Comparative Efficacy of High-Intensity Training Versus Conventional Training in Individuals With Chronic Traumatic Brain Injury: A Pilot Randomized Controlled Study

Abstract

Numerous studies have evaluated the efficacy of interventions to improve locomotion after acute-onset brain injury, although most focus on patients with stroke, with less attention toward traumatic brain injury (TBI). For example, a number of studies in patients post-stroke have evaluated the effects of high-intensity training (HIT) attempting to maximize stepping practice, while no studies have attempted this intervention in patients with TBI. The purpose of this blinded-assessor randomized trial was to evaluate the effects of HIT focused on stepping practice versus conventional training on walking and secondary outcomes in individuals with TBI. Using a crossover design, ambulatory participants with TBI >6-months duration performed HIT focused on stepping in variable contexts (overground, treadmill, stairs) or conventional training for up to 15 sessions over five weeks, with interventions alternated >4 weeks later. HIT focused on maximizing stepping practice while trying to achieve higher cardiovascular intensities (>70% heart rate reserve), while conventional training focused on impairment-based and functional exercises with no restrictions on intensities achieved. Greater increases in 6-min walk test and peak treadmill speed during graded exercise testing were observed after HIT versus conventional training, with moderate associations between differences in stepping practice and outcomes. Greater gains were also observed in estimates of aerobic capacity and efficiency after HIT, with additional improvements in selected cognitive assessments. The present study suggests that the amount and intensity of stepping practice may be important determinants of improved locomotor outcomes in patients with chronic TBI, with possible secondary benefits on aerobic capacity/efficiency and cognition. Clinical Trial Registration-URL: https://clinicaltrials.gov/; Unique Identifier: NCT04503473.

Introduction

Recent estimates suggest the incidence of traumatic brain injury (TBI) is approaching 333/100,000 individuals in the United States,¹ potentially resulting in long-term physical and cognitive deficits. In patients with moderate to severe TBI, limitations in mobility and reduced physical activity often contribute to reduced independence and increased risk of cardiovascular diseases.^2,3 Rehabilitation interventions that improve physical function and mobility should lead to additional gains in cardiovascular health and independence,⁴ although data delineating the comparative efficacy of interventions to improve mobility after TBI are limited.

A recent systematic review identified only 23 trials that assessed the comparative efficacy of any physical training strategy, with substantial variability in outcomes and interventions such that cumulative findings were inconclusive.⁵ In studies comparing body-weight supported treadmill training^6,7 or robotic-assisted walking training⁸ to alternative interventions, there were only minimal differences in locomotor outcomes (independence, speed, or distance walked).

Other studies indicate interventions attempting to target higher neuromuscular or cardiovascular intensities during cycling,⁹ cycling or walking combined with strength training,¹⁰ or ballistic strength training¹¹ also elicited limited gains in walking function.

Interestingly, previous work in animal models of TBI suggest aerobic training focused on locomotor activities may improve specific measures of both cognition and mobility,^12,13 particularly through enhanced neurotrophin expression (i.e., brain-derived neurotrophic factor, or BDNF). In patients with moderate to severe TBI, however, evidence of gains in mobility and cognition with aerobic locomotor activities is scarce.

Conversely, studies performed in other acute-onset neurological diagnoses, including stroke and incomplete spinal cord injury (iSCI), may provide insight into potentially effective strategies for individuals with TBI. Specific research suggests large amounts of stepping practice, particularly at higher cardiovascular intensities (70–80% heart rate [HR] reserve), can improve gait speed and endurance.^14–16 Such training appears to consistently improve walking outcomes compared with more conventional strategies or lower-intensity stepping practice.¹⁷

In addition, high-intensity training (HIT) in variable contexts, in which patients perform varied walking tasks (e.g., multi-directional stepping, walking over uneven, narrow, or compliant surfaces) in different environments (overground, treadmill, stairs) can elicit gains in locomotor function,^18–20 as well as aerobic capacity and/or efficiency,^21,22 balance, balance confidence, and quality of life.^20,23

In other aerobic training studies, gains in cognitive assessments have also been observed.^24,25 While some researchers have argued that changes in function after acute-onset brain or spinal injury may be because of similar neuroplastic mechanisms in spared neural pathways,^17,26 there are no similar studies in patients with TBI.

