ARNOLD, BRENT L.1; DE LA MOTTE, SARAH2; LINENS, SHELLEY1; ROSS, SCOTT E.1
Functional ankle instability (FAI) is common after ankle sprains and is characterized by feelings of "giving way" at the ankle and recurrent ankle sprains (35). Thirty to forty percent of patients with ankle sprains report recurrent sprains (11) or residual symptoms of instability (31,40). FAI has been shown to prevent approximately 6% of patients from returning to their occupation (40), and due to residual symptoms, 5-15% of patients remain occupationally handicapped from at least 9 months to 6.5 yr, respectively (31,40). Single-leg balance impairments, additionally, have been associated with FAI (1,3,4,6,15-17,20,21,38,39) and have predicted ankle sprain injury in physically active individuals (23,34,37,41). As a result of this association between balance deficits and ankle sprain injury, single-leg balance tests have been used as clinical and research examinations to assess postural instabilities associated with FAI.
One reason balance tests are used to evaluate postural instabilities associated with FAI is due to the work of Freeman et al. (9), who reported that FAI impaired static single-leg balance as measured by the Romberg test. Freeman et al. (9) proposed that disrupted sensorimotor pathways associated with FAI diminished postural reflex responses, causing single-leg balance deficits. Thus, clinicians and researchers have used noninstrumented static single-leg balance tests to assess FAI (6,17). To more objectively assess balance, researchers have used instrumented force plates to quantify static single-leg balance deficits associated with FAI (1-4,12,15-17,19-21,25,29,36-39,44). However, the balance literature on FAI lacks consistency in reporting balance deficits associated with FAI, as some researchers have indicated that balance impairments exist with FAI, and other researchers have reported that balance deficits are not associated with FAI (24,27). Reasons for this disparity in the FAI literature are unclear. Several factors, however, may be affecting the outcomes of studies, making it difficult to conclude how FAI affects balance. Factors that may confound the balance literature on FAI include subject characteristics, inclusion/exclusion criteria, type of balance test used to examine postural stability, and/or type of balance measure (instrumented vs noninstrumented or static vs dynamic).
Balance tests that are more functional may be better than static single-leg balance tests at detecting balance impairments associated with FAI. Dynamic single-leg jump-landing tests and the Star Excursion Balance Test (SEBT) are alternative assessment techniques that may challenge balance greater than static single-leg balance tests. Researchers have reported dynamic balance deficits associated with FAI using time to stabilization (TTS) (2,29,30,43). The SEBT has also quantified dynamic single-leg balance, and researchers have reported balance impairments associated with FAI (14,26). Like static balance measures, however, some researchers have reported dynamic balance deficits with TTS and SEBT measures, whereas others have not (2,14,26,29,30,43). These conflicting results contribute to the confusion in determining if balance deficits exist with FAI.
Inconsistencies reported in the literature on the effects of FAI on static and dynamic balance measures may result from different balance assessment measures or tests. In examining the balance literature related to FAI, we cannot definitively conclude if balance deficits exist with FAI nor can we determine the degree to which various balance tests or balance measures impact previously reported results. Thus, the purpose of this meta-analysis was 1) to pool studies to determine an overall effect size difference between ankles with instability and uninjured ankles, 2) to determine whether effect sizes differed depending on the type of measure used, and 3) to determine whether effect sizes differed between studies with clear inclusion and exclusion criteria and those without criteria.
On the basis of the above, we identified five hypotheses: 1) balance is impaired in subjects with ankle instability; 2) dynamic balance tests will be associated with greater balance impairments than static tests of balance; 3) within the categories of dynamic and static balance, different measures will produce different effect sizes; 4) studies with clearly stated inclusion criteria will report greater balance deficits than those without stated inclusion criteria; and 5) studies with clearly stated exclusion criteria will report greater balance deficits than those without exclusion criteria.
Description of outcomes.
