Changes in Spinal and Corticospinal Excitability in Patients with Chronic Ankle Instability: A Systematic Review with Meta-Analysis

The objective of this systematic review with meta-analysis was to determine alterations in spinal and corticospinal excitability of ankle muscles in patients with chronic ankle instability (CAI) compared to uninjured controls. Independent researchers performed comprehensive literature searches of electronic databases and included studies that compared groups with and without CAI and investigated neural excitability with Hoffmann reflex (H-reflex) and/or transcranial magnetic stimulation (TMS). A fixed-effect meta-analysis was conducted to determine group differences for (1) soleus and fibularis maximal H-reflex (Hmax)/maximal M-wave (Mmax)-ratios, and (2) soleus and fibularis longus cortical motor thresholds (CMTs). Seventeen studies were included in the current meta-analysis. They showed that the Hmax/Mmax-ratios of the soleus and the fibularis longus in the CAI group were significantly lower than those in the uninjured control group (soleus: d = −0.41, p < 0.001; fibularis longus: d = −0.27, p = 0.04). There was no evidence for changes in the CMT. This systematic review is the first to demonstrate evidence that patients with CAI present decreased spinal reflex excitability in the soleus and fibularis longus. However, there is no evidence of changes in supraspinal excitability when considering only the CMT. The latter result needs to be interpreted with caution as all except one study demonstrate some changes at the supraspinal level with CAI.


Introduction
Chronic ankle instability (CAI) is one of the most prevalent debilitating conditions in athletic populations [1,2]. A high percentage (23% to 61%) of athletes are identified as having CAI, with soccer, basketball, and volleyball being the most represented team sports [3][4][5][6]. In addition, recent research has highlighted that CAI is a public healthcare burden affecting not only athletic but also general populations [1]. CAI is often characterized by feelings of unstable ankles, frequent episodes of the ankle giving way, prolonged symptoms, and/or recurrent ankle injuries [2]. There is growing evidence that CAI significantly affects patient-oriented outcomes such as physical activity level [7,8] and health-related quality of life [9]. More clinically concerning, there is emerging evidence that CAI may be the primary source of early onset of post-traumatic osteoarthritis [10][11][12][13]. Despite research efforts made over the past six decades, the etiology of CAI remains unclear.

Experimental Section
The current systematic review with meta-analysis was conducted according to the guidelines listed in The Preferred Reporting Items for Systematic Reviews and Meta-Analysis Statement (PRISMA) [41].

Eligibility Criteria
Studies were considered eligible for inclusion if they had at least one group with CAI and one without CAI (uninjured control group) and investigated neural excitability with the H-reflex and/or transcranial magnetic stimulation (TMS). The electrically evoked H-reflex, which is an analog to the stretch reflex but bypasses the muscle spindles, allows assessment of spinal reflex excitability [42].
TMS is a non-invasive technique to assess corticospinal excitability by generating a magnetic field over the targeted brain area where cortical cells are electrically activated [43]. Studies were excluded if they (1) recruited patients with acute ankle sprains or other ankle and foot pathologies, (2) examined neural function with electroencephalography or used other neurophysiological measures (i.e., stretch reflex), (3) were published in a non-English language, or (4) were reported as abstracts due to the risk of duplicating data.

Literature Search Strategy
We performed comprehensive literature searches to identify peer-reviewed journal articles on the neural excitability of ankle muscles in patients with CAI. Electronic databases (Web of Science, PubMed, CINAHL, and SPORTDiscus) were searched from their inception until April 2019. Two librarians with expertise in developing a search strategy in the medical and sports science fields were consulted to assist in developing comprehensive search strategies. Two primary investigators (K.M.K and J.S.K) then established search terms using keyword searching and Medical Subject Headings (MeSH) vocabulary. The search strategies were applied using various combinations of the following key terms: ("ankle sprain" OR 'joint instability" OR "functional ankle instability" OR "chronic ankle instability" OR "unstable ankle") and ("neural excitability" OR "supraspinal excitability" OR "corticospinal excitability" OR "motor neuron pool excitability" OR "motor evoked potential" OR "motor threshold" OR "transcranial magnetic stimulation" OR "TMS" OR "spinal excitability" OR "spinal reflex activity" OR "Hoffman reflex" OR "H-reflex" OR "arthrogenic muscle response" OR "arthrogenic muscle inhibition"). In addition to database searches, we performed manual searches of the reference lists of relevant studies.

