Interventions for preventing ankle ligament injuries (Review)

Transcription

1 Handoll HHG, Rowe BH, Quinn KM, de Bie R This is a reprint of a Cochrane review, prepared and maintained by The Cochrane Collaboration and published in The Cochrane Library 2005, Issue 3 1

2 T A B L E O F C O N T E N T S ABSTRACT SYNOPSIS BACKGROUND OBJECTIVES CRITERIA FOR CONSIDERING STUDIES FOR THIS REVIEW SEARCH STRATEGY FOR IDENTIFICATION OF STUDIES METHODS OF THE REVIEW DESCRIPTION OF STUDIES METHODOLOGICAL QUALITY RESULTS DISCUSSION AUTHORS CONCLUSIONS ACKNOWLEDGEMENTS POTENTIAL CONFLICT OF INTEREST SOURCES OF SUPPORT REFERENCES TABLES Characteristics of included studies Characteristics of excluded studies ADDITIONAL TABLES Table 01. MEDLINE - SilverPlatter search strategy used up to August GRAPHS Comparison 01. Any intervention v no intervention/control Comparison 02. External ankle support v control Comparison 03. Ankle disk training v control Comparison 04. Warm-up exercises and ankle disk training v control Comparison 05. Stretching of leg muscles before warm-up v control Comparison 06. Cushioned insole v standard insole Comparison 07. Health education v control Comparison 08. Prophylactic programme v control Comparison 09. Adhesive tape v cloth wrap Comparison 10. High top shoes and chamber v high top shoes Comparison 11. Semi-rigid ankle orthosis v ankle disk training Comparison 12. Urethane insole v special mesh insole Comparison 13. Rehabilitation: supervised programme (emphasis on balance training) v control Comparison 14. Rehabilitation: wobble board (ankle disk) exercises v control INDEX TERMS COVER SHEET COMMENTS AND CRITICISMS GRAPHS AND OTHER TABLES Fig. 1. Comparison 01. Any intervention v no intervention/control Ankle sprain by intervention type Fig. 2. Comparison 01. Any intervention v no intervention/control Ankle sprain: previous history Fig. 3. Comparison 01. Any intervention v no intervention/control Ankle sprain: no previous history Fig. 4. Comparison 02. External ankle support v control Ankle sprain Fig. 5. Comparison 02. External ankle support v control Ankle sprain by previous history - relative risk Fig. 6. Comparison 02. External ankle support v control i

3 03 Ankle sprain by previous history - odds ratios Fig. 7. Comparison 02. External ankle support v control Severity of ankle injury Fig. 8. Comparison 02. External ankle support v control Incidence of other leg injuries Fig. 9. Comparison 02. External ankle support v control Subjective instability of the ankle Fig. 10. Comparison 03. Ankle disk training v control Ankle sprain Fig. 11. Comparison 04. Warm-up exercises and ankle disk training v control Ankle sprains (NB. Spuriously narrow confidence intervals) Fig. 12. Comparison 04. Warm-up exercises and ankle disk training v control Other leg injuries (NB. Spuriously narrow confidence intervals) Fig. 13. Comparison 05. Stretching of leg muscles before warm-up v control Ankle sprain (NB spuriously narrow confidence intervals) Fig. 14. Comparison 05. Stretching of leg muscles before warm-up v control Other lower limb injuries (NB spuriously narrow confidence intervals) Fig. 15. Comparison 06. Cushioned insole v standard insole Lateral ankle sprain Fig. 16. Comparison 06. Cushioned insole v standard insole Attendance of medical facility Fig. 17. Comparison 06. Cushioned insole v standard insole Activity restriction Fig. 18. Comparison 06. Cushioned insole v standard insole Medical discharge Fig. 19. Comparison 06. Cushioned insole v standard insole Insole uncomfortable Fig. 20. Comparison 07. Health education v control Ankle sprain (lateral ankle injury) Fig. 21. Comparison 08. Prophylactic programme v control Ankle sprain (NB spuriously narrow confidence intervals) Fig. 22. Comparison 09. Adhesive tape v cloth wrap Ankle injury Fig. 23. Comparison 10. High top shoes and chamber v high top shoes Ankle sprain Fig. 24. Comparison 11. Semi-rigid ankle orthosis v ankle disk training Ankle sprain Fig. 25. Comparison 12. Urethane insole v special mesh insole Lateral ankle sprain Fig. 26. Comparison 12. Urethane insole v special mesh insole Attendance of medical facility Fig. 27. Comparison 12. Urethane insole v special mesh insole Activity restriction Fig. 28. Comparison 12. Urethane insole v special mesh insole Medical discharge Fig. 29. Comparison 12. Urethane insole v special mesh insole Insole uncomfortable Fig. 30. Comparison 13. Rehabilitation: supervised programme (emphasis on balance training) v control Re-injury (recurrent ankle sprain) Fig. 31. Comparison 14. Rehabilitation: wobble board (ankle disk) exercises v control Recurrent ankle sprain Fig. 32. Comparison 14. Rehabilitation: wobble board (ankle disk) exercises v control Subjective functional instability ii

4 Handoll HHG, Rowe BH, Quinn KM, de Bie R Status: Commented This record should be cited as: Handoll HHG, Rowe BH, Quinn KM, de Bie R. Interventions for preventing ankle ligament injuries. The Cochrane Database of Systematic Reviews 2001, Issue 3. Art. No.: CD DOI: / CD This version first published online: 23 July 2001 in Issue 3, Date of most recent substantive amendment: 01 April 2001 A B S T R A C T Background Some sports, for example basketball and soccer, have a very high incidence of ankle injuries, mainly sprains. Consequently, ankle sprains are one of the most commonly treated injuries in acute care. Objectives To assess the effects of interventions used for the prevention of ankle ligament injuries or sprains in physically active individuals from adolescence to middle age. Search strategy We searched the Cochrane Musculoskeletal Injuries Group s specialised register, MEDLINE, PubMed, EMBASE, CINAHL, the National Research Register and bibliographies of study reports. We also contacted colleagues and some trialists. The most recent search was conducted in July Selection criteria Randomised or quasi-randomised trials of interventions for the prevention of ankle sprains in physically active individuals from adolescence to middle age were included provided that ankle sprains were recorded. Interventions included use of modified footwear, external ankle supports, co-ordination