The outcomes used in this study were restricted to two broad categories of balance measures: dynamic and static. Static measures were defined as measures that required subjects to stand either single-legged or double-legged and maintain a quiet posture. We defined dynamic measures as those that required subjects to perform movement as the task (e.g., jump landings). The outcome(s) selected for each study is in Table 1. In both cases, the outcomes were restricted to measures of central tendency (e.g., not root mean squares, SD, etc.) and to those measured on a firm/stable surface.
Type of study designs used.
We did not restrict our analysis to any particular study design or research question. Rather, we used any study that met our inclusion/exclusion criteria. The typical study included was a case-control study comparing subjects with stable and FAI ankles. However, experimental/randomized control trials and ex post facto designs were included when pretreatment data were available.
As with study designs, no restriction was placed on study populations. Demographics for the study populations are in Table 2.
Search Strategy and Manuscript Selection
Our effort to include all available studies consisted of four components: 1) the initial search and evaluation of the studies; 2) a forward search of the included articles from step one; 3) a hand search of included articles from steps 1 and 2; and 4) a direct contact with the corresponding authors of the included articles. The process flow is in Figure 1. In theory, steps 2 and 3 were iterative to the point that no further articles were located. In practice, only one repetition was necessary to identify the included studies.
FIGURE 1-Flow chart ...Image Tools
We searched the literature using PubMed (National Library of Medicine, Bethesda, MD), CINAHL, pre-CIHAHL, and SPORTDiscus through November 2007. The three latter databases were searched simultaneously using EBSCOhost (EBSCO Industries, Inc., Birmingham, AL). The search strategy and results are presented in Table 3. The literature search was directed by our senior research team member (BLA) who has 14 yr of research experience and expertise in the area of FAI. The search was assisted by two doctoral students in the area of rehabilitation and movement science and specializing in FAI. The search strategy was limited to the English language (five foreign language articles were excluded).
Article inclusion and exclusion.
After the initial electronic search was complete (n = 145), three reviewers reviewed titles and abstracts and selected articles for detailed review. Articles were selected or eliminated based on the consensus of the three reviewers. For the detailed review, we used two criteria to determine whether the article should be included in the analysis: 1) tabular means and SD must have been reported for an injured group (or ankle) and an uninjured group (or ankle), or sufficient statistical detail (e.g., P values and group n) was reported to calculate an appropriate effect size; and 2) stated inclusion criteria required injured ankles to have episodes of giving way or frequent sprains, or "functional ankle instability" was described as the target pathology.
We did not include abstracts in this analysis. However, we did review theses and dissertations as part of the overall selection process. Provided the thesis/dissertation met the inclusion criteria and had not been previously published, they were included in the analysis.
Forward and hand search.
Using the articles selected for inclusion from the initial search, we conducted a forward search using the Science Citation Index (The Thomson Corp., New York, NY). This search produced two additional articles for inclusion in the analysis, and we consequently conducted forward searches on these articles. From the articles selected from the initial and both forward searches, we conducted a hand search of each articles cited references. The hand search produced no additional articles.
Contact with authors.
Additionally, the corresponding author of each of the included articles was contacted by letter or e-mail. A listing of their included articles was presented in the correspondence, and they were asked to identify any additional articles that might be eligible for inclusion.
Three investigators extracted relevant data from each study. Specifically, we identified the means, SD, and group sample size for the FAI and the stable ankle groups (or ankles in the case of a contralateral comparison) for the selected outcomes. If only statistical test values were provided, those were extracted. In two cases (4,20), the median and the range were reported. These were converted to estimated means and SD (18). Each investigator extracted data independently. If discrepancies existed between investigators, we resolved those differences by reexamining the study and by agreeing on the final data by consensus. For studies that included a treatment, only the pretreatment values were used in the analysis.
Assessment of confounding.