The purpose of this study was to assess the comparative efficacy of HIT focused on stepping in variable contexts (i.e., different tasks and environments) on locomotor function in individuals with TBI. In this pilot, randomized crossover design, changes in outcomes after 15 1h sessions of HIT in variable contexts were compared with gains after an equivalent number of conventional training sessions providing gait and balance training, strengthening and aerobic training exercises at lower intensities. Primary outcomes were changes in gait speed and distance, with secondary measures of balance, aerobic capacity and efficiency, quality of life, and measures of cognition. Consistent with data from other populations, we hypothesized that HIT focused on walking in variable contexts would lead to greater increases in gait speeds and distances compared with conventional interventions.

Methods

Study sample and design

Individuals with a history of acquired (non-vascular) brain injury of at least six months duration were recruited from outpatient clinics of local rehabilitation systems. While definitions of TBI vary,²⁷ we utilized the Brain Injury Association of America criteria,²⁸ which defines TBI as an “alteration of brain function or other evidence of brain pathology caused by an external force.” Additional criteria consisted of the following: 18–75 years old, ability to walk with minimal or no physical assistance at self-selected speeds <1.4 m/sec with or without below-knee braces and assistive devices as needed.

Exclusion criteria included uncontrolled cardiorespiratory or metabolic disease that limited exercise participation, active heterotopic ossification, recurrent history of lower extremity fractures, previous orthopedic or other neurological injury that may affect locomotor function, history of botulinum toxin injection in the upper leg <3 months prior, and current enrollment in physical therapy. All participants provided informed consent, and all procedures were approved by the local ethics committee.

Participants were randomized to receive either HIT focused on stepping tasks in variable contexts or conventional interventions followed by the alternate training paradigm initiated at least one month later. The crossover design was used previously in pilot studies in individuals post-stroke and iSCI^21,29,30 and employed here to increase the efficiency of the data collected and to minimize potential confounders of TBI-related co-morbidities on outcomes (i.e., impaired cognition, attention, motivation, or vision).

Given the lack of studies in TBI, effect sizes and targeted sample size were modeled after previous crossover studies comparing HIT with alternative interventions.^21,29,30 Namely, results from two SCI studies indicate approximate gains of 40 ± 26 m in the 6-min walk test and 0.14 ± 0.12 m/sec in gait speed with HIT, whereas more conventional, lower-intensity training resulted in gains of 10 ± 1 8 m and 0.04 ± 0.06 m/sec, respectively. Power calculations using these differences revealed 16 individuals were necessary to achieve significant between-group differences (82–98% power). Given the uncertainty of anticipated changes in TBI, a slightly higher sample size was targeted.

Following informed consent, participants completed baseline testing, were stratified by walking speeds (<0.5 m/sec >0.5 m/sec) and block randomized (four per block) via a concealed allocation sequence to receive HIT in variable contexts or conventional training first.

Interventions

Participants received up to 15 1h training sessions over 4–5 weeks of either intervention, with training terminated after five weeks even if 15 sessions were not achieved. The number of sessions completed during the first training epoch was targeted during the second epoch. All interventions were supervised by licensed therapists. Stepping training activity was measured using ankle-worn accelerometers on the more impaired limb³¹ (StepWatch, Modus Inc, Washington DC).

HIT in variable contexts

The primary goals of HIT in variable contexts were to maximize stepping activity at higher cardiovascular intensities for up to 40 min of scheduled 1h sessions, with rest breaks as needed, and efforts to increase the difficulty of walking tasks as tolerated.^18–20 Targeted HR ranges were >70% HR reserve calculated using the Karvonen formula and utilizing age-predicted HR maximum (HR_max) rather than peak observed HR (observed HR_max) during graded exercise testing (see Outcomes below); this strategy was performed secondary to blunted HR responses during graded exercise testing (GXT) observed in individuals with neurologic injury from neuromuscular impairments that limit peak locomotor and cardiovascular capacity.^14,19

Ratings of Perceived Exertion (RPEs) of 15 (“hard”) or greater were also targeted as secondary measures of intensity when participants could not achieve the desired HR ranges secondary to autonomic dysfunction, medications, or individual variations.³² The HRs were recorded using HR monitors (H10, Polar, Inc, Finland) and documented with RPEs every 3–5 min.

The HIT sessions were divided into ∼10-min increments between four different tasks including speed-dependent (forward) treadmill training, skill-dependent treadmill training, overground training, and stair climbing as detailed previously.¹⁸ Briefly, speed-dependent treadmill training attempted to achieve the highest-possible treadmill speed to achieve the targeted HR ranges during forward walking. Patients wore a safety harness system and were provided body-weight support or physical assistance only as needed to ensure successful stepping, defined as achieving positive step lengths, limited stance-phase limb collapse, and sagittal/frontal plane stability.