Study Selection and Data Extraction
All articles retrieved through database searches were assessed for eligibility after duplicates were identified and excluded. Two independent reviewers (K.M.K and J.S.K) screened the titles and abstracts of all the articles identified from database searches against the selection criteria, with all authors involved in resolving any conflicts. The two reviewers then further screened the full text of all remaining articles and discussed any discrepancies to reach a consensus. Following determination of the studies to be included in the review, the reviewers independently extracted the following data from each included study: author, publication year, study design, inclusion criteria for CAI (experimental) and comparison (control) groups, participant characteristics, stimulation settings, and specific neural excitability outcomes. The reviewers extracted the means and standard deviations of all neural excitability outcomes from each study as well as sample sizes of the CAI and comparison groups. Authors were contacted where data were not reported in the articles [44]. The extracted data were cross-checked to verify their accuracy and any disagreements between reviewers were resolved through discussion.

Assessment of Methodological Quality
Two primary reviewers (K.M.K and J.S.K) independently assessed the methodological quality of each included study using the modified version of the Quality Index instrument, initially developed by Downs and Black [45]. The original Quality Index, consisting of 27 items, has been modified to include only 14 items to enhance its utility. This modified version has been commonly used in the meta-analysis studies on foot and ankle injuries [46][47][48][49]. For the evaluation, each item can be scored with 1 point upon a "Yes" response, except for one item regarding the distribution of principal confounders that can be scored with 2 points upon a "Yes" response or 1 point with a "partially" response. Thus, a total of 15 points is the possible maximum score. Subsequently, the points are expressed as percentage values (points obtained divided by 15 points and multiplied by 100). The following guideline was used to determine the level of methodological quality: 0%-59% was considered as poor, 60%-74% as moderate, and 75%-100% as high quality [49].

Data Analysis
We utilized the Comprehensive Meta-Analysis software (Version 3.0, Biostat, Inc., Tampa, FL, USA) to analyze and perform a meta-analysis to determine differences in neural excitability between groups with and without CAI using a fixed-effect model. Effect size estimates using Cohen's d were employed to quantify the group differences. In accordance with Cohen's guideline, the magnitude of the effect size was interpreted as follows: d = 0.2 small, d = 0.5 medium, and d = 0.8 large [50]. Between-study heterogeneity was determined with Cochran's Q statistic and the I 2 index. In accordance with Higgins' guideline, the I 2 index was interpreted as follows: >25% low, >50% moderate, and >75% high heterogeneity [51]. Finally, Egger's test was performed to examine the risk of bias across studies. The level of significance for all statistical analyses was set a priori at p ≤ 0.05. Figure 1 illustrates the search findings. The initial searches through the electronic databases identified a total of 206 studies. After removing 89 duplicate studies, 117 studies remained, and their titles were screened, resulting in 75 studies being excluded. Abstracts of the remaining 42 studies were further screened, causing an additional 26 studies to be excluded. The full-text screening performed over the remaining 19 studies determined 17 studies for final inclusion [27,28,[32][33][34][35][36][38][39][40]44,[52][53][54][55][56][57], with two studies [31,58] excluded because they reported the same subject data as two other studies that were already included [52,53]. No additional studies were identified based on the references in previously included studies. Table 1

