Gait Impairment in Traumatic Brain Injury: A Systematic Review

Introduction: Gait impairment occurs across the spectrum of traumatic brain injury (TBI); from mild (mTBI) to moderate (modTBI), to severe (sevTBI). Recent evidence suggests that objective gait assessment may be a surrogate marker for neurological impairment such as TBI. However, the most optimal method of objective gait assessment is still not well understood due to previous reliance on subjective assessment approaches. The purpose of this review was to examine objective assessment of gait impairments across the spectrum of TBI. Methods: PubMed, AMED, OVID and CINAHL databases were searched with a search strategy containing key search terms for TBI and gait. Original research articles reporting gait outcomes in adults with TBI (mTBI, modTBI, sevTBI) were included. Results: 156 citations were identified from the search, of these, 13 studies met the initial criteria and were included into the review. The findings from the reviewed studies suggest that gait is impaired in mTBI, modTBI and sevTBI (in acute and chronic stages), but methodological limitations were evident within all studies. Inertial measurement units were most used to assess gait, with single-task, dual-task and obstacle crossing conditions used. No studies examined gait across the full spectrum of TBI and all studies differed in their gait assessment protocols. Recommendations for future studies are provided. Conclusion: Gait was found to be impaired in TBI within the reviewed studies regardless of severity level (mTBI, modTBI, sevTBI), but methodological limitations of studies (transparency and reproducibility) limit clinical application. Further research is required to establish a standardised gait assessment procedure to fully determine gait impairment across the spectrum of TBI with comprehensive outcomes and consistent protocols.


Introduction
Traumatic brain injury (TBI) is defined as mild, moderate (modTBI), or severe (sevTBI) injury that results in symptoms that can persist across an acute (days to weeks) or chronic (months to years) time-period [1]. Mild TBI (mTBI), commonly known as concussion, has had predominant focus as it is the most common type of TBI (i.e., mTBI accounts for up to 84% of TBI) [2,3]. TBI can cause deficits in motor and non-motor functions, such as impaired cognitive function, headaches, fatigue, depression, anxiety, and irritability [4]. American Congress of Rehabilitation Medicine [5] describes mTBI as a "mild insult to the head that results in a brief period of unconsciousness followed by impaired cognitive function". Alternatively, moderate and severe TBI are described as traumatic brain injuries of increased severity lasting a longer period of time [6]. Individuals who present with modTBI express

Search Strategy
This review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The key search terms were "traumatic brain injury" and "gait". A catalogue of synonyms was formulated for each key term ( Figure 1). Relevant Boolean and medical subject subheadings (MeSH) were applied as seen in Figure 1. The search strategy compromised of four electronic databases: AMED, CINAHL, PubMed and OVID, from 1960 to February 2021. Studies were considered relevant if they incorporated terminology which focussed on gait assessment in TBI and healthy control subjects in the title, abstract or keywords. An initial title screen for relevant articles was performed by the reviewer (AD) once the searched database results had been combined. After initial title screen, both the titles and abstracts of the selected articles were reviewed by two independent reviewers (AD, DP). A review of full text was required if it was not clear from the title or abstract whether the study met the review criteria.

Inclusion and Exclusion Criteria
Articles were included if they reported use of a digital device to measure gait in people with TBI. Studies were included only if they included a control group for comparison to TBI cohorts, so that TBI specific differences could be identified. Articles were excluded if they involved children (<18 years old), participants who had sustained a previous TBI, or a TBI group that did not have any information on the diagnosis (i.e., self-reported history of TBI with no current symptoms), did not provide specific objective gait outcomes from a digital device (i.e., only reported subjective outcomes) and involved a rehabilitation or intervention of some form. Only articles written in English were considered for review and any abstracts, case studies, conference proceedings, reviews, commentaries, discussion papers, or editorials were excluded.

Data Extraction
Data were extracted by the reviewer (AD) then synthesised into table format, with a second reviewer (DP) confirming the data. Data included demographic, instrumentation, study protocol, outcome measures and key findings.

