INTRODUCTION

The severity of a traumatic brain injury (TBI) is clinically categorized as mild, moderate, or severe (1-4), which is traditionally based upon clinical measures for acutely altered consciousness combined with imaging findings (5, 6). The Department of Defense/Veterans Administration definition of TBI severity (which does not include penetrating injury) was used for this study (2). The incidence of mild-to-moderate TBI (mmTBI) accounts for the great majority (>80%) of global TBIs (7), and typically both mild and moderate injuries have better outcomes than severe TBI (2, 8); however, nearly one quarter of individuals suffering from a mild TBI have persistent postconcussive symptoms (9). Symptoms including cognitive, somatic, and emotional problems may persist ≥1 year post injury, interfering with daily activities and resulting in disability, lost workplace productivity, and a significant burden on the individual and to society (9).

Current treatment options for balance impairment after mmTBI are limited. Approximately 14.6 million people, globally, suffer from a chronic balance deficit after mmTBI (10-12), stressing the unmet need in this patient population. Traditional treatment involves physical therapy (PT), often vestibular based. Evaluations of this approach have indicated that it can result in full recovery (i.e., return to normal balance and gait function) in many, but not all, patients; therefore, controlled studies are needed to draw definitive conclusions with regard to optimal treatment (13).

For many patients, the signs and symptoms of mmTBI resolve with time, and patients can return to normal daily activities; however, 10%–50% of patients experience one or more chronic symptoms, such as difficulty with memory, attention, and executive function; dysregulation of sleep, speech, mood, eye movement control; and increases in headache frequency/magnitude (14-19). Postural instability or imbalance can also persist after TBI (10, 20), which has a major negative impact on functional status, capacity to return to work, and quality of life (17, 21-25).

Recently, neuromodulation targeting cranial nerves via noninvasive neurostimulation has emerged as a potential treatment approach for patients with chronic pain and neurological disorders (26). Neurostimulation devices targeting the trigeminal nerve or vagus nerve by application to the forehead or neck, respectively, have been approved by the US Food and Drug Administration to treat migraine and/or cluster headache (26-29). Both noninvasive neurostimulation and motor training/physical exercise have been shown to independently promote neuroplastic changes (30). When considering combining these approaches to enhance outcomes, it has been proposed that neurostimulation may prime neuronal networks, recruit task-specific synaptic connections, and/or consolidate new signal pathways when used before, during, or after a physical rehabilitation task (30).

Translingual neurostimulation (TLNS) is a noninvasive method used to elicit neural changes by stimulating cranial nerves via the anterior third of the tongue (31, 32). The tongue is a desirable organ for stimulation because it is highly innervated, provides an environment with a constant pH level and temperature, and has a low excitability threshold (31). TLNS delivers sequenced patterns of electrical stimulation to sensory fibers approximately 300–400 μm from the tongue surface to create action potentials through multiple existing synaptic connections (31). This stimulation engages a flow of signals through naturally occurring neural networks via activation of the lingual branch of the trigeminal nerve (cranial nerve Vc) and the chorda tympani branch of the facial nerve (cranial nerve VII) (33, 34). Although the primary function of the chorda tympani is to relay information regarding taste (31), there is considerable intermingling of cranial and other nerves, which can provide an opportunity for chorda tympani activation to occur alongside stimulation of facial or other nerve branches (e.g., trigeminal nerve) (35). Using high-density electroencephalography, Frehlick et al. reported that TLNS significantly changed resting-state brain activity (36). Additionally, high-resolution image processing has allowed observation of changes in functional activity in the brainstem, pons, and cerebellum, which are major sensory integration and movement control centers of the brain (34, 37-39). Together, these data help explain the potential link between the physiological mechanism of neurostimulation, the underlying neural changes, and any clinical effects that may be observed.