One concern for every meta-analysis is confounding of the results by factors and/or variables outside the focus of the included studies. For example, balance is affected by age. Mixing age groups without accounting for this confounder may bias the results in a particular direction or produce a null result. How to assess confounding can be a challenge in that many confounders may exist simultaneously. We assessed confounding from two separate perspectives: anthropometric similarity of injured and uninjured subjects and comparability of FAI definitions, inclusion criteria, and exclusion criteria. The former was assessed as an item in our quality assessment with studies statistically comparing subjects' anthropometrics being given a higher score. The latter was assessed with regard to the studies reporting inclusion criteria/FAI definitions and exclusion criteria. Each of these were coded separately as "yes" or "no" and included in the analysis as separate moderator variables. Due to highly variable reporting of inclusion/exclusion criteria by the respective studies, a more detailed analysis of these was not considered possible.
Quality was assessed using a questionnaire developed for this analysis (Fig. 2). The questionnaire was developed from the threats to construct, internal, and external validity identified by Cooke and Campbell (5) and tailored to issues associated with ankle instability research. We elected to use our own scale rather than a similar scale developed for randomized control trials, for example, the PEDro scale, because these scales specifically penalize for not randomizing. For this meta-analysis, all of the data were extracted from observational (i.e., not randomized) studies or from time points before randomization. All studies were assessed by three reviewers on a 20-point scale. Three items were eligible for the rating "not applicable." Thus, the final quality score was calculated as the percentage of points possible, that is, 20 minus any not applicable items. Initially, we reviewed all studies independently. After the initial review, we compared scores and identified large discrepancies among the reviewers. Studies with large discrepancies (scores greater than ±1 SD above the mean) were again independently reviewed. We used these final results as the quality score. The study quality was then regressed on the mean outcome standard difference of the mean (SDM) for each study to determine the effect of quality on SDM.
FIGURE 2-Quality ass...Image Tools
All statistical analyses were completed using Comprehensive Meta-analysis version 2.2.034 (Biostat, Inc., Englewood, NJ). Depending on the data available from each study, data were entered as means and SD (n = 21), as SDM (n = 1), or as means and P values (n = 1). From these data, the SDM was calculated and used for the analysis. The statistical significance of individual and category SDM was tested using the Z statistic, which follows the normal distribution and tests whether the observed SDM differs from zero (13).
Test of the overall effect.
The first hypothesis was initially assessed for heterogeneity using the Q statistic. The Q statistic approximates the chi-square distribution and tests whether the observed effect size variance is greater than chance expectations, that is, heterogeneous (32). If the Q statistic was significant, the random-effects analysis was used to account for this heterogeneity in the analysis.
Comparisons between categories.
For hypotheses 2-5 (e.g., dynamic versus static balance), if the Q statistic was significant, the random-effects model was extended to a mixed-effects model. The mixed-effects model combines studies within each category using the random-effects model, and the categories are subsequently compared using the fixed-effects model.
Treatment of multiple outcomes.
All of the studies included in this analysis used multiple outcomes. Because we viewed hypotheses 1 through 3 as largely exploratory and believed it important to study the effects of all of the potential variables used, we elected to treat each variable as independent. The effects of this are 1) the SE (and confidence interval [CI]) for the overall effect is too small, 2) the statistical test for the overall effect is likely to be liberal, and 3) the statistical comparisons between/among outcomes are conservative. For hypothesis 3, we tested the static and dynamic measures separately. On the basis of the outcomes' units, we grouped the static outcomes into five categories: area, linear, time, velocity, and other. Similarly, we divided the dynamic measures into two categories: SEBT and TTS.
For hypotheses 4 and 5, we averaged the effect sizes across outcomes within each study to produce one effect size per study. We felt this was appropriate because 1) it is likely that the inclusion and exclusion criteria would affect all of the outcomes within a study, and 2) the hypotheses were directed at differences between studies rather than between outcomes. Bias was assessed using Duval's (7) fixed effect trim and fill as well as Sterne and Egger's (33) regression intercept on the averaged outcomes.
