Gait-Training Interventions for Individuals With Chronic Ankle Instability: A Systematic Review and Meta-Analysis

Search Strategy

This systematic review with meta-analysis was registered in the International Prospective Register of Systematic Reviews (CRD42022357526) database on September 12, 2022. Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines were followed while conducting this systematic review and meta-analysis.²⁹ A health sciences librarian was consulted for the development of a systematic search of electronic databases. The search was performed in the online search engines PubMed, CINAHL, SPORTDiscus, and MEDLINE from database inception through September 15, 2022, using the following search terms: ([Chronic ankle instability OR CAI OR functional ankle instability OR recurrent ankle sprain] AND [gait training OR gait devices OR biofeedback OR feedback] AND [biomechanics OR kinetics OR kinematics OR electromyography]). Searches were limited to studies published in English with full text available. After the initial search, literature screening and data extraction were completed. Two authors (C.O. and R.M.K.) independently screened all titles, abstracts, and full-text records for eligible studies (Figure 1). If conflicts existed, the authors discussed the study to reach consensus. If consensus was not achieved, a third author (J.D.S.) was consulted. Manual reference-list screening was performed to identify any additional studies.

View Figure

• Download as Powerpoint
• Download Figure

Study Selection Criteria and Quality Assessment

Studies were included if they met the following criteria: (1) individuals with CAI (as determined using the International Ankle Consortium guidelines) were included; (2) a gait-training intervention was administered using devices or biofeedback methods; (3) outcome measures included gait kinetics, kinematics, muscle activity during walking, or a combination; (4) the study was published in a peer-reviewed journal; and (5) the full text was published in English.⁶ Randomized controlled trials, studies with a crossover or a quasiexperimental design, and descriptive laboratory or field studies were included. Studies were excluded if individuals with CAI were not included, interventions did not involve gait training, biomechanical outcomes were not measured, they were not available in English, or the full text was unavailable.

The Downs and Black quality-assessment checklist was used to evaluate the included studies (Table 1).³⁰ The checklist consists of 27 questions within 5 sections (reporting, external validity, internal validity, internal validity—confounding [selection bias], and power) and was designed to assess the methodological quality of randomized and nonrandomized comparative studies.³⁰ Questions were scored as yes (1), no (0), or not applicable (NA) with the exception of question 5, which was scored as yes (2), partially (1), no (0), or not applicable (NA), with a maximum total score of 28 points.³⁰ Higher scores indicated higher methodological quality.³⁰ Two authors (C.O. and R.M.K.) independently scored all included studies. If scores did not align, a third author (J.D.S.) was consulted.

Table 1.Downs and Black³⁰ Quality Assessment of Included Research Studies

View Fullscreen

Data Extraction

Study design, participant characteristics, sample sizes, intervention type (device or biofeedback), intervention length, and biomechanical outcome measures (kinematics, kinetics, and electromyography [EMG]) were extracted by 1 author (R.M.K.) for all included studies (Tables 2 through 4). Authors were contacted if values were not reported in the text or were presented as graphs. When authors of 3 or more studies reported on the same outcomes, the means and SDs were extracted for potential meta-analyses.