training and health education. These could be applied as a supplement to treatment provided that prevention of re-injury was the primary objective. Data collection and analysis At least two reviewers independently assessed methodological quality and extracted data. Wherever possible, results of outcome measures were pooled and sub-grouped by history of previous sprain. Relative risks (RR) and 95% confidence intervals (95% CI) are reported for individual and pooled data. Main results In this review update, a further nine new trials were included. Overall, 14 randomised trials with data for 8279 participants were included. Twelve trials involved active, predominantly young, adults participating in organised, generally high-risk, activities. The other two trials involved injured patients who had been active in sports before their injury. The prophylactic interventions under test included the application of an external ankle support in the form of a semi-rigid orthosis (three trials), air-cast brace (one trial) or high top shoes (one trial); ankle disk training; taping; muscle stretching; boot inserts; health education programme and controlled rehabilitation. The main finding was a significant reduction in the number of ankle sprains in people allocated external ankle support (RR 0.53, 95% CI 0.40 to 0.69). This reduction was greater for those with a previous history of ankle sprain, but still possible for those without prior sprain. There was no apparent difference in the severity of ankle sprains or any change to the incidence of other leg injuries. The protective effect of high-top shoes remains to be established. 1

5 There was limited evidence for reduction in ankle sprain for those with previous ankle sprains who did ankle disk training exercises. Various problems with data reporting limited the interpretation of the results for many of the other interventions. Authors conclusions This review provides good evidence for the beneficial effect of ankle supports in the form of semi-rigid orthoses or air-cast braces to prevent ankle sprains during high-risk sporting activities (e.g. soccer, basketball). Participants with a history of previous sprain can be advised that wearing such supports may reduce the risk of incurring a future sprain. However, any potential prophylactic effect should be balanced against the baseline risk of the activity, the supply and cost of the particular device, and for some, the possible or perceived loss of performance. Further research is indicated principally to investigate other prophylactic interventions, their cost-effectiveness and general applicability. S Y N O P S I S External ankle supports may reduce ankle sprains when used in some high-risk recreational activities Ankle sprains, involving damage to ankle ligaments, are one of the most common sports injuries. Various methods of prevention are in common use for active athletic people. These include taping, the use of external ankle supports and co-ordination training using ankle disks. This review of trials indicates that supporting ankles with semi-rigid braces during high-risk activities can prevent sprains, especially in people who have had previous ankle sprains. Ankle disk training may also be useful, particularly for people with a history of ankle sprain, but needs more research. Evidence for other interventions was inconclusive. B A C K G R O U N D Inversion injuries, primarily sprains, of the ankle are one of the most commonly treated injuries. For instance, in the UK alone 5000 injuries per day are treated, mainly by primary health care doctors (Kannus 1991). In people with a sedentary lifestyle such injuries can be painful but relatively non-disruptive. However, in athletes and those whose work is of a more demanding nature, such injuries may have important effects. Moreover, some sports (e.g. basketball, soccer and volleyball) have a very high incidence of ankle injuries. The most common mechanism of ankle injury is inversion of the plantarflexed foot. Injury occurs to the anterior talo-fibular ligament first, followed to a varying degree by the calcaneofibular ligament. The posterior talofibular ligament is usually uninjured unless there is a frank dislocation of the ankle. Together, these ligaments form the lateral ligament complex. While isolated medial ligamentous injuries do occur, lateral ligament tears are far more frequent. Traditionally, lateral ligament injuries are graded I, II or III (or 1, 2 or 3). Grade I (mild) represents a stretch (or sprain), II (moderate) a partial tear, and III (severe) a complete tear (Kannus 1991). However, there is some evidence that the severity of an ankle sprain may not predict the recovery rate (De Bie 1998). Most injuries resolve but some ankles will have chronic instability (mechanical and / or functional instability). Subsequent ankle injury is also more likely. A recent study of ankle injuries in basketball found that players with a history of ankle injury were nearly five times more likely to sustain an ankle injury (McKay 2001). Prevention of ankle injuries has the potential to play an important role in maintaining health for those people who engage in highrisk sports and those who have suffered a previous injury to the ankle ligament complex. Methods of prevention of ankle ligament injuries include use of modified footwear and associated supports, ankle taping, adapted training regimens including ankle exercises, and injury awareness. Prevention of injury recurrence may include interventions such as ankle disk exercises aimed at enhancing coordination and retraining proprioception (sense of muscular position). Secondary prevention, the prevention of re-incurrence, is a common treatment goal for many studies of ankle sprain treatment (Eiff 1994a). In this review, we limited the inclusion of treatment trials to those evaluating preventive interventions, such as supplementary balance and coordination exercises, applied as part of rehabilitation following an ankle sprain. Such trials would aim to reduce the risk of re-injury as well as providing data on re-injury. This review represents an update of a previous Cochrane review (Quinn 1997). As before, we have restricted our scope to randomised and quasi-randomised trials dealing with the prevention of ligament injuries, and have also applied the method of metaanalysis. Previously, we were unable to locate a published systematic review addressing this issue. However, a systematic review which focuses on sports injuries but also includes non-randomised studies has now been published (Thacker 1999). 2