The Evidence Base
The search strategy yielded 156 articles, we excluded 48 duplicates (Figure 2). An initial screen identified 108 articles of interest, but 75 articles were excluded at title screen for not meeting the inclusion criteria and a further 19 were excluded during the full-text screen, with a further five removed at the final review stage. In total, 13 articles were included by consensus from the screening reviewers (AD, DP, and SS). Most of the removed articles were excluded because they included adolescents (Under 18 s), participants who suffered a previous TBI or did not include a healthy control group (full list of excluded articles and reasons located in Supplementary Material Table S1).

Particpants
The reviewed articles (n = 13) investigated individuals who suffered a TBI over acute and chronic time periods across a range of severity from mTBI to sevTBI (Table 1). Most studies (n = 6) examined participants with mTBI, with modTBI (n = 1) [21] and sevTBI (n = 1) [22] less studied. Several studies examined across a range of different TBI severities, specifically; one study investigated modTBI to sevTBI [23], another examined sevTBI and very sevTBI [24], and three studies examined gait in general TBI (combining mTBI, modTBI, sevTBI into one group) [25][26][27]. Five studies examined participants within an acute stage (<7 days), one study was conducted at a sub-acute stage (>7 days) [25] and seven studies examined participants at a chronic stage (>12 weeks). Only one study examined participants across a range of TBI stages from sub-acute to chronic (time since injury ranging from 2 months to 28 months post injury) [23].
In terms of demographic characteristics, the majority of the studies included both males and females, with ages that ranged from 18 to 53 years. One article did not provide specific demographic information for age [25]. There were various inclusion and exclusion criteria for TBI participants (Table 1). Table 2 shows that there was a lack of standardisation in instruments used to assess the characteristics of gait that were assessed in the reviewed studies, with inertial measuring units (IMUs), instrumented gait mats, force plates or motion capture systems all used. Majority of the articles used IMU devices (n = 5) to monitor spatiotemporal gait features, which were placed at various locations (i.e., feet, lumbar region, sternum, forehead etc.).

Equipment
The sampling frequencies used to quantify gait performance using IMU's appears consistent (128 Hz), while motion capture varied between 60 Hz [27] and 120 Hz [22,28,29]. Three studies used force plates at a sampling frequency ranging from 960 to 1080 Hz [21,22,27]. One study used a smartphone to quantify gait speed [30]. Table 3 shows that there was a lack of consistency in the specific study protocols, but the majority of the studies included in this review investigated both single and dual-task gait conditions (n = 6), while some studies investigated single task (n = 4), dual-task (n = 1) and complex task (n = 2) parameters alone. In terms of dual-task paradigm, eleven articles used a question-and-answer task, including serial subtraction in sevens (n = 5), spelling a 5-letter word backwards (n = 2), reciting months of the year in reverse order (n = 3). Additionally, the audio Stroop test (n = 1) and modified Stroop test (n = 1) and reading aloud a piece from a newspaper article were used (n = 1). Complex gait tasks used obstacle crossing (n = 2) with obstacles individualised according to the participants height.

Outcome Measure
There was a lack of standardisation of outcomes reported with reviewed articles providing various outcome measures on spatiotemporal, kinetic, and kinematic markers of gait. The majority of the articles included examined spatiotemporal parameters of gait with the most consistent measures being gait speed (n = 9) and measurements surrounding stride (i.e., stride length or stride time) (n = 6). Similarly, centre of mass displacement (n = 4) was the most common outcome measure used when considering kinematic assessment of gait. Furthermore, regarding kinetic parameters, ground reaction forces (n = 3) were reported.