Data Analysis
We utilized the Comprehensive Meta-Analysis software (Version 3.0, Biostat, Inc., Tampa, FL, USA) to analyze and perform a meta-analysis to determine differences in neural excitability between groups with and without CAI using a fixed-effect model. Effect size estimates using Cohen's d were employed to quantify the group differences. In accordance with Cohen's guideline, the magnitude of the effect size was interpreted as follows: d = 0.2 small, d = 0.5 medium, and d = 0.8 large [50]. Between-study heterogeneity was determined with Cochran's Q statistic and the I 2 index. In accordance with Higgins' guideline, the I 2 index was interpreted as follows: >25% low, >50% moderate, and >75% high heterogeneity [51]. Finally, Egger's test was performed to examine the risk of bias across studies. The level of significance for all statistical analyses was set a priori at p ≤ 0.05. Figure 1 illustrates the search findings. The initial searches through the electronic databases identified a total of 206 studies. After removing 89 duplicate studies, 117 studies remained, and their titles were screened, resulting in 75 studies being excluded. Abstracts of the remaining 42 studies were further screened, causing an additional 26 studies to be excluded. The full-text screening performed over the remaining 19 studies determined 17 studies for final inclusion [27,28,[32][33][34][35][36][38][39][40]44,[52][53][54][55][56][57], with two studies [31,58] excluded because they reported the same subject data as two other studies that were already included [52,53]. No additional studies were identified based on the references in previously included studies. Table 1 details the characteristics of the included studies.

Methodological Quality
Two reviewers (K.M.K and J.S.K) independently scored a total of 238 methodological quality items (14 items × 17 studies) and initially agreed on 228 items (96%). Cohen's kappa inter-rater reliability analysis demonstrated almost perfect agreement between the two reviewers (K = 0.915, 95% confidence interval of 0.860-0.970). Final consensus for all items (100%) was achieved upon further discussion. Table 2 shows the quality index scores of individual studies. The average score (65.9 ± 7.4%) indicated that overall, study quality was moderate. Fourteen studies (82.4%) clearly described study purposes, primary outcomes, main findings, estimates of random variability, and actual probability values. However, only eight studies (47.1%) explicitly described participant characteristics. While most studies were clear about reporting information about data dredging, statistical tests, the reliability and validity of outcome measures, and the participant recruitment pool, no studies reported information that helped to determine if participants were representative of the entire population and if participants were recruited over the same time period.

Participants
A total of 643 participants were enrolled in the included studies, with 311 participants with CAI, 272 uninjured controls, and 60 copers. Inclusion criteria for CAI were similar across studies including at least one previous ankle sprain, current feelings of ankle instability and/or episodes of the ankle giving way, and self-reported ankle dysfunction. Uninjured controls or copers were similarly recruited and were matched with the CAI group by age, height, and weight. All participants were young adults, with group mean ages ranging from 19.8 to 26.5 years and with 259 males and 384 females.

Outcome Measures
All 17 studies primarily investigated neural excitability in patients with CAI using peripheral nerve stimulation (PNS) to elicit H-reflexes and/or TMS to evoke MEPs: there were twelve PNS studies [27,28,32,38,39,44,[52][53][54][55][56][57], three TMS studies [35,36,40], and two studies [33,34] that used both techniques. The technique for eliciting H-reflexes appeared to be consistent across all studies, involving 1 ms square-wave pulses that were delivered 10 to 20 s apart to prevent post-activation depression. All H-reflex studies reported Hmax/Mmax ratios as an outcome for estimating "spinal reflex excitability". For the TMS technique, almost all studies (four out of five) used the double-cone coil to generate magnetic stimuli up to 1 or 1. 4 Tesla that were delivered 10 to 15 s apart, except for one study [40] which utilized the figure-8 coil to deliver magnetic stimuli 5 s apart. All five TMS studies reported cortical motor threshold (CMT) as the primary outcome for quantifying corticospinal excitability. However, additional TMS parameters were used including MEP [33,34,36], cortical silent period [36,40], and/or corticomotor map parameters [33].
All TMS studies, and most PNS studies, collected data when participants were either in a lying position or were seated. Only three studies assessed "spinal excitability" (H-reflexes) while standing [32,53,54]. A zero score, as reflected by the negative sign (-) in the table, was given to an item that was not satisfied, while the items that were satisfied scored one point, as reflected by the positive sign (+); two points could be earned for item 5 as reflected by the double positive sign (++).