Skill-dependent treadmill training was performed by applying perturbations to challenge relevant biomechanical deficits, including stance, postural stability, propulsion, and limb swing, while maintaining the targeted HRs.¹⁸ Tasks included walking in multiple directions, over inclines and obstacles, and/or with weighted vests and leg weights as tolerated. Task-difficulty was increased when participants were able to perform activities successfully and was reduced when participants were not successful for 3–5 consecutive steps.

Overground training focused on walking at faster speeds and/or performing variable stepping tasks (e.g., walking over obstacles, uneven or compliant surfaces), with use of a gait belt or overhead suspension system. Stair climbing was performed over static or rotating stairs (Stairmaster, Vancouver, WA) with attempts to use reciprocal gait patterns and progression to higher speeds and reduced handrail use as able.

Conventional training

Conventional training included various traditional rehabilitation exercises performed over 40 min of 1h sessions. The types and amounts of activities were derived from observational data of clinical therapy sessions,^33–35 and included lower limb strengthening activities (∼10 min), balance activities/pre-gait (∼10 min), locomotor and cycling activities (i.e., cardiorespiratory training, ∼10 min), and combined stretching and transfers (∼10 min).

Strengthening exercises focused on bilateral hip, knee, and ankle flexors and extensors, using single and multi-joint exercises. Increased weights and sets/repetitions (2–4 sets, 8–20 repetitions) were performed and progressed as tolerated. Balance activities included static (i.e., non-walking) tasks such as standing on unstable surfaces, stepping in place, and reaching tasks.

Cardiopulmonary exercises included some walking practice consistent with conventional interventions, but with focus on normalizing gait patterns, with additional recumbent cycling activities performed. During transfer training, bed, sit-to-stand, lateral, or floor transfers were practiced, with stretching performed as warranted per patient report of tightness. While HRs and RPEs were monitored throughout training, there were no attempts to maintain specific intensity ranges, because HRs are rarely monitored during clinical therapy.³⁶

Outcomes

Participants were assessed before and after each training epoch. Primary measures included self-selected (SSS) and fastest speeds (FS) over short distances (e.g., 10 m walk test), distance covered during the 6MWT and peak treadmill speed (peak TM) during a GXT.^19,21,30

Using blinded assessors, SSS and FS were performed with instructions to “walk at your normal comfortable pace” and “walk as fast as you safely can,” respectively, over two trials with footfall patterns recorded by a pressure-sensitive mat (Zeno Walkway, Protokinetics, Havertown, PA). The 6MWT test was performed around a continuous hallway with instructions to “cover as much ground as possible.” During all tests, participants used their customary below-knee bracing and assistive devices, which was similar for all assessments.

Evaluation of peak TM was not blinded and assessed during modified GXTs with electrocardiography (ECGs) and cardiorespiratory data collected (Quark-C12 and K5, Cosmed, Chicago, IL). During GXTs, participants began walking on a motorized treadmill at 0.1 m/sec for 1 min with speed increased 0.1 m/sec every minute until the subject experienced significant gait instability, requested to stop, or if significant ECG abnormalities were observed.³⁷

Participants with balance deficits often needed to hold onto the handrails but could only minimally support their body weight. Participants wore a safety harness without body weight support, with HRs measured continuously and blood pressures measured before and after testing. Peak TM was defined as the highest speed achieved for 1 min.

Secondary measures included both objective and subjective clinical measures, metabolic responses during the GXTs, and selected cognitive assessments. Additional blinded clinical measures included the Berg Balance Scale (BBS) and 5-times sit-to-stand (5XSTS), and subjective measures included the Activities-specific Balance Confidence (ABC) scale and the Medical Outcomes Short Form-36 questions (SF-36).

Metabolic assessments focused on evaluation of peak aerobic capacity (VO₂peak) and changes in gait efficiency,^21,22 because both can change with exercise training and changes in efficiency can mask gains in capacity. The VO₂peak was calculated as the average VO₂ during the last 30-sec at the peak TM. Changes in gait efficiency were estimated by comparing VO₂ at the highest treadmill speeds at baseline to VO₂ at matched speeds achieved at post-testing (VO₂match). The net gain in VO₂ improvements (VO₂gain) was calculated as the difference between these variables (ΔVO₂peak-ΔVO₂match), and is correlated with walking outcomes.^3,29

In addition, cognitive assessments evaluated before and after each training epoch included the Trail Making Tests A and B (TMT-A, TMT-B),^38,39 as well as the Digit Symbol Substitution Test (DSST).⁴⁰ The TMT-A requires the patient to connect numbered circles in ascending order as quickly as possible and can assess cognitive processing and visuomotor function. The TMT-B was utilized to assess similar constructs in addition to task-switching and cognitive flexibility, and evaluates the speed with which patients connect sequential numbers and letters alternately as quickly as possible.