6 O B J E C T I V E S The objective of this review is to compare the types of intervention for the prevention of ankle ligament injuries in individuals from adolescence to middle age. Where possible, those with no prior ankle ligament injury were analysed separately from those with previous ankle ligament injury. Also where possible, those undergoing rehabilitation for ankle sprain were analysed separately. The specific null hypotheses tested were: 1. No differences exist in outcomes between any intervention aimed at prevention of ankle ligament injuries versus no intervention. 2. No differences exist in outcomes between different methods aimed at prevention of ankle ligament injuries. C R I T E R I A F O R C O N S I D E R I N G S T U D I E S F O R T H I S R E V I E W Types of studies Any randomised or quasi-randomised (methods of allocating participants to a treatment which are not strictly random, e.g. by date of birth, hospital record number, alternation) clinical trials of preventative interventions meeting the specifications below were considered. Types of participants Individuals, from adolescence to middle age, at risk of or who have had a previous ankle ligament injury were eligible. This review also includes people who were undergoing rehabilitation after an ankle sprain. Types of intervention Any intervention, including use of modified footwear and associated supports, taping, adapted training programmes including ankle exercises, and injury awareness, which were applied to prevent ankle ligament injury were eligible. Trials were excluded which involved testing of preventative devices in laboratory conditions or which only reported intermediate outcome measures that have no proven relationship to clinical outcomes. Trials involving the rehabilitative treatment of ankle injuries were included provided the interventions under test were specifically intended to reduce the risk of re-injury, and that re-injury data were actively collected and reported. Trials whose primary objective was treatment rather than secondary prevention were excluded even if re-injury was recorded. Types of outcome measures Evaluation of Prevention 1. Incidence of ankle ligament injury. 2. Severity of ligament injuries to the ankle (grade, surgery considered). 3. Incidence of other lower limb injuries. 4. Complications (e.g. fitness deficit, skin abrasions, other injuries). 5. Measures of service utilisation or resource use (e.g. medical centre visits, cost of bracing). 6. Subjective assessment of instability (giving out), performance inhibition. S E A R C H S T R A T E G Y F O R I D E N T I F I C A T I O N O F S T U D I E S See: Bone, Joint and Muscle Trauma Group search strategy The search for trials was extended from March 1997 to July 2000 in this update. Papers were identified by the following search strategy. No language restriction was applied. a. We searched the Cochrane Musculoskeletal Injuries Group s specialised register (April 2000), MEDLINE (from 1966 to August 1999), PubMed (January 1999 to July 2000); EMBASE (from 1980 to September 1996), CINAHL (1982 to April 2000) and the National Research Register (Issue 2, 2000). In MEDLINE (SilverPlatter), the general search strategy for the prevention and treatment of ankle ligament injuries employed the first two levels of the optimal MEDLINE search strategy (Clarke 2001) in conjunction with the following specific search terms presented in Table 01. This search has now been revised and the following search strategy for MEDLINE (OVID Web) will be used for future updates: 1. Ankle Injuries/pp, pc, rh, th [Prevention & Control, Rehabilitation, Therapy] 2. Athletic Injuries/pp, pc, rh, th [Prevention & Control, Rehabilitation, Therapy] 3. exp Sprains and Strains /pp, pc, rh, th [Prevention & Control, Rehabilitation, Therapy] 4. ANKLE/in [Injuries] 5. or/ (ankle adj5 (sprain$ or stabili#ation or support$)).tw. 7. or/ randomized controlled trial.pt. 9. controlled clinical trial.pt. 10. Randomized Controlled Trials/ 11. Random Allocation/ 12. Double Blind Method/ 13. Single Blind Method/ 14. or/ Animal/ not Human/ 3

7 not clinical trial.pt. 18. exp Clinical Trials/ 19. (clinic$ adj25 trials$).tw. 20. ((singl$ or doubl$ or trebl$ or tripl$) adj25 (blind$ or mask$)).tw. 21. Placebos/ 22. placebo$.tw. 23. random$.tw. 24. Research Design/ 25. or/ not not or/16, and/7,28 In PubMed the following all inclusive search was conducted in July 2000: 1999 to July ankle/injuries [all fields] 2. ankle* 3. injur* OR sprain* 4. #2 AND #3 5. ruptur* OR tear* OR torn 6. ligament* 7. #2 AND #5 AND #6 8. ankle sprain [all fields] OR ankle stabilisation [all fields] OR ankle stabilization [all fields] OR ankle supporting [all fields] OR ankle supports [all fields] 9. #1 OR #4 OR #7 OR #8 / Limits: Publication Date from 1999/01/01 to 2000/07/31 In CINAHL, the following search was conducted in July 2000 for the update. 1. ankle.mp 2. exp Lateral ligament, ankle/in [Injuries] 3. 