Rehabilitation and Improvement of Health-Related Quality-of-Life Detriments in Individuals With Chronic Ankle Instability: A Meta-Analysis

Search Strategy

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to perform a systematic search to locate studies in which investigators evaluated the effect of a conservative rehabilitation intervention on PROs in individuals with CAI.¹² We searched PubMed and EBSCO Host (CINAHL, MEDLINE, SPORTDiscus) from their inception through January 27, 2016. Electronic databases were searched using combinations of key words related to the research question (Table 1). The Boolean operators OR, AND, and NOT were used to combine search terms, and the search was limited to human studies and manuscripts written in English. The reference lists of articles screened during the systematic search were hand searched for additional publications. The constructed Boolean phrase, systematic search, and hand search were completed by 2 investigators (C.J.P., M.C.H.).

Table 1.  Search Strategy

View Fullscreen

Selection Criteria

The eligibility of articles identified by the systematic search was determined by 2 authors (C.J.P., M.C.H.) using specific inclusion and exclusion criteria. Initially, the titles and abstracts of all articles were screened for eligibility. When eligibility could not be determined during the initial screen, the full text of the manuscript was examined.

Methodologic Quality

The Physiotherapy Evidence Database (PEDro) scale¹³ was used to assess the methodologic quality of the included studies. This scale has demonstrated acceptable reliability (intraclass correlation coefficient = 0.68).¹³ The PEDro is a 10-item scale designed to determine the methodologic quality of randomized control trials by assessing their internal validity. Each item is scored as yes or no. Studies that scored equal to or greater than 60% of the PEDro items as yes were deemed high-quality evidence.¹⁴ Included studies were initially scored independently by 2 reviewers (C.J.P., M.C.H.). After independent scoring, the 2 reviewers met to resolve any disagreements. If a disagreement could not be resolved, a third reviewer (J.M.H.) was consulted. The percentage of agreement among the reviewers was calculated for each PEDro item.

Data Extraction

During the initial review of the included studies, 2 independent reviewers (C.J.P., M.C.H.) extracted data, including study aims, study designs, study quality, inclusion criteria, participant characteristics, clinician details, intervention procedures, outcome assessments, statistical techniques, conclusions, and relevant methodologic limitations. Discrepancies in interpretation were resolved by discussion until a consensus was achieved. If a consensus could not be achieved, a third reviewer (J.M.H.) was consulted.

The primary outcome of interest for this systematic review was PRO scores. Only preintervention and postintervention PRO scores were extracted for intervention groups. During the extraction of PRO scores, the Foot and Ankle Disability Index (FADI)–Sport subscale identified in the literature was reported as the Foot and Ankle Ability Measure (FAAM)–Sport in this review because the 2 instruments contained the same questions.

To further classify the included studies, we created a moderator variable to examine specific types of rehabilitation reported in the literature. The moderator variable rehabilitation type refers to the nature of interventions that were conducted. Four levels were coded for rehabilitation type: balance training, manual therapy, strength training, and combination. Balance-training rehabilitation was used to describe studies that included rehabilitation protocols involving tasks that challenged the participants' ability to maintain static or dynamic balance. Manual-therapy studies investigated interventions in which hands-on manual-therapy techniques (eg, mobilizations, massage) were applied to the lower extremity. Strength-training studies examined interventions primarily designed to strengthen the lower extremity. Finally, studies classified as combination used a mixture of rehabilitation approaches in which participants underwent conservative rehabilitation that consisted of 2 or more of the interventions. Studies included in this review may have incorporated multiple intervention groups within the study; therefore, the outcomes were categorized according to the different rehabilitation types.

Statistical Analysis

Separate meta-analyses were performed for the overall effect and each rehabilitation type. For each meta-analysis, we used a random-effects model in which individual study measures were pooled using bias-corrected Hedges g ESs and 95% confidence intervals (CIs) to determine the magnitude of change in patient-oriented outcomes in individuals with CAI from preintervention to postintervention. The Hedges g ES is a unitless measure that is corrected for sample size to represent an effect that exists on a parametric distribution.¹⁵ A positive ES indicated that PRO scores improved from preintervention to postintervention. In most studies, investigators used both the FAAM/FADI activities of daily living (ADL) subscale and the FAAM–Sport. The average FAAM/FADI–ADL ES (0.88) and the average FAAM–Sport ES (0.88) of the included studies were very similar; therefore, when both instruments were reported, the values were pooled for analysis to reduce sample-size inflation. For studies in which multiple rehabilitation types were examined, each group was treated independently in the analyses. All meta-analysis procedures were performed in Comprehensive Meta-Analysis software (version 2.0; BioStat, Englewood, NJ). Effect sizes were interpreted as weak (≤0.40), moderate (0.41−0.69), or strong (≥0.70).¹⁵ The α level was set a priori at .05. Further data analysis was performed via a qualitative assessment of ES estimates between rehabilitation types and determining if CIs crossed zero.