Neural Excitability Meta-Analysis
Only one outcome measure per study is recommended to avoid violation of the assumption of independence in the meta-analysis [51]. Standard practice is to choose a common neural excitability measure in a study reporting multiple measures. For spinal excitability, an outcome measure of Hmax/Mmax ratio at rest (i.e., lying or sitting) was selected because it was commonly reported in 11 out of 14 studies that used H-reflex. The other three studies [32,53,54] reported the same measure, but in the bipedal stance condition; therefore, bipedal Hmax/Mmax ratios were selected. For cross-over H-reflex studies [38,39,52] reporting two baseline measures on different days, random selection was used to include one of the baseline measurements. Furthermore, one H-reflex study [57] had three subgroups with CAI: (1) perceived instability, (2) recurrent ankle sprains, and (3) a combination of both. The group with perceived instability and recurrent ankle sprains was selected because CAI participants in other studies reported both perceived instability and recurrent ankle sprains. For supraspinal excitability, a measure of cortical motor threshold (CMT) was selected because it was used in all five TMS studies. Regarding the target muscles, several lower extremity muscles were tested but all included studies examined either the soleus [19,32,38,[53][54][55][56][57], the fibularis longus [28,[33][34][35]39], or both [27,40,44,52]. Thus, the current study performed a meta-analysis to determine the difference between the CAI and uninjured control groups for each outcome: (1) soleus Hmax/Mmax ratio, (2) fibularis longus Hmax/Mmax ratio, (3) soleus CMT, and (4) fibularis longus CMT. A negative effect size of the Hmax/Hmax ratio represents decreased spinal reflex excitability while a positive effect size of CMT reflects decreased corticospinal excitability.

Fibular Longus Hmax/Mmax Ratio
A total of seven studies on fibularis longus spinal excitability yielded data for 235 (118 CAI and 117 uninjured control) participants. The meta-analysis determined that the Hmax/Mmax ratio of the fibularis longus was significantly reduced in the CAI group compared to that of the uninjured control group (d = −0.27, 95% CI = −0.53 to −0.01, p = 0.04; see Figure 3). The effect size of this reduction was small to medium. The heterogeneity of the effect sizes was not significant (Q(6) = 10.47, p = 0.11, I2 = 42.71%).

Fibularis Longus Cortical Motor Threshold
A total of four studies investigated the corticospinal excitability of the fibularis longus. The meta-analysis of data of 123 (61 CAI and 62 uninjured control) participants determined that the CMT of the fibularis longus in the CAI group was not significantly different from that of the uninjured control group (d = −0.14, 95% CI = −0.22 to 0.50, p = 0.45; see Figure 5). The heterogeneity of the effect sizes was not significant (Q(3) = 6.11, p = 0.11, I2 = 51%).

Fibularis Longus Cortical Motor Threshold
A total of four studies investigated the corticospinal excitability of the fibularis longus. The metaanalysis of data of 123 (61 CAI and 62 uninjured control) participants determined that the CMT of the fibularis longus in the CAI group was not significantly different from that of the uninjured control group (d = −0.14, 95% CI = −0.22 to 0.50, p = 0.45; see Figure 5). The heterogeneity of the effect sizes was not significant (Q(3) = 6.11, p = 0.11, I2 = 51%).

Risk of Bias across Studies
Egger's regression test was performed to examine the risk of bias across studies for each metaanalysis. Applying this test, we found no significant publication bias for any outcome: soleus Hmax/Mmax ratio (intercept = 0.71, p = 0.64), fibularis longus Hmax/Mmax ratio (intercept = −5.40, p = 0.30), and fibularis longus CMT (intercept = 4.70, p = 0.36). Egger's test was not performed for the soleus CMT due to the small number of studies (n = 2).

Discussion
The current systematic review with meta-analysis is the first to synthesize data from individual studies investigating the neural excitability of lower extremity muscles in patients with CAI. This review found moderate-quality evidence indicating decreased spinal reflex excitability of the soleus and fibularis longus muscles in young adults with CAI when compared to age-matched controls without a history of ankle sprain. These findings appear to be in line with a growing body of evidence showing that neural adaptations occur in other musculoskeletal conditions [37,59,60] and support the emerging therapeutic strategies addressing neuroplasticity in the field of sports medicine [61][62][63][64]. However, we did not see a group difference in corticospinal excitability of either muscle between CAI and healthy control subjects. The lack of group differences in corticospinal excitability needs to be interpreted with caution for the reasons discussed in the following sections.