Both assessments have a ceiling duration of 180-sec to complete, although additional time was provided given the cognitive deficits of the patient population; data were analyzed both with and without the 180-sec cut-off duration (values >180 sec set to 180 sec). The DSST was also used to assess processing speed, attention, visuoperceptual function, as well as motor performance associated with drawing as many symbols that matched a numeric code within 90 sec.

Finally, given the pilot nature of the study and lack of previous data applying HIT in TBI, measures of safety and feasibility were also documented. Safety was evaluated by tallying rates of serious adverse events (i.e., death, falls with injury, or cardiovascular events) or minor adverse events (musculoskeletal pain, falls without injury outside of training, light-headedness or dizziness, loss of consciousness, excessive shortness of breath, or episodes of hypertension or angina that limited training). Measures of feasibility included the ability to attain higher HRs and stepping amounts and rates consistent with previous HIT studies in other diagnoses.

Analysis

Demographic and training parameters were collated and presented as averages or frequencies as appropriate. Training parameters of interest included average number of sessions, minutes of stepping per sessions, steps/session, and steps/min. For measures of intensity, we identified both average HRs and RPEs per session, with HRs reported as percentage HRRs and calculated using both age-predicted and observed HR_max during the first GXT. Differences in training parameters were compared between interventions using paired t tests.

For statistical analysis, data were normally distributed (Kolmogorov-Smirnov) with the exception of Trail-Making Tests A and B, although parametric descriptive statistics were provided for simplicity of data presentation, and both parametric and non-parametric statistics were used for Trail Making Tests (Wilcoxon Signed Rank of change scores between training interventions).

On-protocol and intent-to-treat analysis for primary assessments were similar for all participants because all completed the study. One individual presented with surgical removal of a tumor, and data analysis both with and without this individual were similar, and the full data set is presented. Data are presented as mean ± standard deviation in the text and tables with standard errors in the figures.

Baseline demographics and clinical characteristics were presented for participants in their first training epoch (HIT or conventional). Statistical analyses of primary outcomes (SSS, FS, 6WMT, peak TM) were performed using mixed model analyses of variance (ANOVAs), with main factors of time (baseline and post-training) and training intervention (HIT and conventional; both repeated), with order not repeated (HIT vs. conventional first).

Bonferroni corrections were performed for those primary outcomes (α = 0.0125). For these measures, analysis of outcomes for only the first training paradigm was performed using a two-way repeated measure ANOVA (time × training). For secondary clinical, metabolic, and cognitive measures, three-way ANOVAs were performed as described above without Bonferroni corrections because of the pilot nature of the investigation (α = 0.05).

The secondary assessments with missing data included multiple cognitive assessments in those with substantially impaired cognition, and two metabolic assessments during GXTs because of equipment (O₂ sensor) malfunction. Missing metabolic data were imputed by regression imputation utilizing observed changes in VO₂peak versus baseline VO₂peak.

For cognitive assessments, only data from those individuals who completed at least three of the four assessment time periods were included, with missing data from single tests also replaced by regression imputation. Serious and minor adverse events were categorized separately, with χ2 analyses to compare between-group differences.

Associations between differences in stepping or intensity metrics in each intervention and changes in primary outcomes were evaluated using Spearman correlation analyses. Specific differences in steps/session were calculated by subtracting the average steps/session during HIT versus conventional training; similar calculations were made for %HR reserve using age-predicted HR_max. These values were plotted against differences in changes in outcomes after each training epoch²¹ (e.g., Δ6MWT after HIT minus Δ6MWT after conventional training).

Results

Seventeen of 21 individuals who were screened and consented fulfilled all inclusion criteria, completed initial assessments, and were randomized. All randomized participants completed the study (i.e., no dropouts). Figure 1 provides a CONSORT diagram, with demographic and baseline characteristics provided in Table 1 for those randomized to HIT or conventional training first.

FIG. 1.

CONSORT flow diagram of randomized crossover design. All participants enrolled and randomized completed the study.

Table 1.

