalysis in this study was also different from those in the previous review.

Ligament stability
While investigations into ligament stability seem promising, with a high RR for talar tilt tests in particular, these values did not reach signicance. This is most likely due to the variability of testing and scoring methods applied. Furthermore, it should be noted that these studies did not include the injury history associated with ligament laxity. In all likelihood, individuals with increased ligamentous laxity have a history of injury. Future studies in this area must use a standardised testing and scoring methodology and include injury history as a covariate.

The inuence of study quality on effect size of results

Study quality measures did not show any inuence on effect size, with the mean QI score of studies within each meta-analysis not related to the size of the d+ score. Likewise, the meta-analyses whose p value was 0.05 did not differ signi cantly in their mean QI score, from those with a p value above 0.05. The quality of study design and reporting did not inuence the outcomes of these metaanalyses. However, since low-quality studies are to some extent excluded from the meta-analysis by their lack of methodological detail and incomplete data reporting, it is likely that the systematic review process has controlled the standard of quality in the selected papers.

CONCLUSIONS
Meta-analyses demonstrate that higher postural sway, being in the lower postural stability group, lower eccentric eversion strength at slower speeds, higher concentric PF strength at faster speeds and lower inversion JPS in women are all associated with the risk of ankle injury. These ndings will enable clinicians and sports trainers to measure and train specic decits in sports people for the prevention of ankle injury. Since the risk of ankle injury is dependent on multiple variables, 56 57 a single clinical predictive algorithm58 is unlikely to be established for the risk of ankle injuries. Determination of appropriate threshold scores for high-risk versus low-risk levels for injury in each variable may allow screening to target whichever decits are present in an individual. It would also be useful to assess the mechanism by which intervention to modify these risks may work, to enable more targeted, efcient and effective intervention to be applied. Further study is also required to further de ne elements that best predict the risk of ankle injury for specic sports, playing positions and participant subgroups.
Acknowledgements Jeremy Witchalls was supported by an APA scholarship administered by the University of Canberra and the Australian Institute of Sport during the writing of this review. Competing interests None. Provenance and peer review Not commissioned; externally peer reviewed. Br J Sports Med 2012;46:515523. doi:10.1136/bjsports-2011-090137

Muscle reaction time

Meta-analysis did not show any association between muscle reaction times and future injury occurrence. These negative ndings are in keeping with the ndings relating to individuals with FAI, and suggest that the muscle reaction time following an inversion perturbation is too slow to prevent the movement occurring, and that the protective role of reactive muscle contraction is more a saving response for balance, than to contract to prevent the injuring movement itself. 53 54 It is probable that, similar to those individuals with FAI, muscle function in preparation for ankle loading may be more important to injury prevention than reactive muscle reaction times. 55
522