Table 2.Summary of Articles Related to Kinetic Outcome Measures

View Fullscreen

Table 3.Summary of Studies Related to Kinematic Outcome Measures

View Fullscreen

Table 4.Summary of Articles Related to EMG Outcome Measures

View Fullscreen

Data Analysis

When authors of 3 or more studies reported on the same outcome measure using consistent units or units that could be derived for equivocal comparisons, we conducted meta-analyses. We performed meta-analyses using a random-effects model in JASP software (JASP Team 2023, version 0.17.2.1; University of Amsterdam) to compare differences before and during or immediately after administration of gait training for studies involving a single session for the following variables: kinetics (COP gait line, contact area, contact time, peak pressure, pressure-time integral [PTI]) and EMG (root-mean-square [RMS] amplitude before initial contact [IC] and post-IC). The α level was set at .05. We did not find enough multisession gait-training studies for meta-analyses to be conducted on any variables. Meta-analysis effect sizes (ES) and associated 95% confidence intervals (CIs) were displayed using forest plots (Figures 2 through 4). The ES and standard error of ES using the pooled SD were calculated to determine the magnitude of difference between time points (before versus after gait training) or between groups (gait training versus no gait training). These ESs were interpreted as very small (≤0.20), small (0.21–0.39), medium (0.40–0.79), and large (≥0.80).³¹ Heterogeneity was analyzed using the I² test statistic and summarized the variation across studies due to difference rather than chance as recommended by the Cochrane Handbook for Systematic Reviews of Interventions.³² We used the following guidelines to interpret the I² test statistic: 0% to 40%, might not be important; 30% to 60%, may represent moderate heterogeneity; 50% to 90%, may represent substantial heterogeneity; and 75% to 100%, considerable heterogeneity.³² When heterogeneity was considerable (I² ≥ 75%), studies showing the same direction of effect were still considered appropriate for meta-analysis.³³ Publication bias was assessed using funnel plots and associated Egger regression tests for variables identified as statistically significant by the meta-analyses. Publication bias was considered present when P < .05 for the Egger regression test.³⁴

Study Selection and Characteristics

Our initial search yielded 358 studies (Figure 1). After removal of duplicates, abstract screening, and full-text review, 13 studies were included.^{25–28,35–43} Of the studies included, authors of 11 studies^{25,26,28,35–40,42,43} reported on kinetic outcome measures, authors of 3 studies^25,39,41 reported on kinematic outcome measures, and authors of 7 studies^{25–28,35,36,39} reported on muscle activity outcome measures. Authors of 5 studies^25–28,36 used a gait-training device, such as a destabilization sandal or boot, and authors of 8 studies^35,37–43 used a form of biofeedback (visual, auditory, or haptic). Summaries of the study characteristics; outcome measures; and results for kinetics, kinematics, and muscle activity are presented in Tables 2 through 4, respectively.

Methodological Quality Assessment

Downs and Black scores ranged from 16 to 25 points out of a maximum of 28 points. The 3 studies with a randomized controlled trial design had the highest overall scores, ranging from 24 points⁴³ to 25 points.^25,39 Reviewers scored all studies yes or not applicable for all questions within the Reporting section of the checklist except for “adverse events that may be a consequence of the intervention reported” (question 8 [Q8]). For the external validity section, all studies were scored yes for “subjects representative of the entire population from which they were recruited” (Q11) and “subjects who were prepared to participate representative of the entire population from which they were recruited” (Q12). All studies were scored no for “staff, places, and facilities where the patients were treated, representative of the treatment the majority of patients receive” (Q13). The studies were scored no for Q13 because the gait-training methods used in the research studies were not representative of treatments in common use in clinical practice settings for individuals with CAI. In addition, gait-training visits were conducted under the supervision of a research team using unique equipment for administering gait training that is not currently available to clinicians or individuals with CAI. Considering the internal validity subscale, authors of no studies attempted to blind the study participants to the intervention (Q14). In the randomized controlled trials only, attempts were made to blind those measuring the main outcome measures of the intervention (Q15).^25,39,43 All studies were scored yes or not applicable for the remaining internal validity questions (Q16–Q20). Considering the internal validity—confounding (selection bias) subscale, in all studies, participants in intervention groups were recruited from the same population (Q21), and authors of all studies accounted for participants lost to follow-up (Q26). In all randomized controlled trials, participants were randomized into intervention groups (Q23), randomization was concealed (Q24), and adequate adjustment for confounding in the analyses for main findings was made (Q25).^25,39,43 Authors of only 3 studies reported a sample-size estimate needed to meet the power calculation requirement for Q27.^26,27,39

Heterogeneity

Heterogeneity ranged from 0% to 40% and was interpreted to be not important for the 10% increments of the COP gait line at all time points (range, 0%–35.9%), contact time for the medial forefoot (37.0%), peak pressure for the hallux (17.6%), and PTI for the lateral heel (3.8%) and hallux (3.6%). Heterogeneity was >75% and was interpreted as considerable for peak pressure and PTI in the lateral midfoot (87.4% and 90.8%, respectively).