Spinal Reflex Excitability Associated with CAI
The small-to-medium effect size of the group difference in the soleus Hmax/Mmax ratio found in the present review is somewhat consistent with individual studies, except for the latest two studies [32,56]. The group difference in the fibularis longus Hmax/Mmax ratio was of small-to-medium effect size. However, individual studies reported varying effect sizes. Collectively, these results imply that the spinal reflex excitability of ankle muscles was significantly decreased in patients with CAI. Reflex inhibition has also been observed in other musculoskeletal conditions such as chronic low back pain [65], and hip [66] and knee injuries [67,68].

Risk of Bias across Studies
Egger's regression test was performed to examine the risk of bias across studies for each meta-analysis. Applying this test, we found no significant publication bias for any outcome: soleus Hmax/Mmax ratio (intercept = 0.71, p = 0.64), fibularis longus Hmax/Mmax ratio (intercept = −5.40, p = 0.30), and fibularis longus CMT (intercept = 4.70, p = 0.36). Egger's test was not performed for the soleus CMT due to the small number of studies (n = 2).

Discussion
The current systematic review with meta-analysis is the first to synthesize data from individual studies investigating the neural excitability of lower extremity muscles in patients with CAI. This review found moderate-quality evidence indicating decreased spinal reflex excitability of the soleus and fibularis longus muscles in young adults with CAI when compared to age-matched controls without a history of ankle sprain. These findings appear to be in line with a growing body of evidence showing that neural adaptations occur in other musculoskeletal conditions [37,59,60] and support the emerging therapeutic strategies addressing neuroplasticity in the field of sports medicine [61][62][63][64]. However, we did not see a group difference in corticospinal excitability of either muscle between CAI and healthy control subjects. The lack of group differences in corticospinal excitability needs to be interpreted with caution for the reasons discussed in the following sections.

Spinal Reflex Excitability Associated with CAI
The small-to-medium effect size of the group difference in the soleus Hmax/Mmax ratio found in the present review is somewhat consistent with individual studies, except for the latest two studies [32,56]. The group difference in the fibularis longus Hmax/Mmax ratio was of small-to-medium effect size. However, individual studies reported varying effect sizes. Collectively, these results imply that the spinal reflex excitability of ankle muscles was significantly decreased in patients with CAI. Reflex inhibition has also been observed in other musculoskeletal conditions such as chronic low back pain [65], and hip [66] and knee injuries [67,68].

Functional Consequences of Reflex Inhibition in CAI Patients
Sefton et al. [58] found that when considering performance during static balance tasks and the amount of spinal reflex excitability, over 86% of participants could be correctly classified as CAI patients. Similarly, Terada et al. [57] discovered that the combination of reduced spinal reflex excitability and self-reported disability could classify 72% of CAI participants. The authors of both aforementioned studies [57,58] as well as other authors [55,56] have consistently suggested that the consideration of spinal reflex excitability may be more important than consideration of other well-known classifying factors such as impaired joint kinesthesia or ankle joint laxity. In addition, altered spinal reflex excitability was associated with postural control deficits [31] and self-reported disability [29,30]: CAI patients with greater changes in reflex excitability also felt more disabled by their poor balance. All these observations point towards the fact that spinal reflex excitability plays an important role in CAI and constitutes a strong indicator for functional impairments in patients with CAI.