Identification of Subject Characteristics
Study and subject characteristics are reported in Tables 1 and 2, respectively. These qualities were extracted either from reported subject characteristics or from reported inclusion/exclusion criteria. As the tables indicate, multiple factors were identified by the investigators and recorded. It is also apparent from the tables that there is no consistency in reporting of subject characteristics. Although we did not make direct comparisons for equality among studies, studies received credit for statistical comparisons of subjects' anthropometric characteristics as part of our quality assessment.
To compare our agreement among reviewers, we calculated the intraclass correlation coefficient (ICC), form (2,1). The ICC value was 0.735 with a 95% CI ranging from 0.547 to 0.868 and is higher than the reported value for the PEDro scale (22). The mean quality score was 24.3% ± 11.0% with a range from 0% to 43.7%. We found no relationship between study quality and SDM (slope = −0.011, P = 0.09).
Due to the large heterogeneity between balance measures (Q = 257.0, P < 0.001), we used a random-effects analysis to determine whether balance was impaired in FAI subjects. This revealed that FAI was associated with poorer balance (SDM = 0.455, 95% CI = 0.334-0.577, Z = 7.34, P < 0.001).
Differences between dynamic and static measures.
For the comparison between static and dynamic measures, there was significant within-group heterogeneity (Q = 254.6, P < 0.001). Thus, we computed a mixed-effects analysis and found no difference between dynamic and static measures (Q = 3.44, P = 0.063).
Comparison among static measures.
There was significant within-group heterogeneity (Q = 147.71, P < 0.001) for the static measures. Thus, we computed a mixed-effects analysis to compare the static measurement groups (i.e., area, linear, time, velocity, and other). The overall SDM for the static measures was significant (0.324, 95% CI = 0.189-0.459, Z = 4.71, P < 0.001), indicating that FAI ankles had impaired balance as measured by static balance. Furthermore, we found that the static measure groups' SDM were significantly different from each other (Q = 12.85, P = 0.012). The individual SDM and the group values are presented in Figure 3.
FIGURE 3-Forrest plo...Image Tools
Comparison among dynamic measures.
For the dynamic measures (i.e., SEBT and TTS), no significant within-group heterogeneity was found (Q = 26.86, P = 0.836). Thus, a fixed-effects analysis was completed. The overall SDM was significant (0.336, Z = 7.28, P < 0.001, 95% CI = 0.246-0.410), indicating that FAI ankles had impaired balance as measured by the dynamic measures. The comparison between SEBT and TTS was significant (Q = 6.22, P = 0.013), indicating a significant difference between the two categories of dynamic measures. The individual SDM and the group values are presented in Figure 4.
FIGURE 4-Forrest plo...Image Tools
Effects of inclusion criteria.
Due to significant within-group heterogeneity (Q = 84.4, P < 0.001), we conducted a mixed-effects analysis. We found no significant difference (Q = 0.428, P = 0.513) between studies with inclusion criteria versus those without inclusion criteria. The individual SDM and the group values are presented in Figure 5.
FIGURE 5-Forrest plo...Image Tools
Effects of exclusion criteria.
Due to significant within-group heterogeneity (Q = 62.7, P < 0.001), we conducted a mixed-effects analysis. We found no significant difference (Q = 3.20, P = 0.074) between studies with exclusion criteria versus those without exclusion criteria. The individual SDM and the group values are presented in Figure 6.
FIGURE 6-Forrest plo...Image Tools
The bias funnel plot including filled studies is presented in Figure 7. Egger's regression intercept was 3.00 (P = 0.048, two tailed), suggesting that bias was present. The trim and fill procedure identified six studies to be trimmed and filled. The SDM before filling was 0.594 (95% CI = 0.489-0.387) and after filling was 0.361 (95% CI = 0.237-0.486). This suggests that unpublished or "fugitive" (28) studies may exist and that if they were included, the SDM would approximate the filled value.