Key Findings
This review identified a variety of methods associated to measuring gait following a TBI (Table 2). For example, when measuring gait using a single task paradigm, this review identified gait speed as a distinguishing factor between TBI participants and controls [22,24,25,31,32]. However, Fino et al. (2016) reported that single-task gait was not different between TBI and controls and suggested that dual-task paradigms are needed to elicit gait deficits [33]. This was seen in studies that examined dual-tasks, as gait impairments were found during dual-task compared to single-task walking across the spectrum of TBI and different acute and chronic stages of the injury. Furthermore, complex gait tasks were examined in several studies and showed that deficits can be found using these protocols. For example, Vallée et al. (2006) determined that TBI participants were slower while performing the Stroop task when avoiding the wide obstacle and walked more slowly for narrow and wide obstacle conditions [23]. Furthermore, McFadyen et al., (2003) also showed an increased lead-limb clearance margins for TBI group throughout all conditions and TBI spectrum [25]. Overall, despite the differences in methodologies between studies, participants with TBI had impairment in gait with single-task, dual-task, and complex task performance in the reviewed studies, which was regardless of severity or stage, but the deficits were selective to particular outcomes within studies and lacked consistency across studies. • Trained sports medicine physician. Temporal: Step Test: • Significant differences exhibited in the three motor tasks between control group and both severe groups and between severe and very severe group. • Statistically significant differences seen between control group and severe TBI were found in spatiotemporal parameters of Fukuda Step Test.

•
No differences noted in terms of lateral/forward displacements among the three groups.

•
Or among amount of rotation or side rotation among the three groups. • Significant differences in walking speed noted between control group and both severe groups and between severe and very severe group.
Fino et al. [33] To determine the local dynamic stability of athletes who recently suffered a TBI during single and dual-task gait.

•
One year follow up 363 ± 42 days. • Barefoot on a wooden floor.
Single Task:  Oldham et al. [30] Examine whether changes between baseline and acute post-TBI single task and dual-task tandem gait performance differed between male and female athletes.  • TBI group demonstrated greater tandem gait impairments (i.e., a positive change in time) between Time 1 and Time 2 than the healthy controls.
Parker et al. [21] Examine the relationship between measures of dynamic motor performance (single and dual-task walking) and neuropsychological function following concussion over the course of 28 days.
Gait Stability Testing.
• All TBI athletes were tested 48 h, 5 days, 14 days and 28 days post injury.

•
Control Group tested at the same time points of the study. • All participants were tested barefoot and walked on a 10 m walkway at a preferred walk speed. • TBI group had significantly greater sway for the dual-task condition on days 5 and 28.

•
The dual-task condition produced significantly faster sway than the single-task condition for both groups, even at 28 days following initial testing • Maximum anterior COM-COP separation distance revealed a task effect with the dual-task producing a smaller separation distance than the single-task for the TBI group on all days • Visual memory-TBI group showed significant improvement from day 2 to 5 and from day 5 to 14. • Group differences were detected for the testing days 2 and 5 with the TBI group performing worse than controls.

•
The TBI group mean processing speed was significantly faster on day 5 compared with day 2 but did not change significantly after day 5 • the maximum separation between the COM and COP in the anterior direction.
Parrington et al. [32] Evaluate the recovery of gait and balance in concussed athletes to account for changes in trends following return to play.
• Inertial sensors attached bilaterally on anterior and distal aspect of each shank and posterior pelvis at L5. • Participants were assessed during 9 testing periods over the course of an 8-week period. • 2 testing session in week 1 followed by weekly testing for the next 7. • Speed did not differ between groups. Gait speed over time was more pronounced in TBI participants. • Gait speed stopped increasing at RTP time point in both groups with greater change being seen in TBI group.
Dual-task Condition: • No initial differences between groups for dual-task speed.

•
Overall gait speed was increased with a more prominent increase in TBI group.

•
After RTP gait speed stopped increasing in both groups.
Pitt et al. [34] Provide an objective description of angular velocity and acceleration profiles along orthogonal axes from one IMU situated on L5 vertebrae.  • Peak angular velocities during early single leg support distinguished TBI from healthy participants across the 2-month period.
Shan Chou et al. [27] Determine the possibility of quantitatively assessing dynamic stability that did not have an obvious neuromuscular origin in individuals who suffered a TBI.
• Unobstructed level walking. • Performed barefoot and a 6 m walkway.