1 or 2 4. sprain.mp or sprains.mp or sprained.mp 5. injury.mp or injuries.mp or injured.mp 6. exp athletic injuries 7. exp dislocations 8. exp leg injuries 9. exp rupture 10. exp sprains and strains 11. exp tears and lacerations 12. Wounds and injuries or 5 or 6 or 7 or 8 or 9 or 10 or 11 or prevention.mp or prevent.mp or preventing.mp or protect.mp or protection.mp or protecting.mp and 13 and 14 National Research Register ( National/nrr-frame.html) entries, 2000 Issue 2, using the following search: (((((ANKLE* and BRAC*) OR (ANKLE AND INJUR*)) OR ANKLE-INJURIES*.ME) OR (PREVENTI* AND INJUR*)) b. The bibliographies of all papers identified by these strategies were searched. c. A search for unpublished or unlisted studies was made by contacting relevant organisations, for example the Medical Departments of Defence Forces, or appropriate sporting organisations. M E T H O D S O F T H E R E V I E W Retrieval of Studies For the first version of this review (Quinn 1997), the trials identified by the specific search strategy and from a general lowerlimb injury search strategy were scrutinised for consideration for the review by three reviewers. Copies of those deemed eligible were obtained, and independently assessed by all five reviewers previously listed on the byline of the review (Quinn 1997) for inclusion into the review. Disagreement was resolved by discussion. In this update, trials were identified by one reviewer (HH). The screening process for selecting trials to put forward for consideration by the review group was tighter in that studies, such as those conducted in a laboratory setting, located via the literature search update were usually not put forward for consideration by the other members of the research team if there were clearly no data on injury prevention. These aside, all other eligible trials were independently assessed by all four listed reviewers for inclusion into the review. Disagreement was resolved by discussion. Assessment of methodological quality: In the first version of this review, methodological quality for each study was independently assessed, without masking, by at least four reviewers from the group using a piloted, subject-specific modification of the generic evaluation tool used by the Cochrane Musculoskeletal Injuries Group (CMSIG). Any disagreement was resolved by discussion. In this update, methodological quality for each new study was independently assessed, without masking, by at least two reviewers from the group. One reviewer (HH) assessed all the trials for active and systematic follow-up and looked through the scores for consistency for item B; the other reviewers provided feedback on the results and a consensus was reached. The scoring scheme for the eleven aspects of internal and external validity covered by this methodological quality assessment tool plus brief notes of coding guidelines for selected items, and the specific scores for items J, K and L for the primary outcome of ankle sprains, are given below. In the update, the scores for individual items of the methodological quality assessment scheme were changed from 3,2,1 to 2,1,0 in keeping with the revised CMSIG scoring scheme. 4

8 A. Was the assigned treatment adequately concealed prior to allocation? 2 = method did not allow disclosure of assignment. 1 = small but possible chance of disclosure of assignment or unclear. 0 = quasi-randomised or open list/tables. Cochrane code: Clearly Yes = A; Not sure = B; Clearly No = C B. Were the outcomes of patients who withdrew described and included in the analysis (intention to treat)? 2 = withdrawals well described and accounted for in analysis. 1 = withdrawals described and analysis not possible; or no loss implied from the trial results (all participants included in analyses). 0 = no mention, inadequate mention, or obvious differences and no adjustment. C. Were the outcome assessors blinded to treatment status? 2 = effective action taken to blind assessors. 1 = small or moderate chance of unblinding of assessors. 0 = not mentioned or not possible. D. Were the treatment and control group comparable at entry? The principal confounders were considered to be previous ankle injury, previous ankle surgery, current ankle injury, level of activity, age and sex. 2 = good comparability of groups, or confounding adjusted for in analysis. 1 = confounding small; mentioned but not adjusted for. 0 = large potential for confounding, or not discussed. E. Were the subjects blind to assignment status after allocation? 2 = effective action taken to blind subjects. 1 = small or moderate chance of unblinding of subjects. 0 = not possible, or not mentioned (unless double-blind), or possible but not done. F. Were the treatment providers blind to assignment status? 2 = effective action taken to blind treatment providers. 1 = small or moderate chance of unblinding of treatment providers. 0 = not possible, or not mentioned (unless double-blind), or possible but not done. G. Were care programmes, other than the trial options, identical? Examples of clinically important differences in other interventions which could act as active measures for prevention of ankle ligament injuries, or possible risk factors, were considered to be: footwear, training programmes, advice on activity, other devices. 2 = care programmes clearly identical. 1 = clear but trivial differences. 0 = not mentioned or clear and important differences in care programmes. H. Were the inclusion and exclusion criteria clearly defined? 2 = clearly defined. 1 = inadequately defined. 0 = not defined. J. Were the outcome measures used clearly defined? (by outcome measure) E.g. for ankle sprain: was the algorithm for diagnosis clearly described? 2 = clearly defined. 1 = inadequately defined. 0 = not defined. K. Were diagnostic tests used in outcome assessment clinically useful? (by outcome) 2 = optimal. 1 = adequate. 0 = not defined, not adequate. For ankle sprain: 2 = use of radiological tests/ magnetic resonance imaging (MRI). 1 = use of talar tilt/anterior draw; application of specified clinical algorithm. 0 = observation only or not defined. L. Was the duration of surveillance clinically appropriate, with active and systematic follow-up? (by outcome measure) 2 = optimal (including active and systematic follow-up). 1 = adequate (including active and systematic follow-up). 0 = not defined, not adequate (and not active and systematic follow-up). For ankle sprain in prevention trials (see Description of studies): 2 = observation is extended beyond the period of intervention (including active and systematic follow-up). 1 = observation is confined to the period of intervention (including active and systematic follow-up). 0 = not stated/unclear (not active and systematic follow-up). For ankle sprain in rehabilitation trials (see Description of studies): 2 = one year and above (including active and systematic followup). 