Assessment of Publication Bias

Assessment of the robustness of the observed overall effect on PRO change was completed using the Orwin fail-safe N test. This test determines the number of studies with trivial findings that would be needed to nullify the pooled ES of the included studies. A funnel plot of all included comparisons was generated to assess the likelihood of publication bias. To further assess publication bias, we also used the trim-and-fill method of imputing missing studies.

Sensitivity Analysis

To assess the sensitivity of the results to the influence of a single participant group, we completed a 1-study-removed analysis. This method repeats the meta-analysis multiple times, with 1 participant group removed each time to determine if the observed effect was affected. A 1-study-removed analysis was completed for the overall effect and each rehabilitation type.

Level of Evidence

We used the Strength-of-Recommendation Taxonomy (SORT) to assess the level of evidence and provide an overall grade of recommendation.¹⁶ Level 1 evidence was considered good-quality, patient-oriented evidence; level 2 evidence was considered limited-quality, patient-oriented evidence; and level 3 was considered other evidence.¹⁶ The SORT guidelines for strength of recommendation consider grade A to be consistent, good-quality, patient-oriented evidence; grade B, inconsistent or limited-quality, patient-oriented evidence; and grade C, consensus, disease-oriented evidence; and so on.¹⁶ For this systematic review, 3 or more high-quality studies were required to provide grade A evidence.

Literature Search

Using the initial search strategy, we identified 446 potential articles (Figure 1). A hand search of references identified an additional 3 potential articles. Of the 449 articles screened, 399 were excluded based on the title or abstract, and 36 were excluded based on relevance or inadequate data reporting. Fifteen articles met our inclusion criteria and provided 24 participant groups for analysis.¹⁷⁻³¹ One participant group²³ was included after we hand measured the mean and standard deviation from a figure. The 15 articles were classified into the following categories based on rehabilitation type: balance training,^{20,21,23,25,28,30} manual therapy,^{17,19,22,26,31} and combination.^{18,24,25,27,29,30} Strength training was not included as a rehabilitation type because the authors of only 1 study²⁷ investigated the isolated effects of a strength-training protocol. Several participant groups were included in the analysis for each moderator variable: balance training (n = 8), manual therapy (n = 7), and combination (n = 7). A methodologic summary of the included studies is presented in Table 2.

View Figure

• Download as Powerpoint
• Download Figure

Table 2.  Methodologic Summary of the Included Studies Continued on Next Page

View Fullscreen

Authors of the included articles had evaluated HRQL using only region-specific PROs that assessed the physical domain of HRQL. Specifically, 14 combined FAAM–ADL and FAAM–Sport, 3 FAAM–Sport, 3 combined FADI–ADL and FAAM–Sport, and 4 Cumberland Ankle Instability Tool (CAIT) participant groups were included in the analysis.

Methodologic Quality

The 2 reviewers initially agreed on 141 of 150 (94%) PEDro items. All but 1 disagreement was resolved through discussion between the reviewers. Overall, quality scores of the included studies ranged from 10% to 80%, with a median of 50%. A total of 7 studies^{17−20,25,29,31} were high quality, and 8 studies^{21−24,26−28,30} were low quality. Six studies were classified as level 1 evidence,^{17,19,20,25,29,31} and 9 as level 2 evidence.^{18,21–24,26–28,30} The individual items, quality scores, and levels of evidence can be found in Table 3.

Table 3.  The Physiotherapy Evidence Database (PEDro) Individual Items and Quality Index Scores for the Included Articles

View Fullscreen

Data Synthesis

Overall Summary Effect

Across all the included studies and subgroups, the overall effect of the preintervention-to-postintervention comparisons was 1.11 (95% CI = 0.76, 1.46; P < .001), indicating that individuals with CAI demonstrated strong improvements in HRQL after rehabilitation. The individual ESs and the cumulative effects are shown in Figure 2 and Table 4, respectively.