Demographics and Baseline Characteristics

	Conventional first (n = 8)	HIT first (n = 9)
Demographics
age (years)	40 ± 22	46 ± 15
gender (male/female)	7/1	8/1
race (white/other)	6/2	7/2
BMI (kg/m2)	31 ± 2.1	27 ± 5.6
duration post-injury (years)	3.2 ± 3.5	2.8 ± 1.7
mechanism of injury
motor vehicle accident	7	8
gunshot wound	1	0
surgical tumor resection	0	1
Clinical characteristics
ankle foot orthosis (yes/no)	1/7	2/7
assistive devices (yes/no)	4/4	4/5
baseline SSS (m/s)	0.67 ± 0.29	0.65 ± 0.32
baseline FS (m/sec)	0.88 ± 0.42	0.97 ± 0.52
baseline 6MWT (m)	261 ± 132	289 ± 195
baseline treadmill speed (m/sec)	1.1 ± 0.48	0.97 ± 0.62
Medication use
anti-spastics (yes/no)	3/5	3/6
antidepressants (yes/no)	3/5	6/3
antihypertensives (yes/no)	4/4	3/6

BMI, body mass index; SSS, self-selected speed; FS, fastest possible speed.

Training parameters

Table 2 details average training parameters for HIT and conventional training. Despite A similar number of sessions completed between groups, differences were observed for all stepping metrics (all p < 0.01). Sessions that focused on maximizing stepping practice averaged 1109 ± 582 greater steps/session during HIT versus conventional training with greater durations of stepping activity (10 ± 5.2 min difference) and higher stepping rates (15 ± 6.5 steps/min difference).

Table 2.

Training Parameters

	Conventional training	HIT in variable contexts	p
Number of sessions (no.)	14 ± 1.2	14 ± 1.4	0.84
Steps/sessions (no.)	1538 ± 613	2647 ± 1095	< 0.001
Stepping minutes/session	30 ± 10	40 ± 11	< 0.001
Stepping rate (steps/min)	47 ± 13	62 ± 15	< 0.001
Average HRs (beats/min)	120 ± 18	151 ± 13	< 0.001
Average HRR (% predicted HRmax)	36 ± 20	71 ± 14	< 0.001
Average HRR (% observed HRmax)	62 ± 32	124 ± 34	< 0.001
Average RPE (a.u.)	13 ± 2.1	16 ± 1.2	< 0.001

HIT, high-intensity training; HR, heart rate.

Intensity metrics were also higher during HIT versus conventional training, with a mean difference of 31 ± 15 beats/min in HRs and 2.7 ± 2.0 difference in RPEs. Average HR reserves calculated using age-predicted HR_max revealed participants achieved 71 ± 14% HR reserve during HIT versus 36 ± 20% during conventional training, with larger values when HR reserve is calculated using observed HR_max during the first GXT (124 ± 34% HR reserve during HIT; 62 ± 32% during conventional training).

Primary outcomes

Analyses of primary outcomes are presented in Table 3 and Figure 2, with changes in walking outcomes revealing selective differences between groups. No significant time × training interactions were observed for changes in SSS or FS (ΔSSS: 0.08 ± 15 vs. 0.02 ± 14 m/sec and ΔFS: 0.08 ± 14 vs. -0.01 ± 18 m/sec, respectively both p > 0.05; Fig 2A,B), although there was a trend for a significant time × training × order effect for FS favoring HIT completed first (p = 0.04).

Table 3.

Baseline and Post-Training Data for Conventional Training vs HIT in Variable Contexts