1 = three months up to one year (including active and systematic follow-up). 0 = not stated/unclear, under three months (and not active and systematic follow-up). Methods used to collect data from included trials: In the first version of the review, data were independently extracted by at least four review authors using a pre-piloted data extraction tool. Disagreement was resolved by discussion. In this update, data were independently extracted by at least two review authors. All reviewers were asked to check the data entry in RevMan by HH for the trials they were allocated. Where we were unable to include a trial because of lack of usable data an attempt was made to contact the trialists. Methods to synthesise data: Where appropriate, the results of comparable studies were pooled using fixed and random effects models; the individual and pooled statistics are reported as relative risks (RR) with 95% confidence 5

9 intervals (95% CI). The change from Peto odds ratios (OR), presented in the original version of this review, to relative risks was in accordance with the revised statistical policy, announced in March 2000, of the Cochrane Musculoskeletal Injuries Group. Where data were subgrouped according to history of previous sprain, less weight was given to the results of individual trials by calculating 99% confidence intervals (99% CI). Interaction tests, based on Peto odds ratio results, were calculated to determine if the results for these subgroups were significantly different. Heterogeneity in pooled estimates was tested using chi-square statistics within RevMan. Where data allowed and where appropriate, a number needed to treat (NNT) statistic for prevention of ankle sprain was calculated for individual studies. Some of the trials included in the review update used cluster randomisation. When allocation is by a group of participants, such as platoon or soccer team, unit of analysis errors are likely to result from the presentation of outcome by the individual participants. The risk of injury of such individuals can not be considered independent of the cluster unit (team / platoon). Using statistical methods which assume for example that all participants chances of injury are independent ignores the possible similarity between outcomes for participants within the same team or platoon. This may underestimate standard errors and give misleadingly narrow confidence intervals, leading to the possibility of spurious positive findings (Bland 1997). Whilst we have presented the overall results of these trials where available, we have indicated these as cluster randomised trials and suggest cautious interpretation; we have not undertaken pooling or sub-group analysis. D E S C R I P T I O N O F S T U D I E S In this first update of the review (Quinn 1997), there were nine new included trials (Bensel 1986; Ekstrand 1983; Holme 1999; Pope 1998; Pope 2000; Simon 1969; Van Mechelen 1993; Wedderkopp 1999; Wester 1996), five new excluded trials (Brooks 1981; Eiff 1994; Freeman 1965; Janda 1988; Uh 2000) and two trials awaiting assessment (Milgrom 1992; Soderman 2000). Additional reports were found for two trials (Amoroso 1998; Sitler 1994) that were included in the original version of this review. A full publication plus another conference abstract was located for Amoroso 1998, previously listed as Ryan The new abstract for Sitler 1994 provided no new information. Overall, 27 randomised studies were considered for inclusion in this review. Of these, 14 studies (Amoroso 1998; Barrett 1993; Bensel 1986; Ekstrand 1983; Holme 1999; Pope 1998; Pope 2000; Simon 1969; Sitler 1994; Surve 1994; Tropp 1985; Van Mechelen 1993; Wedderkopp 1999; Wester 1996) involving a total of 8279 participants were included, 11 trials were excluded and two await further assessment. Eight of the 11 excluded studies were excluded because they contained no data on injury prevention (see Table of Excluded Studies). Of these eight trials, three (Gross 1987; Hughes 1983; Bennell 1994) were laboratory based studies which focused on biomechanical properties and postural control; two (Burks 1991; Robinson 1986) were concerned with performance loss with the use of ankle orthoses; one (Uh 2000), muscle strength in the contralateral leg; and two (Brooks 1981; Freeman 1965) were treatment trials with no re-injury data. Also excluded was Garrick 1973, which tested the use of high or low top shoes, with tape, J-Flex (elasticated wrap) or no tape. However, combined results for injury outcomes of this trial were presented for the two-year study which was randomised for only the second year; in the first year the participants self-selected treatment. Data for this second year were sought from the trialists without success. The evaluation of quick-release bases in softball by Janda 1988 was judged to be outside the scope of this review and thus this trial was also excluded. A treatment trial (Eiff 1994) comparing early mobilisation with mobilisation was excluded for similar reasons. Further details are required for one trial (Soderman 2000) presently only available as a conference abstract and thus placed in studies awaiting assessment. A request for further information has been sent for the other study (Milgrom 1992) awaiting assessment; this was prompted by the recent discovery of an abstract giving details of ankle sprains that were not referred to in the full report of the trial. Individual details of 14 included studies are presented in the Characteristics of Included Studies Table. These studies fall into two main categories. Namely, prevention trials where participants are not undergoing treatment for a current acute ankle sprain, and rehabilitation trials where participants undergo additional preventive interventions such as balance training during rehabilitation for an acute ankle sprain. All of the prevention studies involved active, predominantly young, adults mainly participating in organised activities. Participants were restricted to males in seven studies and were probably all male in Simon 1969; two studies (Amoroso 1998; Barrett 1993) included a small proportion of female participants (4% and 8% respectively), whereas all participants were female in two trials (Bensel 1986; Wedderkopp 1999). Where reported, the mean ages of trial participants ranged from 19 to 24 years old; however the female participants in one trial (Wedderkopp 1999) were between 16 and 18 years old, and it is likely that the civil servants in another trial (Van Mechelen 1993) formed an older population. The proportion of participants with history of ankle sprain varied from one tenth (Sitler 1994; Tropp 1985) to around one half (Amoroso 1998; Surve 1994). Two studies (Pope 1998; Pope 2000) excluded people with a history of significant injury; it is not clear if this included minor ankle sprains. One study (Bensel 1986) excluded those with ankle surgery within the previous six months and another (Simon 1969) excluded those with a history of chronic ankle instability. 6