The results from clinical reports and preliminary studies have indicated that TLNS plus targeted PT can significantly improve patient outcomes (31, 34, 37, 40-48). Improvements in balance and gait function have specifically been noted in patients with cerebellar degeneration (42), chronic multiple sclerosis (MS) (43, 44), spinal cord injury (45), and stroke (46). In addition, the results from two open-label TLNS studies indicated that noninvasive cranial nerve neurostimulation can improve dynamic gait index (DGI), activities-specific balance confidence scale, and dizziness handicap inventory scores in patients with impaired balance or peripheral and central vestibular loss (38, 47). Interpretation of results from these studies, although informative, is limited by small sample sizes, lack of randomization, the absence of blinding, and variability in treatment schedules and follow-up (48).

A 26-week study of 44 participants was performed at the University of Wisconsin–Madison, comparing the use of high- or low-frequency TLNS plus PT in an mmTBI population that was greater than or equal to one year from their injury and continued to demonstrate a balance deficit after conventional PT (41). Participants engaged in a two-week in-clinic training program, followed by 12 weeks of at-home progressive training and then a 12-week withdrawal period (no training or TLNS). This study reported significant improvement in both treatment groups from baseline to weeks 2, 5, 14, and 26 for balance (based on sensory organization test [SOT] composite score and DGI). The scores during the withdrawal period demonstrated that the improvements achieved during the 12 weeks of training intervention were sustained. Interestingly, although both groups demonstrated significant improvement from baseline, the outcomes between the high- and low-frequency groups were not significantly different.

As described above, promising results have been reported from studies using TLNS plus targeted PT across various neurological conditions; however, many of these investigations are small studies or case reports or had participants with varying neurological conditions. To investigate the role of TLNS plus PT in a well-defined homogeneous population of patients with mmTBI, two studies were designed: the 26-week study noted above (41) and the large, multicenter, double-blind randomized study (NCT02429167) described here. The goal of the latter was to evaluate the safety and effectiveness of TLNS plus targeted PT in people with a chronic balance deficit after an mmTBI who had plateaued with a prior conventional PT program.

MATERIALS AND METHODS Study Design

This multicenter, double-blind, randomized study (NCT02429167; clinicaltrials.gov) was conducted at seven sites in the United States and Canada from August 2015 to September 2017. Institutional review boards at each site approved the study protocol, and the trial was conducted in accordance with the Declaration of Helsinki and the International Council for Harmonization Harmonized Tripartite Guidelines for Good Clinical Practice (see Supporting Information Methods). All participants provided written consent before participation. Randomization was performed by the sponsor-designated clinical research organization according to generally accepted standards to ensure blinding of the participants and treating clinicians. The participants, TLNS trainers, and investigators were all blinded to treatment group. A blinding assessment was conducted at the end of the study to assess success.

Participants

Participants were eligible if they were aged 18–65 years, had a documented mmTBI diagnosis greater than or equal to one year prior to enrollment, and had a balance deficit with a NeuroCom® SOT composite score ≥16 points below normal (based on normative data). All participants had to have undergone a targeted PT program focused on balance/gait deficit due to TBI and reached a plateau in recovery (according to their healthcare providers). Prior PT varied, and participants were not required to submit specifics of the previous program, only that they had a persisting balance deficit (as described above) after undergoing PT. Enrolled patients also had to have a “negative” neuroradiologic scan and report greater than or equal to one year after their most recent TBI. “Negative” neuroradiologic scans were defined as no evidence of refractory subdural hematomas, tumors, significant anatomical anomalies, or loss of gray matter, per clinical judgment of the neuroradiologist.

Participants were excluded if they had oral health problems, nonremovable metal orthodontic devices, oral cavity piercings that could interfere with TLNS use, chronic infectious disease, uncontrolled hypertension, diabetes, other neurological disorders (not attributed to their primary diagnosis), cancer treatment (other than basal cell carcinoma) within the past year, penetrating injury, craniotomy, or refractory subdural hematoma. Participants with long-term use of psychoactive medications that would compromise the participant's ability to comprehend and perform study activities per investigators and those with pacemakers or elevated cardiovascular risk were also excluded. Individuals with severe symptoms (see Supporting Information Methods) were also excluded. Any changes in medication, either for TBI-related or other symptoms, or any use of other investigational device or medication lesser than or equal to three months prior to enrollment were temporary exclusions and subject to study physician discretion.