	Conventional training		HIT in variable contexts		p
	BSL	POST	BSL	POST	time × training	time × training × order
Primary assessments
SSS (m/sec)	0.72 ± 0.38	0.73 ± 0.37	0.67 ± 0.30	0.75 ± 0.41	0.31	0.25
FS (m/sec)	0.98 ± 0.56	0,99 ± 0.54	0.92 ± 0.48	1.00 ± 0.55	0.21	0.04
6MWT (m)	298 ± 182	296 ± 177	277 ± 162	325 ± 182	0.002*	0.72
Peak TM (m/sec)	1.12 ± 0.59	1.10 ± 0.63	1.05 ± 0.57	1.28 ± 0.69	<0.001*	0.24
Secondary clinical assessments
BBS (a.u.)	39 ± 13	38 ± 14	39 ± 14	38 ± 14	0.84	0.05
5XSTS (s)	20 ± 22	18 ± 18	18 ± 11	16 ± 11	0.80	0.03*
ABC (a.u.)	68 ± 22	75 ± 15	67 ± 21	73 ± 18	0.72	0.22
SF-36 (a.u.)	45 ± 9.6	45 ± 9.3	47 ± 11	46 ± 8.9	0.77	0.72
Secondary metabolic assessments
VO2peak (ml/kg/m)	17 ± 6.3	17 ± 7.5	17 ± 8.2	19 ± 9.2	0.11	0.30
VO2match (mL/kg/m)	17 ± 6.3	16 ± 6.8	17 ± 8.2	15 ± 6.7	0.60	0.30
VO2gain (mL/kg/m)	-	1.0 ± 1.6	-	3.7 ± 3.7	<0.001*	0.97
Secondary cognitive assessments
Trail Making A (s)	86 ± 53	72 ± 53	67 ± 45	62 ± 46	0.10	0.03*
Trail Making B (s)	128 ± 60	102 ± 58	104 ± 56	103 ± 61	0.08	0.07
DSST (#)	25 ± 16	23 ± 18	24 ± 15	29 ± 18	0.03*	0.14

HIT, high-intensity training; BSL, baseline; POST, post-training; SSS, self-selected speed; FS, fastest speed; TM, treadmill.

Significant differences for interaction effects of time × group and time × group × order are provided.

FIG. 2.

Differences in primary locomotor outcomes of (A) self-selected speed (SSS), (B) fastest possible speed (FS), (C) 6-min walk test (6MWT), and (D) peak treadmill speed (peak TM); baseline (BSL) and post-training (POST) indicated for both first and second training interventions in order received; dark lines indicate high-intensity training (HIT) in variable contexts, dashed lines indicate conventional training, filled squares denote HIT first, conventional second; open triangles denote conventional first, HIT second.

Conversely, significant time × training interactions were observed for 6MWT (Δ6MWT: 48 ± 41 vs. -2.0 ± 40 m; p < 0.01; Fig 2C) and peak TM (Δpeak TM: 0.23 ± 0.17 vs.-0.02 ± 0.13 m/sec; p < 0.01; Fig 2D), with no order interaction effects. Analysis of parallel-group changes in outcomes from the first training conditions only indicated only peak TM was significantly different (p < 0.01), whereas changes in 6MWT (p = 0.14), SSS (p = 0.72), and FS (p = 0.40) were not.

Evaluation of potential associations between primary outcomes and training parameters are depicted in Figure 3. Significant moderate to low correlations were observed for differences in Δ6MWT versus steps/session and %HR reserve (Fig. 3A,B; both p = 0.02). Associations between peak TM and these training metrics were lower and approached significance, while other primary outcomes were not well correlated.

FIG. 3.

Correlations between differences in steps/sessions (A) and heart rate reserve (B) between training conditions (high-intensity training [HIT]—conventional) vs. differences in changes ( ) in selected outcome measures (i.e., D6MWT = changes after HIT minus changes after conventional); both p < 0.05).

Secondary outcomes

Secondary clinical measures are also presented in Table 3, with no significant time × training or time × training × order interactions, with the exception of a potential order interaction for 5XSTS favoring HIT performed first (p = 0.03). Significant differences were, however, observed for specific metabolic parameters during GXTs. Namely, differences in changes in VO₂peak and VO₂match were not different between training paradigms (both p > 0.05), although significant between-group differences in VO₂gain were observed. Specific changes included greater improvements after HIT (3.7 ± 3.6 mL O₂/kg/min) with minimal changes after conventional training (1.0 ± 1.6 mL O₂/kg/min; p < 0.01).

For cognitive assessments, differences were noted after HIT versus conventional training, although responses varied. For TMT-A and TMT-B, only 14 and 11 participants, respectively, could finish the assessments within 180 sec. Analysis of all tests regardless of completion duration suggest no time × training interactions, with significant time × training × order interactions for both TMTs (both p = 0.03) when participants performed conventional training first. These changes were characterized by large improvements (>40 sec decreased time) in 2/5 participants in this subgroup.

Similar differences were observed when using the 180 sec cutoff duration for TMT-A (p = 0.03), but not TMT-B (p > 0.05; Table 3). Subsequent post hoc Wilcoxon Signed Rank Tests using change scores in TMT-A and TMT-B reveal no differences between HIT and conventional training. For the DSST, three individuals did not perform the test secondary to impaired cognition, although available data suggest significant time × training interactions favoring HIT (DSST:2.3 ± 5.6 points) vs. conventional training (-1.2 ± 5.8 points, p = 0.03), with no order interaction effects.