Mechanisms of Reflex Inhibition in CAI Patients
The results of our meta-analysis indicate chronically reduced H-reflexes in CAI patients. At the spinal level, several mechanisms have been proposed to (task-specifically) adjust and modulate the size of the H-reflex such as presynaptic, reciprocal, recurrent, and Ib inhibition (for a review see [69]). Although the underlying mechanisms of reduced H-reflexes in CAI patients are not well understood, there is evidence for disturbed reflex propagation at both the pre-and postsynaptic level [32,53]. Sefton et al. [53] demonstrated that recurrent inhibition was enhanced in CAI patients, leading to a general depression of the α-motoneuron pool independent of whether subjects were standing in a single-or double-legged stance. Apart from these postsynaptic mechanisms, Sefton et al. [53] also found differences in presynaptic transmission at the spinal level between healthy controls and CAI-patients. When healthy controls switched from a double-to a single-legged stance, they reduced the amount of paired reflex depression (PRD). In contrast, CAI-patients did not modulate the amount of presynaptic inhibition when switching to the more demanding single-legged stance condition.
In contrast to Sefton and colleagues [53], a recent study [32] proposed that CAI patients demonstrate disinhibition of spinal reflexes due to reduced presynaptic inhibition. At the same time, the authors did not find any differences in recurrent inhibition between participants with CAI and healthy controls. Although the study seems well-conducted it is difficult to interpret their results (i.e., facilitated H-reflexes in CAI) as they contradict most other studies in this field, as well as the outcome of the present meta-analysis demonstrating a reduction in spinal reflex excitability in patients with CAI. However, there is one important aspect in common: Thomson et al. [32] confirmed that CAI patients do not show task-specific modulation of presynaptic inhibition. Previously, it has been suggested that presynaptic inhibition provides an effective means for rapidly adapting to sudden environmental changes whereas postsynaptic inhibition may provide a more generalized and longer-lasting change in the spinal reflex circuitry [70]. In this sense, the postsynaptic downregulation of the motoneuron excitability by increased recurrent inhibition may contribute to the reduced H-reflexes seen in CAI-patients [53], probably irrespective of the postural condition in which the patients are measured. In contrast, the impaired modulation of presynaptic inhibition may only be apparent in paradigms in which H-reflexes were assessed in at least two different postural conditions. In doing so, Kim et al. [31,44] observed less modulation of the H-reflex (soleus and fibularis longus) in CAI patients than in healthy controls when switching from lying prone to a unipedal stance. The authors assumed that presynaptic mechanisms were altered in these CAI-patients. However, more importantly, it was demonstrated that the inability to modulate the H-reflex was significantly correlated with impaired balance control [31].

Reduced Spinal Reflex Excitability as a Sign of "Arthrogenic Muscle Inhibition" in CAI Patients?
Reduced spinal reflex excitability in CAI patients, determined by lower amplitudes of H-reflexes, has been considered arthrogenic muscle inhibition (AMI), defined as an on-going reflexive inhibition of uninjured muscles surrounding the injured joint [71]. It was speculated that AMI is a protective mechanism of the central nervous system (CNS) to secure the injured ankle joint at the time of injury but does not seem to resolve completely as injury recovery proceeds [72]. As a result, prolonged inhibition of the dynamic ankle stabilizers may contribute to the chronic nature of poor neuromuscular control at the ankle joint in CAI patients. In this sense, AMI may be considered as an underlying neurophysiological mechanism of sensorimotor impairment following joint injury in general, and ankle joint injuries in particular [15,68,71]. Although many authors [27,28,34,38,39,44,52,55,56] may agree with this view because they consider reduced H-reflexes not only as an essential but as a sufficient criterion to classify (undoubtedly) AMI, this classification is at least questionable. It is well known that balance training reduces spinal reflex excitability too (for a review see [69]). Although most studies reported training-induced reductions in the H-reflex only when measured during postural task execution [73,74], there are also balance training studies that observed reduced spinal reflex excitability at rest [75]. Due to the fact that most chronic CAI patients have undergone balance exercises as part of their rehabilitation program, reduced H-reflexes as the only criterion might considerably bias this classification. Furthermore, it has to be considered that the inter-individual variability in the size of the H-reflex, and also the ratio of the H-reflex normalized to the maximal M-wave (Hmax/Mmax ratio), is extremely high [76] and further depends on the age of the subjects [77]. In addition, populations that display impaired motor control compared to healthy young adults such as elderly people, children born preterm, or persons with pathological conditions like Huntington's disease, cerebral palsy or spasticity often demonstrate reduced inhibitory capacity at both the spinal [77][78][79][80] and cortical levels [81][82][83]. These examples illustrate that a reduction of spinal reflexes-as the data from the systematic analysis of CAI patients in the present study propose-should not be considered as a maladaptation per se and should not automatically be classified as AMI. In order to get a better idea about the functional consequences of reduced spinal reflex excitability in patients with CAI and to establish a better link between cause and effect, it is strongly recommended that future studies concentrate not only on reflex (and TMS) measurements at rest but also during the execution of postural (functional) tasks. Measurements that are performed at rest (or in a less-demanding control conditions) and during activity allow the modulatory range to be assessed by comparing adjustments during activity with the "baseline" level at rest. For both the spinal [80,84,85] and cortical levels [83,86] it was shown that aging as well as certain pathologies reduce the modulatory range, and thus the ability of the CNS to adapt to the task specific requirements. It may therefore be assumed that a chronic reflex inhibition such as AMI would result in a reduced capacity to task-specifically adjust and modulate the H-reflex. Recently, such a reduced H-reflex modulation was indeed shown in CAI patients [31]. Interestingly, the reduced ability to task-specifically modulate the H-reflex between lying prone and one-legged standing in CAI patients was associated with reduced stance stability (r = 0.578, p = 0.049). Furthermore, as mentioned above in Section 4.1.2, CAI-patients demonstrated an impaired capacity to modulate presynaptic inhibition between a single-and a double-legged stance [32,53]. These studies provide strong evidence for chronic malfunctioning of spinal reflex circuitries in CAI patients.