Original article
REFERENCES
1. Hootman JM, Dick R, Agel J. Epidemiology of collegiate injuries for 15 sports: summary and recommendations for injury prevention initiatives. J Athl Train 2007;42:31119. Hertel J. Sensorimotor decits with ankle sprains and chronic ankle instability. Clin Sports Med 2008;27:35370. Braun BL . Effects of ankle sprain in a general clinic population 6 to 18 months after medical evaluation. Arch Fam Med 1999;8:1438. Anandacoomarasamy A , Barnsley L. Long term outcomes of inversion ankle injuries. Br J Sports Med 2005;39:14. van Rijn RM, van Os AG, Bernsen RM, et al. What is the clinical course of acute ankle sprains? A systematic literature review. Am J Med 2008;121:324331. Hiller CE, Kilbreath SL, Refshauge KM. Chronic ankle instability: evolution of the model. J Athl Train 2011;46:13341. Riemann BL , Lephart SM. The sensorimotor system, part I: the physiologic basis of functional joint stability. J Athl Train 2002;37:719. Lephart SM. Preface. In: Lephart S, Fu F, eds. Proprioception and Neuromuscular Control in Joint Stability. Champaign, Human Kinetics 2000:xvxvi. Ashton-Miller JA . Proprioceptive thresholds at the ankle: implications for prevention of ligament injury. In: Lephart SM, Fu FH, eds. Proprioception and Neuromuscular Control in Joint Stability. Champaign, Human Kinetics 2000:27989. Bochdansky T, Kollmitzer J, Ebinbechler G. The role of electromyography in the assessment of neuromuscular control. In: Lephart SM, Fu FH, eds. Proprioception and Neuromuscular Control in Joint Stability. Champaign, Human Kinetics 2000:14560. Konradsen L , Beynnon BD, Renstrom PA. Techniques for measuring sensorimotor control of the ankle: evaluation of different methods. In: Lephart SM, Fu FH, eds. Proprioception and Neuromuscular Control in Joint Stability. Champaign, Human Kinetics 2000:13944. Waddington G, Adams R. Discrimination of active plantarexion and inversion movements after ankle injury. Aust J Physiother 1999;45:713. Perrin DH, Shultz SJ. Models for clinical research involving proprioception and neuromuscular control. In: Lephart SM, Fu FH, eds. Proprioception and Neuromuscular Control in Joint Stability. Champaign, Human Kinetics 2000:34974. Cumps E, Verhagen E, Romain M. Efcacy of a sports specic balance training programme on the incidence of ankle sprains in basketball. J Sports Sci Med 2007;6:21219. Verhagen EA , Bay K. Optimising ankle sprain prevention: a critical review and practical appraisal of the literature. Br J Sports Med 2010;44:10828. Verhagen EA , van Tulder M, van der Beek AJ, et al. An economic evaluation of a proprioceptive balance board training programme for the prevention of ankle sprains in volleyball. Br J Sports Med 2005;39:11115. Beynnon BD, Murphy DF, Alosa DM. Predictive Factors for Lateral Ankle Sprains: A Literature Review. J Athl Train 2002;37:37680. de Noronha M, Refshauge KM, Herbert RD, et al. Do voluntary strength, proprioception, range of motion, or postural sway predict occurrence of lateral ankle sprain?. Br J Sports Med 2006;40:8248. Holmes A , Delahunt E. Treatment of common decits associated with chronic ankle instability. Sports Med 2009;39:20724. Verhagen EA , van Mechelen W, de Vente W. The effect of preventive measures on the incidence of ankle sprains. Clin J Sport Med 2000;10:2916. Arnason A , Sigurdsson SB, Gudmundsson A, et al. Risk factors for injuries in football. Am J Sports Med 2004;32:516. Bahr R , Bahr IA. Incidence of acute volleyball injuries: a prospective cohort study of injury mechanisms and risk factors. Scand J Med Sci Sports 1997;7:16671. McKay GD, Goldie PA, Payne WR, et al. Ankle injuries in basketball: injury rate and risk factors. Br J Sports Med 2001;35:1038. Murphy DF, Connolly DA, Beynnon BD. Risk factors for lower extremity injury: a review of the literature. Br J Sports Med 2003;37:1329. Downs SH , Black N. The feasibility of creating a checklist for the assessment of the methodological quality both of randomised and non-randomised studies of health care interventions. J Epidemiol Community Health 1998;52 :37784. Irving DB , Cook JL, Menz HB. Factors associated with chronic plantar heel pain: a systematic review. J Sci Med Sport 2006; 9 :1122; discussion 234. Munn J, Sullivan SJ, Schneiders AG. Evidence of sensorimotor decits in functional ankle instability: a systematic review with meta-analysis. J Sci Med Sport 2010;13:212. Hopkins WG, Marshall SW, Batterham AM, et al. Progressive statistics for studies in sports medicine and exercise science. Med Sci Sports Exerc 2009;41:313. 