FIGURE 7-Funnel plot...Image Tools
Our initial analysis of outcomes measures revealed that ankles with FAI exhibited poorer balance performance than stable ankles. Although from a historical perspective the literature has been equivocal (27), these results clearly indicate that balance is impaired and that the average effect across outcomes is rather large (SDM = 0.455). What remains unclear is whether these differences preexisted or were the result of injury. Comparisons between the FAI ankles and the contralateral normal ankle have failed to produce differences (36,38,39). Tropp (36) suggested that the lack of differences between contralateral ankles may represent a preexisting condition or a central organizational change due to pain and immobilization. Because our data set predominantly includes studies with case-control designs rather than prospective designs, we cannot answer this specific question.
Differences between Dynamic and Static Balance
One of the clear distinctions in the outcomes was the use of either a dynamic or a static measure of balance. For our purposes, static balance was defined as balancing on a stable surface without intentional movement by the subject, for example, postural sway on a force plate. Conversely, dynamic measures were defined as balance tasks that required the subject to perform some movement (e.g., leg reaching with the SEBT) or a task (e.g., jumping with TTS). We believed it important to determine whether differences existed between these two broad categories of outcomes. In fact, there was no statistical difference, but the P value was quite low (P = 0.063) and close to the criterion value of 0.05. It is worth noting that for this part of the analysis, the outcomes from each study were treated as being independent of each other. This is very likely not the case but was done because we wanted to make direct comparisons among different balance measures. When comparing differences in SDM among groups, this will tend to increase the SE and CI and make the statistical comparison conservative. Thus, we suspect that a difference between dynamic and static measures does exist with dynamic measures producing a smaller SDM than static measures.
Examination of the static measures (Fig. 3) reveals that three of the four categories of measurement (linear, time, velocity, and other) produced significant SDM. The area group did not produce a significant difference, nor were any of the individual study SDM for area significant. This was surprising to us because this was one of the first stabilometry measures used to study ankle injury and has been described as indicative of ankle instability (39). However, other work by Tropp et al. (37) found no differences. Thus, our results are consistent with the latter but not the former findings.
Our examination of the SDM revealed that the greatest SDM (1.818) was for the time measures (time in balance, i.e., postural equilibrium test , and time to boundary ). Of these, the highest SDM was produced by the time in balance, eyes open condition (4). We believe that this is an interesting finding because the time-in-balance measure is simple and could be easily used clinically. However, these results should be viewed cautiously because the mean difference had to be estimated using the median and range, and no reliability of the measure was reported. We recommend further study of this measure to establish its reliability and determine a clinical cutoff score using the receiver operating characteristic (ROC) analysis.
The Other category also produced the second largest SDM greater than one (1.035). Of these measures, the number of foot lifts of Hiller et al. (17) produced the largest SDM (4.845). Similar to time in balance, foot lifts can be easily counted in the clinical setting. Furthermore, the reported reliability of this test was fair to good (ICC = 0.73, 95% CI = 0.40-0.89). Thus, we would suggest that future investigations should include this measure to better establish its usefulness and that ROC curves with cutoff scores should be developed.
On the basis of the group SDM, the TTS measure produced the greatest SDM (0.555) with time to stabilization (TTS) in the anterior/posterior direction producing the greatest effects (2,29). In contrast, the Star Excursion Balance Test (SEBT) produced an SDM (0.287) approximately one half that of TTS, suggesting that TTS is the better of the two measures. However, TTS requires availability of a force plate and processing software to generate the measures. This makes TTS less clinically convenient than the SEBT. On the basis of this and the very similar SDM between static and dynamic measures, we would recommend future studies focus on convenient static measures (e.g., foot lifts) for clinical use.