Corticospinal Excitability Associated with CAI
There is no conclusive evidence that the supra-spinal excitability of ankle muscles in patients with CAI is altered relative to that of uninjured controls, as CMT measures across individual studies provide conflicting results. Pietrosimone and Gribble [35] first investigated the corticospinal excitability of the fibularis longus, and they reported higher CMT in individuals with CAI when compared with uninjured controls. Based on this observation, the authors [35] assumed that patients with CAI may encounter more difficulty in activating the fibularis longus muscle via corticospinal tract fibers, which may in turn lead to inadequate control of the ankle joint, resulting in ankle dysfunction and increased risk of recurrent ankle injuries. In addition, higher CMT was significantly associated with higher self-reported disability, indicating that the level of corticomotor excitability influences the patient's perception of their ankle function [35]. In contrast, all subsequent studies [33,34,36,40] failed to confirm higher CMT's in either the fibularis longus or soleus muscle in patients with CAI. However, three out of these four studies [33,34,36] nevertheless found alterations in different TMS-based parameters. McLeod et al. [34] observed lower amplitudes of the motor-evoked potential in the fibularis longus when measured during weak contractions. This supports the initial findings from Pietrosimone and Gribble [35] and suggests decreased ability to activate lower leg muscles via the corticospinal tract in patients with CAI. The reason for this reduced corticospinal connectivity might be related to observations from Kosik et al. [33], who discovered decreased corticomotor map area and volume of the fibularis longus muscle. In this study, a TMS mapping technique was used to estimate the size of the corticomotor representation, which was apparently smaller in CAI patients than in uninjured controls. Another mechanism, which might also contribute to reduced corticospinal connectivity in CAI patients, is the observation of Terada et al. [36] who reported greater cortical silent periods in the soleus muscle of CAI patients. The silent period is an indicator of GABAb-mediated intracortical inhibition and originates largely from activation of cortical inhibitory interneurons [87] although spinal mechanisms are also involved in the early part of the silent period [88]. Thus, the longer silent periods indicate increased intracortical inhibition, which may impede activation of lower leg muscles via the corticospinal tract in CAI patients. It is noted, however, that there is another TMS study [40] which failed to find alterations in the silent period of the fibularis longus.
Collectively, these TMS studies suggest that supra-spinal adaptions may very well occur in patients with CAI and probably contribute to their sensorimotor impairments [33][34][35][36]. Supporting this, a recent magnetic resonance imaging study [89] assessed the white matter microstructure of the superior cerebellar peduncle by means of diffusion tensor imaging in patients with CAI and found lower white matter microstructure in CAI patients. At the same time, the CAI patients demonstrated worse postural control than healthy subjects.
In summary, the limited number of TMS studies, the different test situations (measurements at rest versus measurements during voluntary muscle contractions while sitting or lying), and the rather non-specific muscle activations during testing may have prevented finding conclusive evidence, so far. In particular, the last point seems crucial: it is very well known that corticospinal excitability as well as inhibitory processes are task-specifically modulated [90][91][92]. It would, therefore, seem plausible that differences in supraspinal control between CAI patients and uninjured controls should be most obvious during tasks in which CAI patients reveal impaired motor control such as balance tasks. This may help to more clearly indicate the underlying neural mechanisms associated with CAI at the supraspinal level.