Treatment

As previously described (31, 49), the TLNS device delivers 19-volt amplitude-controlled, pulse-width modulated, unbalanced biphasic pulses to the anterior superior surface of the tongue through 143 gold-plated electrodes on a polyimide substrate (see Supporting Information Methods). A zero net direct current minimizes the potential for tissue irritation. Participants (N = 122) were randomized in a 1:1 ratio to receive TLNS with either high-frequency pulse (HFP) or low-frequency pulse (LFP) stimulation in combination with PT. The HFP emitted triplets of 0.4 μs to 60 μs wide pulses (five ms intervals; 150 pulses/s), and the LFP used a modified stimulus waveform pattern of 12.5 s intervals to each electrode (0.08 pulses/s). The HFP/LFP stimulation ratio is 1875:1. The stimulus intensity on the device was set to a comfortable level according to the feedback provided by each participant and was fixed at this level for the study duration. The sensation of the stimulus has been reported to feel similar to a carbonated beverage, and reports of taste sensations are infrequent (31).

The treatment program consisted of five weeks of training using the TLNS device in 20-minute sessions for 100–120 minutes/day over two stages. Stage 1 was a two-week in-clinic training during which participants worked with a therapist who was certified in use of the device: twice daily for one-hour sessions to perform the different training modules (two balance, two gait, one warm-up, one movement control exercise, one breathing and awareness training [BAT]), followed by a BAT session completed independently at home. The same regimen was performed in stage 2, with the exclusion of the morning BAT session. Stage 2 was a three-week at-home training program and participants returned to the clinic for weekly supervised training during which the PT program was adjusted. Participants were assessed at baseline and after two and five weeks. All participants followed the same TLNS use and PT regimen; training intensity was customized to the individual's presentation and abilities. Adherence was automatically monitored through the TLNS device by logging usage; these data were downloaded and verified weekly, and compliance rates were determined.

Endpoints

A summary of endpoints and measurements is shown in Table 1. The primary efficacy endpoint was the proportion of SOT responders (defined by an improvement in SOT composite score of ≥15 points) after five weeks of treatment. An increase of greater than or equal to eight points in SOT composite score is considered to be clinically significant (50). There were two secondary efficacy endpoints: change in SOT composite score from baseline to the end of in-clinic training (week 2) and to the end of at-home training (week 5). The primary safety endpoint was the frequency of falls, as recorded by the participant on an electronic case report form (eCRF), which was filled out either at/near the time of each clinic visit or when test results were received; data were de-identified on the eCRF. The secondary safety endpoint was the frequency of headaches at baseline and after five weeks of treatment, measured by the headache disability index (HDI). Additional outcomes of interest were the six-minute walk test (6MWT), DGI, quality of life measure index (QLMI), and sleep quality index (SQI). The incidence of adverse events (AEs) were collected throughout the study, assessed and recorded at each study center visit. Severe AEs are those that are usually incapacitating and prevented the participant from performing normal daily activities; they typically required a systemic drug or other treatment. A serious AE was defined as any reaction that is life-threatening at the time of the event or that results in death, hospitalization, or significant/persistent disability.

Table 1. A Summary of Endpoints and Measurement. Primary effectiveness endpoint The proportion of responders at week 5 Secondary effectiveness endpoint Increase in SOT from baseline at weeks 2 and 5 Primary safety endpoint Frequency of falls during the five-week study* Secondary safety endpoint Frequency and severity of headaches measured by HDI at baseline and week 5 Other assessments of interest

6MWT distance at baseline, weeks 2 and 5

DGI at baseline, weeks 2 and 5

SQI at baseline and week 5

QoL measure at baseline and week 5

AEs throughout treatment

6MWT, six-minute walk test; DGI, dynamic gait index; HDI, headache disability index; QoL, quality of life; SOT, sensory organization test; SQI, sleep quality index. Data Analysis

The intent-to-treat population, the primary population for efficacy analyses, included all participants who were randomized into HFP or LFP treatment groups (see Supporting Information Methods). The safety population included all participants receiving greater than or equal to one treatment in either group.