For estimates of safety, frequencies of adverse events were tallied, revealing no serious adverse events and only selected minor adverse events in each training epoch. Specific events included similar number of falls between groups (nine outside of HIT vs. eight outside of conventional training), complaints of dizziness (one during HIT vs. three during conventional training), hypertensive episodes (one vs. zero), and substantial fatigue (four vs. three). A greater number of complaints of soreness (21 vs. 9; p = 0.03) was observed, however, with HIT versus conventional training, and most often related to increased muscular soreness during HIT.

Discussion:

The present findings suggest HIT in variable contexts elicited greater gains in 6MWT and peak TM compared with conventional training in participants with chronic TBI. The HIT paradigm was characterized by substantially greater amounts and rates of stepping practice at higher HRs and RPEs, with moderate correlations between differences in walking gains and stepping metrics. Additional findings included greater improvements in VO₂gain with HIT, and a potential impact on cognitive function.

Estimates of feasibility in this study included evaluation of the amount of stepping practice and cardiovascular intensities achieved during HIT compared with previous trials in other diagnoses (i.e., stroke, SCI). Mean stepping activities achieved during HIT (∼2500 steps/session) approximated stepping amounts observed in ambulatory patients post-stroke (2300–2700 steps/session)^19,29 and incomplete SCI (2100–2300 steps/session).^21,30 In addition, the average HR reserves calculated with both age-predicted HR_max (71 ± 14 % HR reserve) and observed HR_max (124 ± 34%) were similar to published data in patients post-stroke (67 ± 8.5% and 111 ± 47%, respectively).¹⁹

Further gains in locomotor outcomes presented here are consistent with previous research providing HIT. For example, changes in SSS and FS after HIT in individuals with stroke or SCI may be observed, although gains in 6MWT (∼50 m) and peak TM (0.20 m/sec) are more consistently demonstrated.^21,29,30 The smaller gains after conventional training were also consistent with previous studies,^20,41 with significant associations between differences in training parameters (stepping, intensity) and changes in selected locomotor outcomes.

The present findings appear to support the hypothesis that individuals >6 months after an acute-onset neurologic injury respond similarly when provided similar training paradigms.^17,26

While most secondary assessments were not different between training paradigms, greater gains in VO₂gain are of interest, because previous studies in individuals with TBI reveal inconsistent gains in estimates of aerobic capacity with exercise training.^9,10,42 Differences are similar, however, to studies with individuals post-stroke and incomplete SCI after HIT compared with conventional or lower-intensity walking training.^21,22

Participants in the present study increased their VO₂peak by 2–3 mL O₂/kg/min at higher peak TMs with HIT, but also improved VO₂match by 1–2 mL O₂/kg/min (i.e., decrease in VO₂ at matched speeds). Achieving greater peak TM is typically associated with increased propulsive, stance, or limb swing powers in neurologically intact individuals and increased metabolic output,^43,44 although only limited data have been presented in patients with neurologic injury.

Few studies have evaluated biomechanical mechanisms that subserve changes in gait efficiency, and changes may be because of altered spatiotemporal or joint kinematic patterns and/or changes in neuromuscular timing.²² Delineation of biomechanical contributions to changes in aerobic capacity and efficiency could identify mechanisms for improved gait function.

Improvements in the DSST after HIT versus conventional training are also of interest given the prevalence of cognitive deficits in patients with TBI.⁴ Gains observed after HIT without a significant order interaction suggest these changes were likely related to the interventions rather than a testing effect. While one study has reported gains in DSST after a single aerobic training session,⁴⁵ other studies in patients post-stroke have not shown similar changes with sustained training with DSST or similar measures.^46,47

Potential reasons for improvements in cognition include increased BDNF secretion and utilization as suggested in animal studies,¹² and higher intensity activities are associated with elevated plasma BDNF levels in patients with SCI⁴⁸ and stroke.⁴⁹ No similar studies, however, have evaluated these responses in individuals with TBI. Notably, other cognitive assessments, including the TMTs, did not show consistent differences and additional assessments of cognition are needed to evaluate the consistency of these findings.

Evaluation of safety using frequency and type of adverse events revealed no serious events consistent with previous studies using HIT.^19,21,29,30 An increased incidence of minor adverse events was observed and was related primarily to lower extremity muscle soreness with HIT.

