Ankle Injury Prevention Programs for Soccer Athletes Are Protective: A Level-I Meta-Analysis

Grimm, Nathan L. MD; Jacobs, John C. Jr. BS; Kim, Jaewhan PhD; Amendola, Annunziato MD; Shea, Kevin G. MD

Journal of Bone & Joint Surgery - American Volume:
doi: 10.2106/JBJS.15.00933
Scientific Articles

Background: Soccer has one of the highest rates of ankle injury in sports for both males and females. Several injury prevention programs have been developed to address this concern. The purposes of this study were to conduct a meta-analysis of ankle injury prevention programs for soccer players, assess the heterogeneity among the studies, and evaluate the reported effectiveness of the prevention programs.

Methods: A systematic search of the literature was conducted in PubMed (MEDLINE), Embase, CINAHL (Cumulative Index to Nursing and Allied Health), and the Cochrane Central Register of Controlled Trials (CENTRAL) database. Studies were limited to clinical investigations of injury prevention programs specific to the ankle in soccer players. Title, abstract, and full-text review were utilized to identify articles that met the inclusion criteria. The Cochrane Q test and I2 index were independently used to assess heterogeneity among the studies. Sensitivity analyses were performed to assess heterogeneity. The pooled risk difference was calculated by random-effects models with use of the DerSimonian-Laird method. Publication bias was assessed with a funnel plot and Egger weighted regression technique.

Results: Ten studies met the inclusion criteria as randomized controlled trials. A total of 4,121 female and male soccer athletes were analyzed for ankle injuries. Significant heterogeneity was found among studies of ankle injury prevention (p = 0.002), with an I2 index of 65.2%. For studies of ankle injury prevention programs, the risk ratio was 0.60 (95% confidence interval, 0.40 to 0.92) and a significant reduction in the risk of ankle injury was found in the prevention group (p = 0.002). No evidence of publication bias was found among the included studies.

Conclusions: This meta-analysis of studies regarding ankle injury prevention programs identified a significant reduction in the risk of ankle injury. Future high-quality research designs with a low risk of bias are necessary to further evaluate the effectiveness of specific exercises and the optimal timing and age at intervention for the prevention of ankle injuries in the athletic soccer player.

Level of Evidence: Therapeutic Level I. See Instructions for Authors for a complete description of levels of evidence.

Author Information

1Department of Orthopaedic Surgery, Duke University Medical Center, Durham, North Carolina

2Department of Orthopaedic Surgery (J.C.J. Jr.), and Division of Public Health, Study Design and Biostatistics Center (J.K.), University of Utah School of Medicine, Salt Lake City, Utah

3St. Luke’s Intermountain Orthopaedics, Boise, Idaho

E-mail address for N.L. Grimm:

Article Outline

With an estimated 265 million players worldwide as of 2006, soccer is the most popular sport in the world1. The sport has grown immensely in the United States over the past 3 decades, due in large part to increased participation among youth and female competitors, and participation continues to grow annually2. The ankle is one of the most frequent sites of injury in soccer players, with ankle sprain being the most common specific injury in athletes of both sexes at high school, college, and professional levels of play3-6. In fact, soccer has one of the highest rates of ankle injury in sports for both males and females. According to the National Center for Injury Prevention and Control of the Centers for Disease Control and Prevention, the estimated total lifetime medical cost, in the U.S. in 2010, for the treatment of a sprain or strain of the lower leg or ankle was $672,972,000 for females between 10 and 39 years old and $623,549,000 for males between 10 and 39 years old7. For these reasons, ankle injury prevention has become an increasingly important concern for athletic and medical communities alike.

Many different injury prevention programs have been used in an effort to reduce injuries of the lower extremity. Clinical trials have attempted to demonstrate the efficacy of these various intervention programs; however, studies of this nature are time-consuming, expensive, and difficult to conduct. These limitations can lead to study design and methodology flaws that subsequently increase the risk for bias and reduce the quality of evidence provided by the study.

The purposes of this systematic review and meta-analysis were to identify Level-I studies of ankle injury prevention programs for soccer players, assess the internal validity of these studies, and evaluate the quantitative pooled effectiveness of these prevention programs.