Given reports of positive responses to both high- and low-level stimulation in vagus nerve studies (51, 52), transcutaneous electrical nerve stimulation (TENS) (53-56), and transcranial magnetic stimulation (57, 58), it was considered possible that both the HFP and LFP treatment groups would respond to TLNS treatment in this study. If there was no significant between-group difference for the primary efficacy endpoint, the two treatment arms were to be pooled for preplanned statistical analyses of the primary and secondary efficacy endpoints.

One interim effectiveness analysis was planned for when 30 participants in each study arm had completed the five-week treatment and was conducted by an unmasked independent statistician. If the results were significantly higher in the HFP + PT than in the LFP + PT group (p ≤ 0.005), the study would be terminated. If p > 0.005, the study would be continued until all participants had been enrolled.

A post hoc statistical analysis of additional variables of interest (i.e., 6MWT, QLMI, SQI) was conducted. Least squares mean values were determined at baseline and at week 5 for each measure in each treatment group. The difference in these values was calculated along with a p value to assess significance between the two time points. A score decrease in HDI or SQI or a score increase in QLMI was considered a clinical improvement.

To evaluate the effect of different covariables on changes in SOT composite score, two post hoc analyses were performed: a logistic regression analysis of 88 covariates (see Supporting Information Methods) and a multivariate logistic regression analysis (MVA). Covariates in the MVA included age, sex, body mass index, prior PT duration, time since TBI, baseline 6MWT distance, and baseline postconcussion total score.

Sample Size Determination

Determination of sample size was based on results from a small pilot study of participants who received HFP + PT. There were no available data to predict the magnitude of the difference between results for the HFP + PT and LFP + PT groups when this randomized study began; therefore, it was assumed that 90% of patients in the HFP + PT group and 60% of those in the LFP + PT group would be responders. Under these assumptions, a two-group continuity-corrected χ2 test with a 0.05 two-sided significance level indicated that 90% power to detect a between-group difference in responder rate for the HFP + PT and LFP + PT groups could be achieved with a sample size of 49 in each arm. The target enrollment for each arm was set at 60 to account for a projected 20% dropout rate.

RESULTS

A total of 4303 participants from across seven sites in the United States and Canada were screened via phone interview followed by 210 live screens. Inclusion criteria were met by 122 participants, who were randomized into two TLNS treatment groups: HFP + PT (n = 59) and LFP + PT (n = 63; Fig. 1). The median time from the qualifying injury to enrollment in the overall population was 5.7 years (range, 1–33 years). Prior to study entry, participants underwent a mean 8.8 months (range, 1–60 months) of outpatient PT for TBI-related symptoms. Rehabilitation programs varied in their inclusion of components such as exercise, balance training, gait training, vestibular therapy, work adaptations, ocular training, and neuromuscular reeducation. Detailed demographic characteristics of study participants are summarized in Table 2. At the study start, participants had common ongoing symptoms that included headache/migraine (n = 74), dizziness (n = 45), fatigue (n = 40), and concentration/memory difficulties (n = 29). A total of 115 participants had data available at the five-week assessment. Overall compliance measured by the TLNS device was a mean of 94% across weeks 2 through 5 of the study.

CONSORT flow diagram. HFP, high-frequency pulse; LFP, low-frequency pulse; PT, physical therapy.

Table 2. Patient Demographic and Clinical Characteristics. Characteristics HFP + PT (n = 59) LFP + PT (n = 63) Mean age, years (range) 48.9 (21–63) 43.8 (18–64) Male, n (%) 16 (27) 23 (37) Female, n (%) 43 (73) 40 (63) Race, total population (N = 122), n (%) African American/Black 6 (5) American Indian/Alaskan Native 1 (<1) Asian 3 (2) Native Hawaiian/Pacific Islander 2 (2) White 110 (90) HFP, high-frequency pulse; LFP, low-frequency pulse; PT, physical therapy. Efficacy Endpoints

The proportion of participants who were SOT responders was significant in both HFP + PT (71.2%, n = 42) and LFP + PT (63.5%, n = 40) groups compared with baseline (p < 0.0005). There was no significant difference between the responder rates for the HFP + PT and LFP + PT groups (p = 0.37). In accordance with the predefined statistical analysis plan, the two treatment groups were pooled, which resulted in an overall responder rate of 67.2% (n = 82) compared with baseline (p < 0.0005; Fig. 2).