10 Sporting activity varied widely. Interventions were applied in field/ court sports such as soccer (Ekstrand 1983; Surve 1994; Tropp 1985), basketball (Barrett 1993; Sitler 1994), American football (Simon 1969) and European handball (Wedderkopp 1999). Other interventions were applied for high-risk training such as parachute jumping (Amoroso 1998) and army basic training (Bensel 1986; Pope 1998; Pope 2000). All of these were regarded as high-risk activities for ankle sprains. Whilst ankle sprains may occur during recreational running (Van Mechelen 1993), soft-tissue knee injuries are generally more common. The two rehabilitation trials (Holme 1999; Wester 1996) involved 153 patients with an acute ankle sprain. No restriction on sprain severity was placed in one (Holme 1999), whereas only patients with grade II sprains were included in the other (Wester 1996). All patients had previously stable ankles and were active in sports before their injury. Both trials only provided patient characteristics of those that were successfully followed up; these showed approximately twice as many males as females and mean ages between 25 and 27 years. Four of the included studies compared use of an ankle brace/orthosis with control. One trial (Amoroso 1998) looked at outside-theboot bracing, while three others (Sitler 1994; Surve 1994; Tropp 1985) investigated semi-rigid ankle orthoses. One trial (Barrett 1993) compared use of high-top shoes lined with inflatable air chambers, with high-top shoes and with low-top shoes. Another trial (Simon 1969) compared adhesive taping with a cloth wrap. These interventions were applied for both practice sessions and games for four trials (Simon 1969; Sitler 1994; Surve 1994; Tropp 1985), but only for games in the remaining trial (Barrett 1993). One trial (Tropp 1985) also compared ankle disk training with control in participants with previous ankle problems. Ankle disk training as well as warm-up exercises via functional activities for specific muscle groups were compared with control in another trial (Wedderkopp 1999). The inclusion of muscle stretching during warm-up before physical training sessions was tested in two trials (Pope 1998; Pope 2000). Calf muscle stretching was compared with arm muscle stretching in the earlier study (Pope 1998); whereas the stretching of six leg muscles, including the calf muscles, was compared with control in the later one (Pope 2000). Two types of cushioning inserts (insoles) were compared with standard inserts in standard US army boots in one study (Bensel 1986). Another study (Van Mechelen 1993) compared a health education intervention, aimed at injury awareness and with instructions for appropriate preparation (warm-up) before running, activities after running (cool-down) and regular stretching exercises, with control. A multi-component tailored prophylactic programme including the correction of training errors, information about risk avoidance and, for those with ankle instability or previous ankle injury, ankle taping was compared with control in another study (Ekstrand 1983). The programme of interventions applied in Ekstrand 1983 also included controlled rehabilitation for those with lower limb injuries. All of the participants in two studies (Holme 1999; Wester 1996) were being treated for acute ankle sprains. One study (Holme 1999) compared supervised physiotherapy with an emphasis on balance training, using a balance board, with control; all patients received standard information about early ankle mobilisation. A 12-week wobble board (ankle disk) training programme was compared with the usual treatment in the other treatment study (Wester 1996). M E T H O D O L O G I C A L Q U A L I T Y Overall, the methodological quality of the 14 included papers was rated as poor to moderate; the difficulties in blinding of trial participants and care providers to most trial interventions meant that the top score was difficult to attain. Of a total possible quality score of 22, the range of overall scores was 3 to 13 with a mean score of 8.1. Table of the individual and overall methodological quality scores A B C D E F G H J K L Total Trial Amoroso Barrett Bensel Ekstrand Holme Pope Pope Simon Sitler Surve Tropp Van Mechelen Wedderkopp Wester 1996 Criteria J, K and L relate to the primary outcome of ankle sprain. Ten of the 14 studies provided some details of the method of randomisation (see under Methods in the Characteristics of Included Studies Table). Treatment allocation was clearly not concealed (item A) in one trial (Simon 1969). None of the other trial reports provided sufficient information to determine whether treatment allocation was adequately concealed. However, allocation concealment was confirmed upon the receipt of additional information for one trial (Ekstrand 1983). Teams or groups of people rather than individuals were randomised in four trials (Ekstrand 1983: soccer teams; Pope 1998 and Pope 2000: platoons; Wedderkopp 1999: handball teams). Although previously we indicated that randomisation was by soccer team in Tropp 1985, this may not have been the case and in this update we have provisionally accepted the 7