Back to Top | Article Outline

Materials and Methods

Criteria for Selecting Studies
Types of Studies

Inclusion and exclusion criteria were established a priori. Only randomized controlled trials (RCTs) of injury prevention programs with Level-I evidence, as defined by the Cochrane Handbook8, were included in this review. The criteria described by the Oxford Centre for Evidence-Based Medicine (http:// were used for level-of-evidence assessments. Therefore, Level-II (prospective, nonrandomized studies), Level-III (retrospective studies), Level-IV (case series), and Level-V (expert opinion) publications were not included in the analysis. Articles not published in English were also excluded from the study.

Back to Top | Article Outline
Types of Participants

Of the studies included in the meta-analysis, only those involving soccer players were included in this analysis. Sex, athletic skill level, or age group did not influence the inclusion of a study in the analysis.

Back to Top | Article Outline
Types of Interventions and Comparison Groups

Only studies that utilized interventions to prevent ankle injuries were included. The interventions utilized included neuromuscular, proprioceptive, strengthening, and stretching exercises. If a study used an exogenous modality as a means of prevention (e.g., bracing or taping), it was excluded. One study compared multiple treatment groups including proprioceptive exercises, strengthening exercises, and orthosis use; however, only the data from the proprioception exercises group and the strengthening exercises group were extracted for inclusion in the meta-analysis, while the orthosis group was excluded9.

Back to Top | Article Outline
Outcome Measures

Studies that had “ankle injury” as an outcome measure were eligible for inclusion. Ankle injury was defined as any acute or overuse injury or one that was due to either contact or noncontact trauma, including both ligamentous and nonligamentous injury, occurring to the ankle. Acute injuries were those with sudden onset with obvious trauma, and overuse injuries were those occurring with an insidious onset10.

Back to Top | Article Outline
Search Method and Strategy

Four major medical databases were searched from inception through February 15, 2015: MEDLINE, Embase, CINAHL (Cumulative Index to Nursing and Allied Health), and Cochrane Central Register of Controlled Trials (CENTRAL) database. A medical library search strategist was consulted to develop a sensitive and comprehensive search strategy, as has been done in previous studies (see acknowledgment)10. Briefly, the following search criteria were used: (injuries OR injury OR injur*) AND (prevention OR prevented OR prevent OR prevents OR prevent*) AND (ankle) AND (soccer OR football).

Supplemental bibliographic reference searches of articles to identify potentially missed, relevant studies were also conducted (a detailed list of the search strategy is available in the Appendix).

Back to Top | Article Outline
Data Collection and Analysis
Study Selection Procedure

The study selection process was conducted independently by 2 reviewers (N.L.G. and J.C.J. Jr.). If a study passed an initial screening on the basis of the title and abstract review, more information was needed to determine if the trial met inclusion criteria; therefore, the full text was retrieved and reviewed in detail. If a disagreement arose between the 2 reviewers, a third reviewer (K.G.S.) helped to establish a group consensus agreement.

Back to Top | Article Outline
Data Extraction

Data were extracted from studies that met the above-mentioned inclusion criteria independently by the 2 reviewers (N.L.G. and J.C.J. Jr.). From each study, the following data were extracted: journal of publication, title, author(s), publication year, subject sex, subject age, number of subjects in intervention and treatment arms, type of intervention(s), characteristics of intervention(s), study design, follow-up time, and outcome data. Additionally, the following outcome data elements were also extracted from each study: type of injury, frequency of injury occurrence, duration of follow-up, diagnostic methods for outcomes of interest, number of dropouts, and compliance rate. Authors of a study were contacted for clarification if any of the data elements of interest were missing or unclear.

Back to Top | Article Outline
Risk-of-Bias Assessment

The internal validity of all RCTs and cluster-RCTs that met the inclusion criteria were assessed by 2 independent reviewers (N.L.G. and J.C.J. Jr.) using the assessment algorithm described by van Tulder et al.11. A risk-of-bias score was assigned to each study on the basis of its assessed internal validity. The van Tulder scale is a critical appraisal tool for the assessment of RCTs developed by the Cochrane Collaboration Back Review Group11 and has been used in previous meta-analyses12. Many methodological criteria that have empirically been shown to create an exaggeration of treatment effects are included in this critical appraisal tool13-19.

As with the data extraction mentioned above, if any of the specific criteria of the van Tulder scale were not explicitly stated in the text of a study, the study authors were contacted for clarification. A third reviewer (K.G.S.) was consulted if disagreements arose. This process of assessment is consistent with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement for conducting systematic reviews20.