The Effects of Inclusion and Exclusion Criteria
As can be seen from Tables 1 and 2, the reported inclusion criteria and subject characteristics vary greatly from study to study. Our initial examination of the selected studies suggested that in fact very few studies actually reported true inclusion criteria and instead reported only subject characteristics. On the basis of this, we hypothesized that those studies that reported clear inclusion criteria would find larger effects than those that did not. However, further examination of the studies revealed that in lieu of clearly identified inclusion criteria, most studies at least had clear definitions of FAI. On the basis of this, we subsequently included FAI definitions as inclusion criteria and grouped studies as either having FAI definitions/inclusion criteria or not. We found only four studies that did not use inclusion criteria or had a clear definition of FAI. This was far fewer than we had initially expected. Furthermore, the comparison between the two groups of studies failed to find a significant difference, which was also counter to our expectation. We doubt that this finding means that inclusion criteria are not important. Rather, we suspect that studies not stating inclusion criteria were actually reporting inclusion criteria as part of their subject characteristics. If that was the case, then in fact all studies had inclusion criteria. Nevertheless, our experience was that the lack of clearly stated inclusion criteria and/or the use of a separate definition of pathology as part of the inclusion criteria made our task more difficult. We would recommend to all authors that future studies should use a priori inclusion criteria and that these criteria should be stated as such and listed together in the methods section of the manuscript.
Exclusion criteria were more clearly reported in all studies but similar to the inclusion criteria varied across studies. Similar to the inclusion criteria, we found no difference between those studies that reported exclusion criteria and those that did not. However, the P value (P = 0.074) was close to the criterion value, and the SDM for the studies without exclusion criteria was nearly three times larger than that for studies with exclusion criteria. This suggests to us that researchers should give more careful attention to exclusion criteria because they may have an important influence on the results.
Quantitative Assessment of Bias
On the basis of Egger's regression intercept, our data demonstrated some degree of bias, and the trim and fill procedure identified six studies for trimming. The adjusted SDM was smaller after the fill procedure, indicating that the SDM was sensitive to this bias. These results suggest that the smaller studies in our data set have larger SDM than expected and that there may be missing studies. This may be because smaller studies without significant differences tend to be under represented or missing in the reported literature (i.e., publication bias). However, publication bias is not the only cause of bias (8). It is also possible that smaller studies truly have larger effect sizes. This may be due either to more potent treatments or to better quality measures that can be provided in smaller studies. Because we did not focus this analysis on treatment effects, the former cause is not relevant. Whether measurement quality is associated with the existing bias is unclear. On the basis of the variety of measures used to assess balance, this seems possible, but we could not discern a pattern indicating that smaller studies used different or better quality measures than larger studies.
Assessment of Quality of Included Studies
Our rating of quality was lower than we had expected with a range that did not go above 44 (out of 100). However, our meta-regression revealed that study quality did not impact SDM size. We should emphasize that study quality as we measured it was dependent on at least two factors: study quality and study reporting. All three reviewers felt that it was often difficult to distinguish between poor quality and poor reporting. Thus, we suspect that several studies probably merited higher quality scores but were penalized due to poor reporting. This fact emphasizes the importance good reporting of study details, especially the methods, on the part of authors, reviewers, and editors.
On the basis of our results, it appears that individuals with ankle instability have deficits in their balance. These deficits appear to exist regardless of whether balance is assessed with static or dynamic tests. However, our analysis is probably conservative based on our processing of the data, and it may be that static measures actually perform better than dynamic measures. Because our data is from observational studies, it is not possible to determine whether these differences were preexisting or the result of injury.
Select dynamic and static measures produced larger SDM than others. Within the static measures, time-based measures (time in balance  and time to boundary ) performed best as well as measures within the miscellaneous Other category. Further analysis of the Other category suggested that the number of foot lifts (17) produced the largest SDM. Because time in balance and foot lifts are measures that can be easily completed in the clinical setting, we believe further study should focus on these measures to establish their clinical validity.
Within the dynamic category, TTS measures performed better than the SEBT and suggest TTS would be preferred to the SEBT. However, although TTS may be useful in the laboratory setting, its setup is more complicated than the SEBT and requires a force plate and appropriate analytical software. Thus, it may be viewed as less desirable for the clinical setting.
Finally, neither inclusion nor exclusion criteria affected the results. However, there was inconsistent reporting of inclusion and exclusion criteria making comparisons difficult. Furthermore, some inclusion criteria were reported as part of the definition of FAI. We would suggest that future reports combine FAI definitions and inclusion criteria into a specific section of the methods.