View Figure

• Download as Powerpoint
• Download Figure

Table 4.  Effect Size and 95% Confidence Intervals of Participant Groups

View Fullscreen

Summary Effects for Rehabilitation Type

We observed no differences among the 3 levels of rehabilitation type (Q = 0.086, P = .96; Figure 3). Studies labeled as balance training demonstrated a strong effect with a CI that did not encompass zero (ES = 1.22; 95% CI = 0.79, 1.65; P < .001). Studies labeled as manual therapy demonstrated a strong effect (ES = 1.10; 95% CI = 0.09, 2.11; P = .03). Lastly, studies labeled as combined demonstrated a strong effect with a CI that did not encompass zero (ES = 1.14; 95% CI = 0.67, 1.60; P < .001).

View Figure

• Download as Powerpoint
• Download Figure

Publication Bias

The likelihood of publication bias was assessed using a funnel plot (Figure 4). Based on the relative symmetry and even distribution of the studies within the funnel plot, it is unlikely that publication bias played a role in the results of the meta-analyses. This was further indicated via the trim-and-fill method, as no studies were inputted or removed. The results of the Orwin fail-safe N test indicated that a range of 214 to 451 additional studies, based on a trivial effect range of the Hedges g from 0.10 to 0.05, would be needed to nullify the overall summary effect. Based on the aforementioned results, publication bias was highly unlikely.

View Figure

• Download as Powerpoint
• Download Figure

Sensitivity Analysis

Level of Evidence

Overall, grade A evidence supported region-specific, physical domain HRQL improvements in individuals with CAI after conservative rehabilitation. This recommendation was based on consistent findings from 6 level 1^{17,19,20,25,29,31} and 9 level 2^{18,21−24,26−28,30} studies. For balance training, grade B evidence indicated that this rehabilitation type improved HRQL based on consistent findings from 2 level 1^20,25 and 5 level 2^{18,21,23,28,30} studies. For manual therapy, grade B evidence supported its effectiveness in improving HRQL based on inconsistent findings from 3 level 1^17,19,31 and 2 level 2^22,26 studies. For combination interventions, grade B evidence supported its use to improve HRQL based on consistent findings from 2 level 1^25,29 and 4 level 2^18,24,27,30 studies.

Balance Training

We found grade B evidence that a balance-training rehabilitation protocol effectively improved HRQL, as measured by patient-oriented outcomes, in individuals with CAI. Moderate to strong ESs (pooled = 1.22; range = 0.59−2.10) indicated improvement when comparing preintervention and postintervention outcomes. Three^25,28,30 of the 7^{18,20,21,23,25,28,30} studies were based on a program developed by McKeon et al.²⁸ These interventions were 4 weeks long and involved progressive single-limb balance and hopping tasks. The remaining balance interventions^18,20,21,23 used progressive exercises for 4 to 8 weeks. The largest ES (2.10) was demonstrated by Cruz-Díaz et al,²⁰ who also used one of the longest intervention durations (6 weeks). The lowest ESs (0.59 and 0.61) were reported in the only studies^18,21 with home-based balance-training programs. These results may imply that supervised balance-training interventions more effectively improve HRQL, as measured by PROs, than unsupervised home-based programs.

Manual Therapy

Grade B evidence indicated that a manual-therapy–focused intervention program improved patient-oriented outcomes in individuals with CAI.^{17,19,22,26,31} This finding should be interpreted with caution, however, as a single participant group substantially influenced the summary ES for this intervention type. When the Cruz-Díaz et al¹⁹ study was removed (ES = 5.41), the summary ES was moderate (ES = 0.44). Therefore, it is more likely that PROs moderately improve after isolated manual-therapy interventions.

The manual-therapy techniques investigated included talocrural anterior-to-posterior Maitland grade III joint mobilizations,^26,31 Mulligan talocrural mobilizations with movement,^19,22 tibiofibular manipulations,¹⁷ and plantar massage.³¹ Tibiofibular manipulation demonstrated the weakest ESs (range = −0.03 to 0.23)¹⁷ compared with the 4 studies^19,22,26,31 in which researchers investigated talocrural mobilizations (range = 0.45 to 5.41). Therefore, from the patient's perspective, manual-therapy techniques for improving talar mobility and positioning may be more effective than other manual-therapy techniques. The talocrural mobilization protocols included 3²² (ES = 0.45) to 6^19,26,31 (ES = 0.64–5.41) mobilization sessions completed during periods ranging from 2 weeks^22,26,31 (ES = 0.45–0.86) to 3 weeks¹⁹ (ES = 5.41). Mobilizations-with-movement protocols varied, with participants completing 2 sets of 10 repetitions in 1 study¹⁹ and 2 sets of 4 repetitions with 30-second holds at the end range of dorsiflexion in another.²² Anterior-to-posterior Maitland grade III joint-mobilization techniques were implemented using 2³¹ and 4²⁶ sets of 2-minute applications. In addition, plantar cutaneous massage had a moderate effect on PROs (ES = 0.54).³¹ This finding, however, was accompanied by a 95% CI that crossed zero. Together, these results indicated a continued need to determine the patient characteristics, manual-therapy techniques, and treatment volume and dosage that optimize HRQL improvements in individuals with CAI.