Back to Top | Article Outline
Meta-Analysis and Statistical Procedures

The ankle injury rates concerning intervention and control groups from each study were evaluated and summarized in tabular form. Relative risk (RR) was used to estimate the intervention effect on the prevalence of ankle injuries, by comparing the ankle injury rate in the intervention group (numerator) with the rate in the comparison group (denominator).

The analysis of publication bias was performed with STATA statistical software (release 13; StataCorp) using the Harbord test, Peters test, and Begg test, which are general approaches to test publication bias in a meta-analysis. The Egger test has traditionally been widely used, but has been shown to be biased for binary outcomes. The Harbord test is more appropriate as a modified test for small-study effects in a meta-analysis of RCTs8,21,22. The Harbord and Peters tests are regression-based tests, which are parametric. However, the Begg test is based on a rank correlation method, which is nonparametric. The analysis of publication bias was presented qualitatively using a funnel plot derivation.

With the use of STATA, an inverse variance method was employed to calculate the weight for each study included, with random-effects meta-analyses with the DerSimonian-Laird method used to estimate the between-study variance23. The Cochrane Q test, with a p value of 0.10 considered significant, and the Higgins and Thompson I2 index were independently used to assess heterogeneity among the studies8,24.

Back to Top | Article Outline


Search Findings

Inclusion and exclusion criteria identified 9 RCTs of ankle injury prevention programs for soccer players within the medical literature databases that were searched. However, 1 study assessed 2 different prevention strategies and therefore is treated as 2 separate studies for this meta-analysis, which can be considered to have 10 RCTs9. A total of 1,634 articles were screened on the basis of the title and/or abstract, and of those studies, 50 full-text articles were retrieved and reviewed. This review resulted in 10 RCTs that met the inclusion criteria (Fig. 1). The primary reasons for not including a study in this review were a nonrandomized study design, the lack of a comparison group, and study subjects who were not soccer-playing athletes.

Back to Top | Article Outline
Search Characteristics

A summary of all studies included in the systematic review and meta-analysis is provided in Table I. A total of 4,121 athletes in the 10 RCTs were analyzed. Six studies included only male subjects9,25-28, 3 studies included only female subjects29-31, and 1 study included both male and female subjects32. The study design of each trial was either a prospective cluster-RCT or a prospective RCT. All studies had ankle injury as an outcome measure (Table II).

Back to Top | Article Outline
Risk of Bias

The average risk-of-bias score of the studies included was 6.4 (range, 5 to 10) of a maximum score of 11 (Table III). Authors of the 9 articles, or 10 studies with 1 article considered as 2 studies, were contacted for clarification on at least 1 of the internal validity elements collected in the van Tulder scale assessment that was not apparent or clearly stated in the manuscript. One study accurately and completely described within the text all elements of internal validity that were assessed9. The dropout rate was found to be acceptable in 6 of the 9 studies, was not reported in 2 studies, and was found to be unacceptable in 1 study (Table III). The other elements that were not adequate included avoidance of cointerventions, compliance, group similarity at baseline, concealment of treatment allocation, adequate randomization method, conducting of an intention-to-treat analysis, and subject blinding. However, subject blinding in particular is difficult if not impossible in this type of study design.

Back to Top | Article Outline
Study-Specific and Quantitative Analysis

The overall estimated RR for ankle injuries was <1, indicative of a protective effect, with 8 studies individually having an estimated RR of <1 (4 studies had a significant difference). The pooled effect size of the injury prevention interventions, using a random-effects meta-analysis technique, showed a significant protective effect against ankle injuries (RR of 0.60; 95% confidence interval [CI], 0.40 to 0.92; p = 0.002) (Fig. 2). Sensitivity analyses were conducted, and the data were analyzed with fixed-effects meta-analyses using both the Mantel-Haenszel method and an inverse-variance method. These analyses found similar results with respect to the RR, indicating the intervention is effective at preventing ankle injuries.

Back to Top | Article Outline
Publication Bias

In assessing for publication bias as described above, the results of the Harbord test identified weak evidence for small-study effects (p = 0.06). However, both the Peters test and the Begg test were also performed and found no evidence of publication bias (p = 0.165 and p = 0.07, respectively). A contour-enhanced funnel plot was created to qualitatively represent this result (Fig. 3).