11 statement: men were allocated at random. The unit of analysis issue associated with cluster randomisation is expanded on below. Intention to treat analysis (item B) was considered very likely in one trial (Pope 2000) where monitoring continued for those participants who were reassigned to a later platoon (backsquadded) or withdrew. Loss to follow-up including post-randomisation exclusions was also reported in seven other trials (Amoroso 1998; Barrett 1993; Holme 1999; Pope 1998; Tropp 1985; Van Mechelen 1993; Wester 1996). Thirty-two participants were excluded in Amoroso 1998 for non-fulfilment of study requirements; none were injured. In Barrett 1993, nine participants were cancelled from the study for non-compliance, and a further 44 droppedout or were excluded by the end of the two-month trial period. Twenty five participants (27%) were lost from follow-up in Holme 1999; this high proportion (27%) and the failure to provide baseline characteristics for all randomised patients resulted in a zero score for this trial. Unlike in their later trial, Pope (Pope 1998) did not monitor the outcomes after discharge, backsquadding or withdrawal of 162 recruits. In Tropp 1985, baseline data were available for 439/450 participants, 11 having been excluded (nine from the control group; two from the ankle disc training group). Further intention to treat problems in this study (Tropp 1985) resulted from the omission of data from analyses for the 64 non-compliers out of the 124 who were allocated an ankle orthosis. Ninety-four runners (22%) were lost to follow-up in Van Mechelen Three of the 13 patients lost to follow-up in Wester 1996 were inappropriately excluded from the analyses; since the numbers initially allocated to each group were not provided, a zero score was given. Three trials (Bensel 1986; Sitler 1994; Surve 1994) did not report loss to follow up but all participants seemed to have been included in the analysis. Three trials failed to provide information for either the numbers of participants randomised (Simon 1969) or the numbers included in the final analyses (Ekstrand 1983; Wedderkopp 1999); all scored zero. Thus, serious intention to treat problems were likely in some trials as well as attrition/exclusion bias. None of the four trials which employed cluster randomisation made explicit reference to the associated unit of analysis problems or carried out appropriate analyses. As explained in methods, unit of analysis errors resulting from the analysis of the results of individual rather than the cluster unit (platoons, sports teams) can give misleadingly narrow confidence intervals. The use of interventions directed at the individual participants and the likely comparability in the characteristics of the clusters may have reduced the clustering effect in these trials, but no information was available to confirm this. Lastly, Pope s two trials (Pope 1998; Pope 2000) are still cluster randomised trials irrespective of preliminary quasi-randomised allocation of recruits into platoons; however, this procedure could have enhanced the independent nature of the individuals within the platoons. For trials involving team sports, the assessment of outcomes appeared to be only by coach or trainer at the time of injury in two studies (Barrett 1993; Tropp 1985), but further clinical examination occurred in three trials (Ekstrand 1983: orthopaedic surgeon; Surve 1994; no information; Wedderkopp 1999; physician). Later diagnosis of injury was by an orthopaedic surgeon in one study (Sitler 1994). No information was provided for another study (Simon 1969). Assessment of outcome was more varied in the rest of the prevention trials. Later diagnosis of injury away from the drop zone was by an orthopaedic surgeon in one study (Amoroso 1998). Regular examination at set times by medical personnel as well as sick-call data and a compliance and comfort questionnaire were used in another (Bensel 1986). Only non-trivial injuries reported to medical assistants in the two Pope trials (Pope 1998; Pope 2000) were forwarded for assessment by the regimental medical officer, and / or two researchers (Pope 1998). Running diaries were kept by the participants in Van Mechelen 1993; notified injuries were examined by the study physician. Re-injury was assessed by telephone (Holme 1999) and in-person interviews (Wester 1996) in the two rehabilitation trials. The blinding of some assessors (item C) occurred in only two trials (Amoroso 1998; Pope 2000). It is possible that independent assessment may have occurred in other trials, particularly in Sitler Most trials provided insufficient information concerning the equivalence of the so called care programmes (consisting of other preventative measures such as training, instructions, and other protective devices) which could have been employed preferentially in one group rather than another (item G). However, the basic activities performed by the participants of prevention trials, who were mainly from organisations (military; football leagues) which would impose some basic uniformity, appears similar. Thus we are fairly confident that the performance bias inherent within assessment of care programme equivalence in the trials was likely to be small. However, since this could not be accurately determined a score of zero was given. Three trials, all involving army training, provided confirmatory information of comparability (Bensel 1986; Pope 1998; Pope 2000). Some evidence of comparability was also provided in the two rehabilitation trials (Holme 1999; Wester 1996). R E S U L T S TRIAL IDENTIFICATION Twenty-seven randomised trials were considered for this updated review. Of the 14 included studies, six (Barrett 1993; Ekstrand 1983; Holme 1999; Surve 1994; Van Mechelen 1993; Wedderkopp 1999) were obtained from electronic database searches of MEDLINE, one (Pope 2000) from CINAHL and one (Wester 1996) from EMBASE. Two trials (Amoroso 1998; Sitler 1994) were located as a result of the hand-searching of journals, and four (Bensel 1986; Pope 1998; Simon 1969; Tropp 1985) from bibliographic checking of relevant journal articles and trial reports. Of 8