SOT responders at five weeks. SOT responder rate in the HFP + PT, LFP + PT, and pooled groups at five weeks. Responders were defined as participants who had an improvement of ≥15 points in SOT score from baseline. *p < 0.0005 vs baseline. HFP, high-frequency pulse; LFP, low-frequency pulse; PT, physical therapy; SOT, sensory organization test.

Results for the combined HFP + PT and LFP + PT populations indicated a clinically and statistically significant improvement in mean SOT composite scores from baseline to two and five weeks of treatment (Fig. 3). Mean SOT composite score (± standard deviation) for the pooled population was 18.3 ± 16.8 points (p < 0.0005) at two weeks and 24.6 ± 18.8 points (p < 0.0005) at five weeks. Mean DGI scores in both treatment groups were significantly increased from baseline at weeks 2 and 5 (p < 0.0001; Fig. 4).

Mean increase in SOT composite score from baseline in the pooled population (HFP + PT and LFP + PT). Mean changes (+SD) from baseline to weeks 2 and 5 in SOT composite score for the pooled group. *p < 0.0005 vs baseline. HFP, high-frequency pulse; LFP, low-frequency pulse; SD, standard deviation; SOT, sensory organization test.

Mean increase in DGI score from baseline. The mean DGI score at weeks 2 and 5 was calculated for both the HFP + PT and LFP + PT treatment arms. *p < 0.0001 vs baseline. DGI, dynamic gait index; HFP, high-frequency pulse; LFP, low-frequency pulse; PT, physical therapy.

The change in least squares mean values of the 6MWT, QLMI, and SQI from baseline to week 5 was measured (Supplemental Table 1). HDI and SQI scores significantly decreased and QLMI scores significantly increased from baseline to week 5 in both treatment groups.

A post hoc logistic regression analysis identified seven covariates with potential to predict an SOT response. No logistic regression estimates were >1.0, and most odds ratio CIs contained 1.0 (Supplemental Table 2). An MVA was also performed. This analysis showed that the variables (i.e., age, sex, body mass index, baseline 6MWT distance, and baseline post-concussion total score), including time since TBI, had little or no predictive power for those who responded to treatment (Supplemental Table 3). The duration of prior PT also did not have any predictive power for improvement in SOT composite score (Fig. 5).

Change in SOT composite score from baseline over the duration of prior PT. The pooled population of HFP + PT and LFP + PT groups were analyzed through linear regression analysis (orange line) and locally weighted scatter plot smoothing (LOESS; blue line). The corresponding shaded areas represent confidence bands for the linear regression and LOESS lines; fit of the solid line within the associated confidence band suggests there is no likely relationship between the observed change in SOT composite score from baseline and the duration of prior PT. LOESS, locally weighted scatterplot smoothing; PT, physical therapy; SOT, sensory organization test.

Safety

Both the HFP + PT and LFP + PT groups had a reduction in the number of falls from baseline (Fig. 6a) and a reduction in HDI scores from baseline (Fig. 6b) over the course of treatment.

Reduction in falls and in HDI scores from baseline. Reduction in total number of falls (a) and changes in mean HDI (+ SD) scores (b) from baseline and at five weeks were calculated for each treatment arm. HDI, headache disability index; HFP, high-frequency pulse; LFP, low-frequency pulse; PT, physical therapy; SD, standard deviation.

A total of 841 AEs were reported, 279 of which were device related. Twenty-two AEs across 12 participants were deemed “definitely” related to the TLNS device and were all considered mild (Table 3). There were 24 severe treatment-related AEs reported across nine participants during the study; these AEs resolved, and each participant continued in the study. None of these severe AEs was considered related to the device. The most common severe AEs were multiple episodes of vomiting (n = 5), nausea (n = 5), fainting (n = 2), and worsening of nausea (n = 2). No serious AEs were reported.