Back to Top | Article Outline


In the sport of soccer, the ankle is one of the most frequently injured areas for both males and females33. Many injury prevention programs have been developed in an attempt to reduce the rate of ankle sprains in the athletic population. This meta-analysis was initiated to identify the highest level of evidence for injury prevention programs developed for soccer players, to assess the quality of this body of literature, and to use empirical meta-analytical approaches to assess the true effectiveness of such programs. Overall, we found a significant protective effect in support of injury prevention programs for reducing ankle injuries in this athletic population.

Previous analysis of studies that assess the efficacy of ankle injury prevention programs is limited. Valovich McLeod34 reviewed the efficacy of reducing ankle sprains in adolescent athletes using balance training programs. In that investigation, only 4 studies (2 RCTs and 2 cohort studies) were identified, and the author concluded that these programs were moderately effective in reducing ankle sprains in this population. Thacker et al.35 conducted a comprehensive systematic review of the evidence on different interventions for preventing ankle sprains. They identified several major areas of prevention, including external ankle support, risk factors, and methods to prevent ankle sprains (including training). The authors concluded that some recommendations made in injury prevention studies could not be supported by the evidence; however, they stated that certain prevention exercises such as those targeting proprioception and strengthening are likely to reduce certain injuries. Recently, Schiftan et al.36 conducted a systematic review and meta-analysis that assessed the efficacy of proprioceptive training in preventing ankle sprains in athletes of multiple sports. An overall significant reduction in ankle injuries was found with use of proprioceptive training. The results of the study by Schiftan et al. are comparable with those of our current study, which found a significant protective effect of ankle injury prevention programs for soccer players. However, we are unable to comment on the role of individual elements of the injury prevention programs, the role of sex, or player skill level in our analysis. Our study was designed a priori to have an appropriate statistical power to detect differences in the rate of ankle injuries between athletes managed with injury prevention programs and those who were not. Subgroup analyses on these individual covariates would not have sufficient power for detection. Furthermore, several studies have advised against the use of subgroup analyses as the multiplicity of these analyses leads to spurious results and are a risk for a false-positive result that may mislead the reader37. Additionally, many authors have raised concern for outcome reporting bias in publications when multiple subgroups are analyzed but are not the primary outcome of a study design38. Therefore, these analyses were not performed in our study.

The current study had several limitations. In regard to the types of studies analyzed in this meta-analysis, there was minimal methodological heterogeneity; however, one can never control all elements of clinical heterogeneity. We attempted to control for this by limiting our cohort to a single sport—soccer. There was also a wide variation among age groups studied, ranging from children to adults. Furthermore, the competition level also varied from recreational to professional. One variable that we did try to control for was study design. Prospective cohort (Level-II) studies also are available in the literature, but they were excluded in this analysis because of the lower level of evidence. Although our statistical analysis suggested no evidence of publication bias, there did appear to be qualitative asymmetry of the funnel plot. There are various explanations for this asymmetry, including the degree of statistical heterogeneity, the number of studies included in the meta-analysis, and small-study effects39. The data abstractors for the current study were not blinded to the studies they were reviewing. However, this is not a standard advised by the PRISMA protocol, and there is no evidence that this extra step affects outcomes20.

In conclusion, this meta-analysis of ankle injury prevention programs in soccer players found an overall, significant protective effect for reducing ankle injuries in soccer athletes. To our knowledge, this is the first meta-analysis of Level-I RCTs of ankle injury prevention programs for soccer players. The findings of this meta-analysis support the use of these interventions to prevent ankle injuries in these athletes.

The clinical trials that assess the efficacy of injury prevention programs are difficult to conduct, and the researchers who do so are to be congratulated. We encourage further research in this area, and we urge future researchers to follow the guidelines established in the CONSORT (Consolidated Standards of Reporting Trials) statement.

Back to Top | Article Outline

Appendix Cited Here...

A summary of the search strategy for studies on ankle injury prevention programs for soccer players is available with the online version of this article as a data supplement at

NOTE: The authors thank Mary McFarland of the Eccles Health Science Medical Library for assisting with the development of a sensitive search strategy for this study.