29. Deeks JJ, Higgins JPT, Altman DG. Analysing data and undertaking metaanalyses. In: Higgins JPT, Green SL, eds. Cochrane Handbook for Systematic Reviews of Interventions. John Wiley & Sons, Ltd, Chichester, UK: 2008:24396. Arnold BL , Linens SW, de la Motte SJ, et al. Concentric evertor strength differences and functional ankle instability: a meta-analysis. J Athl Train 2009;44:65362. Hopkins WG. The Condence Interval and Statistical Signicance, 2007 [cited 2011 25 February]. http://www.sportsci.org/resource/stats/index.html. (accessed 15 Mar 2011) Hopkins WG. A Scale of Magnitudes for Effect Statistics, 2006 [updated 7 August 2006; cited 2011 25 February]. http://www.sportsci.org/resource/stats/ index.html. (accessed 15 Mar 2011) Sterne JAC, Egger M, Moher D. Addressing reporting biases. In: Higgins JPT, Green SL, eds. Cochrane Handbook for Systematic Reviews of Interventions. John Wiley & Sons, Ltd 2008:297333. Baumhauer JF, Alosa DM, Renstrm AF, et al. A prospective study of ankle injury risk factors. Am J Sports Med 1995;23:56470. Beynnon BD, Renstrm PA, Alosa DM, et al. Ankle ligament injury risk factors: a prospective study of college athletes. J Orthop Res 2001;19:21320. Engebretsen AH, Myklebust G, Holme I, et al. Intrinsic risk factors for acute ankle injuries among male soccer players: a prospective cohort study. Scand J Med Sci Sports 2010;20:40310. Hiller CE, Refshauge KM, Herbert RD, et al. Intrinsic predictors of lateral ankle sprain in adolescent dancers: a prospective cohort study. Clin J Sport Med 2008;18:448. Hrysomallis C, McLaughlin P, Goodman C. Balance and injury in elite Australian footballers. Int J Sports Med 2007;28:8447. McGuine TA , Greene JJ, Best T, et al. Balance as a predictor of ankle injuries in high school basketball players. Clin J Sport Med 2000;10:23944. McHugh MP, Tyler TF, Tetro DT, et al. Risk factors for noncontact ankle sprains in high school athletes: the role of hip strength and balance ability. Am J Sports Med 2006;34:46470. Trojian TH, McKeag DB. Single leg balance test to identify risk of ankle sprains. Br J Sports Med 2006;40:61013; discussion 613. Tropp H, Ekstrand J, Gillquist J. Stabilometry in functional instability of the ankle and its value in predicting injury. Med Sci Sports Exerc 1984;16:646. Wang HK , Chen CH, Shiang TY, et al. Risk-factor analysis of high school basketballplayer ankle injuries: a prospective controlled cohort study evaluating postural sway, ankle strength, and exibility. Arch Phys Med Rehabil 2006;87:8215. Watson AW. Ankle sprains in players of the eld-games Gaelic football and hurling. J Sports Med Phys Fitness 1999;39:6670. Willems TM, Witvrouw E, Delbaere K, et al. Intrinsic risk factors for inversion ankle sprains in male subjects: a prospective study. Am J Sports Med 2005;33:41523. Willems TM, Witvrouw E, Delbaere K, et al. Intrinsic risk factors for inversion ankle sprains in females a prospective study. Scand J Med Sci Sports 2005;15:33645. Payne KA , Berg K, Latin RW. Ankle injuries and ankle strength, exibility, and proprioception in college basketball players. J Athl Train 1997;32:2215. Hiller CE, Nightingale EJ, Christine Lin C-W, et al. Characteristics of people with recurrent ankle sprains: a systematic review with meta-analysis. Br J Sports Med 2011;45:66072. Hsu LJ, Lin SI, Li SW, et al. Effect of weight bearing on ankle joint position sense. J Biomech 2006;39:484. Refshauge KM, Fitzpatrick RC. Perception of movement at the human ankle: effects of leg position. J Physiol (Lond) 1995;488 (Pt 1):2438. Refshauge KM, Kilbreath SL, Raymond J. The effect of recurrent ankle inversion sprain and taping on proprioception at the ankle. Med Sci Sports Exerc 2000;32:1015. Grigg P. Peripheral neural mechanisms in proprioception. J Sport Rehabil 1994;3:217. Af Klint R, Nielsen JB, Sinkjaer T, et al. Sudden drop in ground support produces force-related unload response in human overground walking. J Neurophysiol 2009;101:170512. Duysens J, Beerepoot VP, Veltink PH, et al. Proprioceptive perturbations of stability during gait. Neurophysiol Clin 2008;38:399410. Wikstrom EA , Tillman MD, Chmielewski TL, et al. Measurement and evaluation of dynamic joint stability of the knee and ankle after injury. Sports Med 2006;36:393410. Hertel J. Functional Anatomy, Pathomechanics, and Pathophysiology of Lateral Ankle Instability. J Athl Train 2002;37:36475. Bahr R, Krosshaug T. Understanding injury mechanisms: a key component of preventing injuries in sport. Br J Sports Med 2005;39:3249. Childs JD, Cleland JA. Development and application of clinical prediction rules to improve decision making in physical therapist practice. Phys Ther 2006;86:12231.

30.

2. 3. 4. 5. 6. 7. 8. 9.

31.

32.

33.

34. 35. 36.

10.

37.

38. 39. 40.

11.

12. 13.

41. 42. 43.

14.

15. 16.

44. 45.

17. 18.

46.

47. 48.

19. 20. 21. 22.

49. 50. 51. 52. 53.

23. 24. 25.

54. 55.

26.

56. 57. 58.

27.

28.

Br J Sports Med 2012;46:515523. doi:10.1136/bjsports-2011-090137

523

Copyright of British Journal of Sports Medicine is the property of BMJ Publishing Group and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.