One caution that should be added is that most of the studies included in the analysis were conducted on relatively physically active individuals. This is because most of the research on FAI is conducted by sports medicine specialists, either medical or allied health. Whether these results apply to more sedentary populations is unknown. Thus, additional FAI research may want to focus on nonphysically active populations. It remains possible that different measures may better apply to different populations.
The results of this study are not an endorsement by the American College of Sports Medicine.
The authors have no conflict of interests to report.
Disclosure of funding: no funding was provided to support this project.
1. Baier M, Hopf T. Ankle orthoses effect on single-limb standing balance in athletes with functional ankle instability. Arch Phys Med Rehabil
2. Brown C, Ross S, Mynark R, et al. Assessing functional ankle instability with joint position sense, time to stabilization, and electromyography. J Sport Rehabil
3. Brown CN, Mynark R. Balance deficits in recreational athletes with chronic ankle instability. J Athl Train
4. Chrintz H, Falster O, Roed J. Single-leg postural equilibrium test. Scand J Sport Sciences
5. Cook TD, Campbell DT. Quasi-experimentation. Design and Analysis Issues for Field Settings
. 1st ed. Boston: Houghton Mifflin Company; 1979. p. 405.
6. Docherty CL, McLeod TCV, Shultz SJ. Postural control deficits in participants with functional ankle instability as measured by the balance error scoring system. Clin J Sport Med
7. Duval S. The trim and fill method. In: Rothstein HR, Sutton AJ, Borenstein M, editors. Publication Bias in Meta-analysis: Prevention, Assessment, and Adjustments
. Chichester: John Wiley & Sons, Ltd.; 2005. pp. 127-44.
8. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. Brit Med J
9. Freeman MR, Dean ME, Hanham IF. The etiology and prevention of functional instability of the foot. J Bone Joint Surg
10. Gauffin H, Tropp H, Odenrick P. Effect of ankle disk training on postural control in patients with functional instability of the ankle joint. Int J Sports Med
11. Gerber JP, Williams GN, Scoville CR, Arciero RA, Taylor DC. Persistent disability associated with ankle sprains: a prospective examination of an athletic population. Foot Ankle Internat
12. Hale SA, Hertel J, Olmsted-Kramer LC. The effect of a 4-week comprehensive rehabilitation program on postural control and lower extremity function in individuals with chronic ankle instability. J Orthop Sports Phys Ther
13. Hedges LV. Fixed effects models. In: Cooper H, Hedges LV, editors. The Handbook of Research Synthesis
. New York: Russel Sage Foundation; 1994. pp. 285-99.
14. Hertel J, Braham RA, Hale SA, Olmsted-Kramer LC. Simplifying the Star Excursion Balance Test: analyses of subjects with and without chronic ankle instability. J Orthop Sports Phys Ther
15. Hertel J, Olmsted-Kramer LC. Deficits in time-to-boundary measures of postural control with chronic ankle instability. Gait Posture
16. Hiller CE, Refshauge KM, Beard DJ. Sensorimotor control is impaired in dancers with functional ankle instability. Am J Sports Med
17. Hiller CE, Refshauge KM, Herbert RD, Kilbreath SL. Balance and recovery from a perturbation are impaired in people with functional ankle instability. Clin J Sport Med
18. Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Med Res Method
[Internet]. 2005 [cited 2005 April 20];5:13. Available from: http://www.biomedcentral.com/1471-2288/5/13
19. Hubbard TJ, Kramer LC, Denegar CR, Hertel J. Contributing factors to chronic ankle instability. Foot Ankle Int
20. Konradsen L, Ravn JB. Prolonged peroneal reaction time in ankle instability. Int J Sports Med
21. Lee A, Lin W, Huang CH. Impaired proprioception and poor static postural control in subjects with functional instability of the ankle. J Exerc Sci Fitness
22. Maher CG, Sherrington C, Herbert RD, Moseley AM, Elkins M. Reliability of the PEDro scale for rating quality of randomized controlled trials. Phys Ther
23. McGuine TA, Greene JJ, Best T, Leverson G. Balance as a predictor of ankle injuries in high school basketball players. Clin J Sport Med
24. McKeon PO, Hertel J. Systematic review of postural control and lateral ankle instability: part I. Can deficits be detected with instrumented testing? J Athl Train
25. Michell TB, Ross SE, Blackburn JT, Hirth CJ, Guskiewicz KM. Functional balance training, with or without exercise sandals, for subjects with stable or unstable ankles. J Athl Train
26. Olmsted LC, Carcia CR, Hertel J, Shultz SJ. Efficacy of the Star Excursion Balance Tests in detecting reach deficits in subjects with chronic ankle instability. J Athl Train
27. Riemann B. Is there a link between chronic ankle instability and postural instability. J Athl Train
28. Rosenthal MC. The fugitive literature. In: Cooper H, Hedges LV, editors. The Handbook of Research Synthesis
. New York: Russel Sage Foundation; 1994. pp. 85-94.