Publication Bias Assessment

Funnel plots and associated Egger regression test results for the meta-analyses are reported in Supplemental Figures 1 through 5. Publication bias was present for the location of the COP during 0% to 10% of the stance phase (P = .02), peak pressure in the lateral midfoot (P < .001), and the PTI in the lateral midfoot (P < .001). We did not find publication bias for any other measures included in our meta-analyses.

Gait-Training Approaches

Authors of 5 studies^25–28,36 used gait-training devices, and authors of 8 studies^35,37–43 used biofeedback for gait training. Among the studies in which authors used gait-training devices, authors of 2 studies^25,27 used destabilization boots and sandals, authors of 1 study²⁶ used a wearable multiaxis destabilization device, and authors of 2 studies^28,36 used a custom-built gait-training device with resistance bands. Among the studies in which authors used biofeedback gait training, authors of 3 studies^37,39,42 used visual biofeedback, authors of 2 studies^35,43 used auditory biofeedback, and authors of 3 studies^38,40,41 used haptic biofeedback. For visual biofeedback, authors of 1 study⁴² used a shoe-mounted crossline laser with instructions to “walk in a manner in which the vertical laser line aligns with the piece of tape on the wall,” authors of 1 study³⁷ used real-time 2-dimensional video from the posterior aspect of the treadmill with instructions to “walk in a manner where you can no longer view the outside or inside of your foot on the television screen while you walk,” and authors of 1 study³⁹ used a custom real-time display of ankle-inversion angles that turned red for steps with ankle inversion above the set threshold (too much inversion) or green for steps within the desired range for ankle inversion with instructions to “avoid walking on the outside of your foot so as not to exceed the inversion threshold.” For auditory biofeedback, authors of 2 studies used a custom device that was created to set a pressure threshold under the lateral aspect of the foot and provide an auditory tone when participants’ vertical force exceeded the set threshold.^35,43 For haptic biofeedback, authors of 3 studies used a custom device similar to that used in the auditory biofeedback studies; however, instead of delivering an auditory tone, the device provided vibration on the lateral malleolus of the test limb when participants’ vertical force exceeded the set threshold under the lateral aspect of the foot.^38,40,41

Kinetic Outcomes

Eleven studies in which authors examined kinetic outcomes met the inclusion criteria for this systematic review (Table 2).^{25,26,28,35–40,42,43} Authors of 6 studies^{26,28,36,40,42,43} reported on the COP gait line, which was defined as the location of the COP from the most medial border of the foot at 10% increments in 5 studies^{26,28,36,42,43} and the location of the COP in the lateral-medial direction from the position of the marker at the fifth metatarsal head with the foot modeled as a rectangle at 10% increments in 1 study.⁴⁰ Of these studies, 4 studies were single session, and results were pooled for meta-analyses (Figure 2).^26,28,40,42 The meta-analyses revealed small to large medial shifts in the location of the COP at each 10% increment from 0% to 90% (ES range, −0.35 to −0.82; I² range, 0–35.911; P range, <.001–.04; Egger regression P range, .02–.13) for the COP gait line.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Authors of 7 studies reported on traditional plantar-pressure measures (contact area, contact time, peak pressure, PTI, and time to peak pressure), and results were pooled for meta-analyses (Figure 3).^{26,28,35–37,42,43} Contact area was defined as how large of an area of each region of the foot was in contact with the ground during the stance phase and was measured in centimeters squared.^{28,35–37,42} Contact time was defined as how much time each region of the foot was in contact with the ground during the stance phase and was measured in milliseconds.^{28,35–37,42} Peak pressure was defined as the highest amount of pressure in a given region of the foot during the stance phase of gait and was measured in kilopascals.^{26,28,35–37,42,43} Pressure-time integral was defined as the total plantar pressure applied to a specific region of the foot multiplied by the time spent in the stance phase of gait and was measured in kilopascals multiplied by seconds.^{28,35–37,42} Time to peak pressure was defined as the percentage of stance when peak pressure occurred for the specified region of the foot.^28,35,36 Meta-analyses revealed that contact time was decreased in the medial forefoot (ES = −0.43 [95% CI = −0.86, 0.00]; I² = 36.997; P = .049; Egger regression P = .26). Peak pressure was decreased in the lateral midfoot (ES = −1.18 [95% CI = −2.24, −0.12]; I² = 87.438; P = .03; Egger regression P < .001) and increased in the hallux (ES = 0.59 [95% CI = 0.21, 0.96]; I² = 17.624; P = .002; Egger regression P = .16). Pressure-time integral was decreased in the lateral heel (ES = −0.33 [95% CI = −0.66, 0.00]; I² = 3.775; P = .050; Egger regression P = .07) and lateral midfoot (ES = −1.22 [95% CI = −2.43, 0.00]; I² = 90.757; P = .049; Egger regression P < .001) and increased in the hallux (ES = 0.63 [95% CI = 0.30, 0.97]; I² = 3.556; P < .001; Egger regression P = .14). No other differences from the meta-analyses were found for any other kinetic parameters. Authors of 2 studies reported on internal joint moments and found no differences after gait training.^25,39 Authors of only 1 study reported on impact peak; time to impact peak; impact loading rate; propulsive peak; time to propulsive peak; propulsive loading rate; and ankle-joint contact force peak, impulse, and loading rate.³⁸