Table 3. TLNS Device-Related AEs. TLNS-related AEs with >1 episode* Number of episodes Burning sensation on the tongue 6 Tongue tingling 4 Sore tongue 3 Decrease in tongue sensation 2 AE, adverse event; TLNS, translingual neurostimulation. DISCUSSION

This study is the largest multisite randomized trial in patients with a chronic balance deficit after mmTBI who had plateaued on their previous rehabilitation program. Results from this study indicate that TLNS plus targeted PT with either HFP or LFP stimulation had a significant benefit in this population. Improvements from baseline in both the SOT composite score and DGI from both groups were statistically significant and clinically meaningful, defined as a minimal eight-point increase in the SOT composite score and a minimal 2.9-point increase in DGI score, respectively (59, 60). Responder analysis in this study used 15 points as a cutoff threshold to demonstrate improvement in the SOT, almost double what has been demonstrated to be a clinically meaningful change (50, 59).

The statistical analysis plan stipulated that if the results for the two groups were not statistically different, they would be pooled for analysis. The responder rate for the pooled group significantly improved from baseline (67.2%, p < 0.0005). Furthermore, the combined analysis showed that the use of TLNS + PT resulted in significant improvements in SOT composite scores after two and five weeks of treatment (both p < 0.0005).

The lack of significant difference for SOT composite scores between the dosage groups was, in part, attributed to high response rates occurring in both the LFP + PT group (63.5%) as well as the HFP + PT group (71.2%). At the time of study development, there were no previous data for participants with LFP stimulation (the 26-week study [41] had just been initiated), and three potential outcomes were considered for group comparison in this study: neither group responded, one responded significantly better than the other, or the groups produced comparable responses. Having a treatment arm with a low-level stimulation would be preferred to a placebo device without stimulation, as a device with no perceived stimulus may unblind a participant as being in the control group. However, although a low-level stimulation group is necessary in neurostimulation studies, positive responses in this group may limit the statistical superiority of high-level stimulation (29).

After this study protocol was developed and registered, investigations with other neurostimulation devices reported difficulties with low, minimally perceived sham stimulation and little or no difference in treatment effect between high- and low-frequency stimulation (52-58, 61). These findings parallel the positive results observed with the LFP + PT group and suggest that neural activity and clinical responses may still occur despite stimulation parameters being considerably lower than the hypothesized active treatment. Future research will help assess the dosing parameters of TLNS, the degree to which PT contributes to the outcomes, and long-term benefits of the treatment.

Placebo effect, Hawthorne effect, and nonspecific attention and care should be discussed here as they are often recognized as factors that may impact the outcomes of neurological disorders in response to treatments (62). In fact, a heightened placebo effect is often recognized with medical devices (63). Studies with TENS reported 40%–60% placebo response rates (64-66), which challenge the accurate interpretation of the efficacy of TENS treatment. Placebo effect influences the statistical power of a study because of the elevated baseline measurement; therefore, sufficient trial sample size calculated according to realistic estimates of the differences between the expected effects of the treatment and control arms is always recommended (67). The sample size was powered by the statistical assumption that 90% of patients in the HFP + PT group and 60% of those in the LFP + PT group would be responders.

Placebo effect may also result from a favorable view of one's disease condition. By believing that the post-TBI symptoms are transient or improving, patients may put forth greater effort in therapy and continue social activities or form expectations based on experiences/knowledge of previous treatment with a similar device (63). In this study, all participants had chronic post-TBI symptoms that had not completely resolved with prior PT. Yet, with TLNS plus targeted PT, many participants had highly meaningful improvements in balance and gait. In addition, an assessment was conducted at the end of the study and demonstrated adequate blinding. Furthermore, in animal studies, in which “participant” expectations regarding potential treatment benefit are unlikely to influence study results, trigeminal nerve stimulation of different electrical intensities positively impacted outcomes with stroke, TBI, and hemorrhagic shock and hypoxia in rats, with seemingly little difference between different-intensity groups (68, 69). In summary, the placebo effect in this study, if any, was likely negligible.