Investigation performed at the Department of Orthopaedic Surgery, Duke University Medical Center, Durham, North Carolina

Disclosure: No external sources of funding were obtained for this study. The Disclosure of Potential Conflicts of Interest forms are provided with the online version of the article.

Back to Top | Article Outline


1. Doherty C, Delahunt E, Caulfield B, Hertel J, Ryan J, Bleakley C. The incidence and prevalence of ankle sprain injury: a systematic review and meta-analysis of prospective epidemiological studies. Sports Med. 2014 ;44(1):123–40.
2. American Academy of Pediatrics. Committee on Sports Medicine and Fitness. Injuries in youth soccer: a subject review. Pediatrics. 2000 ;105(3 Pt 1):659–61.
3. Agel J, Evans TA, Dick R, Putukian M, Marshall SW. Descriptive epidemiology of collegiate men’s soccer injuries: National Collegiate Athletic Association Injury Surveillance System, 1988-1989 through 2002-2003. J Athl Train. 2007 ;42(2):270–7.
4. Dick R, Putukian M, Agel J, Evans TA, Marshall SW. Descriptive epidemiology of collegiate women’s soccer injuries: National Collegiate Athletic Association Injury Surveillance System, 1988-1989 through 2002-2003. J Athl Train. 2007 ;42(2):278–85.
5. Gaulrapp H, Becker A, Walther M, Hess H. Injuries in women’s soccer: a 1-year all players prospective field study of the women’s Bundesliga (German premier league). Clin J Sport Med. 2010 ;20(4):264–71.
6. Leininger RE, Knox CL, Comstock RD. Epidemiology of 1.6 million pediatric soccer-related injuries presenting to US emergency departments from 1990 to 2003. Am J Sports Med. 2007 ;35(2):288–93. Epub 2006 Nov 7.
7. Centers for Disease Control and Prevention. Data & statistics (WISQARSTM): cost of injury reports. Accessed 2016 June 13.
8. Higgins JPT, Green SE. Cochrane handbook for systematic reviews of interventions, version 5.0.2. The Cochrane Collaboration, 2009. 2009.
9. Mohammadi F. Comparison of 3 preventive methods to reduce the recurrence of ankle inversion sprains in male soccer players. Am J Sports Med. 2007 ;35(6):922–6. Epub 2007 Mar 22.
10. Grimm NL, Shea KG, Leaver RW, Aoki SK, Carey JL. Efficacy and degree of bias in knee injury prevention studies: a systematic review of RCTs. Clin Orthop Relat Res. 2013 ;471(1):308–16. Epub 2012 Sep 8.
11. van Tulder M, Furlan A, Bombardier C, Bouter L; Editorial Board of the Cochrane Collaboration Back Review Group. Updated method guidelines for systematic reviews in the Cochrane Collaboration Back Review Group. Spine (Phila Pa 1976). 2003 ;28(12):1290–9.
12. Grimm NL, Jacobs JC Jr, Kim J, Denney BS, Shea KG. Anterior cruciate ligament and knee injury prevention programs for soccer players: a systematic review and meta-analysis. Am J Sports Med. 2015 ;43(8):2049–56. Epub 2014 Dec 1.
13. Poolman RW, Struijs PA, Krips R, Sierevelt IN, Marti RK, Farrokhyar F, Bhandari M. Reporting of outcomes in orthopaedic randomized trials: does blinding of outcome assessors matter? J Bone Joint Surg Am. 2007 ;89(3):550–8.
14. Schulz KF, Altman DG, Moher D. Allocation concealment in clinical trials. JAMA. 2002 ;288(19):2406–7; author reply 2408-9.
15. Schulz KF, Chalmers I, Altman DG. The landscape and lexicon of blinding in randomized trials. Ann Intern Med. 2002 ;136(3):254–9.
16. Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials. JAMA. 1995 ;273(5):408–12.
17. Schulz KF, Grimes DA. Blinding in randomised trials: hiding who got what. Lancet. 2002 ;359(9307):696–700.
18. Schulz KF, Grimes DA. Allocation concealment in randomised trials: defending against deciphering. Lancet. 2002 ;359(9306):614–8.
19. Wood L, Egger M, Gluud LL, Schulz KF, Jüni P, Altman DG, Gluud C, Martin RM, Wood AJ, Sterne JA. Empirical evidence of bias in treatment effect estimates in controlled trials with different interventions and outcomes: meta-epidemiological study. BMJ. 2008 ;336(7644):601–5. Epub 2008 Mar 3.
20. Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. J Clin Epidemiol. 2009 ;62(10):1006–12. Epub 2009 Jul 23.
21. Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Comparison of two methods to detect publication bias in meta-analysis. JAMA. 2006 ;295(6):676–80.
22. Harbord RM, Egger M, Sterne JA. A modified test for small-study effects in meta-analyses of controlled trials with binary endpoints. Stat Med. 2006 ;25(20):3443–57.
23. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986 ;7(3):177–88.
24. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med. 2002 ;21(11):1539–58.
25. Ekstrand J, Gillquist J, Liljedahl SO. Prevention of soccer injuries. Supervision by doctor and physiotherapist. Am J Sports Med. 1983 ;11(3):116–20.
26. Engebretsen AH, Myklebust G, Holme I, Engebretsen L, Bahr R. Prevention of injuries among male soccer players: a prospective, randomized intervention study targeting players with previous injuries or reduced function. Am J Sports Med. 2008 ;36(6):1052–60. Epub 2008 Apr 3.
27. Tropp H, Askling C, Gillquist J. Prevention of ankle sprains. Am J Sports Med. 1985 ;13(4):259–62.
28. van Beijsterveldt AM, van de Port IG, Krist MR, Schmikli SL, Stubbe JH, Frederiks JE, Backx FJ. Effectiveness of an injury prevention programme for adult male amateur soccer players: a cluster-randomised controlled trial. Br J Sports Med. 2012 ;46(16):1114–8. Epub 2012 Aug 9.
29. Söderman K, Werner S, Pietilä T, Engström B, Alfredson H. Balance board training: prevention of traumatic injuries of the lower extremities in female soccer players? A prospective randomized intervention study. Knee Surg Sports Traumatol Arthrosc. 2000;8(6):356–63.
30. Soligard T, Myklebust G, Steffen K, Holme I, Silvers H, Bizzini M, Junge A, Dvorak J, Bahr R, Andersen TE. Comprehensive warm-up program to prevent injuries in young female footballers: cluster randomized controlled trial. BMJ. 2008 ;337:a2469.
31. Heidt RS Jr, Sweeterman LM, Carlonas RL, Traub JA, Tekulve FX. Avoidance of soccer injuries with preseason conditioning. Am J Sports Med. 2000 ;28(5):659–62.
32. Emery CA, Meeuwisse WH. The effectiveness of a neuromuscular prevention strategy to reduce injuries in youth soccer: a cluster-randomised controlled trial. Br J Sports Med. 2010 ;44(8):555–62.
33. Shea KG, Grimm NL, Ewing CK, Aoki SK. Youth sports anterior cruciate ligament and knee injury epidemiology: who is getting injured? In what sports? When? Clin Sports Med. 2011 ;30(4):691–706.
34. Valovich McLeod TC. The effectiveness of balance training programs on reducing the incidence of ankle sprains in adolescent athletes. J Sport Rehabil. 2008 ;17(3):316–23.
35. Thacker SB, Stroup DF, Branche CM, Gilchrist J, Goodman RA, Weitman EA. The prevention of ankle sprains in sports. A systematic review of the literature. Am J Sports Med. 1999 ;27(6):753–60.
36. Schiftan GS, Ross LA, Hahne AJ. The effectiveness of proprioceptive training in preventing ankle sprains in sporting populations: a systematic review and meta-analysis. J Sci Med Sport. 2015 ;18(3):238–44. Epub 2014 Apr 26.
37. Schulz KF, Grimes DA. Multiplicity in randomised trials II: subgroup and interim analyses. Lancet. 2005 ;365(9471):1657–61.
38. Dwan K, Altman DG, Arnaiz JA, Bloom J, Chan AW, Cronin E, Decullier E, Easterbrook PJ, Von Elm E, Gamble C, Ghersi D, Ioannidis JP, Simes J, Williamson PR. Systematic review of the empirical evidence of study publication bias and outcome reporting bias. PLoS One. 2008;3(8):e3081.
39. Mavridis D, Salanti G. How to assess publication bias: funnel plot, trim-and-fill method and selection models. Evid Based Ment Health. 2014 ;17(1):30.
Copyright 2016 by The Journal of Bone and Joint Surgery, Incorporated