12 the 11 excluded trials, two (Burks 1991; Eiff 1994) were retrieved from MEDLINE, seven (Brooks 1981; Freeman 1965; Garrick 1973; Gross 1987; Hughes 1983; Janda 1988; Robinson 1986) were obtained from bibliographic checking, one (Bennell 1994) as a result of hand searching and one (Uh 2000) was provided by the CMSIG Trial Search Co-ordinator. Although no language restrictions were applied, English language publications were available for all identified trials. INCLUDED STUDIES The stated outcome measures were sought for all 14 included studies. Six studies (Amoroso 1998; Barrett 1993; Sitler 1994; Surve 1994; Van Mechelen 1993; Wedderkopp 1999) reported incidence of ankle sprains within the context of the activities undertaken but in various ways (e.g. injuries per player minute or per 1000 athlete exposures); the differences in the outcome data provided for trials testing external ankle supports prevented pooling. Some of the others (Bensel 1986; Ekstrand 1983; Pope 1998; Pope 2000; Simon 1969) provided information on the length of time or number of days of exposure to the activities rather than the actual exposure. Aside from the two rehabilitation trials (Holme 1999; Wester 1996), the exposure, whether expressed in terms of player minutes (Barrett 1993), player / running hours (Surve 1994; Van Mechelen 1993; Wedderkopp 1999), player games/ athletic exposures (Sitler 1994), match or practice days (Simon 1969), length of training programme (Bensel 1986; Pope 1998; Pope 2000), duration of soccer season (Ekstrand 1983) or parachute jumps (Amoroso 1998), was comparable for the intervention groups for each study. Thus, the results for incidence of ankle sprains can be considered as consistent with those for frequency of ankle sprains. Data for severity of ankle injury were only available in three studies (Amoroso 1998; Sitler 1994; Surve 1994); only the inversion ankle sprains were graded in Amoroso Other leg injuries, usually knee, were reported in all but two prevention studies (Simon 1969; Tropp 1985), but were not recorded in the two rehabilitation trials. Other potential complications of the interventions were not reported, nor perhaps assessed within these studies, with the exception of Bensel 1986 where various complaints such as blisters and dermatitis were recorded. Of the prevention trials, only one (Barrett 1993) reported on subjective instability (just one case). Long term instability was recorded in one study (Wester 1996). None of the studies discussed performance inhibition which was the focus of some of the excluded trials. The general acceptability of the interventions which would be relevant to their general use was rarely discussed; comfort and compliance was assessed by Bensel 1986; Sitler 1994 reported on the attitude towards bracing but only in the braced group; and Van Mechelen 1993 assessed attitudes to preventing running injuries. No study reported costs or enumerated resource use, thus detailed economic analyses are not possible. However, Amoroso (Amoroso 1998) provided an overall estimate of projected savings. In presenting the results, comparisons have been arranged broadly into three sections: preventive interventions versus control; one preventive intervention versus any other preventive intervention; and preventive interventions versus control within rehabilitation of ankle sprains. Additional analyses where data were sub-grouped by previous history of ankle sprain are presented for some of the comparisons within the first section. As stated in the Methods section, even if data had been available, sub-group analyses by previous history were not planned for trials where cluster randomisation was employed. Given that the results from both random and fixed effects methods were similar, we have opted to present these outcomes using the fixed effects method. PREVENTIVE INTERVENTIONS VERSUS CONTROL Any preventive interventions versus control The results for the number of ankle sprains incurred by intervention and control groups arranged according to intervention type in the first analysis table show a significant decrease in the ankle sprains in the ankle orthosis (relative risk (RR) 0.51, 95% confidence interval (CI) 0.38 to 0.67) and ankle disk training groups (RR 0.28, 95% CI 0.13 to 0.62). A significant decrease is also presented for a composite intervention of warm-up exercises and ankle disk training (Wedderkopp 1999) and a multi-component prophylactic programme (Ekstrand 1983). However, the results from these two trials (Ekstrand 1983; Wedderkopp 1999) are potentially misleading, given that both employed cluster randomisation. The next two analyses show the number of ankle sprains incurred by those with or without a previous history of ankle sprain. This shows a greater effect size for those with previous ankle sprain using ankle orthoses and disc training. However, it should be noted that ankle disc training (Tropp 1985) was only provided for those with a prior history of ankle sprain. The various interventions are considered separately in the following sections. External ankle support versus control A separate analysis focussing on interventions providing direct external support to the ankle shows a significant reduction in the number of ankle sprains in the intervention group (RR 0.53, 95% CI 0.40 to 0.69; 5 trials; 228 sprains; 3682 participants). Pooling using Peto odds ratios gives a similar result (OR 0.49, 95% CI 0.37 to 0.66). Whereas we could claim that there was no significant heterogeneity in the pooled results based on Peto odds ratios (p = 0.24) in the original version of this review, the heterogeneity is marginally significant (p = 0.07) when these results are viewed using relative risks. Nonetheless, we consider that the pooling of these results remains valid given the general similarity in participant characteristics, the activities undertaken in terms of risk of ankle injury and the nature of the interventions. An exploratory analysis where each of the five trials was removed in turn did not prompt any new hypothesis based on the special nature of any given trial. Data from the four studies providing outcome data grouped by previous history of ankle sprain are presented in the next analysis, 9