Limitations
Our findings indicate an overall impairment (reduction) of spinal reflexes with CAI, but there are inconsistencies when regarding supraspinal changes. Although our approach, analyzing solely CMT as one common TMS outcome parameter, is in line with the recommendation to avoid violation of the assumption of independence in the meta-analysis [51], the outcome or interpretation may change when considering other TMS measures. However, it has to be noted that each TMS measure of either soleus or fibularis longus, other than CMT, was only used in a single study, making it difficult to perform a meta-analysis. The findings of the present systematic review are limited to young adults with ages ranging from 19.8 to 26.5 years and thus, cannot be uncritically translated to other age groups. Furthermore, the present review did not incorporate non-English studies.

Recommendations for Future Research
The present systematic review of the literature found several points to be considered in future research investigating neural excitability in patients with CAI using TMS and/or H-reflex techniques. First, while studies included in the current investigation suggested that CAI may be associated with decreased spinal reflex excitability [27,28,44,52,[55][56][57], the retrospective nature of these studies does not provide information on whether the reduced excitability is related to the development of CAI or is simply a consequence of CAI. Prospective studies are warranted to elucidate this link. In addition, it is not known how the neural changes occur following an initial ankle injury. It is recommended that this time course be established in order to make comparisons with knee-injury patients. In knee patients, reduction of quadriceps reflexes appears to be most severe in the acute stages of knee injury or surgery and these are slowly restored over time, although it is noted that the diminished level of reflexes may be still clinically significant in the long term (i.e., 18-33 months) [72]. Thus, a longitudinal study is needed to understand how neural excitability following initial ankle injury plays a role in developing CAI. Secondly, reporting outcome measures of spinal excitability should be better standardized to allow for between-study comparisons. A report of a recruitment curve may be preferred, from which multiple parameters of neural excitability can be extracted. Thirdly, studies investigating corticomotor excitability with TMS should be aware that the MEP is influenced by both spinal and supraspinal (cortical) excitability [93]. Thus, studies that want to assess cortical measures should combine TMS with other measures. Apart from this, it is strongly recommended that spinal as well as cortical changes accompanying CAI be assessed, not only at rest but also during activity (see text above for more details). Finally, a wide range of CAI populations should be examined in future studies. A majority of the studies included in the review recruited young adults in a university setting, which certainly limits the generalizability of the findings. This seems important considering the fact that CAI affects not only young but also older populations [1,4,94].

Conclusions
The current systematic review with meta-analysis is the first to demonstrate moderate-quality evidence that patients with CAI presented decreased spinal reflex excitability in the soleus and fibularis longus, as determined with Hmax/Mmax ratios. The diminished spinal reflexes may contribute to the occurrence of AMI. However, limited data about reflex modulation and task-specific adjustments of spinal reflexes makes it impossible to clarify this question at this stage, as reflex inhibition is not maladaptive per se (for details see the text above). On the other hand, there is limited evidence to determine whether supraspinal excitability is altered in CAI patients when considering only CMT. However, when taking into account a more global picture of all the supraspinal changes that were detected by means of TMS (and diffusion tensor imaging), it seems reasonable to assume that descending drive from the motor cortex to ankle muscles is reduced in patients with CAI.