View Figure

• Download as Powerpoint
• Download Figure

Kinematic Outcomes

Three studies in which authors examined kinematic outcome measures met the inclusion criteria for the systematic review (Table 3).^25,39,41 Authors of 3 studies^25,39,41 measured 3-dimensional (3D) ankle-joint angles, and authors of 2 of those studies^25,39 measured 3D joint angles at the knee and hip. Authors of all studies reported 3D ankle kinematics at IC and throughout the loading phase (first 10% of stance), but given that only 1 study was a single-session gait-training study, meta-analyses were not performed.⁴¹ Decreased ankle inversion during the loading response was reported by authors of 2 studies,^39,41 and authors of 1 study²⁵ found no differences in ankle inversion. Authors of only 1 study reported on hindfoot- and forefoot-joint angles and found increased forefoot abduction during the loading phase in the laboratory and real-world settings and increased forefoot eversion during the loading phase in the real-world setting.⁴¹ Authors of 2 studies reported on ankle, knee, and hip kinematics throughout the stride cycle (0% or 1% to 100%).^25,39 Authors of 1 study³⁹ reported increased external rotation at the knee during terminal swing with a medium ES, and authors of the other study²⁵ found no differences. No differences were identified by authors of either study for hip-joint angles.^25,39

Muscle Activity Outcomes

Seven studies in which authors measured muscle activity met the inclusion criteria for the systematic review (Table 4).^{25–28,35,36,39} Of those included, authors of 4 studies^26–28,35 reported EMG RMS amplitudes for the 50 to 200 milliseconds pre-IC and the 200 milliseconds post-IC, authors of 2 studies^25,39 reported EMG RMS amplitudes throughout the stride cycle (0% or 1% to 100%), and authors of 1 study³⁶ reported EMG RMS amplitudes during the stance phase (0% to 100%). Meta-analyses were conducted for the EMG RMS amplitudes pre- and post-IC for the tibialis anterior, fibularis longus, and gluteus medius muscles. During the 200 milliseconds post-IC, muscle activity was increased during gait training for the fibularis longus muscle (ES = 0.83 [95% CI = 0.43, 1.22]; I² = 0; P < .001; Egger regression P = .99; Figure 4). No other differences were identified by the meta-analyses for any other muscle activity variables. Before IC, authors of 2 studies^27,28 reported increased fibularis longus activity with large ESs, and authors of 2 studies^26,35 reported no differences for fibularis longus activity. During the stance phase, authors of 1 study³⁶ reported increased fibularis longus activity with medium to large ESs, authors of 1 study²⁵ reported decreased fibularis longus activity with large ESs, and authors of 1 study³⁹ reported no differences. For the tibialis anterior muscle activity, authors of 1 study²⁶ reported increased activity pre-IC with a large ES, and authors of 3 studies^27,28,35 reported no differences. Authors of 1 study³⁶ reported decreased gluteus medius activity during late stance with medium to large ESs, and authors of 2 studies^25,39 reported no differences.