•
Participants were allowed to lead over obstacles with preferred leg. • TBI suffers walked with significantly lower gait speed and presented with a shorter stride length in comparison to matched controls. • TBI elicits greater and faster medio-lateral centre of mass motion and significantly maintained medio-lateral separation distance between centre of mass and centre of pressure when compared to their matched controls.  • 11 m walkway stepping over a narrow obstacle and over a wide obstacle.

•
Obstacle dimensions set to ratio of participants maximum step height and length (Individualised difficulty). • Calculated over 2/3 steps with depth and height of obstacles set to 30% of the respective data.
3 trials were performed.
• Stroop when seated. • Participants familiarised themselves with walkway × 2/3 trials. • Participants were exposed to 5 trials of each physical condition (unobstructed, narrow, and wide obstacles). • 10 trials of each physical condition with visual stimuli randomly presented.
Williams et al. [22] Identify the most common gait abnormalities following a TBI and determine their rate of incidence.

reflective
• Pelvis and lower limb.
• Used to define joint centre location. • Participants performed walked over a 12 m walkway at a self-selected pace. Spatiotemporal.
• Individuals with TBI demonstrated significantly slower walking speed.

•
Additionally, TBI sufferers demonstrated differences in cadence, step length, stance time on affected leg, double support phase, width of base of support.

•
Biomechanically abnormalities were noted with TBI suffers exhibiting excessive knee flexion at initial foot contact. Kinetic.
• Push off Terminal Stance.
• Significantly increased trunk anterior/posterior amplitude of movement, increased anterior pelvic tilt, increased peak pelvic obliquity, reduced peak knee flexion at toe-off, and increased lateral centre of mass displacement were seen in TBI suffers. Table 3. Objective gait task paradigm.

Article Single Task Dual-Task Complex Task
Basford et al. [26] Belluscio et al. [24] Fino et al. [33] Fino, [29] Martini et al. [31] McFadyen et al. [25] Oldham et al. [30] Parker et al. [21] Parrington et al. [32] Pitt et al. [34] Shan Chou et al. [27] -obstacle crossing Vallée et al. [23] -obstacle crossing Williams et al. [22] A notable methodological limitation was found when considering gait impairment across the spectrum of TBI. Specifically, none of the reviewed studies examined gait deficits in TBI across the full spectrum of the injury (mTBI to sevTBI), with several studies combining TBI severities into a single TBI category rather than defining and assessing specific sub-groups. Therefore, there was no evidence on how gait differs between different severity levels of TBI, or if there are consistent deficits that become worse with increased TBI severity.

Discussion
To the authors knowledge, this review represents the first systematic synthesis of the literature examining gait impairment across the spectrum of TBI. Here we examined 13 studies that reported gait assessments in healthy controls and TBI participants specifically (i) how gait was measured; (ii) gait outcome measures and equipment used; (iii) how does TBI severity impact upon gait metrics.

Instrumentation
There was a lack of standardisation of instruments used to examine gait, therefore gait performance was quantified using several different technologies, but largely motion capture systems or IMUs were favoured over instrumented gait mats or force plates. Motion capture systems are a traditional approach to gait assessment, which are expensive, time consuming to set up, require specialist training and are often limited to specialist research centres or supervised laboratory surroundings, which may not be scalable to low resource settings [35]. Therefore, findings and conclusions drawn cannot be applied, relate or be replicated in other real-life contexts. Alternatively, IMUs have been suggested to overcome this challenge as they are easily implemented, low cost and portable [36], with excellent validity and reliability for gait assessment [37]. Progression to use of IMUs was seen in the majority of reviewed articles, as most used IMU's to measure gait in TBI [24,[30][31][32][33][34]38] which were reported to be a viable and reliable method of gait assessment. IMU's (and 3D motion capture) were shown to detect abnormalities in gait and provide an overall account as to an individual's gait cycle following a TBI, as all of the reviewed studies showed gait differences between those with TBI and healthy controls.