These observations are consistent with the phenomenon of delayed onset muscle soreness (DOMS), which may be expected in sedentary populations with or without neurological injury when exposed to higher frequencies or intensities of exercise training. In general, DOMS is an indicator of microtrauma of muscle and surrounding connective tissues with repeated use,^50,51 which initiates an inflammatory cascade that subsequently results in tissue repair and increased strength. As such, these findings are expected, but may ultimately lead to improved neuromuscular power subserving gains in locomotor function.⁴³

Limitations of this study include the small sample size and lack of blinded assessors for the GXTs, although testing protocols utilized specific criteria for testing termination. Given limited data in individuals with TBI, this pilot study was designed to determine whether any potential effect would be observed (see also ^21,30), and the current data will assist with power analyses for potential further evaluation in larger populations.

While the use of the crossover design does increase the efficiency of the subject sample recruited, the potential for carryover effects remains a concern in rehabilitation trials.⁵² In this patient population, however, specific behavioral or cognitive deficits may influence physical performance and responsiveness to, or compliance with, testing instructions or training interventions. As such, the crossover design may help control for these population-specific factors.

Another limitation is the lack of ability to discern how specific training parameters, such as intensity or specificity of stepping practice, influenced the observed differences in outcomes. Previous studies in both stroke and SCI have attempted to delineate the relative contributions of these different variables,^19,21,30 although there are no comparable studies in individuals with TBI. For example, it is unclear whether greater amounts of stepping practice contributed to the observed gains in outcomes, or whether working at higher intensities was responsible.

Future studies should attempt to understand whether these findings are reproducible and how different training parameters affect physical, physiological, and cognitive outcomes in this patient population.

Conclusions

The present study represents the first investigation to delineate the comparative efficacy of HIT versus conventional training in individuals with chronic TBI. Significantly greater gains in 6MWT and peak TM were observed after HIT and related to stepping metrics achieved during each intervention. Additional findings included greater gains in combined measures of peak O₂ capacity and efficiency, as well as potential benefits cognition.

The present and previous results suggest that such training interventions may be important components of rehabilitation interventions to improve walking and possibly other secondary outcomes for individuals with chronic TBI, consistent with data in patients with other acute-onset neurological injuries.

Transparency, Rigor and Reproducibility Summary

Details of this randomized cross-over study design were preregistered on ClinicalTrials.gov (NCT04503473). Calculation of the estimated sample size required to observe significant differences in two primary outcomes (fastest walking speed and 6-min walk test). revealed 16 individuals would provide 82-98% power. Estimates of outcomes were derived from previous studies in patients with neurologic injury affecting bilateral lower extremities (incomplete SCI). Additional subjects were recruited and enrolled secondary to the uncertainty of changes in outcomes in TBI. Twenty-one individuals were recruited and screened, with 4 individuals unable to complete the study due to inability to provide informed consent (n = 1), was not interested in participating (n = 1), and inability to evaluate medical records (n = 2). A total of 17 subjects were enrolled and completed training and testing sessions. Participants were stratified by self-selected gait speed (<0.5 m/s or >0.5 m/s), and randomization was performed by a random number generator using block of 4 individuals. Baseline differences between groups for all demographics and primary and secondary outcomes were similar (Tables 1 and 3). Changes in outcomes were not observed for gait speeds given smaller changes and larger variability than expected with HIT. Gains for 6MWT and peak TM, however, were larger than observed in previous studies, and significant differences were observed. All primary and secondary clinical outcomes were assessed by investigators blinded to the intervention assignment with the e27ception of peak TM. All materials required to perform the assessments and interventions may be available upon request. The Kolmogorov-Smirnov test were used to assess normality and revealed all variables except Trail-Making Tests A and B had normal distributions. Correction for multiple comparisons was performed for the primary clinical outcomes. The findings have not yet been replicated or externally validated. Data are available upon reasonable request.

Authors' Contributions

AP: Investigation; Methodology; Writing – review & editing. CEH: Conceptualization; Formal analysis; Investigation; Methodology; Writing – review and editing. JKL: Investigation; Methodology. LHS: Investigation; Methodology. EI: Investigation; Methodology. MS: Investigation; Methodology. CJV: Investigation; Methodology. SKC: Conceptualization (supporting). TGH: Conceptualization; Formal analysis; Investigation; Methodology; Supervision; Writing – original draft; Writing – review and editing.

Funding Information

Funding was provided by the Indiana Brain and Spinal Cord Injury Research Foundation and the National Institute of Health R01-NS079751.

Author Disclosure Statement

TGH is co-owner and CEH is employed by the Institute for Knowledge Translation. For the remaining authors, no competing financial interests exist.