View Figure

• Download as Powerpoint
• Download Figure

Methodological Quality

Studies included in our systematic review and meta-analysis were critically appraised using the Downs and Black scoring system. Study quality using the Downs and Black scoring has previously been categorized as excellent (26–28), good (20–25), fair (15–19), and poor (<15).⁴⁴ The scores of the included studies ranged from 16 to 25 points out of a possible 28 points, demonstrating fair to good methodological quality (Table 1). The randomized controlled trials^25,39,43 had the highest methodological quality (24–25 points), followed by the quasiexperimental trial³⁶ (17 points) and descriptive laboratory studies^{26,28,35,37,38,40,42} (16–18 points). Studies did not satisfy all criteria because information was not included or explicitly stated within the published manuscript and, therefore, could not earn points for that question. For the Reporting section, most studies scored yes or not applicable for all questions, except for whether adverse events were reported. None of the authors of the included studies reported or mentioned any adverse events associated with gait training, which may suggest that the gait-training techniques used by those authors are not high risk for the given population. The included studies scored yes to all questions in the external validity section except for whether the staff, places, and facilities were representative of treatments patients receive. This is not surprising as all studies took place in a laboratory setting, and all authors used techniques that are not currently available to most practicing clinicians. In addition, gait-training methods explored in the research studies were not representative of current treatments used for individuals with CAI. Whereas study methods do not reflect current treatment methods in the clinical setting, the authors of these studies provided foundational evidence to support that gait biomechanics in individuals with CAI may be improved through various gait-training methods. To improve external validity, future studies should be done to explore gait-training methods that can be easily implemented in clinical practice for individuals with CAI. Scores were high within the internal validity section, but no authors blinded participants to the intervention. This would be a considerable challenge given that the primary modes of gait training involve wearing devices or responding to some form of immediate biofeedback. When considering the confounding or selection bias within the internal validity—confounding section, scores were low. Authors of most studies did not report the time frame in which participants were recruited, did not randomize participants into intervention groups, and did not conceal randomization apart from the randomized controlled trials.

Heterogeneity

For peak pressure and PTI in the lateral midfoot, heterogeneity was considerable (87.4% and 90.8%, respectively); however, authors of all studies showed the same direction of effect, and all studies were, therefore, still considered appropriate for inclusion in the meta-analyses. Higher levels of heterogeneity may indicate that authors were measuring different underlying effects or methodological differences existed between studies. On further inspection of the individual studies included in the meta-analyses for peak pressure and PTI, authors of all studies used the Pedar-X (Novel Inc) plantar-pressure system to measure and analyze plantar-pressure outcomes, but the gait-training interventions varied greatly between studies.^28,35,37,42 For example, Donovan et al used an auditory biofeedback device placed under the fifth metatarsal, Feger and Hertel created a custom gait-training device using resistance bands, Ifarraguerri et al projected a live video of a posterior view of the foot in front of the treadmill, and Torp et al placed a crossline laser on top of the foot.^28,35,37,42 Authors of studies using auditory feedback and the custom gait-training device found reductions in peak pressure and PTI in the lateral midfoot, whereas authors of studies using the live video and crossline laser found no differences while participants were receiving gait training. The substantial variations in gait-training methods possibly contributed to the considerable heterogeneity found by the meta-analyses.

Publication Bias Assessment

Publication bias was evaluated using funnel plots and Egger regression tests in our meta-analyses. Notably, publication bias was detected regarding the location of the COP from 0% to 10% of the stance phase, peak pressure in the lateral midfoot, and PTI in the lateral midfoot. These findings indicate a potential overstatement of results pertaining to these measures within our meta-analyses. Such bias may skew the meta-analysis ES upward, potentially inflating the results and inaccurately suggesting a stronger ES that may be attributable to random chance. The detection of publication bias suggests an overrepresentation of studies in which authors reported positive outcomes for these measures may be present. This bias may distort the overall findings, leading to an inflated perception of the ES and potentially resulting in misleading conclusions.