Additionally, it is improbable that the significant benefits observed in both groups in this study were associated with spontaneous recovery. Study participants plateaued on a previous PT program and had a long interval between their most recent TBI and study enrollment (a mean interval of 5.7 years for the total population); because PT is based on individual response, there was variance in the amount of time patients participated in their respective therapeutic programs. Given the significant outcomes presented here and the lack of improvement in participants in the lengthy period prior to study enrollment, we speculate that PT enhanced with TLNS contributed to the clinically meaningful benefits in both groups; although the degree of individual contribution of TLNS or PT to the participant remains to be determined in future investigations.

Another point of discussion pertains to whether the effects of physiotherapy without TLNS can be estimated. Because patients had failed to achieve normal balance in previous PT programs, it would have been difficult to justify enrolling a sufficient number of participants for a PT alone arm; therefore, to estimate the effect of PT alone, results from the present study were compared to those in a study by Badke et al. (see Supporting Information Additional Efficacy Analyses), in which participants were treated with vestibular therapy alone (70). The goal of this comparison was to estimate the difference in observed improvement in SOT scores between a combination of TLNS plus targeted PT and PT alone. In the study by Badke et al., participants with central or peripheral vestibular dysfunction, as well as with balance dysfunction due to central nervous system pathology, were enrolled (70). They received vestibular therapy similar in duration and intensity to the PT received in the present TLNS study. The analysis showed that the magnitude of SOT improvement achieved in the present study was significantly higher than those reported by Badke et al. Although the intervention used in the latter study by Badke et al. was specifically vestibular rehabilitation, this treatment approach is used often for balance therapy (71) and is considered a reasonable approximation of the PT-only approach.

In the Badke et al. study, results were not statistically different before and after therapy (70): an average increase of 6.4 ± 3.8 in the SOT composite score was observed for the peripheral group and 11.2 ± 9.9 for the mixed/central group. In contrast, in our TLNS study, the SOT composite score was increased by 26.5 ± 20.8 in the HFP + PT group and by 22.9 ± 16.5 in the LFP + PT group. Statistically significant differences in the improvement of SOT composite scores were observed for both HFP + PT (p = 0.0026) and LFP + PT (p = 0.0133) groups when compared with Badke's treatment group. In support of this analysis, systematic reviews of the effectiveness of PT put forth limited evidence of the positive effects of exercise intervention alone for the improvement of balance and gait in patients after TBI (13, 72). Furthermore, the independent therapeutic effect of TLNS has been observed in patients with MS at various stages. In a pilot study by Leonard et al., participants either received TLNS treatment similar to the HFP + PT group described here or received PT in combination with a placebo device (control group) with no perceptible stimulation (43). Results showed significant improvements in SOT score from baseline to week 14 in the HFP + PT group but not in the control group.

To analyze the potential association between certain baseline patient characteristics and study outcomes, two post hoc statistical analyses were conducted. Results indicated that there was no single variable with large predictive value for SOT response in either treatment group; in addition, previous PT duration seems to have the weakest explanatory power for the response observed.

TBI treatment efficacy and its variability is another point that merits discussion. Investigation of TBI treatment efficacy in clinical trials is hindered, in part, by recruitment of a heterogeneous population that includes participants with various TBI severity classifications and diverse levels of function (72). Additionally, TBI studies often report on treatment of acute and chronic TBI symptoms (72). Strict enrollment criteria were used for this study to provide a more homogeneous population of participants who had a clearly defined functional level. Importantly, this study included participants suffering from persistent symptoms, which is a population believed to be less likely to fully recover and who are often underrepresented in TBI studies (73). Although this inclusion did result in several screened patients being ineligible for this study, over 100 participants were enrolled and treated. Hopefully, this study will lay the groundwork for additional investigations with this treatment modality, which can possibly include a broader population of patients with TBI or other neurological deficits associated with balance impairment.

Of additional note, in this study, there were two times more female than male participants. In the general population, men typically experience TBIs at an incidence rate 1.4-times that of women (74); however, for sports-related concussions specifically, when comparing the