Outcome Measures
Gait can be characterised into spatiotemporal, kinematic, or kinetic outcome measures that are underpinned by selective neurological mechanisms [39]. There was a lack of consensus on the approaches used in assessing and reporting gait impairment in TBI, but studies generally reported spatiotemporal and kinematic outcomes. There were a wide range of gait outcomes reported between studies, but most studies reported on a limited amount of selected gait characteristics. The lack of standard assessment and reporting limits the generalisability of the findings, and does not support the use of quantitative methods of review reporting (i.e., meta-analysis) due to risk of bias. The most consistently reported outcome was gait speed (or velocity/pace). Gait speed is a measure of global walking performance [30,40], and is essentially an accumulation of multiple gait features that cannot be accurately quantified with a single outcome measure (e.g., speed [41]). As a result, gait speed in isolation is not a disease specific outcome and it does not reflect the subtle and precise underlying neural mechanisms involved in gait, which requires a more comprehensive examination of multiple gait outcomes. Despite gait speed being used in several studies we are unable to definitively report that it is useful at differentiating TBI groups, as the differences in methodologies (i.e., instruments, protocols, outcomes etc.) mean that we cannot directly compare outcomes across studies, and future work is needed to standardise procedures. Different underlying brain regions control different aspects of gait [42,43] and therefore with TBI of different regions and severity there is a need for comprehensive gait outcome measure assessment and reporting. Gait is underpinned by a complex system of neural cortical and sub-cortical networks [44] and impairment of any of the specific elements of the networks involved can result in impairment. Comprehensive reporting of gait in TBI literature is limited by the cohort sizes that have been examined, as there are many outcome measures that can be assessed and reported, but small samples sizes limit reporting capabilities and may lead to statistical errors. Many of the reviewed articles in this review had small TBI cohorts (n < 20) and as a result the number of outcomes reported may have led to inappropriate statistical analysis or reporting, due to the number of statistical comparisons [45]. There have been attempts to control for the number of comparisons made by using gait models within TBI cohorts (i.e., statistical analysis to reduce data in order to avoid statistical error issues [15]). However, only one of the reviewed articles [31] used a data reduction technique to assess gait outcomes, which highlights the emerging nature of gait assessment and reporting in this field. The development of an outcome measure framework would enable a hypothesis-driven research plan aiming to explain gait disturbance and examine the effect TBI on gait performance across the spectrum. Thus, leading to a greater consensus on most sensitive and accurate gait measure within TBI.