Kinetics

Kinetic variables were the most frequently reported by authors of studies included in this systematic review. The COP gait line has been described as the mediolateral location of the COP at 10% increments during the stance phase and was the most frequently reported kinetic variable.¹⁰ Authors of all studies^{26,28,36,42,43} used the Pedar-X plantar-pressure system to measure and analyze the location of the COP except for Migel and Wikstrom.⁴⁰ Gait-training strategies to target the COP gait line included a custom gait-training device with resistance bands,^28,36 a multiaxis destabilization device,²⁶ visual biofeedback,⁴² haptic biofeedback,⁴⁰ and auditory biofeedback.⁴³ The meta-analyses revealed that, from 0% to 90% of the stance phase, gait training shifted the COP gait line medially while individuals received gait training. The pooled ES ranged from −0.35 to −0.82 throughout the stance phase, suggesting small to large improvements. Authors of studies involving multiple gait-training sessions tended to show greater medial shifts in the COP gait line as seen with the larger mean differences after gait-training sessions (Table 2).^36,43 The medium to large medial shift in the COP is considered beneficial because, when the center of gravity approaches or exceeds the lateral boundary of the foot, an episode of giving way or LAS may occur.¹⁴ Various gait-training strategies were effective at reducing laterally deviated COP and should be implemented when indicated for individuals with CAI.

Traditional plantar-pressure measures (contact area, contact time, peak pressure, PTI, and time to peak pressure) were often reported for 9 specified regions of the foot, including the medial heel, lateral heel, medial midfoot, lateral midfoot, medial forefoot, central forefoot, lateral forefoot, hallux, and toes 2 to 5 in 7 studies.^{26,28,35–37,42,43} Gait-training strategies to target the traditional plantar-pressure measures included a custom gait-training device with resistance bands,^28,36 multiaxis destabilization devices,²⁶ visual biofeedback,^37,42 and auditory biofeedback.^35,43 Generally speaking, traditional plantar-pressure measures were reduced in the lateral aspect of the foot and pressure was shifted medially, which is the desired outcome for gait training in individuals with CAI. Authors of individual studies reported decreased contact area for the lateral midfoot^28,35 or increased contact area in the medial midfoot,^36,42 suggesting a medial shift in pressure area may exist after gait training.

Peak pressure was considered the maximum loading in an area under the foot.^{26,28,35–37,42,43} The meta-analyses revealed decreased pressure in the lateral midfoot with a large ES and a medium increase in peak pressure for the hallux. Increased peak pressure for the total foot was reported by authors of 2 studies.^28,35 Although not investigated among patients with CAI, the overall increase in peak pressure may be a beneficial adaptation regarding PTOA. Authors of studies have found that greater mechanical loading during walking is associated with less type II collagen turnover among patients who underwent anterior cruciate ligament reconstruction.^45,46 Similar to peak pressure, PTI was described as the total amount of pressure for a specific region of the foot multiplied by the time spent in stance.^{28,35–37,42,43} The meta-analyses revealed that PTI decreased in the lateral heel and lateral midfoot and increased in the hallux, again suggesting a shift from lateral to medial plantar pressure. The results for the peak pressure and PTI in the lateral midfoot should be interpreted with caution. Considerable heterogeneity and publication bias were identified in the meta-analyses for these outcomes and suggest that the larger ESs for these outcomes may be due to chance rather than an actual observed change. Future studies involving larger sample sizes in which authors assess the effects of gait training on these plantar-pressure outcome measures are warranted.

The results from our meta-analyses suggest that several plantar-pressure measures are improved by various gait-training methods involving devices or biofeedback techniques. Many individuals with CAI demonstrated increased plantar pressure along the lateral column of the foot, which may be associated with an elevated risk of LAS and could contribute to the earlier onset of ankle PTOA in individuals with CAI.^13–15 Shifting the pressure medially reduces the risk of the COP approaching the lateral boundary of the foot and potentially resulting in an LAS. This altered ankle position can also result in abnormal stress distribution throughout the talar cartilage, thereby influencing the development of ankle PTOA.^17,18 Therefore, restoring gait patterns in individuals with CAI is crucial to maintaining long-term ankle-joint health, which appears to be possible using gait training.