29. Ross SE, Guskiewicz KM. Examination of static and dynamic postural stability in individuals with functionally stable and unstable ankles. Clin J Sport Med
30. Ross SE, Guskiewicz KM, Yu B. Single-leg jump-landing stabilization times in subjects with functionally unstable ankles. J Athl Train
31. Schaap GR, de Keizer G, Marti K. Inversion trauma of the ankle. Arch Orthop Trauma Surg
32. Shadish WR, Haddock CK. Combining estimates of effect size. In: Cooper H, Hedges LV, editors. The Handbook of Research Synthesis
. New York: Russel Sage Foundation; 1994. pp. 261-81.
33. Sterne JAC, Egger M. Regression methods to detect publication bias and other bias in meta-analysis. In: Rothstein HR, Sutton AJ, Borenstein M, editors. Publication Bias in Meta-analysis: Prevention, Assessment, and Adjustments
. Chichester: John Wiley & Sons, Ltd.; 2005. pp. 99-110.
34. Trojian TH, McKeag DB. Single leg balance test to identify risk of ankle sprains. Brit J Sports Med
35. Tropp H. Commentary: functional ankle instability revisited. J Athl Train
36. Tropp H. Pronator muscle weakness in functional instability of the ankle joint. Int J Sports Med
37. Tropp H, Ekstrand J, Gillquist J. Stabilometry in functional instability of the ankle and its value in predicting injury. Med Sci Sports Exerc
38. Tropp H, Odenrick P. Postural control in single-limb stance. J Orthop Res
39. Tropp H, Odenrick P, Gillquist J. Stabilometry recordings in functional and mechanical instability of the ankle joint. Int J Sports Med
40. Verhagen RAW, de Keizer G, van Dijk CN. Long-term follow-up of inversion trauma of the ankle. Arch Orthop Trauma Surg
41. Wang HK, Chen CH, Shiang TY, Jan MH, Lin KH. Risk-factor analysis of high school basketball-player ankle injuries: a prospective controlled cohort study evaluating postural sway, ankle strength, and flexibility. Arch Phys Med Rehabil
42. Wayne HA. A comparison of the balance error scoring system with selected forceplate measures in individuals with functionally unstable ankles [master's thesis]. Chapel Hill (NC): University of North Carolina at Chapel Hill; 2004. p. 132.
43. Wikstrom EA, Tillman MD, Borsa PA. Detection of dynamic stability deficits in subjects with functional ankle instability. Med Sci Sports Exerc
44. You SH, Granata KP, Bunker LK. Effects of circumferential ankle pressure on ankle proprioception, stiffness, and postural stability: a preliminary investigation. J Orthop Sports Phys Ther
45. Zinder SM. Influence of ankle bracing and load on factors of stability in functionally unstable and normal ankles [dissertation]. Charlottesville (VA): University of Virginia; 2002. p. 330.