Combined Interventions

We found grade B evidence that rehabilitation programs involving 2 or more targeted interventions improved PRO measures in individuals with CAI. The summary effect (ES = 1.14) indicated that combined interventions had a strong effect on PROs from preintervention to postintervention. The combined interventions consisted of stretching,^24,30 strength training,^18,24,27,29 balance training,^{18,24,25,27,29,30} vestibular-ocular reflex training,²⁵ soft tissue mobilization,³⁰ dry needling,²⁹ and strain-counterstrain.¹⁸ All combined rehabilitation protocols included a balance-training component; 2 protocols^25,30 required participants to complete a balance-training program based on McKeon et al.²⁸ In 5 studies, researchers investigated the combined effect of 2 interventions^{18,25,27,29,30}; in 2 studies, the combined effect of 3 interventions.^24,29 Combined interventions demonstrated a slightly lower summary effect than isolated balance training, which may indicate that adding other interventions to balance training may not result in greater HRQL gains for individuals with CAI.

Practical Implications

This systematic review with meta-analysis demonstrated that the available rehabilitation strategies improve ankle-specific PROs for individuals with CAI. We observed a strong overall effect for improving scores on region-specific PROs, specifically the FAAM–ADL, FAAM–Sport, FADI, FADI–Sport, and CAIT. Whereas improvements were noted using each of the aforementioned PROs, the CAIT seemed to be sensitive to rehabilitation, as it displayed the largest ESs (1.03−5.41). Of the included PROs, the CAIT was the only instrument that was not designed to capture changes in self-reported disability because it is primarily used as a diagnostic tool to identify CAI.³² The high responsiveness of the CAIT may have resulted from the relevance of its items to the activity limitations of individuals with CAI, whereas the FAAM and FADI were designed for broader patient populations. In future research, investigators should evaluate this instrument as an outcome measure.

Despite variations in rehabilitation strategies, dosages, and rehabilitation lengths, region-specific improvements were consistently demonstrated. Of the available rehabilitation strategies, supervised balance-training programs demonstrated the greatest PRO improvements in individuals with CAI. This was true when balance training was used in isolation or in combination with other treatment modalities. In addition, balance training used in combination with other rehabilitation strategies demonstrated summary effects similar to those of isolated balance training. This indicates that supplementing balance training with other interventions may not further improve HRQL when compared with isolated balance training.

Limitations

Whereas we followed the PRISMA guidelines,¹² our study had limitations. Our electronic search was conducted to find articles written in English in databases thought to be most relevant to journals that frequently publish articles related to CAI. However, we may not have identified and included all articles relevant to this review. In addition, evidence was limited regarding the isolated effects of strength-training interventions despite their common use in clinical practice. Therefore, we could not make recommendations about the effect of strength training on HRQL. Lastly, individuals with CAI have shown decreased HRQL as measured using region-specific, dimension-specific, and global outcome measures.¹¹ Thus, a multidimensional profile of HRQL should be used to evaluate the effectiveness of CAI rehabilitation strategies. The evidence presented in this review included only studies in which the researchers used region-specific PROs due to a lack of evidence assessing other domains of HRQL. In future work, investigators should use a multidimensional HRQL profile to examine the effectiveness of common CAI rehabilitation strategies from the whole-person perspective.

Authors of the included studies all used similar inclusion (a history of ankle sprains and subsequent episodes of giving way) and exclusion criteria, but none implemented intervention protocols with designs based on clinician-oriented, measured impairments. For all included studies, the interventions were delivered using blanket procedures regardless of the presence of measureable deficits. This cookie-cutter approach to CAI rehabilitation is contradictory to developing CAI treatment protocols.³³ In their rehabilitation paradigm, Donovan et al³³ suggested that CAI rehabilitation should be conducted using an assess, treat, and reassess model. They theorized that, by treating individual-specific deficits, greater health improvements may be attained.³³ By focusing on deficits specific to the individual, clinicians may create an environment most conducive to achieving HRQL improvements from the patient's perspective. To promote patient-centered care and to mimic a realistic model of clinical care, researchers studying CAI interventions should consider adopting impairment-based treatment protocols. Furthermore, to mimic true clinical care, investigators need to move from laboratory-based intervention studies to point-of-care research.