Kinematics

Kinematics were the least reported outcome measures, with only 3 studies meeting the inclusion criteria for the systematic review.^25,39,41 Gait-training strategies to target the kinematic measures included destabilization devices,²⁵ visual biofeedback,³⁹ and haptic biofeedback.⁴¹ Authors of 2 studies^39,41 found that gait training with biofeedback (visual and haptic) reduced ankle inversion by 2.5° to 7.3°, and authors of 1 study²⁵ using destabilization devices found no changes in ankle inversion after gait training but found increased ankle dorsiflexion by 5.4° during mid to late stance. Of those included, authors of only 1 study specifically targeted the reduction of ankle inversion as part of the gait-training protocol.³⁹ Because authors of only 3 studies using gait training to improve biomechanics in individuals with CAI measured kinematic outcomes, understanding the utility of gait training for improving ankle kinematics is difficult, but the medial shift in the COP gait line and additional plantar-pressure outcome measures likely could be associated with shifting from an inverted to everted ankle position. Walking with the foot in an everted position has been shown to create more contact under the medial aspect of the foot, and thus, the COP was located on the medial aspect of the foot.⁴⁷ Future gait-training studies in which authors measure kinematic outcomes in individuals with CAI should be done to consider techniques specifically targeting ankle inversion.

Muscle Activity

Muscle activity was measured in 7 studies^{25–28,35,36,39} using EMG, and RMS amplitude was reported for all included studies; however, the timing during the stride cycle for which data were reported differed among studies, making meta-analyses possible only for short periods pre- and post-IC.^{26,28,35–37,42,43} Gait-training strategies were not specifically used to target muscle activity, but authors of several studies measured muscle activity as a primary outcome measure and included a custom gait-training device with resistance bands,^28,36 destabilization devices,^25–27 visual biofeedback,³⁹ and auditory biofeedback.³⁵ Our meta-analyses revealed a large increase in fibularis longus activity during the 200 milliseconds post-IC while individuals received gait training. Increased fibularis longus activity immediately post-IC during the loading response may contribute to increased ankle stability and the medial shift in plantar pressure.^10,48 Individuals without a history of LAS have been shown to activate their fibularis longus during midstance to assist with pronation and stabilize the first ray during propulsion.⁴⁹

Limitations

Several limitations should be considered when interpreting the results of this study. The sample size in each study was relatively small and only included 10 to 27 participants. Results from these studies should be interpreted with caution, and further research is needed in this area. The timing in which biomechanical outcomes were measured varied among studies. Authors of several studies measured gait outcomes while participants were wearing devices^26,28 or receiving biofeedback,^37,38,42,43 and authors of some studies measured outcomes after gait training had ended.^{25–27,36,39–41,43} Gait-training protocols differed substantially between studies. For example, studies in the visual biofeedback category involved a variety of techniques including projecting real-time ankle kinematics in front of the treadmill,³⁹ projecting real-time video of the posterior aspect of the ankle,³⁷ and using a crossline laser attached to the dorsal aspect of the foot.⁴² In addition, the number of gait-training sessions implemented for each study protocol ranged from 1 to 12 total sessions, which may have influenced the effects of gait training on biomechanical outcomes. Lastly, the gait-training methodology that authors of many studies used is not currently clinically accessible, which makes implementation unrealistic for athletic trainers or other health care professionals treating individuals with CAI. Future studies should be done to consider gait-training techniques that would be feasible for clinical implementation.

Future Directions

Several future directions should be considered for gait-training implementation for individuals with CAI. New gait-training strategies should attempt to transition concepts from laboratory-based interventions to strategies using minimal or no equipment to increase the feasibility of implementation in the clinical setting. Future studies should also be done to consider assessing long-term outcomes, dosage, measures of joint health, and the risk reduction of subsequent LAS associated with gait training. Whereas, in this study, we have established that gait training can be used to improve various lower extremity gait biomechanics immediately and for a short duration (up to 1 week), long-term outcomes are not yet understood. Another component of gait training to consider is the total number of gait-training sessions and the length of sessions needed to improve and maintain desired gait changes. This information may be useful in determining if additional sessions are needed as a booster or refresher after the cessation of gait-training programs to maintain desired gait changes. The overarching goal of gait training should be not only to improve biomechanics but also to improve ankle-joint health and reduce the risk of future LAS. Future research should be done to address these critical areas to continue facilitating gait training and its broader adoption in clinical practice for patients with CAI.