Protocols
There was a lack of consistency when reporting basic methodological procedures in classifying TBI severity, time scale (acute, sub-acute and chronic) and inclusion and exclusion criteria, which limits the generalisability and understanding of results. However, findings suggest that gait is impaired in TBI across the spectrum from mild (concussion) to severe injury status. Despite gait impairments being found, there was a lack of standardisation of procedures that limit the future implementation of gait assessment protocols.
All articles included in this review (n = 13) were undertaken in a gait laboratory setting. While laboratory assessments allow for complete experimental control that may uncover gait deficits, the environment may lack functional validity as it may not reflect 'real-world' gait [46]. Specifically, assessment of gait in a laboratory setting may fail to capture subtle deficits due to TBI that may ensue within usual environments (e.g., home, community, clinic, work, sports pitch/field, etc.), where there are multiple distractions and a vast array of environmental information to process to complete tasks effectively and safely [21,47]. None of the reviewed studies made the progression to examine gait outside of the laboratory within free-living environments, which has been conducted with physical activity and turning characteristics in previous TBI studies [48], which limits the understanding of the functional impact of potential gait impairments following a TBI.
There is no 'gold-standard' protocol for assessment of gait in TBI, as studies used a variety of tasks in an attempt to uncover deficits (e.g., single-task, dual-task, complex tasks etc.). The variety of experimental protocols employed across the included tasks of various complexity that sought to uncover specific TBI-related deficits. For example, single-task gait along a straight path was used in the majority of studies as this is thought to be a 'baseline' task that is controlled subcortical processing with minimal executive control [46], which can then be used to compare with more complex gait tasks that may elicit subtle deficits following a TBI. Dual-task gait was commonly used as a more difficult gait task that requires simultaneous cognitive and motor processing involving the executive function [49], which was compared to single task gait and healthy control gait to uncover TBI deficits. The least common gait assessment protocol was complex gait tasks, such as obstacle crossing, which require higher order cortical planning in order to plan and execute obstacle avoidance during walking [50]. The lack of a standardised protocol limits the generalisability of results across studies (i.e., even single-task walking was conducted for different times and distances), which means that quantitative analysis of the outcomes across studies is inappropriate until a common protocol is developed.
Gait provides a simple marker for an individual's overall health and is a widely accepted predictor of quality of life, decline of cognitive proficiency and falls [51]. Due to neurological decline and difficulties with age, gait becomes a more difficult task to perform efficiently and economically causing a transfer from an automatic to a cognitive level of control in order to execute and perform within a complex environment [39,52]. Increased task complexity for gait assessment was thought to increase the sensitivity of gait analysis for discriminating participants with TBI from healthy controls. While dual-task and complex task gait were not assessed within the same study protocol, there were similar outcomes when walking with these additional tasks (i.e., slower gait in TBI groups), which may indicate that adding any additional task could highlight impairments. However, despite the reviewed studies finding gait differences in TBI with the increased cognitive (or cortical) demand of dual-tasks and complex tasks, the benefit over using single-task gait (a simpler and quicker task) remains unknown, as gait deficits were detected across the TBI spectrum (mTBI, modTBI, SevTBI) using single-task. This is further complicated by the lack of consistency in the type of dual-task, and the set-up of the complex task (obstacle crossing) makes it difficult to directly compare outcomes across studies, and therefore difficult to make any clinical assessment recommendations. Future studies should consider whether their protocols require increased task complexity in order to detect gait deficits, as performance of a single-task walk may be sufficient to detect deficits when comprehensively investigating gait with data-driven digital technologies.

Outcome Interpretation
While the reviewed studies found differences in gait in those with TBI compared to controls, or within TBI when examined using tasks of increasing complexity, there were substantive methodological limitations that impact the interpretation of the reported outcomes. Specifically, none of the reviewed studies examined gait impairment differences between the various severity levels of the injury (mTBI, modTBI, sevTBI), with few studies examining modTBI. This is likely a result of the difficulties in defining the various levels of TBI, as there were variations in the reported diagnostic criteria (i.e., some acute diagnosis was 7 days, others only hours, and chronic ranged from months to years post-injury) and the specific individuals involved in the diagnostics within the reviewed studies (i.e., athletic trainer or a team physician, or merely medical recorded screen). Without being able to clearly define the severity and stage of TBI using a standardised criteria and then examine gait across these sub-groups, it is difficult to determine whether gait could be an effective biomarker for determining diagnosis, severity level, prognosis or monitoring of this neurological condition. Additionally, none of the reviewed studies included area under the curve or receiver operating characteristic curve analysis for specificity or sensitivity of gait characteristics in determining TBI gait differences with controls, which limits the interpretation of results (i.e., there may be differences but they may have low diagnostic value [53]. Therefore, future studies are needed to develop standard procedures for examining gait impairment in TBI, which will aid in the determination of gait as a marker of TBI.

Conclusions
Gait was shown to be impaired in TBI within the reviewed studies regardless of the severity or stage of the injury, but the specific impairments and the outcomes of clinical relevance are yet to be fully established across the spectrum of the condition. Further research is required to establish standardized methods for gait assessment in TBI, which will help to determine the gait deficits at each severity level of injury (mTBI, modTBI, sevTBI) in larger well-defined cohorts to establish findings.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10.3 390/s22041480/s1, Table S1: List of articles that did not meet inclusion criteria and reason for exclusion.